# MetalicaRapBaby

Overwiev description of the prototype, actual builders should go to builders only section below.

MetalicaRapBaby in it's current form has 8 main elements which form the printer hardware proper:

• Triode electron gun within a 4 way reducer cross.
• Electron optics assemply including deflection coils and magnetic lenses.
• A metal powder dispensing trough hopper, with a cartesian 8cm diameter topological pick up ring.
• Build platform in a 14 inch dia. build platform tube within outer glass tube chamber.
• SEM for vision system for feedback regarding the shape of the part, using the 12kg trough, hopper pick up ring.
• One roughing pump capable of 10-4 Torr
• Switch Mode Power Supply -63kV up to 5kW series parallel resonant converter topology.
A low power gun build example
Low power gun in operation.. See [1]
Electron Gun Elements
Coil wiring diagram

### Electron Gun

Static electron gun 1-5kW max (-63kV accelerating voltage) output approximately 1m above the build platform. The gun has a small filament chamber and a Whenelt electrode, i.e. a triode gun. A triode gun contains an anode with a hole in it, to accelerate the electrons from the filament. This anode will function as an electrostatic lens and form an image of the filament close to the gun. This image is known as the cross over. Out current design of the electron gun is based on an analytical model from Potapkin (2008). With our chosen parameters the the diameter of the cross over is about 120µ. This cross over will have to demagnified and reimaged onto the work surface by electron optics.

• Electron Gun constructed from self printed Stainless steel 304L 6" 4.5" inch 4 way crosses reducer with ID 100.4mm beam pipe. Beam pipe terminates in joining adapter to glass chambers.

The poteltial of the Wenelt electrode will be set by feedback through a resistor to the cathode to make the output form the electron gun self-stabilising. To set the dimension for the power supply and the feedback resistors we need a rough estimate of the perveance (current-voltage relation I=P*U^{3/2}) of the electron gun. The gun is to first order a triode, one of the mainstays of electronics in the first half of the 20th century (amplifier tubes). Hence there excists lots of literature on how to calculate the perveance of a triode. The number of electrons emitted from the cathod themionicaly is given by the Dushman Equation which is exponential with the temperature of the cathode. But must of these themionically emited electrons are pushed back by the electric field from space charge, so the important quantity is the space charge limited current, which is must often descriped with the Child-Langimur equation. This equation strightly only applys to diodes and we have to adapt it for our pourpose with is a triode with the Whenelt beeing the grid. Currently the distance from the cathode to the Wenelt is 1.175mm and the distance form the Whenelt to the anode is 29.3mm. The diameter of the cathode is 2mm.(Check I think the thickness of Wenelt is 1.175mm and its 0.3mm from cathode?)

In the paper (VanDeBroek1986)on the emitting disk ratio the emmiting disk ratio is given in terms of the drive which is 1-(V_g1/V_co) (equation(1) VanDeBroek1986). Where V_co is the cut off potential (zero field strength right at the center of the filament, emitting area zero) which can be measured experimentially for a given electron gun. It can also be calculated from the Durchgriff, but the paper does not give the coefficeints for a triode with the ratios that we have designed. But the Durchgriff is probably somewhere between 0.02 and 0.01 times V_acc (Table 1 VanDeBroek1986). Some where between 600 and 1200V. At half the V_co we can expect to have approximately half of the emmitter area available, from which we can calculate the space charge limited current form Child's law.

Child's law says that for a diode the current density is given by j[A/cm^2] = 2.33e-6 V^(3/2)[V]/d^2[cm] [5]. Where d is the distance between the cathode and the anode and V is the voltage across. For our case we get j=4A/cm^2. For a 2mm diameter cathode where all of it emits this gives current about I=0.1A. But for our desired optics which can achieve a small gun cross over diameter of 120µ Potapkin(2008) requires a wehnelt electrode nearly at cutoff voltage(i.e 97%) [Important note U_c in Potapkins paper probably means the cut-off voltage]. This means that emitting area of cathode is reduced, from fig 3(VanDeBroek1986) drive of 0.03 so emting disk ratio is 0.1, So that at our operating conditions we have about 10 percent of the emitting radius avaliable we get a radius of 0.1mm which corresponds to an area of 0.0003cm^2 and a limiting current of about 0.00125A allowing for 15% loss of beam in to anode. So the 47µ spot size beam would be at 65W. Assuming we desire this 1000V cut off we need a 2.4KOhm resistor between wehnelt and cathode.(Accepting L/R=10.99 in fig3 VanDeBroek1986))

Spangenberg page 416, formular 15.4 gives a way of calculating the amplification mu of en electron gun. Doing the calculation with the numbers of the current design we get mu=367. In formula 15.3 Spangenberg gives that the ratio of the Whenelt (V_2) and anode (V_3) potentials is related to mu at the cut-off mu=-V_3/V_2. Given this mu and V_3=62.5kV we get a cut-off potential of -170V.

Currently 2x 430K ohm in parrallel or 215Kohm for 940W gun power level and 4x 430Kohm in series or 1720Kohm for 100W power level ( maintaining 3KV across resistor in both cases).

### Electron Optics Assembly

The electron optics assembly reimage and demagnify the electron gun crossover onto the build platform to create the electron beam focal spot for melting metal powder for additive printing and melting targets for solar cell EBPVD process. The focus spot size should be 100µ for printing and 10µ for SEM imaging (to achive 20µ resoluton ie. the Nyquist limit), with a pointing accuracy 10µ.

Additionally the focus spot size for providing micro vaporization of errors in the build should be in the range 10µ to 40µ, see gun layout page 3. Initially this would be used to flatten every tenth layer of the build through high points/blobs removal. Future software development could provide live modeling of build, so through the adjustment of beam melting path in the subsequent layers errors/blobs could be accounted for as they arise, thereby limiting the need for blob removal to critical surfaces.

The electron optics assembly will take the form of a two lens demagnifying telescope with an interlens deflection system for moving the image. To first order electron optics can be designed as standard geometric optics, with a few modifications. We will need to reimage and demagnify the cross over onto the printing surface. It is important to note that rays around the crossover are not straight lines, only asymtotically so. This means that to consider the crossover in terms of geometric optics, we have to use the asymtotic position of the crossover, that is, the focal point where the asymtotically straight rays meet, when projected backwards.

The smalles spot size need is around 10µ, which means that out optics will have to demagnify by a factor of about 0.1 given the cross over diameter of 120µ. The needed demagnification can be distributed over the two lenses and the total magnification will be given by M = M_1 M_2, where M is found from the thin lens formula 1/s_o + 1/s_i = 1/f and M = -s_i/s_o. We see that for a lens to be demagnifying M<1 the distance to the image (s_i) will be smaller that the conjugate distance to the object (s_o). In practical terms this means that it will be hard to have L2 be de-magnifying, it will have to be neutral (M=1) or magnifiying in order to leave room in the vacuum chamber and not having to bend the beam excessivly to achive a large printing area. Hence we will have to put all the de-magnification on L1. This leads to the complication of having a very divergent beam between L1 and L2, implying that the diameter of L2 will have to very much larger than the diameter of L1. The situation is illustrated in the figure. At precent we do not know how large diameter magnetic lenses are feasible? But figure 4.22 in W.D Riecke (1982) showes a realised lens with a diameter of 8cm.

Overview of the optical design. CO = crossover, L1 is the first lens, the red dots the focal points of L1. L2 is the second lens with blue focal points. Im is the final image. The magnification of L2 is one.
Overview of the quantities used for calculating the necessary magnetic field in the deflection coils.

Another possibility would be an adjustable aperture in the beam. We note that the small spot (10µ) is only needed for SEM imaging, hence it will not require the energy of the full beam, we can afford to lose energy. We could then design the electron optics assembly to have a magnification M on the range 0.1 to 1, and then stop down the beam with the adjustable aperture when the small spot is needed. This procedure could also possibly reduce the aberrations when low aberrations are most need, ie. for SEM imaging. Maybe this could also be done by varying the lens current, hence changing the focal length?

