Duet 3.3V supply quality

Re: Duet 3.3V supply quality

Radian — Sat, 04 Jan 2014 13:30:52 -0500

Quote
Rory166
Dmould

Whilst not disagreeing with you. There is a post somwhere here that I have replicated Radians work using a battery scope and Duet powered from Laptop USB ie no mains at all, not even laptop charger. There was a suggestion about using two scope probes in difference mode AC coupled with the two earth clips joined but not connected. I have yet to try this but I believe Radian has.

Rory

I tried loads of tricks Rory, but didn't really make any breakthoughs. I was investigating the transient response of both the switching and linear regulators and got some pretty wild looking waveforms, but I freely admit the results were hampered by measuring difficulties. As far as I could progress things, it remained the case in my opinion that there were very fast transients happening on the 3.3V rail and that the layout of inductors/capacitors around the microcontroller was not helping but hindering things. However I switched my attention to the cause of the transients when I found that reducing their severity increased printing reliability. Call me pragmatic - or anything you like :P

Re: Duet 3.3V supply quality

Radian — Sat, 04 Jan 2014 13:17:54 -0500

Just to add that I've started a dedicated thread to discuss the Heater switching transients issue.Unfortunately this particular thread about the quality of the 3.3V supply contains an awful lot of discussion about the heater switching transients. The two are indeed related, but there are parallel issues to do with the "not so LDO" regulator and noise immunity that might be better continued here. I hope we can pick up the topic of heater switching transients in the new thread.

Re: Duet 3.3V supply quality

Rory166 — Sat, 04 Jan 2014 05:56:48 -0500

Dmould

Whilst not disagreeing with you. There is a post somwhere here that I have replicated Radians work using a battery scope and Duet powered from Laptop USB ie no mains at all, not even laptop charger. There was a suggestion about using two scope probes in difference mode AC coupled with the two earth clips joined but not connected. I have yet to try this but I believe Radian has.

Rory

Re: Duet 3.3V supply quality

dc42 — Sat, 04 Jan 2014 05:55:57 -0500

Hi Treth, which spike are you talking about? If you mean the spike on the mosfet drain, I really don't think it's a problem as regards reliability and operation of the board. If you mean the spike on the +12V rail, then I agree that this is a problem.

Some possible fixes are:

1. Add capacitors on the +12V input, as close to the Duet board as possible. This is my least-preferred solution, and on its own, probably the least effective. It's more effective if you add the capacitor on the back of the Duet board, between bed heater +ve and the -ve side of the 12V input. I have added a 1uF ceramic capacitor here.

2. Add a resistor in series with the mosfet gate. This is a tricky mod to do and I don't recommend it. I added 100 ohms here, but I think around 200 ohms would be better.

3. Add a capacitor on the back of the Duet board, between bed heater -ve and the negative side (upper terminal) of the +12V input. This is an easy mod to do. I would not normally recommend connecting a capacitor directly across a mosfet, but the mitigating factor here is that the mosfet turns on slowly, so the discharge current is modest. Radian used 1uF, but I would choose 0.47uF because I am less worried by ringing on the mosfet drain than he is.

Fixes (2) and (3) also get rid of the spike on the mosfet drain, or at least reduce its amplitude and sharpness. You could also get rid of the spike on the mosfet drain by adding a Schottky flyback diode, but I think this is pointless unless you tackle the spike on the 12V supply.

Re: Duet 3.3V supply quality

dmould — Sat, 04 Jan 2014 05:31:47 -0500

@Radian, I am sceptical that the spikes you are seeing when unrelated main switching occurs are really there. I suspect that the ringing is on the earth side of your 'scope probe, You might try powering the Duet from just its USB port (or even a 12V battery) and see if you observe similar ringing when your soldering iron is switched.

Re: Duet 3.3V supply quality

Rory166 — Sat, 04 Jan 2014 04:52:26 -0500

Treth

What scope were you considering? I bought a Hantek 1062B but have yet to fully master its capabilities. Are you still quite sure about your sense comment?

All

This thread is getting very long and quite hard to follow. Perhaps a thread for proven conclusions would be beneficial.

Rory

Re: Duet 3.3V supply quality

Treth — Sat, 04 Jan 2014 03:56:45 -0500

This post is getting long so apologies if I missed something.

The switching transients at switch off are obviously a concern on long term reliability. There is an interesting post here:-
[reprap.org]

I used to work in switch mode PSU design back in the 80's when things didn't switch fast and worked just above 20kHz if you were lucky!

From memory and from recent reading this spike will eventually cause failure and it is likely to be bigger than you see due to scope and probe limitations, so I would like to cure it.

Slowing down the MOSFET has been shown to work and is a good idea, but I would prefer not to mod the board at this stage. I also like to keep things switching fast, I spent lots of time trying to make slow devices fast!

The use of a capacitor on the output can be useful, but does affect the FET switch on, so I would like to avoid that as well.

The ribbon cable is a pain (but excellent for a flexible power cable!) as I can't add ferrite beads easily.

So I come back to the clamp methods.
The diode across the load is the easiest choice, but doesn't help the wiring inductance. I think I'll add this first, was this tried and did the spike reduce?

A better solution is to add a diode with R and C, these dissipate the energy in a more controlled way in the "R". This method was used in the good old days of unipolar stepper drives to remove the energy from the de-energised coils. As the spike is very narrow only low power components are required, but the diode must be fast (the lead inductance will actually help here as it slows the prpogation towards the FET switch allowing the diode to turn on).

There are other options such as transient (transorbs) suppressors, but has anyone tried this D,R,C combination at the load?

My scope will be a bit 'iffy' for these measurements, in fact I was either going to buy a scope or a 3D printer! I'm pleased sense prevailed and I bought the Ormerod B)

Re: Duet 3.3V supply quality

dc42 — Fri, 03 Jan 2014 08:00:16 -0500

I agree with you, capacitor to ground is best. With the capacitor across the load, the current drawn from +12V will still drop abruptly to zero, which is what gives the positive spike on the +12V supply in the absence of a local decoupling capacitor.

Re: Duet 3.3V supply quality

Radian — Fri, 03 Jan 2014 06:31:06 -0500

So dc, now I find myslef wondering what the subtle difference is placing the capacitor across the load versus across the switch... both provide a path for the back-EMF and equally well prevent the rise above 12V at the Drain terminal. From a theoretical POV the two networks are identical AC wise due to the bulk capacitance of the PSU. In practical terms I think the difference is the series inductance of the supply leads, so without local decoupling on the 12V, snubber capacitor to ground is the best bet - do you agree? It would however be a very easy mod for the Bed Heater just to solder a leaded cap. in the unused terminal holes provided. Hmm...

Re: Duet 3.3V supply quality

dc42 — Fri, 03 Jan 2014 06:12:16 -0500

It happened again.. your post crossed with my edit to my previous post.

Re: Duet 3.3V supply quality

Radian — Fri, 03 Jan 2014 06:06:56 -0500

Quote
dc42
So I'm thinking that a capacitor of 0.22uF would probably be enough to prevent avalanching, saving 40uJ dissipation at turn-off, but only add 16uJ of dissipation at turn-on.

0.22uF certainly would prevent avalanching, but it would definatley "ring" all the way down. I most recently played around with values on the Extruder heater MOSFET which switches roughly half as much current than the bed and with 0.1uF between Drain and Source there was a significant amount of ringing. 0.22uF reduced it quite a bit, but I finally settled on 0.47uF for smoothness. This scales almost exactly to the Bed heater hence 1uF for the Bed MOSFET.

The Extruder switch is potentially more significant due to being at the opposite end of the PCB to the 12V Supply terminals. It might have been bettern to locate the Extruder connector and Transistor at the other end of the PCB - where the Z-axis motor is now. Anyway, the capacitively coupled transients on the 12V rail from flyback is totally eliminated with the snubbers in place.

Re: Duet 3.3V supply quality

dc42 — Fri, 03 Jan 2014 05:09:46 -0500

Thinking about this a little more, there is another effect. Without the gate resistor, the inductive energy in the bed is dissipated as avalanche energy in the mosfet. There is 45-50V on the mosfet drain for a little under 200ns, and during this time the current will drop smoothly from 10A to zero. So I estimate that around 40uJ of avalanche energy is dissipated in the mosfet. Your capacitor suppresses the positive spike so that the inductive energy is dissipated in the bed resistance + circuit resistance instead. So although your capacitor increases turn-on dissipation by 72uJ, it decreases turn-off dissipation by 40uJ.

However, your capacitor is so effective at decreasing the positive spike that it is clear that a lower value capacitor would suffice. So I'm thinking that a capacitor of 0.22uF would probably be enough to prevent avalanching, saving 40uJ dissipation at turn-off, but only add 16uJ of dissipation at turn-on. Rise time would be about 500ns (double the rise time that my 100 ohm gate resistor achieves). And adding a capacitor between bed heater -ve and ground is a much easier mod than adding a gate series resistor!

[EDIT: your post crossed with mine. Using a 0.22uF capacitor instead of 1uF might mean there is a little residual ringing. If you wanted to stop that as well, perhaps 0.47uF would be a good value. Also 0.1uF across the extruder heater mosfet drain and source.]

Re: Duet 3.3V supply quality

Radian — Fri, 03 Jan 2014 04:58:42 -0500

Hi dc - no worries. I guess it all comes down to priorities.(tu)

My current priority, while trying to track down intermittent microprocessor crashing, is to keep voltage and current transients in the wiring looms to an absolute minimum. Comparing our two sets of scope traces for the Transistor Drain voltage, your sceme nicely rounds off the edges but still leaves us with a 45V transient that rings at around 5MHZ for a uS or two. I much prefer to integrate all this within the same period to keep it from exceeding the supply rail. I think the full analysis of the system plus snubber network is similar to that of a Class E amplifier - the likes of which could actually be developed into providing full analogue control of the heater(s) - but probably wouldn't be worth the extra effort.

Re: Duet 3.3V supply quality

dc42 — Thu, 02 Jan 2014 18:22:32 -0500

Hi Radian,

I appreciate what you are trying to do, but I don't agree with your rationale.

Quote

Without knowing exactly what the present (or future) firmware is going to do with the gate drive, keeping the MOSFET in it's semi-conducting region with 10A flowing could soon turn into excessive dissipation. What I mean is, it depends on the number of on/off cycles in a given amount of time and we don't necessarily have any control over this.

With my added gate resistor, the turn-off time is increased to about 250ns. But the turn-on time of the mosfet is much higher - around 2us due to the gate charging through the 1K resistor to +5V. Given that the number of turn-off events is equal to the number of turn-on events, even if the firmware uses PWM in future, my change only causes a 10% increase in dynamic power dissipation. Ideally, the circuit would use a gate driver IC or similar, so that the turn-on and turn-off times can be kept similar. If I thought the dynamic power dissipation was a problem, I would reduce the 1K resistor to e.g. 470 ohms.

Quote

So a 1uF suppression type capacitor between Source and Drain will have 12V across it when the MOSFET is turned on but the charge it's holding is insignificant given the characteristics of the MOSFET.

I was initially horrified at the thought of discharging a 1uF capacitor through the mosfet, because if the turn-on is fast then this will result in very high peak currents These may not harm the mosfet, but would cause large transients. However, the mitigating factor is that the mosfet turn-on is quite slow.

Let's do some sums:

Energy stored in your 1uF cap, which is dissipated mostly in the mosfet every time you turn it on, is 0.5*C*V^2 = 72uJ.

Energy dissipated in the mosfet due to turning it off in 250ns rather than instantaneously is some fraction (probably well below 0.5) of (6V * 6V /1.3 ohms) * 250ns = 7uJ. [Explanation: if the load is resistive, then the worst case power dissipation is when the mosfet and the bed are both dropping 6V. In practice, the voltage across the mosfet won't be static at 6V but will move from 0V up to 12V, hence "some fraction". I'm ignoring the inductance of the bed, because in the original design, the inductive stored energy is dissipated as avalanche energy in the mosfet anyway.]

So your 1uF cap causes the dynamic power dissipation in the mosfet to increase by more than 10 times more than my gate resistor (maybe a bit less if I take account of the gate resistor slightly increasing the turn-on time as well). I'll stick to my gate resistor!

Re: Duet 3.3V supply quality

Radian — Thu, 02 Jan 2014 17:56:10 -0500

@dc42
I'm not so sure about deliberatley slowing the turn-off time to reduce the back-EMF. Without knowing exactly what the present (or future) firmware is going to do with the gate drive, keeping the MOSFET in it's semi-conducting region with 10A flowing could soon turn into excessive dissipation. What I mean is, it depends on the number of on/off cycles in a given amount of time and we don't necessarily have any control over this.

