High voltage safety and design considerations

[Someone]

as I mentioned before, we have to drive 12 actuators. I think off the shelf amps would be way too expensive to use.

You added that very important information a long way into this. Deciding on a topology to power that requires knowledge of the layout and relative costs/distances to carry the different cables.

Adding information a little at a time you're going to get most people to leave and stop offering assistance.

batteries will always be floating with respect to ground (actual ground, not batteries mid point) so if someone accidentally touches the robot during operation, it shouldn't be a problem.

So how is the 16 bit DAC interfaced? Making systems isolated requires taking account of all the parts, interconnections, connectors, accessibility, for everything including the actuators (which you still haven't touched on).

Take a big step back from your schematic. How do other products drive heavily inductive loads? If current is the control parameter, why aren't you measuring it?

You've focused on a very particular design which the more experienced people here see many issues with. If you wish to add undefined constraints of cost and time onto this then you're on your own. Its not safe, you lack the basic knowledge to make it safe, and need to seek assistance from people around you with the appropriate knowledge rather than fishing for answers on the internet to reinforce your plan.

[max_torque]

The problem with that part is it uses bootstrapping capacitors and presumably he wants it to be work down to DC.

Seems unlikely at the voltages we are talking about that either 0 or 100 % duty will be required. 1.5k requires 37v to drive 25mA, so if the supply is greater than 37V, and it looks like it needs to be around 100V to get the dynamic performance, then at no point will the duty cycle sit for any length of time at the end stops.  There may be some short periods where a high or low duty is required transiently, but in those cases, sizing the bootstrap caps and limiting the duty to say 98% and 2% is going to do the trick

[max_torque]

A +ve and -ve rail architecture does have some safety advantages, especially if the potential of those rails is kept within the normal 50 to 60Vdc working limit for "low voltage".  For example, a pair of 48V batteries could be connected in series, and the centre point taken as the "ground reference". That way, touching just one conductor would be "ok" and you'd have to touch two conductors to get a 96V "shock".  I work with HV DC all the time, ime, realistically (NB: but not legally!!!!) you're unlike to get any kind of significant shock of "just" 96V, unless you are stood in the shower and are jamming the live conductors through your skin.....)

A fast semiconductor fuse on each battery, and probably also on each driver feed rail, will give the necessary short circuit protection to avoid fire risks, those fuses cannot avoid the electrocution risk, but that is low at <100Vdc


NOTE:  the text up their has a lot of words in "". The OP must determine for themselves both the legal and practical requirements (ie best practice) when working with systems above the Uk's "low voltage directive" potential levels.  You universtity tutor and department head should be able to advise as to these requirements.

[Zero999]

I had a look at voltage boost circuits for op amps, to be honest they are way above my level of understanding and each manufacturer recommends a different circuit (I had a look at TI, AD and ST micro app notes) so the ADHV with dual supplies is the simplest choice here. I will use HRC fuses which should be able to deal with this without any problems.

There isn't a single correct answer, regarding which booster to use, which should depend on what the output voltage/current is required, rather than the op-amp or manufacturer.

Isn't there anyone who can help you with this? I thought the purpose of doing a degree was to learn, even if it's material which isn't on the syllabus.

We finally settled on 3 7s and one 4s LiPo packs for a voltage range of 105-80V. This will be doubled for the + and - rails. The problem with the birdge approach is the high part numbers, especially needing 2 op amps. The ADHV costs about 15$ a pop and I need 12 copies of that circuit. I think using batteries will be safe since they will be floating with respect to ground and low current fuses will prevent shorts or other issues from causing any major damage.

The Howland current pump I posted doesn't really need two op-amp amps. It just needs an op-amp and a very basic inverting buffer, with a gain of around 1. I used an op-amp in the design for convenience. A crude inverting amplifier can be made from discreet components. Now only one high voltage op-amp is required, rather than two.

[magic]

I think the original circuit isn't really that bad. I'm not sure what's causing the ringing, I got much better results with my DIY amp below. Perhaps some manipulation with C1 or R2 would solve it.

One thing you certainly need to do when driving inductive loads with BJTs is to add catch diodes (D7-D10 on my schematic). This is regardless of whether it's a DIY discrete power buffer or some audio amp IC (unless it uses MOSFETs and is rated to withstand backdrive). Reason is, when drive is turned off, the inductor will still try to conduct and it will pump current into output capacitance of the switches, slamming them against the opposite rail. See my sim when direction is reversed. This current needs to be allowed to bypass the transistor and flow into supply capacitors, otherwise B-E junction will be destroyed by reverse voltage.

Not sure what's the point of R1 in your schematic. Not sure if R5 really helps against crossover too, that would need to be solved by biasing the output trannies which is a PITA. You probably don't care, it only is a problem during transition from sourcing to sinking, not in steady state. It only is a short blip of distortion, which may not even be visible due to the massive filtering effect of load inductance.

Finally, yes, I have simulated the solution I mentioned yesterday, using one HV rail and a single differential amplifier to kill three birds with one stone:
- compare floating current sense (two shunts are used and their output averaged) with ground-referenced drive signal
- amplify the high voltage swings needed to drive the load
- generate balanced, complementary load drive and maintain load in the middle between supplies

Sadly, I'm not aware of any off-the-shelf chips that do that at 100V. So I made a discrete version with a few 150V BJTs and power darlingtons. I suppose you won't bother building it so I'm not attaching ASC unless somebody asks for it.

[Zero999]

I agree, there really isn't anything wrong with the original circuit. It will work. The disadvantages are the very high voltage supply and expensive op-amp.

I've analysed the Howland current pump again and realised it can be bootstrapped using zener diodes. If the gain is low (0.1 in this case) the input voltages remain within a jelly bean op-amp's common mode range.

The only problem this any the circuit I previously posted is there will be some crossover distortion, as the discrete inverting amplifier is class B, but that's just a matter of biasing if it's an issue, although this isn't for an audio applications, do the distortion is probably a non-issue.