Assuming a two lens design we can write M_1 s_{o1} = - s_{i1} and M_2 s_{o2} = - s_{i2} for the two lenses respectively. (The magnification M is negative). We have some further constrains that we would like to add to the system. First the image from the first lens is the object of the second lens. Assuming a separation l between the two lenses we have l-s_{i1}=s_{o2} and we want a given total magnification M given by M=M_1 M_2. Further we could want to constrain the apature of L2.

Given the maximum deflection of the beam required to realize a printing area of 24cm by 24cm is 14.65 degrees measured at L2 and 16mm is beam diameter of at max deflection 151mm before L2 and the baffel is situated at 30mm after L1.

One appealing solution is M_1 = 0.3 \ M_2=1.25 \ s_{o1}=100mm \ s_{i1}=30mm \ {o2}= 459 mm \ s_{i2} = 573 mm. With this design the aperture radii of L1 and L2 have to be at least 3.8mm and 27.15mm, respectively but due to inter lens XY deflection of beam further space is required prior to L2. So minimum pipe diameter is 83mm thereby giving the space needed to deflect the 16mm diameter beam within to achieve 14.65 degrees deflection at L2. For this Print mode asymtotic focal lengths we get f_1 = 23.1mm and f_2=255mm. This design will give a total magnification of 0.38, which corresponds to a focal spot size of 45µ, which is good enough for printing but too large for the vision requirements. By changing the magnifications to M_1 = 0.08 \ M_2=1.2 with out changing any lengths we will achieve a total magnification of 0.096, corresponding to a spot size of 11µ which is sufficient for the SEM vision system. As lenses and baffel lengths stay the same, in this SEM configuration the intermediate image will be before the baffel instead at it, as the aperture radius of the baffel is 0.12mm, and the size the beam is 0.37mm going in to this aperture, we will lose 67% of beam power in to the baffel, but this should be acceptable or even preferably in the case of SEM imaging. The needed SEM mode asymtotic focal lenghts in this case are f_1 = 7.4mm and f_2 = 261mm.

Compared to optical lenses magnetic lenses are very poor with large abarations. In essence we are building an SEM, and lots of research has been done on magnetic lenses for this pourpose. SEM operates with spot sizes on the order of 1µ or less, that is, 1 or two orders of magnitudes less than what we are aming for. Hence we do not have to optimise the design to the same extend as SEM designers do. One of the fundamental parameters in magnetic lens design is the gap to bore ratio S/D. According to W.D Riecke (1982) p. 179 we are free to choose this ratio is in the range 0.5 to 2, preferably in the range 0.7 to 1.5.

The pole pices in an magnetic lens are tuncated cones the diameter at the truncations sould be 3 to 4 times the diameter of the bore, and the conic half angle should be 55 degrees. W.D Riecke (1982) p. 255.

The magnetic lenses we have their magnetic flux carried on pole pices and cores. One important requirement is that we do not saturate the magnetic cores and theis will couse a large leak in magnetic field which will distrupt the magnetic lens. IN a classical magnetic lens the cores is connected to two flanges (top and bottom) which connects to the magnetic shielding which incases the electric coils. Theese two "flanges" will go through the pipe which contains the EOA. The minimum thickness can becalculated acording to formular 4.119 in Riecke (1982), given the need excitation and the saturation flux of the flange material. h_0 = \frac{\mu_0 NI}{B_0 |ln(tan(\theta_0/2))|}

For soft iron we will assume a saturation B_0 = 1T and the cone angle is on the order of 50 degrees. So all four flanges needs to be at least 3mm thick. The minimum thickness of the encasing can be calculated from 4.125. The inductance of the proposed coil designs, assuming a an iron core, will be on the order of 1H, which is significant. Care will have to be taken to avoid transients, which could generate very high voltages and arching in the coil.

Simple formula to calculate the temperature increase in the coils. NI=\sqrt{C_T h_{0}^3 sigma q \deltaT} \sqrt{eta +1/eta}, where C_T is the cooling efficiency of the coil, sigma is the specific electrical conductivity of the wire material; q=Na/A is the space factor of the coil windings, where a is the actual cross-sectional area of the copper wire employed; h_0 represents a charateristic length for the size of the coil cross-section and is defined by h^2_0 =A; and eta=h/h_0 specifies the shape of the coil cross-section with h as the extension of the coil in the direction of the lens axis. If b is the radial extencion of the cross-section of the (rectangular) coil, its 'aspect ration' h/b can be easily seen to be conected with eta by h/b = eta^2. The value eta=1 indicates a square cross-section. This formula can also be expressed in current density j_w=\sqrt{C_T sigma q (\deltaT/h_{0})} \sqrt{eta +1/eta}.

It is not possible to change the cooling efficiency by very much by changing the aspect ratio of the coil cross-section. The space factor of the windings is generally about q=0.65. The value C_T is determined by the general properties of the coil and the cooling mechanism. If water cooling is incorporated in the external lens casing, C_T of 40-50Wm^{-2}C^{-1} can bee achieved corresponding to an average current density of j_w = 1.8A/mm^2 at a temperature rise of \delta T = 70C.

So for L1 SEM mode we can choose design with S/D=1 and D=8mm, hence f_1/D=0.925 (f_1=7.4mm SEM). So D/f_1 = 1.08 According to Lenz (1982) figure 3.11 we then need an reduced excitation of 8.7AV^(1/2). The reduced excitation is defined as (NI)/(\sqrt{U}), p172, where U is the beam voltage. That is, sqrt{62500}*8.7 = 2175turns.

In L1 Printing mode we need the asymtotic focal length of 23.08mm, this requires less turns hence f_1/D=2.89. So D/f_1 = 0.346 According to Lenz (1982) figure 3.11 we then need an reduced excitation of 5AV^(1/2)Thus we need a reduced excitation of 5*sqrt{62500}=5*250=1250 A-turns.

We choose the demagnifying lens SEM mode design as this demads more turns, yet will also operate in print mode as well but at a lower current .

For L2 in printing configuration we need a focal length of 255mm. At this position we need to accommodate the expanded beam and allow for delection of the beam in the lens. The needed lens diameter has been found to be 67mm at least.plus its 16mm beam thickness (See spread sheet K81) and clearance ie needs 83mm diameter. We would like to make this lens shorter to allow room for the deflection coils hence we will choose a S/D ratio of 0.7 D=90 so S=63mm. For L2 we found D/f_2 = 0.353 according to Lenz (1982) figure 3.11 which results in an reduced excitaiton of 5.0 or 1250A-turns.

So coil dimensions allowing 180% space for wire packing density.

For demagnifying lens SEM mode assume that we need 2000A-turns. The max current density we can accept with a temperature increse of 70C is 1.8A/mm^2. We then need a coil cross-section area of 2000/1.8 = 1052.6mm^2. Corresponding to 32.4mm x 32.4mm. Assuming that we can drive a max of 1A through the coil, we need a wire cross-sectional area of 0.52mm^2 or a side length of 0.73mm. This corresponds to a AWG 20 or 21. [6] If we take AWG 20 each wire will take up an area of 0.81^2mm^2 = 0.65mm^2. For 2000 wires we then need 1312.2mm^2 of coil cross sectional area. This corresponds to 36.2mm by 36.2mm. or with an extra margin 25 mm x 55mm. 1375mm^2.

Demagnifyng lens Assume j_w = 1.8A/mm^2 max current/cm square is 180A/cm sq in coil, Assume wire crossection is square, and max current in an individual wire is 4A, therefore 67 wires per cm sq max., so round down to 8x8 wires in one cm sq diameter 10mm/52 =0.19mm from awg chart AWG 17 with varnish insulation, 2750 wires in probe forming lens so approx cross section area of coil 2750/64=43 sq cm 12 cm x 4cm. round up to 12 x 6 for tolerance.

For demagnifying lens Print mode assume that we need 1500A-turns. The max current density we can accept with a temperature increse of 70C is 1.8A/mm^2. We then need a coil cross-section area of 1500/1.8 = 789.5mm^2. Corresponding to 28.1mm x 28.1mm. Assuming that we can drive a max of 1A through the coil, we need a wire cross-sectional area of 0.52mm^2 or a side length of 0.73mm. This corresponds to a AWG 20 or 21. If we take AWG 20 each wire will take up an area of 0.81^2mm^2 = 0.65mm^2. For 1500 wires we then need 975.0mm^2 of coil cross sectional area. This corresponds to 31.2mm by 31.2mm or with an extra margin 25 mm x 55mm. 1375mm^2.

Probe forming lens Assume max current/cm square is 180A/cm sq in coil,Assume wire crossection is square, and max current in an individual wire is 4A, 67 wires per cm sq max. , , so round down to 8x8 wires in one cm sq diameter 10mm/8 =1.25mm from awg chart [7] AWG 17 with varnish insulation, 600 wires in probe forming lens so approx cross section area of coil 600/64=10 sq cm 4cm x 2.5cm.

(Note Later we should consider whether we can use a permanet magnet design, W.D Riecke (1982) p. 168 writes that permanent magnets are can provide magnetic potentials up about 3000 Oe cm or about 2400 ampere turns, which is sufficient up to a beam voltage of 50kV, ours beam voltage is 62kV, so how does these figure compare with the values for modern neodymium magnets? The magnets must remain under their Curie temperature of 320C.)

(Note reduced excitation is defined as (NI)/(\sqrt{U}) The minus disappears when the multiplication is turned into a division Then a reduced excitation of 8.7 AV^(1/2) corresponds to 2175A-turns.) (Note The asymtotic focus position will be 10.0mm form the lens midplane according to Lenz (1982) figure 3.12 from reduced excitment of 7.5 so from graph Zv/D =1.2, D=8 so Zv=1.2*8=10 . The real focus position will be about 9.1mm from the lens midplane acording to figure 3.10 as at reduced exicetment of 8.7, Zf /D =1.1 given D=8 so Zf=8*1.1=9.1. K: I currently have baffel at 29.3 not 9.1, in SEM mode the image should be before the appature. We have calculated abouve that the image sits at 30mm after the lens midplace in printing mode, so the apature placement should be alright.))

#### Simplificaton of the EOA design

Given that we have taken our design principles form SEM designs, where the spotsize requirements are much more stringent than for our pourpuses we might be able to get by with some further simplifications that will make the machine much easier to build. The pole pieces are introduced to reduce the abarations, but they might not be needed. Also the deflection system we have chosen is selected because it reduces abarations, by making the electron beam pass through the center of L2. We might be able to get by with having the delections coils sit after L2 outside of the pipe for the EOA. We the delection coils outside the pipe and no requirement for the pole pieces and choosing a thin non-magnetic pipe, (about 1mm thinkness, 0.5 should be enough to hold the vacuum, but it would be very hard to weld. 1mm is more feasible.) we could place the magnetic lenses outside the pipe. This would make it much easier to acess and adjust the lenses, i.e. they can then be slid up and down the pipe. The only thing we would need to be inside the pipe is a baffle. If we make the baffle sit at the intermediate corss over and we can isolate the baffle electrically we could use the current from the baffle to ground as a measure of focus (if outside of focus electrons will be absorbed by the baffle), which would make it much easier to focus L1.

#### Design of the deflection coils

Out first choice of deflection systems is called a prelens double deflection system in P.Hawkes & E.Kasper volume 2 page 824. This system consists of two deflectors D1 and D2 at a distance 2a and a in from of L2, respectively. D2 has twice the amount of windings and the opposite polarization of D1. In this system D1 will deflect the beam out of the optical axis and D2 will deflected it back in such a way that the beam passes through the center of L2 to reduce aberrations. The deflectors consists of rings of a high permeability material like ferrite, with windings whos normal is perpendicular to the optical axis. From expanding the magnetic field in the interior of the deflector in Fourier terms one learns that the coefficients can be written as a_k = \frac{4}{\pi}\sum_{i=1}^{n} N_i sin(k\theta_i). By carefully choosing the number and angles of the windings on the ferrite ring the odd coefficients can be made to vanish, reducing the aberrations from the deflectors. Hawkes & Kasper p. 839. For instance n=2 and N_1 = N_2 leads to \theta_1 = 48 degrees and \theta_2 = 72 degrees, see figure 40.13 in Hawkes & Kasper.

Calculation of the necessary field strength. Assuming the z axis to be the optical axis of the EOA, one would need a magnetic field purely in the x direction to deflect the beam in the y direction, i.e. the field is perpendicular to both the velocity of the electron and the direction of deflection. In this case the force on an electron can be written as F=evB. In the case where a force is perpendicular to the velocity one will obtain a circular motion, i.e. centripetal force. The radius of curvature will be given as R=(m/eB)\sqrt{2eV_acc/m}. With the definitions of the various lengths given in the figure to the right we can write the deflecton length x over the distance S as R=S^2 + x^2 / 2x \approx S^2/2x. Combining with the expression for the radius of curvature we get S^2/2x = (1/B)\sqrt{2mV_acc/e}.

In our current design we would like to deflect the beam about 20 degrees over a distance of approximately 2cm, corresponding to x=7mm. Hence we can estimate the required B field from the above formulas. B = (2x/S^2)\sqrt{2mV_acc/e}. Setting V_acc = 62keV and S=2cm and x=7mm we get B=0.015T.

For a circuit with an inductance the instantaneous current and voltage is given by i=I_p*sin(2*pi*f*t) and v=2*pi*f*L*cos(2*pi*f*t) where f is the frequency, I_p the peak current and L the inductance. To drive the coils at the frequency f we will then dissipate the power P=R*I_p^2*pi. And the peak voltage in the system will be V_p = 2*pi*f*L*I_p.

Experiments with a mock up model of the deflection coil configuration has shown that a high permitivity material like ferrite is required. Because of the high frequencies the coil is required operate at a low conductivity material like ferrite should be used. The experiments showed that about 2500A-turns are needed to obtain a field of 15mT in the center of the deflector. A ferrite coil with 2500 windings will have an inductance of at least a few H, which could present a problem. Driving this coil at 10000 Hz would mean we would have to handle peak voltages on the order of 100kV. We might need another design of the defection coils. Maybe akin to a set of Helmholtz coils. This would also mean we should change the deflector design form a one deflector design to a one deflector design like in a tv.

The expression for the field strength in the center of a pair of Helmholtz coils is B = (4/5)^(3/2)\mu_0NI/R. Setting R=4cm and B=0.01T we get NI=500. And because these coils are air coils we get much lower inductances. An air coil with a diameter of 8cm and 500 windings will have an inductance of about 50e-6H.

By making the deflection zones longer, we can lower the requirements for the B field strength. Making the deflection coils 8cm long instead of 2cm we can lower the required field strength by a factor of 4^2 = 16. That is we need a field of 1mT. For a set of Helmholtz coils to give 1mT at the center, when the radius of the coils is 5cm, we need about 100 ampere turns. Such a coil will have an inductance of a few microH. And we would have to handle peak voltages of about 1V.

Measurements in the lab has shown that for a traditional deflector design (not Helmholtz) we need 0.9A*2400 turns to achive a field strength of 1mT in the center 5 cm from the edge of the coils in the case of air coils. If we construct a yoke of laminated silicon iron we need 0.1 A*2400 turns to achive 1mT in the center. In this experiment the windings of the coils was 5cm by 5cm and the yoke was 2.5cm x 2.5cm.

Helmholtz deflection coil assume max current/cm square is 250A/cm sq in coil,Assume wire crossection is square, and max current in an individual wire is 0.5A, 250/. =500 wires per cm sq max. , , so round down to 22x22 wires in one cm sq diameter 10mm/22 =0.45mmsquare from awg chart [105] AWG 25 with varnish insulation, NI required 200 per coil , (400 wires per pair) ,each helmholtz 400 wires so approx cross sction area of coil 400/500=0.8 sq cm thickness 2.cm x width 0.4 cm. coil groove is 2cm by 1cm (skin depth at 20K .9mm diameter so ok). For 1 m from platform to probe forming lens 11 deg interlens deflection required pipe radius 16mm I.D.