I'm currently using a capacitor-only snubber between the Drain and Source. As you know, the turn-on time is relatively slow compared to the turn-off, but the inductance of the heater means the MOSFET is well on before any serious current flows. So a 1uF suppression type capacitor between Source and Drain will have 12V across it when the MOSFET is turned on but the charge it's holding is insignificant given the characteristics of the MOSFET. The capacitor current then falls as the bed current rises but the droop on the 12V rail is still negligable overall. However, at switch-off, the capacitor provides a nice path for the back-EMF letting it down without avalanche. The scope traces show the resulting Drain voltage - roughly 2us rise and fall with no nasties. This keeps 50V transients from developing in all those ribbon strands and, from there, coupling into the adjacent cable looms.

Transistor on, Drain goes to 0V in a couple of uS:
[attachment 23933 2014-01-0222.31.42.jpg]

Transistor off, Drain goes to 12V in a couple of uS:
[attachment 23934 2014-01-0222.31.47.jpg]

Re: Duet 3.3V supply quality

dc42 — Thu, 02 Jan 2014 08:24:51 -0500

Quote
kwikius
The flyback diode would be connected with anode to TR2 drain and cathode to Vin, basically across the bed heater inductance. On switching off TR2 the diode starts to conduct with 10 A running through it into Vin.

Not quite, the current circulates round the bed and the flyback diode only, so your decoupling capacitor isn't involved. However, your decoupling capacitor (or my 1uF one) does help absorb the shock of the PSU load current decreasing abruptly by 10A.

Quote
kwikius
Alternatively without the diode, on switching off TR2, then the drain of TR2 reaches @50 V and then acts basically like a 50 V Zenner. In this case at switch off ( get a bit hazy here) Vin continues to ( smoothly?) supply 10 A and the energy in the inductor is dissipated in the Mosfet.. so ideally a gentler solution, although maybe not working so great in practise?

Yes, although the waveform at the drain is very jagged, suggesting that avalanche breakdown is not happening uniformly. Adding the series gate resistor made it better.

Re: Duet 3.3V supply quality

kwikius — Thu, 02 Jan 2014 06:20:01 -0500

Quote
dc42

Andy, the body diode of a mosfet is in the wrong place in the circuit to act as a flyback diode in a single-mosfet circuit like this. It's different in a mosfet H-bridge, where the body diode of the upper mosfet acts as a flyback diode when the lower mosfet turns off, and vice versa.

Ah Apologies! .. OK I think I'm getting the hang of it. The flyback diode would be connected with anode to TR2 drain and cathode to Vin, basically across the bed heater inductance. On switching off TR2 the diode starts to conduct with 10 A running through it into Vin. hmm.. In this case my large decoupling capacitor at the input would works nicely to sink the 10 A current from the supply inductance as well as the current through the diode, which add up at switch off to 20 A. If we assume a 0.1 ohm ESR from my capacitor stack then Vin rises at switch off by 20 * 0.1 == 2 volts. Without decoupling though... not so good!

Alternatively without the diode, on switching off TR2, then the drain of TR2 reaches @50 V and then acts basically like a 50 V Zenner. In this case at switch off ( get a bit hazy here) Vin continues to ( smoothly?) supply 10 A and the energy in the inductor is dissipated in the Mosfet.. so ideally a gentler solution, although maybe not working so great in practise?

regards
Andy

Re: Duet 3.3V supply quality

dieterzar — Thu, 02 Jan 2014 05:34:38 -0500

Hmmm... it doesn't appear to be doing its job though... I am going to have a good look at it when I get home tonight. I am still waiting for my replacement head which was never packed in to the box but it is built and working from every other point of view, so I can spend some time testing this and trying a few things.

I was thinking the use of the other rails from the psu for 5v and 3v3 may work but they share a common gnd which means the problem will still be there but maybe to a lesser extent. I can quite see how this could cause it to reset or stall during a print. Switch mode PSUs all have recover times, some just have better cap arrangements on the outputs and circuit designs.

The final solution to this that works 100% should be interesting.

