2
\$\begingroup\$

In a special project I'm trying to do PWM current control of a electromagnet like application.
This project is special because it has to pass very strict EMI tests (<30Mhz radiated emissions).
Almost all products with typical PWM using low side mosfets fail (backlight controller, dc-dc, servo's even dc fans fail) it's really strict.

So I need to slow down the slew rates of the low side switch nodes. Attempts with mosfets were futile, they are inherently too fast and slowly controlling them through the linear region exceeds the SOA of almost all I can find. Except special purpose linear (>30$).

Even linear has crossed my mind, it would be very wasteful, I'd need 3 to 4 transistors for that much energy.

After some tests of an initial prototype this was made. I've drawn up the schematic with all relevant parts of the PWM switching.

Setpoint is fed into a comparator with a triangle, with slew-rate limiting properties. This works great, it slowly raises to +15V. And goes down to 0V. (There is no -15V.)

Nominal current is about 3A, I've tested with 0.5A.

schematic

simulate this circuit – Schematic created using CircuitLab
(Circuit as build, does not perform well)

The darlington is an MJH11020. (200V, 15A)
Correctly notices by Kevin White in the comments, I've corrected the schematic

Turn-off

Turn off works great! The voltage on the output does not have any fast edges. turn off

Turn-on

This is more challenging. The turn-on does not appear to react to the slew rate control very well. I've even tried this without Q2, but that does not have much effect. There is also ringing, approx 85 Khz, that worries me most. turn on

There is also severe ringing on the base voltage I have no screenshot of unfortunately.
Something is not right in this circuit, and I'd like to know what.

Some more experienced colleagues pointed to this schematic like how this was done "back in the day" however, that was also done using components you can't get anymore.

What can be done to make this circuit behave nice and softly? Possibly getting rid of the ringing would be nice as well. Maybe the filter causes this? Unfortunately the filter I cannot lose because of noisy slip rings.

Update on comments #1:
Tim Williams asked in the comments for more details on EMC side of things with the environment. Unfortunately, I cannot share photos or diagrams of those.
The board will be in an aluminum housing, only power comes from outside. A filter was placed for this.

Specs wise the maximum current is 5A, nom 3A.
There will barely be any slew rate in the current. The 1.5 Henry inductivity of the load makes sure of that. I've selected a slow diode to prevent ringing from that. I was unable to measure this, so I suspect this is no issue.
The main concern is the rapid changing voltage on the Collector, my target was 100-500 µs for the transitions, so < 1V/µs.

Update:

I've performed some more measurements. And was unable to get a better slew rate using the circuit assembled on the board. I was able to reduce the ringing visible on the scope images.

Best result was obtained with Cx on the output side of the choke removed. When this is added, there is soo much capacitance and the circuit is unstable.

enter image description here
Yellow: current in R1, Blue: current in L1, Green: Vbe, Orange: Vb-gnd (includes R1)

The little yellow hump during turn on (green falling) is a surge through the remaining Cx, this actually helps reducing the slew rate on the output. Add the other Cx back to it and the circuit startings ringing and pulls Vb negative. (not good)

This was the best I could get. Around 70V in 20us (3.5 V/us). Probing with the TinySA does not yield any measurable signals on the bench.

EMC profile

Target EMC profile is <25 dBuV at 1 Meter E-Field. This appears to be very difficult. Even the test room lights must be disconnected.

emc profile
(Yes, Limit A)
Those who recognize this profile now know why I can't post full circuits or pictures.

Update: It appears the above circuit may just work. emc plot

\$\endgroup\$
11
  • \$\begingroup\$ You don't say how much current/voltage you are trying to control. That schematic has faults in it (capacitor to opamp with no bias for example). Are there constraints on size, weight cost etc. Why not take feedback from Q3 emitter? \$\endgroup\$ Commented May 10, 2024 at 16:35
  • \$\begingroup\$ @KevinWhite Yes, there are weight constraints. And costs, I'm replacing super simple and small thyristor circuit. \$\endgroup\$ Commented May 10, 2024 at 18:27
  • \$\begingroup\$ EMC is a holistic topic -- please include any and all details possible about how the system will be arranged, what cable(s) are connected and to where, what can be inside enclosures, what has to be outside, etc. \$\endgroup\$ Commented May 10, 2024 at 18:45
  • \$\begingroup\$ You forgot the most important specs: what is the maximum current you need to drive, and what is the maximum slew rate of current you want when changing the current. Please edit the question to add those in. Maybe you need 3-4 transistors, maybe not. We can’t know unless you tell use the specs! So far we only know it’s a 15Ohm load. As far as I can tell, a basic audio amplifier output stage, or even an IC audio amp, will drive it very happily, with full protection from overheating and dead shorts. Your time is way more valuable than energy wasted here! Make it simple and robust. \$\endgroup\$ Commented May 10, 2024 at 18:50
  • \$\begingroup\$ For all I know you can close a DC control loop around a slightly modified off the shelf old-school linear audio amplifier and it will do the job well. You could buy a used Technics class AA stereo amp, DC couple it, and bridge the load. Whisper quiet, no EMI concerns, and cheap too. \$\endgroup\$ Commented May 10, 2024 at 18:53

1 Answer 1

Reset to default
2
\$\begingroup\$

You'd typically use a Miller integrator to achieve controlled slew rates, like this:

Slew-rate-controlled NPN switch

This circuit uses a pair of modified current mirrors (Q2..Q5) to generate a bipolar drive current, which is then fed into the Miller integrator (Q1/Q6/Cslew) to create a linear voltage ramp at the output.

Note, though, that you shouldn't connect capacitors directly to the output of this circuit as that will cause instability.

\$\endgroup\$
1
  • \$\begingroup\$ It can be a useful technique to connect the feedback capacitor directly to the output to enable the amplifier to drive heavy capacitive loads. See for example: ti.com/lit/ds/symlink/lm8272.pdf?ts=1715603654708 \$\endgroup\$ Commented May 13, 2024 at 18:44

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.