### SEM Vision System

The 3d printed layer can be measured via a sensor pick up ring that's attached to gravity fed powder hopper just above the build surface with 2 powder wiper blades attached. The SEM Vision will be based topological mode SEM vision and provide µm resolution.

### Vacuum Chamber and Pump

• Glass Chambers MetalicaRapBaby (3D printer prototype version) and MetalicaRapLight EBPVD (solar cell printing version)

MetalicaRapBaby 3D printer: 3D print chamber based around large diameter borosilicate glass tube with aluminum plates on the top and bottom with L gasket seals (as used on vacuum bell jars). The bottom aluminum plate can be lowered to gain access to the attached build platform/motors/sensors and includes electrical feed through.A Stand will support the glass tube and the top aluminum plate with the gun and power supply attached via O rings above it, beam tube to gun/filament/anode chamber will need to be welded in mother chamber due to heat when printer is operating. Large diameter removable cement pipe used to shield around it all.

MetalicaRapLight EBPVD ( Solar version) chamber based around a square bottom flask with metal vapor trap. Glass chambers are connected to electron gun tube via top Aluminum plate Tube/Glass adapter; with 5 o rings, 2 for gun tube sealing and 3 for top aluminium plate to square bottomed flask sealing, having 1 o-ring for cushioning end of glass and two for sealing each joint.Large diameter removable cement pipe used to shield around it all.

• Pump is a roughing pump 2 stage Vane pump e.g.Leybold D4 D4B Trivac Rotary Vane Dual Stage 3.4cfm 2K euro new. to 1 x10-4 torr gas ballast off [8] [9][10] 1.0x$10^-$$^4$Torr )
• High vacuum $10^-$$^4$ Torr to $10^-$$^5$
• Electrical feed-troughs ; For tungsten filament AC 6V 15W 50/60Hz heater power supply based around modified microscope flat wound bulb's body(sm8018) giving a -63 KV Wehnelt feedthrough which is within a test tube providing -60KV cathode feed-through, using large diameter glass cylinder chamber with L gaskets TV feedthrough seal with o rings or o ring to copper tube then connect TV neck to copper tube. Glass tube to metal tube connection video[11].
• Mechanical feed-troughs ; will be avoided by using periodically replaced cheap standard Nema motors within the chamber.
• Viewing & illumination windows ; As main chambers will be glass no others are needed beyond.a window in the surrounding shielding box of standard glass 70mm thick can be multiple sheets no seal required as outside main large diameter glass cylinder.
• Pirani vacuum gauge home build version (use google translate) [12]
• Outer box for shielding constructed of cheap material 3cm of concrete (or solidified sand or earth brick).

### Metal powder dispenser

This will be a gravity fed split powder hopper.A small hopper will move in the x direction, intermittently refilled by main hopper situated besides electron gun tube. The small hopper will have a 12Kg powder capacity, this gravity fed powder hopper just above the build surface has two 1cm square section powder wiper blades attached wither side, leveling the roughly dispensed powder. The powder is released using two sliding slotted plates. A second main powder hopper will gravity feed the 12 kg hopper. The 12kg hopper will be sitting in a box created from a single 8cm thick sheet of aluminum self shaped to a box by MetalicaRap subtractive electron beam cutting.

Other dispensing considerations:

• clumping may be an obstacle at high vacuums, A better metal powder flow through the hopper could be achieved through a slight vacuum reduction during gravity hopper operation and then reevacuating again before bean melting resumes.
• SEM pickup PIN diodes protection cover will be used when printing.

### Build platform

A 30cm vertical travel stepper motor driven circular platform within a 14 inch 304 tube printed in parts. Within the build platform is a built in ceramic insulation layer. Two felt or titanium metal brush o- rings seal to keep the metal powder from falling through build platform and build platform cylinder.

### Power supply

Power supply Full Bridge LCC Series Resonant Converter with duty cycle switching variation control and above resonance frequency switching variation control, through this frequency variation maintaining Zero Current Switching during Coolmos switch ON and Zero Voltage switching during Coolmos OFF switching, with Arc sense, arc quench and arc count with output via a gun resistor, this gun resistor combined with the power supply itself will be used for beam current control .

High output series-parallel resonant DC-DC converter (search on this bold text string) -62.5KV 0 to 5KW running at 181.5KHz resonant up to 500Khz at idle/no load, with a -325V to -62.5KV transformer with voltage doubler on output, the transformer is designed for a specific value of inductance and capacitance to operate at the desired resonant frequency, as the load changes the parasitic capacitance of the transformer changes too, this means that at high load the series resonant converter topology dominates and light / no load the parallel resonant converter topology dominates. The change in load changes dramatically due to occasional arcing in the gun, This topology deals with these changes effectively. Each of the 178 PCB transformer secondary converter stages is a pcb with 1 turn coil track outputting -354 V d.c. via cheap rectifiers and voltage doublers/smoothing capacitors, when this is combined with the resonant tank gain of a range from 2.25 times to 6 times, the output voltage of 63KV is obtained. Along the stack of pcb's the voltage increases gradually keeping under the paschen air arc limit. Secondary converter stages are connected in series creating the -62.5 KV output,. Two C shaped transformer N87 ferite cores with 11 turn primary (54A) and a high voltage 178 PCB secondary insulated bobbin 4mm deep of UHMWPE seperates ferite core and PCB secondaries.

Power supply feedback control around transformer measures Output voltage feedback via 208 thick film resistors creating voltage divider. Output current sense feedback via measuring voltage either side of a temperature stable resistor. Input /Primary current sense via 2 "current transformers" in series in each leg of full bridge. Wehnelt voltage feed back via thick film voltage divider. These four signals go via signal reconditioning board then on to Field programmable gate array (FPGA) [13]via four channel analogue to digital converter (ADC) 12 bit 40Msps[14][15]. FPGA process maintains input and output current parity within acceptable current window values. Allows a small error signal from voltage output variations when within a narrower voltage output window values. FPGA drives 4 coolmos FET gates in full bridge topology via driver chip.

Solftware for the FPGA_High_Voltage_side is coding voltage feedback from linear to Log and then coding for sending over fibre optic to FPGA_Low_voltage_side. FPGA_Low_voltage_side software consists of feedback signal decode from optic fiber, Followed by low pass filter and PID controler d_controller and Digital pulse width modulator with 4 outputs to H bridge FETS. Along with a power supply user interface.

A separate housekeeping supply provides bias for all control circuitry, providing a 5V separate stand-by voltage which remains active when the power supply unit is shut down for any reason ( later once efficiency of power supply is assessed a Power Factor adjuster circuit may be added between rectifier and dc to dc converter). FPGA also checks for error conditions.

Beam current modulation options;

• Control via power supply feed back circuit, leaving gun cathode to Wehnelt electrode bias fixed via resistor. When at 1KW power and shut down in half a cycle ( 0.2MHz switch frequency) the resonant energy left in LCC circuit and output capacitance would lead to one extra 55micron cube of melted Titanium or 40 microns cube of melted steel which is workable .

## Self replication

Self-replication of a vacuum chamber runs into the "how to make a match box inside a match box problem".

This will be overcome by purchasing a large diameter glass cylinder 406.4mm diameter 1130mm long and using Vacuum bell jar style Viton L gaskets top and bottom to aluminum top and bottom plates. All build platform parts can be printed in the mother machine in segments then assembled and fitted within the glass cylinder which provides the vacuum seal. This also means TIG welding gear is not needed as all parts can be electron beam welded together in mother machine. Later glass test tubes and ultimately glass cylinder can be manufactured in oversize machine by beam melting glass sand in contact with metal within chamber or in hotter climates by solarsintering[16][www.markuskayser.com].