Dieter

Re: Duet 3.3V supply quality

dc42 — Thu, 02 Jan 2014 05:17:04 -0500

Dieter, I did add a flyback diode, but I removed it when I added the resistor in series with the mosfet gate. The back emf from the bed heater inductance isn't really the problem, the problem is the abrupt removal of a 10A load from the 12V supply along with the lack of local decoupling. Adding the gate series resistor to slow down the turn off speed alleviates both. I added a 1uF ceramic local decoupling capacitor too.

Andy, the body diode of a mosfet is in the wrong place in the circuit to act as a flyback diode in a single-mosfet circuit like this. It's different in a mosfet H-bridge, where the body diode of the upper mosfet acts as a flyback diode when the lower mosfet turns off, and vice versa.

Re: Duet 3.3V supply quality

kwikius — Thu, 02 Jan 2014 04:59:21 -0500

Hi Dieter,

AFAIK Power Mosfets have this diode integrally as part of their structure. See Wikipedia Body Diode

regards
Andy

Re: Duet 3.3V supply quality

dieterzar — Thu, 02 Jan 2014 04:51:31 -0500

I see a diode was mentioned earlier in a post as a solution, was it given a go?

Dieter

Re: Duet 3.3V supply quality

dieterzar — Thu, 02 Jan 2014 04:45:21 -0500

I think that short sharp spike you see is back emf from the heater board, it must be acting like a coil as what you describe is typical of coil back emf. A diode across the heater board should solve this, flyback diode, use a common 1N4007. This is a common issue switching relays. Brushless motors do the same. Over time you will possibly destroy the fet.

I will have a go on mine and see what it does, I only have a DSO pocket scope, so it might be to slow to pick it up.

Dieter

Re: Duet 3.3V supply quality

dc42 — Thu, 02 Jan 2014 04:09:46 -0500

Hi Dieter, I believe the biggest problem is that the bed heater mosfet switches off very abruptly (in <20ns) and there is no local decoupling. This leads to a short, sharp positive spike on the 12V supply in the locality. Switch-on is less of a problem because it is much slower, around 2us. Using PWM to do a soft start and finish would result in many more of these spikes being generated.

I have tackled the problem by adding a gate series resistor and a local decoupling capacitor.

There is also a short-lived spike on the mosfet drain when it switches off, enough to make it avalanche, but the avalanche energy is well within the mosfet rating. Again, a gate series resistor to slow down the turn-off is enough to reduce the amplitude of this spike.

Re: Duet 3.3V supply quality

dieterzar — Thu, 02 Jan 2014 03:10:56 -0500

I assume the heater bed is causing most of the issues, voltage spikes via back emf? Voltage sag when switched on. I have not looked at the circuit yet but I would imagine a diode would be needed to clip the spikes. Also it seems from what I understand, the mosfet it just being switched on? A small rewrite on the software side should sort the spikes out by ramping up the voltage to the heater bed by employing pulse width switching to the bed. May add some caps then on the bed wires to smooth it out? By pulse width switching the bed mosfet, much more accurate control over temp can be managed? I am not an electronics engineer but been a hobby since year dot and I think this can be corrected by some software tricks? My 2c

Dieter

Re: Duet 3.3V supply quality

kwikius — Wed, 01 Jan 2014 11:14:41 -0500

Heres a screen dump which may give you a flavour of it.

[attachment 23854 Duet.png]

regards
Andy

Re: Duet 3.3V supply quality

dc42 — Wed, 01 Jan 2014 10:50:20 -0500

Ah, I haven't installed KiCad yet - looks like I had better do that to better understand the board.

Avalanche breakdown occurs at about 50V for these mosfets. When the mosfet turns off sharply, the inductance of the bed + wiring causes a back emf to be induced across the bed heater, giving rise to a spike at the mosfet drain, even though Vin stays close to 12V.

Re: Duet 3.3V supply quality

kwikius — Wed, 01 Jan 2014 10:44:05 -0500

Quote
dc42
AFAIR, Ian said the bed current was 9A by design, and someone else on this forum reported 11A. I've just measured the resistance to be 1 ohm at the bed, and there will be some resistance in the ribbon cable. So 9 to 11A is believable.

I think the board has a ground plane as one of the inner layers, in which case the stepper driver decoupling capacitors will help suppress the transient when the bed heater turns off, except at very high frequencies - which is why I added a 1uF ceramic capacitor to my board. The stored energy in the lead and bed inductance appears to be dissipated mostly as avalanche energy in the mosfet - which is another reason to add a ceramic decoupling cap between the bed +ve output and a ground point close to the mosfet source terminal.

There will be some shunt capacitance and reverse recovery time in D1, but I believe C1/C3 should be more than adequate to take care of that.

If current is 10 A ... apologies. 10 A is quite a bit to be switching though!
As viewed in KiCAd, there are various small independent ground planes (looks like 1 for MPU and one each fro stpper drivers, 1 for bed mosfet) but the bed mosfet is only connected to the stepper caps by one 2mm trace.. which goes through the input screw terminals.

As for Mosfet Avalanche breakdown.. this requires quite a high voltage doesnt it? My idea is to try to keep Vin as near as possible to 12V. This makes the 5V switchers job easier surely? No reverse bias in D1 == no glitches for switcher

Tricky.. I certainly feel outclassed in elctronics knowledge here though so maybe I should quit arguing while I'm not too far behind :)

regards
Andy.

Re: Duet 3.3V supply quality

dc42 — Wed, 01 Jan 2014 09:35:37 -0500

AFAIR, Ian said the bed current was 9A by design, and someone else on this forum reported 11A. I've just measured the resistance to be 1 ohm at the bed, and there will be some resistance in the ribbon cable. So 9 to 11A is believable.

I think the board has a ground plane as one of the inner layers, in which case the stepper driver decoupling capacitors will help suppress the transient when the bed heater turns off, except at very high frequencies - which is why I added a 1uF ceramic capacitor to my board. The stored energy in the lead and bed inductance appears to be dissipated mostly as avalanche energy in the mosfet - which is another reason to add a ceramic decoupling cap between the bed +ve output and a ground point close to the mosfet source terminal.

There will be some shunt capacitance and reverse recovery time in D1, but I believe C1/C3 should be more than adequate to take care of that.

Re: Duet 3.3V supply quality

kwikius — Wed, 01 Jan 2014 07:44:45 -0500

Quote
dc42

Quote
kwikius
So my proposal is to hang some hefty capacitance off the input leads where they enter the enclosure

I ignore the existaing capacitance at the ATX PSU, which I believe is stated by Rory166 to be 2200 uF somewhere, but its 2 ft away down the leads. I assume the ATX PSU can handle the large extra capacitiive load I plan to add! I also ignore the hot end heater and concentrate on the Bed Heter Lets keep it simple.

So some numbers. Lets say we have a current of 20 A with bed heater on.

Assume we want a max voltage drop of 6 V, when the Bed heater switches on (So 5v reg has enough headroom) Lets also arbitrarily assume we want to supply that for 1 ms.
My math gives a required Capacitance of around 3000 uF.

The max ESR of the Cap allowed is 6 V / 20 A == 0.3 R which I think is achieveable. Just to be sure I'll get 3 1000 uF caps and put them in parallel. ( Maplin dont specify ESR)

Hi Andy,

Your maths looks OK to me, except that if you are worried only about the transient caused by the bed heater turning on, then it is only the bed heater current you need to worry about, not the total 12V current.

There is the drop when the heater bed turns on but I believe the current through the bed heater circuit is 20 A , so when that switch turns off abruptly ( as we know it does!) there is no where for the lead inductance and bed inductance to go, so I guess that is what I am trying to address. (hmm Did I say that before .. no :) ) The existing capacitance on Vin consists of the stepper driver capacitors, but doesnt look like they are close to the ground point (which basically is the screw terminals where the supply enters AFAICS)

Am not too up on diodes etc .. but wont there be some shunt capacitance through D1 as it comes in and out of reverse bias too?

regards
Andy

Re: Duet 3.3V supply quality

kwikius — Wed, 01 Jan 2014 07:11:49 -0500

Well... here is my effort. It feels so right :) Maybe I will write a little app to keep firing the heater and see if it actually does improve matters.
Had room for a little 100 uF capacitor atop the pile and added a discharge resistor for good measure.
[attachment 23837 Inputcap.jpg]
Tested.. Nothing blew up :)

regards
Andy