Stepper drivers - failure modes

Sorry if this isn't the right place (I'm fairly new).
I've heard that there are 'issues' with the stepper drivers letting out their magic smoke if they are disconnected when powered.

Does anyone here know the failure mode that this causes?

I'm _considering_ designing a slightly different stepper driver (although still based on an A49xx or DRV88xx chip) that mounts directly to the back of the stepper motor itself. However, I'll probably be wanting a bit more than the 2A of drive current, so I'm looking at using the 2* internal DMOS full bridges as 'pre-drivers' for secondary (higher current) full bridges (which would still use the sense inputs of the Allegro / TI driver to allow PWM current limiting).
Obviously, I'd be LOSING the internal thermal monitoring ability of the chosen driver chip in order to gain the additional current drive capabilities of the beefier full bridges, but this is something I can accept.
Since the driver will be direct-mounted and hardwired to the motor itself, there's less chance of an inadvertent disconnect of the motor whilst powered up (which might help to keep the smoke in).
I'm leaning more towards TI over Allegro (for the extra 1/32 microstep setting) with something like a pair of DMHC3025LSD as the full bridges.
Ideally, I'd like to use a form factor that enabled use on Nema17, 23 or 34 motors (snap the PCB to the desired size).
1 * power connector (Motor supply)
1 * signal connector (Step/Dir/Gnd being bare minimum - possibly Enable/Fault/Home/MSn too)
1 * Dipswitch or header (to allow local setting of MSn pins)

Not sure what else I've missed yet...

In future, I'm thinking of doing likewise for the thermistor inputs and heater outputs too.
i.e. Simple 'decentralisation' leading to the elimination of the quasi traditional RAMPS type board.

Quote

I've heard that there are 'issues' with the stepper drivers letting out their magic smoke if they are disconnected when powered.

Does anyone here know the failure mode that this causes?

Disconnecting the stepper while powered up induces a high voltage and if you're unlucky, a spark of the just disconnected connector, pins still very close, flies back to the Pololu and ... bye bye Pololu. The same principle is intentionally used to operate spark plugs.

---

Generation 7 Electronics	Teacup Firmware	RepRap DIY

Thanx for the reply!
I 'd have thought that was fairly easy to circumvent by simply throwing some zener diodes across each field winding output to 'safely' conduct away the induced high voltage shocks?
The motor winding is not a HUGE inductance (generally a few mH in our case), so the amount of energy stored in the winding is also not huge (given the 0.5LI² formula), so the zeners wouldn't have to be huge.
Assuming the supply is < 20V, simple 24V zeners should do the trick?

Having said that, I still like the concept of placing the stepper drivers right at the motors themselves and hardwiring them to the motors. I simply don't think the centralised stepper drivers is the ideal choice...
Surely the (comparatively) long conductors to the motor fields act as a decent RFI transmitter (switching a couple of Amps with a 40kHz square wave will surely produce some decent RFI and then there's the PWM on top of that!). I'm also not entirely sure I like the fact that each field winding conductor has to flow through an additional pair of IC socket pins (one pin for Stepstick->RAMPS et al PCB and another pin for the PCB->motor connection). I'm pretty sure header pins aren't designed to handle more than about 1 Amp!
I'm still deliberating with myself what the DOWN side is to routing signal conductors (eg: Step/Dir/En/Gnd) instead of the high current conductors (eg: Out1A/Out1B/Out2A/Out2 all the way to the motors. If anyone can think of a decent argument against it, please let me know...

Steppers can run very hot, and besides their own windings may be emitters perhaps not just the wires forth and back. To put the right on the motors, the thermal model has to math out starting with like 80C still air, and imo the enclosure would need to be shielded (unlike regular away-pcbs. This is probably why the steppers drivers of which i saw pictures mounted on motors have aluminium enclosures, i think both for thermal dissipation and also for shielding. Starting with 80C ambient (instead of 25C) is a big handicap and i think thermal protections is a must, or it may burn very easy. Also, when starting with 80C ambient temp everything has to be derated, for example capacitors are also power devices, would need higher temp capacitors and even so they will have less than nominal values, plus probably the MTBF (med tim be4 failure) will be like halved. And also i think everything else may use derating too, even the simplest diodes. For this to succeed i think important aspects are thermal model, shielding, and good derating on probably all parts, with special attention to values of absolute peak -currents. Because the operating coditions are already sort of extreme, i would try keep the solution somewhat classical and something that is proved to work with good safety margin. So i would not try chaining h-bridges, and i would not use one without thermal protection, and if external mosfets are a must (may not fit on nema17), then myself i'd rather go with ics made for external mosfets, which would probably be a more proven solution than alternatives. Just my opinion ofc - you make your own calls. I will be looking forward to see this working.

Tbh, sounds like you are already set on the scheme and looking for confirmation. I don't think there is a very great need to make the motors hot pluggable, the problem is easily circumvented by turning the power off first. Of course, this will continue to catch out newbies, but whether it is worth making it newbie-proof is up to you.

I think the header pins can easily take 4-5A. Going from 4 wires to 6 wires or more doesn't sound like an improvement, if 4 or 5 sets are needed. Reprap electronics has gone from decentralised electronics to single board solutions, I think these are generally better as they simplify wiring and are easier to use. If RFI is an issue it would simpler to address it with proper shielding of the motor wires.

Regarding RFI:

One of the issues with multi-board solutions (like the Gen3 electronics used on the Makerbot Cupcakes and so on) was that a lot of noise was induced on the data cables between boards, causing odd issues.

There is no reason that you can't go back to the same thing, if you make sure the cabling between boards is shielded, or you use a different signalling standard (eg: current loop rather than standard TTL voltage signalling).

Note: The noise on long 5V lines is bad, and with 3.3V electronics the signal to noise ratio is bound to get worse, so at an absolute minimum, signal shielding is an absolute must!

Just as a 'follow-up' posting...

There's already a company that markets stepper motors with 'built-in' controllers...
Specifically, LinEngineering link.

While some of their offerings aren't too well suited to the Reprap community, they have others which are a very good fit!
The D and DE models should be quite simple to integrate while the C and CE models would be more work.

IMNSHO, they should have included some additional 'protection' into the devices such that they didn't risk 'letting out their magic smoke' if disconnected while powered up. (Since the motor is permanently connected to the associated driver electronics, I would have expected this protection to be intrinsic? However, I've obviously missed something since they included the usual warning against disconnecting while powered on)
Also, they only support 1:1, 1:2, 1:4 and 1:8 micro-stepping on the Nema17 series. (Significantly more micro stepping ratios exist on their Nema23 and Nema34 versions!)
I _do_ like the fact that they've elected to isolate the input signals with opto-couplers