## Specialist Parts

Those which are not self printable or readily available;

• Transformer ferrite core ; Rectangular hollow section shape single phase style [17] but modified by a drill removing corners so circular cross section on secondary high voltage side.
• Roughing pump capable of 10-4Torr
• long large diameter Flexible hose to reduce back streaming of oil from pump to gun (with end CF-16 hose to pump fitting)
• Priani vacuum gauge ( now home build version availible (use google translate) [18])
• Electrically operated CF-16 T valve
• Exhaust Valve
• Transformer varnish insulated wire for coils
• Viton o rings; 7 small for cathode, 3 small & 3 large for gun tube to glass chamber adapter, two large 16.5inch bell Jar L gasket seals with bottom aluminum and top 304 plates.
• Copper OPFC knife edge flange sealing rings; 2x 6.34inch DN(100) , 1 x DN80 anode and 1 x DN80 pump attachment side
• Safety vacuum cleaner ( needed for cleaning up Aluminum and Titanium) Self printed.

Access to ; Vacuum leak detector Spot welder for in chamber electrical connections of OFPC copper conductors having glass beads as conductor insulation. Geiger counter eg Radiation Detector - Pocket Geiger Type 4 [19], Self-indicating instant radiation alert and dosimeter (SIRAD) [20] easily worn, result via comparison with color chart ( 4 - 25 Euro each SIRAD) [21] [22].

Avoid the need for TIG Welder in MetalicaRapBaby by electron gun tube being short enough to be able to fit in chamber of mother machine so mother machine can Electron beam weld parts together, (TIG welder for electron gun tube and tube to flange attachment for MetalicaRapBabe).

.

.

.

.

.

# Builders Section Very detailed technical information for actual builders only

## MetalicaRapBabe elements

Power Supply section is correct but the other sections on the repstrap machine called MetalicaRapBabe is still being edited so not correct yet!

### Electron Gun MetalicaRapBabe for builders

Single electron gun providing 100W for SEM , and 1KW for printing operations. Has two lens and 4 paris of deflaction coils.

Filament based around tungsten microscope bulb ( ie flat wound)( sm8018 6V 15W bayonet square filament towards top BA15D ) with top glass dome removed held on remaining glass, half in and out of chamber by 2 viton O rings in bespoke sealing fitting having 2 o rings on inside sealing to bulb glass and on outside 2 more o rings sealing to inside of 2 inch test tube. These o rings are supported by dual grooved on inner & outer aluminum o ring support cylinder which is situated in the 2 inch test tube and which also supports 3 x 304 pillars that attach and support the tungsten Wehnelt cylinder, via 9 in total 1000+ metling point carbon/ceramic/alumina washers, electrical connection must be maintained by tungsten bypass wire or washers being electrically conductive(one; wehnelt hole OD, support rod ID. 2; larger dia. than wehnelt hole OD, support rod ID )[23]The test tube which is half in and out of the chamber is held in its own compression fitting on the end vacuum flange, the test tube is modified with a hole in its end for the bulb filament to protrude out of. A threaded 2 wall cylinder that is clipped in to bulb stub ends on either side of bulb base at one end and pressed against dual grooved inner & outer aluminum o ring support cylinder at the other, and thus provides adjustment of Z position of filament with respect to Wehnelt, accessible from outside of the chamber. A coil surrounds the test tube and a small coil attached to the bulb completes the filament drive "core-less transformer" electrical circuit, when combined with a dimmer circuit connected to coil out side the test tube it replaces conventional filament current control. Thus using a test tube with a home wound coil round it and a modified light bulb provides; -80KV electrical feed-through, -77KV electrical wehnelt feed-through, filament transformer, filament , heat removal mechanism from bulb to test ube, support for wehnelt disc(adjustable height above filament aim 0.3mm) ( also covered with a cap so no pointy HV parts revealed). A user selectable resistor is also included between wehnelt and filament within the test tube. 6.5Keuro saved here.

Thus providing additive printing through melting at high beam deflection speeds, enabling high build rates along with melting targets behind a metal vapor trap EBPVD solar cell production. The one gun will be stationary situated 1m above the build platform. A sensor pick up ring will be attached to a 12Kg gravity fed powder hopper just above the build surface with 2 powder wiper blades attached, using sliding slotted plates to release the powder (design similar to household air vents, but ours will have sharpened vent edges). A second powder hopper will gravity feed the 12 kg hopper. This one gun will be used for both melting and vision systems. a back-scatter detector ring will be attached to the powder hopper, close to the workpiece mounted on the Cartesian XY system (the hoppers motion itself provided by the X motion of the hopper, Y motion by means of further 2 linear bearings on front of hopper and drive shaft (No automated Z movement of the ring is included in the design ).

Providing build platform metal powder heating and melting, spot size 100µ (may be inaccurate) pointing accuracy 10µ (may be inaccurate). The gun will have a liquid cooling anode system via its support.

Vision gun method of providing Sub µm Topological mode SEM vision system (See [24])

Additionally providing Micro vaporization (See gun layout Page 3 [25] down to 10µ to 40µ spot size), Initially this would be used to flatten every tenth layer of the build through high points / blobs removal. Future software development could provide live modeling of build, so through the adjustment of beam melting path in following layers errors/ blobs could be taken account of as they arise, thereby limiting the need for blob removal to critical surfaces.

