Overvoltage problem on X and Y servos

Unresponsive servos:

I've tried running an existing program to see how the Newker system gets on in its new surroundings. Answer - initially seems to get started but after a couple of fairly "impressive" (= fast) moves, all movement ceases. The controller still thinks it's scything its way through the stock but in reality the motors are inert - and more tellingly are free to move. Given that the PSUs are still running, that sounds like an overvoltage type issue. Cycling the mans supply restores the servos - until you try to follow a toolpath that is,

Looking at the servo drives, these are pulse, direction and 0V, so clearly no error feedback signal (yet) to tell the controller of any problem. I'll be connecting up a servo status line to the controller at some point, once I've figured out how (RTFM etc).

Sure enough, connecting up the comms cable shows "overvoltage". This explains why the servos shut down independently of each other. 

So the rapid moves using the current default values in the Newker are a bit ambitious. I could perhaps fix the issue by use of more appropriate settings but I'm not massively optimistic that this would be the best solution. These Leadshine SMPSs have very little output capacitance, so it will take very little decel (regen) to provoke them into OVP. Given the relatively large kinetic energy stored in a rapidly moving table / saddle / vise / work combination, I don't see a couple of tiny electrolytic caps having the capacity to swallow that without a decent voltage excursion.

What to do?

3 solutions come to mind:

1. Fit some additional bulk capacitance to help the PSUs swallow any regen energy. That's what you'd have on a std 50Hz transformer / rectifier / cap arrangement. Space is a bit tight but it's possibly just about manageable.
2. Fit an overvoltage clamp aka braking module or "power zener" circuit.
3. Swap the X and Y SMPSs for a good old mains transformer and be done with it. Due to the 50Hz supply frequency, they tend to come with large bucket caps anyway.

There are some issues with these options of course:

1. I'd have to hope the SMPSs are happy to start up into large a capacitive load. I suspect that's a low risk. I'd probably want to parallel up the 48V supplies or possibly even remove one of them. They should be fine in parallel although one of them is likely to take the bulk of the load without any form of current sharing scheme.
2. I could make up a braking resistor circuit or possibly just fit a bastard great zener diode / TVS / MOV to gobble the transient. Downside is that I don't know how good the transient response of the SMPSs is following an overvoltage above the set point. Furthermore, the voltage swings will continue to be pretty rapid without any additional bulk capacitance.
3. Yes, I could replace the SMPSs with a bastard great transformer and bulk cap but that would require a significant reworking of the cabinet and its tray. I'd also need a soft start circuit to avoid the old chestnut of the tripping MCBs at turn on. Besides, as an SMPS career engineer, that would be wimping out.

The solution:

I'm going to go with a big fuck off electrolytic with a voltage clamp to prevent transients. For that, initially I prefer the sound of a passive (MOV / TVS) voltage clamp. But an active clamp is not really such a problem - I could go for the comparator / FET / resistor solution easily enough. MOVs and TVSs don't have very tight tolerances and I need something that will act consistently between 48V and whatever the OVP threshold of the DMM Tech servos happens to be. Sounds like 53-55V or so region, with only a +/-3V tolerance.

Looking at the cabinet, the available space is about 140mm x 150mm by 120mm high. That's the spare area on the tray just below the VFD where I fitted a DIN rail terminal block which so far has no use. I don't see any reason not to "repurpose" it. First come, first served and all that:

At first glance, the caps I used on The Shiz for those CNC Drives systems appear to be too bastard big for this space but there again the inertias on that machine are several times bigger. Having said that, they are 70mm dia and appear to be about 100mm long, so would fit OK here. I expect they came from CPC or Farnell - dug through past orders but couldn't see any trace of them. None of Farnell, RS or CPC order histories seem to go back beyond 2018 - WTF??

This looks like the man for the job:

• 68mF, 63V from BHC, 76mm dia x 105mm long
• 76mm clamp
• £38 + vat (gulp)

Active clamp (braking resistor):

That just leaves the slight matter of the active clamp itself. Not much to these - something like this would do the trick:

• Rhy provides hysteresis to the threshold which is presented to the 431 by the pot chain on the left. Don't want to go too wild on the amplitude, as the servos have to cope with the resulting triangular waveform.
• The frequency of switching will be dependent on the regen power coming out of the motor, the hysteresis threshold levels and the value of the braking resistor (Rbk) itself. Don't want that frequency to be too high, as the servos have to live through it.
• In the limit, the value of Rbk needs to be low enough to swallow the max regen power at 100% duty without the voltage exceeding the OVP threshold of the servo drives. With a 400W servo and 60V threshold, that suggests Rbk should not exceed about 10R or so. That's going to be quite conservative, give that I doubt it will be able to regenerate at 400W, even if there were sufficient kinetic energy available.
• I've got loads of aluminium clad resistors and stupid great FETs, so no shortage of choice there.

I hate the idiot approach of trial and error based around simulation but I have a pretty clear idea what I want here, so may as well flash up SIMetrix and check it out.