# Slew Rate Booster

Discussion in 'Electronic Design' started by Haude Daniel, Mar 28, 2007.

1. ### Haude DanielGuest

For the umpteenth time I'm thinking about a way of replacing horrenduously
expensive commercial piezo motor drives that we use with our STMs with something
cheap and simple.

The motors are driven in a slip-stick fashion using a waveform like the one
shown here:

http://www.nanoscience.de/group_r/members/dhaude/waveform_walker.png

(To reverse the direction of motion the polarity needs to be inverted).

The S-shaped part of the curve isn't terrible critical. What is
critical is the steep slope between the slow parts. With a 20n capacitive load,
this is almost 1A of current at 44V/Âµs

The obvious approach to this would be to follow a low-voltage waveform generator
with some big-ass HV amplifier. I've played around a bit with stuff along the
lines of Fig. 3.75 in AoE but never got it anywhere near the needed performance.
Besides, the thing needs to be short-circuit proof in both polarities. The split
supply is a requirement because I need ~400V swing but a maximum voltage of
~200V across the piezos.

So I thought if it weren't possible to exploit the fact that we know exactly
when the steep rise is supposed to happen, and to add a "yanker stage" to the
output of a slow and cheap HV amplifier like this:

+250V supply----+--V.Reg.------+---+ +220V
| | |
| D 1uF
| +-G |
| | S GND
| | |
+------+ +----+---+ | |
|wavefm| | | | |
|gen. +---------+ HV amp +---R1-|---+---R2----+
+---+--+ | | 1k | | 100 |
| +----+---+ | | |
trig _____|_ Gate__/ | 1n..20n
| | | Drv. \ D |
+--L.Shift--+ | ---G GND
| S
| |
-250V supply-----+--V.Reg.------+--1uF--GND
-220V

The slow amp needs to be isolated from the yanker in some way, otherwise the
yanker would probably upset the feedback loop. For this I added R1.
The idea is that once the rising slope comes along, the top MOSFET essentially
shorts the output against some power rail whose voltage is adjusted to the peak
of the amplified waveform. The FET is kept on for some time until the slow amp
has caught up and then turned off, so that the HV amp can take over for the slow
part. A mismatch between the yanker supply voltage and the amp peak output of a
few volts wouldn't matter.

There could be intrinsic short-circuit protection by foldback-limiting the
yanker supplies to the few mA of average current that this thing actually needs.
In case of a short circuit the voltage of the 1uF storage caps would collapse,
and the HV amp would protect itself.

Before I start thinking about actually implementing this beast I'd like to hear
if anybody thinks that there's anything fundamentally wrong with it, and if
there are better ways to solve the problem.

And an adjustable amplitude would be great as well... nah, let's leave it.

--Daniel

2. ### MooseFETGuest

Things like this have been done in the past and work fairly well so
long as you keep them from crashing into the rails.

In the usual form of this, the circuit works as an inverting
amplifier.

The slow amplifier has to be made such that it doesn't attempt to
clamp its inverting input to ground. It also needs to have a high
open loop output impedance.

The fast boosting section tends to make the output do what the input
commands it to do, so the error signal on the slow amplifier's
inverting input is greatly reduced.

If you are careful about preventing oscillations, you can use a couple
of comparitors on the signal at the inverting node to develop the
trigger signal for the booster. In your case, I think you could
simply trip oneshots and a bit of interlock stuff from the
comparitors.

3. ### John LarkinGuest

That looks OK. The gate drive could be some dirt-cheap ASDL
transformers.

+240
|
|
d
+-----g
| s
| |
| |
| |
in------ wimpy hv opamp------------+---R--+-------+------out
| |
| |
| |
| s
+-----g
d
|
|
-240

so the fets help when the current gets big enough to drop a few volts
across R. Some sort of compound feedback would probably be needed, but
that shouldn't be hard. As you suggest, current-limit the hv rails to

For more interesting dynamics, R could be a true current limiter, like
some depletion-mode Supertex fets.

The "wimpy hv opamp" could of course be based on my optocoupler trick,
so the whole thing becomes about a dozen cheap parts.

John

4. ### Haude DanielGuest

Looks good.
You mean, some feedback around the HV amp only, and some more around the
whole thing?
Yeah, but the wimpy amp now (as opposed to my original idea) needs to
have the full 40V/Âs slew rate.