### Vacuum Chamber and pump MetalicaRapBabe for builders

• Pump is a roughing pump 2 stage Vane pump e.g.Leybold D4 D4B Trivac Rotary Vane Dual Stage 3.4cfm 2K euro new. to 1 x10-4 torr gas ballast off [26] [27][28] 1.0x$10^-$$^4$Torr )( a dry pump option is better for EBPVD but currently 10K plus euro so rely on back flow oil filtering.)
• Electron Gun constructed from Stainless steel 304L 4 way crosses reducer Conflat(CF) CF100eu/CF6inchUsa reduce to CF64eu/or63eu-CF4.5inchsUsa flanges in middle of sides, Top to bottom 219.4, left to right 209.8 . Beam tube has TIG welded angled collars of stainless 304 and 1001 Fe working out from centre with Conflat(CF) flanges conecting longer bottom beam pipe section, with ID 8 -80mm beam pipe . Beam pipe terminates in O ring to aluminum top plate of chamber. .Avoid the need for TIG Welder in MetalicaRapBaby by bourgt in cylinder parts of electron gun beam tube being short enough to be able to fit in chamber of mother machine (or with stopper and compression fitting chamber base plate if needed), (TIG welder for electron gun tube and tube to flange attachment for MetalicaRapBabe). TIG filler Dissimilar 304 to 1001 check ENiCRFE-2 (WLD A) [29] [30] ASIS [31] (Or friction Weld 316 to 1010 N.G. but 316 to 1018 ok but too high carbon check[32]).
• Use large diameter Borosilicate glass tube with aluminum plates top and bottom, with L gasket seals ( as used on vacuum bell jars), bottom aluminum plate can be lowered to gain access to attached build platform/ motors / sensors and includes electrical feed through,( thus avoiding the need to cut TV screens hardened glass as in TV solution, or use TIG welding equipment, as mother machine can self print or fit parts requiring welding in her chamber, but initially requiring more machining to produce top and bottom aluminum plates (later we can electron beam machine over size top and bottom plates in mother machine, by first printing O ring collars and grooves inserts, then welding these inserts in to half drilled holes in top or bottom over size plates in mother machine, then through drilling these half drilled holes, finally cutting down circumference of top and bottom plates with hand tools). Beam tube to filament /anode chamber connection can not be oring due to heat, so will need to be welded in mother chamber, though 450C knife edge connection may also be added. The top plates bottom surface will have two protruding ridges each with 2 sealing o rings in grooves on the outside diameter plus one buffer ring to support the edge of glass vessels that can be attached. ( a beaker for evaporation work and one large rounded base jar for metal powder production MetalicaRap Light and MetalicaRapPlay respectively). A extra 5mm thick 75mm diameter shielding ring fitted to beam pipe to aluminum top plate junction needed, if top plate has internal beam exit rounded corners added inside beam pipes continuation in aluminum top plate, to allow beam larger deflection angle with out striking aluminum top plate. The plate will support the weight of the gun and will have rise 1cm and swing 90 degree hing so free access to top of glass cylinder holding build chamber, 4 screw on extension rods will allow alignment of build chamber during vertical removal/replace for build platform maintenance. (Option for lowering the chamber out of bottom of cylinder a 3 legged rectangular cross section pipe stand supports glass tube (45Kg) and top aluminum plate and gun and power supply, gun tube attached (17Kg?) via O rings above it. The stand will have attachable round section feet for the first 15cm from the floor, that will change to square section to align with square holes in bottom plate (with a key channel matching width of round feet for initial front access) . So that this plate with build platform can be aligned correctly so can enter the large diameter tube smoothly. ) Large diameter removable cement pipe used to shield operator and any environmental disturbance removal shielding e.g.electromagnetic shielding so beam is not effected by outside electromagnetic disturbance, e.g. nikel coating, when printer is running.
• ((OLD solution TV Glass Chamber; chamber based around two recycled CRT TV tubes, face to face. The 2 salvaged CRT's with holes cut in screens, after end neck pip breakage release of vacuum, capacitive discharge via anode button and electron gun removal in both CRT's. Lower CRT pip reseal with blowtorch so lower feed-throughs continue to function, upper CRT has electron gun neck of CRT tube shortened. Two o-rings supported on one side by a double grooved Aluminum metal support ring sealing the CRT's face to face and providing access door. (Capacitive discharge via shorting anode cap to earthing band on corners, use own cable as anode cable may have diode in [33][34][35] be safe out there!). Glass chamber is connected to electron gun tube via Aluminum Tube/Glass adapter; with 5 o rings, 2 for gun tube sealing and 3 for CRT (MetalicaRapBabe) , having 1 for cushioning end of glass and two for sealing each joint.(or to square bottomed flask (MetalicaRapLIght) sealing) feed-through motor connections via Cathode RayTube connections in lower recycled Cathode RayTube,. ))
• High vacuum $10^-$$^4$ Torr to $10^-$$^5$
• Electrical feed-troughs ; For tungsten filament AC 10A 3V 50/60Hz heater power supply based around modified microscope flat wound bulbs body(sm8018) giving a -62.5 KV Wehnelt feedthrough which is within a test tube providing -59.5KV feedthrough. Using large diameter glass cylinder chamber with L gaskets TV CRT neck feed-through seal with o rings Or o ring to copper tube then connect TV neck to copper tube by melting CRT glass. Glass tube to metal tube connection video[36])
• Mechanical feed-troughs ; will be avoided by using periodically replaced cheap standard Nema motors within the chamber.
• Pirani vacuum gauge ( avoid cathode gauges as ionization from gun my interfere with them)* Home build Pirani vacuum gauge (use google translate) [37]
• Outer box for shielding constructed of cheap material concrete or solidified sand.
• Viewing windows; Main chamber viewing window with large diameter glass cylinder; Viewing windows are just multiple planes of normal glass in hole in concrete pipe as no sealing is required, total thickness 70mm for shielding. ; Electron Gun viewing window will be a CF 50 flanged type on a Twith extra removable metal plate for shielding

### Metal powder dispenser MetalicaRapBabe for builders

Currently same as above

This will be a gravity fed split powder hopper.A small hopper will move in the x direction, intermittently refilled by main hopper situated in side of electron gun tube.The small hopper will have a 12Kg powder capacity, this gravity fed powder hopper just above the build surface has two 1cm square section powder wiper blades attached wither side, leveling the roughly dispensed powder. The powder is released using two sliding slotted plates, ( design similar to household air vents but ours will have sharpened vent edges). At the end of the hopper travel a small waste powder slot will remove old dispensed powder before the next dispensing sweep takes place. A second main powder hopper will gravity feed the 12 kg hopper, it will be situated up in the side of the 14 inch OD diameter gun tube and will obscure a little part of the build chamber from the beam (a necessary compromise). The vision system feedback can accommodate some roughness in the powder deposition during beam melting phase and subtractive removal phase.This design may lead to clumping with some powders at higher vacuums. The window option overcomes this but introduces other complexities in vision system and beam distortion (see below).

Other dispensing/ considerations include:

• Metal Vapor will be released in to the chamber during the melting of the the powder, a later consideration is protection of the gun, initially the guns distance from the build platform should be adequate, but the possible inclusion of a spinning slotted disc in front of the gun as further protection may need to be considered,
• SEM pickup PIN diodes protection cover will be used when printing,
• Variable layer thickness might (?) be desirable to provide fine control of the vertical resolution using thin layers where desired, while still allowing relatively high build speeds that come with thick layers for other areas.
• The Arcam EMB uses a "textured roller". Presumably a roller is textured with a pattern of holes which is loaded with powder, then the powder falls out of the holes, and a little random redistribution occurs on the way down to the workpiece surface for a reasonably uniform powder layer of repeatable thickness this design will likely operate at ultra high vacuum.
• Leveling the surface of the powder bed or enabling better flow through hopper at ultra high vacuums may be possible using ultrasonic lubrication to briefly fluidize the powder bed. It might cause unacceptable settling though.
• The density of the powder is lower than the density of the solid metal. So if the powder is only deposited using say the textured roller the upper level of the powder bed will get higher than the surface of the part being printed. How is this dealt with in the Arcam printers? More research needed
• Upon beam heating material loss in the form of evaporation and melting is released in to the chamber redepositing itself on windows and light optics and even chemically react with filament ( a minor effect with most materials), use replaceable and disposable glass sheet covers.

### Build platform MetalicaRapBabe for builders

A 30cm vertical travel stepper motor driven circular platform within a 14 inch 304 tube printed in sections which is not vacuum sealed , as it is all surrounded by large radius glass tube. Within the build platform is a built in ceramic insulation layer. Two felt o- rings seal the gap between the vertical motion circular build platform and the surrounding cylinder.

#### Power supply MetalicaRapBabe for builders

Overview

Power supply Full Bridge LCC Series Resonant Converter with duty cycle switching variation control and above resonance frequency switching variation control, through this frequency variation maintaining Zero Current Switching during Coolmos switch ON and Zero Voltage switching during Coolmos OFF switching, with Arc sense, arc quench and arc count. ( see old information for more on beam current control options)

Layout of electron gun power supply PCB's -63KV SMPS on top and 0V to-1KV SMPS floating at the -63KV's output on bottom, providing the beam current modulation via the bias voltage applied to wehnelt electrode, showing bar connector to bulb cathode filament and ring connector to wehnelt electrode (Note that the primary SMPS pcb is left transparent so you can see the connections to the other PCBs) Smps 1KV & 63KV assembly scale 1 to 1 with pcb.FCStd.tar.gz Down load this file from file and part section and unpack and open in Freecad, to get freecad[2] PCB's are also available in download section open in development version of Kicad [3] running on a ubunto utopic linux PC (Due to a freecad bug for the PCB 's to open correctly you must; after opening the file in freecad, scroll down left hand menu to green PCB icon labelled power supply primary, and click so its highlighted, then select placement in edit menu, then tick box apply incremental changes to object placement then click apply then click ok, the pcb's will then appear in correct positions)
More detailed of above showing secondary PCB,s the top -63KV smps secondary transformer shows only the first and last 12 Secondary PCB's of the 178, The lower -1KV smps show the 4 required secondary tramsformer smps PCB's

High Voltage planar transformer PCB
High Voltage planar transformer PCB surrounded by insulator ( ferrite core removed for picture)

Schematic of 1KV Power supply (Download SMPS1KV.pdf for seeing 1KV wiring details in file and parts section 9, this smaller supply has some resistor changes and only 4 secondary pcb's )

High output series-parallel resonant DC-DC converter 5KW 62.5KV (search on this bold text) -62.5 KV 0 to 5KW running at 181.5KHz resonant range 260Khz up to 500Khz at idle/low load, with a -325V to 62.5KV transformer with voltage doubler on output, the transformer is designed for a specific value of inductance and capacitance to operate at the desired resonant frequency, as the load changes the parasitic capacitance of the transformer changes too, this means that at high load the series resonant converter topology dominates and light / no load the parallel resonant converter topology dominates. The change in load changes dramatically due to occasional arcing in the gun, This topology deals with these changes effectively.

