3
\$\begingroup\$

Background: Using 33kHz PWM through an RF isolator [Si8710AC], to drive a power MOSFET [AUIRFZ44N], to control a high-power LED (8-10 watts). Isolator output is open-collector pulled up to +12V, and its output is beautiful. Its leading-edge slows down when attached to MOSFET (Miller effect?) but basic operation seems okay. MOSFET drain is connected to a constant-current bench power supply presently set at about 2 Amps.

Issue: There is significant overshoot/ringing at the MOSFET output drain at turnoff (i.e. when the voltage returns). O-scope is a simple handheld, but think issue is real--from what I’ve read this is not surprising. My question is, could the constant-current power supply be aggravating things, in that it might initially try to react to turn-off by forcing things... and then backing itself off?

screenshot of MOSFET drain

I previously read through posts to help get to where I am... thanks! Any additional guidance would be appreciated.

Edit: 10/12/14 @ 16:40... Thank you all for your comebacks. I need to read through them a few times to make sure I follow.

In the meantime, allow me to provide a little more color—and describe some things I tried in between—so I know I'm not missing something else. I don’t really have much in the way of "stock" components, so each iteration I make involves cycles of ordering online and waiting for stuff to show up, then testing, etc.

Shown below is a rough schematic of the circuit under discussion, followed by screenshots of the waveforms at different points in the circuit. The first screenshot is of the PWM waveform at the output of the isolator by itself, i.e. when it is not connected to the MOSFET. The second screenshot is taken at the same point, but with the isolator connected to the resistor that feeds the MOSFET gate. (The output of the MOSFET at the drain was posted above.)

Rough schematic of MOSFET circuit

Isolator output--floating

Isolator output--connected

Isolator datasheet here: Si8710

Side Notes:

  1. The MOSFET is actually connected to an array of LEDs. But they are separately selected—and only one at a time. To verify that this was not complicating things, I tested the MOSFET against a single high-power LED with short wires and still reproduced the problem. (For some reason, however, a standard 5mm 20 mA LED did not show much ringing (?).)

  2. The PWM is run at 33kHz. The MOSFET needs to be able to handle the full range, i.e. 0% to 100%.

Edit: 10/13/14 @ 20:15

I located some 2N3906 and 2N3904 transistors, added the complimentary BJT circuit suggested by Spehro Pefhany, and put a 10 ohm series resistor in parallel with the existing 120 ohm gate resistor to lower the latter to about 9 ohms. The following 2 screenshots show the waveform at the MOSFET drain:

MOSFET output waveform with complimentary BJT circuit and gate resistor change

MOSFET output waveform with complimentary BJT circuit and gate resistor change--scaled

(I don't really have any options regarding adjustment of the wiring. As noted above, the MOSFET feeds an LED array and the configuration is fixed.) Per Russell McMahon's suggestion, I tried to add a reverse-biased diode between the gate and source. Couldn't find any Schottky diodes locally; all I could find was a 12V Zener diode so I lowered the drive voltage to 9 Volts:

MOSFET output with previous mods plus zener diode

MOSFET output with previous mods plus zener diode--2nd photo

The bottom line is that these suggestions helped, but I'm not there yet. I'll look into getting components to implement some of the other suggestions. Thanks again.

10/15/2014 00:25 EDIT:

Additional test results documentation; tried to be more systematic:

A. BJT follower circuit suggested by Spehro included. LED array system wiring bypassed, to isolate MOSFET. Instead, a single high-power LED was placed on breadboard next to the MOSFET.

  1. Slight improvement in MOSFET drain output waveform from dropping series gate resistor from 120 ohms to ~5 ohms. First two screenshots are for 120 ohm resistor, next two are for 5 ohm resistor:

Results using 120 ohm series gate resistor Results using 120 ohm series gate resistor closeup Results using 5 ohm series gate resistor Results using 5 ohm series gate resistor closeup

  1. Same configuration as 1., but 10 paralleled standard LEDs used in place of single high-power LED. Closer examination shown in second and third screenshots suggests that little actual noise was present for this configuration:

Results for 10 paralleled standard LEDs instead of high-powered one, 1 Results for 10 paralleled standard LEDs instead of high-powered one, 2 Results for 10 paralleled standard LEDs instead of high-powered one, closeup

  1. Same configuration as 1., but 5 series standard LEDs used in place of single high-power LED. No ringing/overshoot apparent:

Results for 5 series standard LEDs instead of high-powered one, 1 Results for 5 series standard LEDs instead of high-powered one, 2

B. Using a high-power LED within system array (i.e. the intended configuration), BJT follower circuit and series gate resistor of ~5 ohms.

  1. MOSFET drain output waveform when using constant current power supply set at 3 Amps:

For LED at 3A drive current; MOSFET drain output waveform through system

Summary: Seems there is no problem running standard LEDs with this circuit configuration. But when using a high-powered LED--even while bypassing the system wiring--the MOSFET drain overshoot/ringing problems become evident. Will continue to work to mitigate their effects.