Actually I have the schematic of a commercial unit before me.
They do use some optocoupler scheme for the positive drive,
but I haven't fully understood it. They have the gate of a
BUP37 IGBT riding at 14V on top of the output voltage, and a
Darlington NPN between the IGBT's emitter and the output whose
base is controlled by the OC.

--Daniel

5. ### Fred BloggsGuest

If you can't adapt the AoE Fig 3.75 ckt to fit your needs, then you
don't know what you're doing. Take your question to SEB...

6. ### Haude DanielGuest

If it is that simple to adapt the circuit to do >40V/Âµsinto
20nF with bipolar short-circuit protection, I'll happily expect
your suggestions to appear on sci.electronics.basic.

--Daniel

7. ### Jim ThompsonGuest

Not to mention that Fig. 3.75 doesn't really address, even with
modifications, what Daniel specifically needs to do.

Daniel doesn't simply need a power booster.

It might be more accurately called a "reset".

...Jim Thompson

8. ### John LarkinGuest

Something like that. The local fb would be fast, and the global fb a
bit slower, especially if the load is capacitive.
Then ignore my suggestion! If tha driver amp isn't already this fast,
the follower can't help, and you will have to go to the pulsed
booster. That has interesting dynamics.
Sounds strange. Post to abse?

John

9. ### John LarkinGuest

Fred, of course, doesn't want to help. He wants to demonstrate that
we're stupid and he's smart. So far, it's not working.

John

10. ### Fred BloggsGuest

The main thing he needs is voltage gain, and that 40v/us seems
suspiciously close to a standard 35v/us interference level. It is best
to just adapt the AoE circuit with suitable protection and current capacity.

11. ### John LarkinGuest

But it's more fun to design stuff.

John

12. ### Haude DanielGuest

An edited portion of the ckt can be found here:

http://www.nanoscience.de/group_r/members/dhaude/sed/hv_amp.png

Note the line labeled "boost". This goes to a logic output in the
wave generator section to help the high side driver to get its
ass in gear on a rising slope. The negative side doesn't seem to
need this.

I omitted the output current monitor / overcurrent shutoff portion.
And I can't make up my mind if the circuit is ingenious or stupid.
All I know is that the device has been working flawlessly for some
15 years now and has taken lots of abuse.

--Daniel

13. ### Haude DanielGuest

Actually I'm beginning to like this circut. I've figured out hoe the
negative side works:

When U3's output is at or above 0V, R21 feeds 4.5 mA through D11 and D13,
causing -490 V to appear at Q5's gate which is close to the -488 V at
Q4's gate, so Q4 and Q5 are off.

Whan U3 goes negative, its output starts robbing about 1.1mA/V from this
current, which makes Q5's gate move towards the negative rail with 4.2 V
per volt change on U3's output. I need to look up the subthreshold
characteristics of the IGBT, but at any rate this turns the Q5 and Q4
cascode-like combo, letting the output go negative. The R13/R28 divider
provide a little negative feedback to Q3's base, probably for stability.

In the positive leg, Q1 and Q2 form a cascode driven by U1, riding on
top of the output voltage. D4 and D5 provide fast turn-off of Q2 on a
falling slope, and they make sure that Q2 can never turn on when Q4/Q5
are on.

At first I wondered why they would use IGBTs instead of MOSFETs. But it
turns out that 1200V MOSFETs are rare (and may be even rarer back in
1990), but a big, cheap, wildly overrated IGBT like the FGA15N120 is easy
to get.

I think I'm gonna prototype this. But first I need to come up with some
good SPICE models for optocouplers and IGBTs. Is Mr subthreshold, Win

--Daniel

14. ### Guest

Actually, fig 3.75 can be used to address the issue. Until
last year we always made our own stick-slip piezo stepping
drivers for our STMs, basically using the MOSFET totem-pole
drive scheme shown in fig 3.75.

The problem with using a "reset" approach is it assumes
you want to take the piezo all the way to the rail, which
you may not want (although some commercial units do this).
We wanted ours to be fully programmable.

What's needed is a high enough current capability and a fast
enough slew rate. Fig 3.75 can easily operate at even 1A or
above currents, because the pulldown and pullup devices are
power MOSFETs. The Institute's AMP-10 circuit is an enhanced
version of fig 3.75 with bipolar capability. An early version
of the AMP-10A that I made in 1991 had a 40V/us slew rate,
the AMP-10A-4 had 60V/us, and some later ones were faster yet,
delivering high currents into the modern lower-voltage higher-
capacitance piezo actuators.