The primary of the high voltage transformer will have 11 winds .The transformer secondary consists of a magnetic flux pole with multiple ac to dc converter stages with planar coils [38][39] (search on this bold text string) but pole is extended in to a traditional single phase core type transformer arrangement [40] or technically defined as a "dual C ferrite core assembly" ( with high voltage secondary side cylindrical cross section not rectangular, so as to avoid corners at high voltage). Each side is surrounded by an insulating HDPE or nylon 6/ 6 insulation cover called a transformer bobbin, one square sectioned and one circular cross sectioned (dual C ferrite core assembly and bobbins will be within a insulated HDPE box, and will run in air with the option of mineral oil insulation later). ( for wehlnet power supply 4 x 354V =1400V as PLA is ok up to 70C has resitivty of 5e10+18 ohms/cm when 0.2MV/cm is applied[41].Break dowm is 0.62MV/mm[42] So for 1400V less than a mm is required for insulation. So pla is ok for this 1400V power supply.

Each of the 178 transformer secondary converter stages is a 2 layer PCB 0.4mm thick, with a 3 turns/winds (A 1 turn PCB layer and a 2 turn PCB layer blind via's not possible) ( 1oz. thickness) 0.6mm wide circular tracks (6mm spaced tracks so same as class2 double insulated) on the PCB looped around the ferrite core and 4mm insulation bobbin connected to 2 rectifying diodes and 2 capacitors in a voltage doubling arrangement, ((0.1mm min pcb thickness for insulation safety) + 0.055mm Mica paper washer (better insulation and heat dissipation(check availability)= 0 .455mm thick) conductive turn/winds of track creates a transformer secondary coil (actually end up with 119mm high ferite core space 8(only 4mm top wall bottom wall Only 2mm space top bottom from ferite core CHHECK IF PRACTICAL) so 178x(0.6mm PCB +0.055mm mica paper )= 117mm . The series parallel resonant converter topology gives voltage gain in this coil from 2.25 times to 6 times dependent on switching freq/duty cycle which is combined with the gain of the o/p voltage doubler arrangement of output rectifying diodes and capacitors, gives a resultant gain of ; 11primary winds,3 secondary wind, times factor of 2 (built in Topology)and times 2 (voltage doubler circuit on each output pcb)(3/11)*2*2*325V= 354V, so outputting -354 V d.c.from each PCB. Each voltage doubling circuit uses fast rectifiers and 2 smoothing capacitors. Along the stack of pcb's the voltage increases in a negative direction gradually, keeping below Paschen or arc limit . Physically each secondary pcb will be clipped to the next (salvaged breadboard contacts in existing bunches of five work are functional clips). The power supply secondary converter stages can be tested (unregulated) as 178 separate -354 V power supplies, before all the secondary converter stages are connected in series creating the -62.5 KV output.

Physically the secondary pcbs end up making a spiral every 8th board will complete a revolution of the spiral, it will have a spacing of above next board of 8x0.55mm=4.4mm.

To reduce the parasitic capacitance each secondary PCB 3 turn track layout is in alternate direction clockwise/anticlockwise, while the connection polarity to the 2 smoothing/doubling capacitors is also alternated so as to maintain a continually decreasing voltage up to the top of the secondary pole across the o/p capacitor stack of -62.5KV. This gain requires the following component values n=17 Cp=15nF Cs=48nF( Series Capacitor requiring reduced ESR [43] so achieved with many film type nonpolar capacitors in parallel ( capacitance additive, ESR resistance reducing each extra parallel)) alpha=0.3125 Ls=16µH. fo= 181.5Khz. Max duty cycle 0..8 at frequencies 260KHz to 500Khz. Temperature rise is under 10deg C for this 0.6mm wide 1 oz (35µm) thick track area 0.21 sq mm rated up to 08A will use 0.08A Fig2[44] At switching frequency 181.5Khz skin effect is 0.15mm depth of circular conductor (each side of track approximately) so effective rating 0.6A [45]. Min track spacing 16mm for 178V. (( Alternative use flexpcb 0.2layer x2 but under 400µ thick insulation issue could test further)).

Transformer ferrite core is splittable in to two "C's" half's horizontally in the middle of vertical sides.Core Cross section 29mm x 20mm Vertical of C 130.0mm Drilled to round 19mm OD cross section is 92mm on each of two c's horizontals.(Diamond Drill recommended outside diameter 3/4" (19.05mm) Inside diameter .748" (19.00mm) this must be smaller than legs in both x and y directions lenght 2.75" (69.85mm)). Min transformer "window" vertical space for coils/PCB is in practice approx 100mm from ((178*(0.4mm PCB + 150µm Mica washer )+8mm insulation ) 8mm insulation is made up of 4mmTop&Bottom Sides of Bobbin. 0.15mm mica washer (Fx. Kapton sheet SIL-PAD K-10.[46] Or Muscovite mica paper MPM1(501) 82g/m2 0.055mm 0.025MV/mm insulation one sheet is 1.3KV insulation. 0.055thick[47]- could doubled up 2KV protection (3/11*325*6=531V max. poss.)) as spacers between PCB's, with Creepage distaance around edge off sheet/paper of Cat III 8mm min. ie 4mm track from edge of paper inside and outside track/coil[48].

Transformer ferrite core In any of the following equivalent materials //Material code (Manufacturer)// CF139 ( Cosmo ferrite),// N87 (epcos),// 3c94 3f3 (ferroxcube ),// R (Magnetics ), //PC44 (TDK),// 2FB ( tomita )
Ferite Core made up of 2 C's with cylindrical high voltage secondary side to the right
.For example Cosmo part no.C ferite core UU 12620 is big enough but square secondary that must be cut down to a cylinder using 7/8" Sintered (metal bond) diamond core drill, drill depth: 3" [49] ((or unlikely add 2 rods of same material 120mm long matching diameter to core square sections, or unlikely to get hold of round sectioned UR version that has 120mm window height for PCB's.))

Primary wire is 11 turn primary (87A continuous)(coolmos short circuit peak current limit of 117A) on a transformer bobbin (10cm flange on HV end) high voltage bobbin 4mm of HDPE or cross linked low density polyethylene XLPE also called PEX or nylon 6/6 insulation.

The -62.5kV Direct Current (DC) after regulation is directly connected to the Welnelt electrode then continues on via a manually selectable gun resistor to the Tungsten filament/cathode. Then the current flow continues down the beam to the target and returns via the chamber body to the power supply via a fixed resistor in the earth path to measure its magnitude. The voltage at the filament/cathode is also measured and fed back to power supply. This feedback via increased beam current leading to more voltage dropped across gun resistor meaning Wehnelt pinches beam current to a lower value and vice-versa means the beam current settles to a stable value. To maintain the best beam shape this simple feed back mechanism should ideally stabalise with 3KV over the gun resistor ie +3KV between Wehnelt and the cathode/ filament.( see spread sheet above for more details of optics)(NB the most negative part of gun is the Wehnelt, the cathode traps electrons in front of it by being up to 3KV more positive than the Wehnelt (-59.5KV). The max beam current possible would be achieved if the cathode-filament became more negative, ultimately as negative as the Wehnelt at the same -63KV, thus the wehnelt would have no limiting effect on beam current. Unfortunately the focusing effect of the wehnelt would also be lost at this point, so cathode being slightly positive with respect to Wehnelt is always desirable. Maintaining this cloud of electrons in-front of the cathode-filament called the space charge effect See spread sheet for calculations)

Power supply feedback control around transformer ;Output voltage feedback via 178 thick film resistors in series creating voltage divider . Output current sense feedback via a temperature stable resistor in ground voltage path back to power supply from earth on chamber. . Input /Primary current sense via 2 "current transformers" in series in each leg of full bridge with 1:125. Cathode voltage feed back via 178 thick film resistors acting as voltage divider. These four signals go via signal reconditioning board then on to Field programmable gate array (FPGA)[50] [51] via four channel analogue to digital converter (ADC) 12 bit 40Msps[52].

Power supply has a primary Board, with a Piggy backed ADC primary board, a pimary FPGA board (offshelf) , primary optic link board, the planar transformer has secondary A (clockwise) & secondary B (anti clockwise) repeated winding's boards, secondary ADC output board , secondary FPGA (offshelf), and secondary optic link board.

FPGA process software will maintains input and output current parity within acceptable current window values. Current measured via output resistor (300W?? rated resistor 50mm x 373mm E12?? values [53]) Allows a small error signal from voltage output variations when within a narrower voltage output window values. FPGA drives 4 coolmos FET gates in full bridge topology via driver circuits. FPGA also checks for error conditions.

Solftware for the FPGA_High_Voltage_side is coding 2 voltage feedbacks either side of ouotput resistor from linear to Log and then 8/10coding for sending over fibre optic to FPGA_Low_voltage_side. FPGA_Low_voltage_side software consists of feedback signal decode from optic fiber, along with max current measurement from either side of H bridge via ADC channel's (from 2 current sense transformers ). This output and input current data along with output voltage is examined in comparison to required voltage and current outputs so as to create an error signal. This error signal is low pass filtered (at 100KHz limited to avoid pulse width modulation ripple error) and sent to PID controller d_controller and finally to Digital pulse width modulator creating 4 outputs to H bridge FETS. Along with the digital pulse width modulation widening the pulse either side of a central point, the time delay between consecutive central points can also be adjusted thereby changing the driving frequency between 260Khz and 500Khz of the H Bridge. Adjusting the frequency will mainly adjust output voltage adjusting the pulse width will mainly adjust the current entering the transformer so affecting power transfer to the output. A power supply user interface will set user parameters.

ADC ADS4222(12 bits)primary side & ADS4242 (14bit)secondary side will oversample at 65MSPS needing 16 clock cycles 32 clock transition to process each sample, taking 240nS per sample (given total processing time availble of 2400nS) then sample rate reduced by a third to under 1Gbits per second for optical link. On the primary side ADC to FPGA via 12 diferential pairs or 24 lines ( though single ended could be used if further 12 FPGA pins are needed later.)2 pairs of clocks 4 lines - main clock from fpga and clock output to enable fpga to decode LVDS , serial clock a single line SCLK , 3 control lines RESET SDATA SEN , So 8 ADC control plus 24 data , FPGA to FETs 7 lines including ready fault and reset, leaving up to 9 pins for optical link. User computer interface via USB link.

The ADS4242 (14bit ) on the secondary side ADC to FPGA via 14 diferential pairs or 28 lines, 2 pairs of clocks 4 lines ( main clock from fpga and clock output to enable fpga to decode LVDS) , a serial clock line SCLK , 3 control lines RESET SDATA SEN , So 9 ADC control plus 28 data , leaving up to 12 pins for optical link.

ADC ADS4222(12 bit) & ADS4242 (14bit) when RESET pin is pushed high by fpga voltage applied from fpga to pin SEN sets data format single or double ended, applied low/ 0V gives required format LVDS DDR twos complement ( LVDS low voltage differential signalling), Normally RESET is LOW so; SCLK with these chips will only act as serial clock input, SDATA is serial data input.( SDOUT is used only for diagnostics of serial communication link as it can read backout what has already been sent via the serial link in to the serial register not connected on primary due to shortage of pins).

For control methodology also see High output series-parallel resonant DC-DC converter 5KW 62.5KV ( via search on this bold text )(NB this paper does not include the real transformer just 2 dummy inductors and 2 Dummy capacitors one series one parallel) . A separate housekeeping supply provides bias for all control circuitry, providing a 5V separate stand-by voltage which remains active when the power supply unit is shut down for any reason ( later once efficiency of power supply is assessed a Power factor adjuster circuit may be added between rectifier and dc to dc converter) The following safe guards will be included; Under/over voltage protection, Short-circuit protection, Output current limit, over temperature protection and line fuse. Adaptive control[54] is applied using Gain Scheduling in feed-forward methodology; Controller verifies the amount of current available (ie how much in tank ) to reach a target voltage at the o/p, adaptive gains are recalculated over each sample , this system identification technique thereby selects the appropriate linear controller from 3 options(low load controller, high load controller, arc controller) , so target voltage is met what ever the operational reality demands ( other people have done with fuzzy logic).

(Useful cheap analyzer for FPGA development [55],[56] , [57] )

Beam current modulation options MetalicaRapBabe;

• by control of the bias on the Wehnelt voltage in gun ; 0V( full beam wehnelt has same voltage as cathode filament No bias ) to - 1KV pinch off (cut off where wehnelt is at the lowest voltage and cathode filament is 1KV more positive ( i.e. more like anode) this bias is provided via 2nd isolated power-supply providing up to 1KV bias.

#### Power supply High Voltage Isolation optical SFP Implementation MetalicaRapBabe for builders

To control the power supply we need to measure the output current and output voltage just before the tungsten cathode, the current is measured via ADC voltage dividers either side of a fixed value thermionic stable ideally 2x 50Kohm in parallel so 100Kohm (rated 340W) in the SMPS, one 1KV SMPS will provide up to 1KV for wehnelt pinch beam current control and the other 63KV SMPS will provide the cathode anode gun voltage.

To isolate them from the user (transformer primary side) and still receive the voltage and current data from the high voltage output, we use two Papillio FPGA's one each side of the transformer connected with an optic fibre which provides the insulation (optic fibre must be non metalic cased type). This transformer Secondary side FPGA measures the voltage and current via ADC's and then converts this parallel data in to a serial bit stream, which is sent over the optic fiber via two Gigabit SFP transceivers [58]. Due to the FPGA's running off non synchronized clocks a buffer is required on the primary side FPGA to resync and recover the data and its associated clock, so the voltage and current information can be decoded by the primary side FPGA. Further details TBA [59]

High speed optical link receiver buffer FPGA logic gate circuit; showing layout of buffering of incoming data signal (data signal sampled at 400MHz/ or 2.5ns period , data bit is 6 bits long, +/- one bit). This gives plenty of scope for tracking the incoming data as its relative phase wanders between the two FPGA's caused by FPGA's master clocks inherent rate variations [4]

cheaper singlemode sfp's (1.25 gigabit) are designed for 10km, but will work dowm zero or yelow LC to LC patch cables [60]( 40km modules and above use multimode and cannot work over low distances without attenuaters).

Implementation example fpga to sfp is seen on [61] [62] [63] pin out [64](50.8 x 260mm Wtover 1mm)

At 1400Volts scondary power supply very high impedance on the welhnelt It will just be voltage driven and resistor only functions in arc so power supply in voltage mode? 1400 V power supply testing at 16mA in 100KOhm resistor rated at180Wmax 1.8KV max Testing of 1400V SMPS one 50KOhm acts as load and the other 50Kohm as current sensing resistor. Test of 63KV smps both will be inseries as current sensing resistors load will be in hgh voltage lab or electron gun.

#### Powersupply Feedack Control Implementation MetalicaRapBabe for builders

The primary side FPGA also runs the PID error signal correction [65] and low pass filter [66] along with the Digital Pulse width modulation outputs [67] to the drive the FET's via a 300V isolation chip and a high low FET driver chip.

### SEM MetalicaRapBabe for builders

We may be able to use 6 PIN diode Topological / 3D imaging sensor also for spot size calibration, as we can solve for spot size knowing distance from gun...etc.