10/17/2014 00:25 EDIT:

Status Update: The signal up to the MOSFET gate looks fine. Read up on snubbers and tried a series resistor-capacitor snubber (R: 27-53 ohms, C: 4-10 nF) between the MOSFET's drain and source. That effectively eliminated almost all of the ringing. Resistor gets somewhat hot, though :-). The overshoot remains.

At this point I'm kinda stuck:

  • Wasn't able to locate a suitable Schottky diode locally, will order something online.

  • I isolated things, and the reverse-biased zener diode didn't seem to have any effect? I'm still puzzled that I get those "70V" spikes; it doesn't seem to make sense in the context of a 3V LED. Any thoughts? (@Russell, @Spehro, @gbulmer)

\$\endgroup\$
  • \$\begingroup\$ What happens when you increase the Power Supply current limit to, say, 3A? What happens when you drive a more modest LED, say 5 x 20mA LEDs? \$\endgroup\$ – gbulmer Oct 12 '14 at 1:52
  • \$\begingroup\$ I don't think I explicitly did this while viewing a single waveform. But I did view the output at different duty cycles (hence different current values). I could only bump up the current significantly at the higher duty cycles, it wouldn't move at the lower ones. I haven't seen much discussion of using a CC power supply on the mosfet output in this way--they usually show some +Vcc. So I was unsure how to think about it. But I tried using the bench supply in a constant voltage mode (and also tried using a +5V AC/DC adapter) and they seemed to show the same issue. \$\endgroup\$ – user56021 Oct 12 '14 at 21:14
  • \$\begingroup\$ [In constant current mode, the P.S. shows +14V, which I don't get, but I think that includes the internal drop of the supply.] \$\endgroup\$ – user56021 Oct 12 '14 at 21:31
  • \$\begingroup\$ @gbulmer, to more directly respond to your questions: 1.)The ringing increases with current level up to a point. It stops growing somewhere between 2 and 3 Amps. 2.) I directly compared a high-power LED against multiple standard ones on a breadboard running short wires to the MOSFET. Results above seem to support the adverse effects of higher current. 3.) (Regarding the +14VDC mentioned above, I looked more carefully;. The constant current power supply voltage shoots up when running in low duty cycle. When near 100% duty cycle, the voltage is closer to what one would expect for LEDs.) \$\endgroup\$ – user56021 Oct 15 '14 at 3:26
3
\$\begingroup\$

The drain appears to be ringing at about 600KHz.. Probably due to inductance in the drain circuit. The simple circuit you are using to drive the MOSFET is very asymmetrical.. the turn-off will be much faster than turn-on because your pull-up is much higher value than the ~50 ohm resistance of the MOSFET in the Si8710AC, so you are seeing the ringing.

I suggest a complimentary BJT emitter follower after the Si8710AC with a series resistor to the gate selected to trade off ringing vs efficiency. You can also minimize inductance in the drain circuit by shortening or twisting the wires arc., add a snubber or even a flyback diode to the supply.

enter image description here

\$\endgroup\$
  • 1
    \$\begingroup\$ The driver is great until I connect it :-) In testing an LED with short leads to the mosfet, I was attempting to isolate/eliminate external wiring, etc. as suspects. Had tried a single BJT to invert the PWM, but pulled my hair out. Will look at this. Thx.. \$\endgroup\$ – user56021 Oct 12 '14 at 21:45
2
\$\begingroup\$

The following are easily tried and may well work:

All are synergistic and can be tried in any mix and order.
In this case the Schottky diode fix seems most liable to be useful.

Connect a small Schottky diode gate to source of MOSFET & reverse biased (Anode to source, cathode to gate). This prevents oscillatory waveforms on the date with negative excursions. If these are present the negative half cycles are clamped by the diode and reduce signal energy very rapidly.

Zener diode gate to source reverse biased (ie zener conducts when gate reaches zener voltage +ve going). Vzener > Vdrive max and below MOSFET Vgs max. Closer to Vdrivemax (but greater) the better.

Series gate drive resistor between drive source and gate - can be "a few Ohms".

Gate driver as Spehro suggests - the above Schottky and zener fixes can both be used with this driver and may be useful in some cases. FWIW - I have used the arrangement Spehro shows in designs of which N00,000 have been made (3 < N < 5?) - cheap, compact and effective. I include the gate to source zener essentially as of right. It might get excluded if space or cost utterly precluded, but it never will :-). Saves you from death by Millar injection due to load inductance transients that you could have sworn could not possibly exist. (In one design it increased survival time from hours to infinite).

It's always useful to include links to data sheets.

MOSFET datasheet here

\$\endgroup\$
  • \$\begingroup\$ Thanks on the datasheet tip. I’d forgotten to mention I already do have a series resistor, but its value is higher, at 120 ohms. I removed the Rgs resistor after I ditched opto-isolator/floating ground. It seemed to do little but pull down the driver amplitude. \$\endgroup\$ – user56021 Oct 12 '14 at 21:37

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy