Ultrafast diodes too slow? / Tesla Coils / Forums

Forums

4hv.org :: Forums :: Tesla Coils

« Previous topic | Next topic »

Ultrafast diodes too slow?

Move Thread

LAN_403

Dr. Dark Current

Fri Dec 25 2009, 10:55PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

I have been scoping around a SSTC circuit before applying bridge power, with just the gate driver powered. I have 10 ohm gate resistors in parallel with ultrafast discharging diodes (UF4007). This is what I get when I connect my scope directly across one of the diodes:

5V/div. As you can see there are something like 8 volt spikes in the *forward* direction.

How can this be ??

EDIT: with 1n4148's its better. weird

doctor electrons

Fri Dec 25 2009, 11:54PM

Registered Member #2390 Joined: Sat Sept 26 2009, 02:04PM
Location: Milwaukee Wisconsin
Posts: 381

I had the same thing happen to me! I was confused about it for quite a while. When i finally worked up the nerve i asked one of the
electrical engineers at my job. This is what he told me. "Because the diode is paralleled with the resistor!" Then he told me to figure out why.
Not really what i had hoped to hear! Anyway, i do not know for sure, but heres my best guess. I think the reason that this happens is when you read directly across the diode you are also reading across that resistor. Mixed signals! Try to read the diode on one side and see what you get. Maybe you are correct in the speed of the diode not being enough! I would love to hear what you figure out!!!!!

Sat Dec 26 2009, 02:28AM

Registered Member #2372 Joined:
Location:
Posts: 62

What is the timebase? it kinda looks like the derivative of a square signal.

Dr. Dark Current

Sat Dec 26 2009, 08:45AM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

dugg, it is 1us/div (f around 260KHz)

Steve Conner

Sat Dec 26 2009, 11:05AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Diodes have forward recovery as well as reverse recovery: it takes a finite time for them to turn on as well as off.

They also have inductance because of their finite size and lead lengths.

Also, your scope probe can pick up signals that aren't really there, because of induction into the loop area of the ground lead, and common-mode pickup in the probe lead. To test this, connect the probe tip to the same place as the ground clip. If the spikes are still there, then they aren't there, if you see what I mean.

Dr. Dark Current

Sat Dec 26 2009, 02:53PM

Registered Member #152 Joined: Sun Feb 12 2006, 03:36PM
Location: Czech Rep.
Posts: 3384

Steve, I tried that, the spikes are real :) and the ground clip and probe tip are connected close to the plastic diode case.
With 1n4148's the spike is only around 2 volts, with schottky rectifiers even less. So it is indeed because of the forward recovery.
But why don't they say this in the datasheet?

tesla500

Sat Dec 26 2009, 05:25PM

Registered Member #347 Joined: Sat Mar 25 2006, 08:26AM
Location: Vancouver, Canada
Posts: 106

This is one thing that you need to learn from experience, it seems. Datasheets don't mention forward recovery for some reason, and few research papers talk about it.

Schottky diodes don't have this problem, but watch out for the large shunt capacitance compared to silicon diodes.

David

GeordieBoy

Sun Dec 27 2009, 02:03PM

Registered Member #1232 Joined: Wed Jan 16 2008, 10:53PM
Location: Doon tha Toon!
Posts: 881

Firstly this sort of waveform is difficult to measure for several reasons:

1. Both ends of the diode are floating, (neither are grounded.) This means that you either have to use a differential voltage probe, two single-ended probes with Y1-Y2 setting on the scope, or use a single-ended probe with one end grounded and float the circuit under test.

2. Both ends of the diode exhibit voltages with fast rise and fall times. This is apt to drive common-mode HF currents down the screen on the probes. Even if the circuit under test is floated from ground in respect to DC signals, fast AC edges will couple across the gate-drive transformer and then cause unwanted cmmon-mode return currents to flow in the measurement probes ground lead. Since this current edge is fast it will drop a voltage across any inductance in the return path and appear as "interference" superimposed onto the signal you are trying to measure.

3. Measuring any signals with fast rise and fall times means that you MUST check the scope probes are properly compensated using a fast edge generator. If the scope probe is not compensated correctly it is easy to see voltage overshoots due to excess differential action in the scope probe, or curvature to otherwise "flat top" pulses due to HF droop. The first thing to do is always to correctly adjust the compensation of each scope probe FOR THAT CHANNEL that it is being used on.

Regarding forward-recovery: In general "fast recovery diodes" that have fast reverse-recovery generally have slower forward turn-on times. Conversely, bog standard LF rectifier diodes like 1N400x series often have a very fast turn on time. I have designed several discontinuous flyback converters that operated in the 100kHz+ range and used 1N400x series output rectifiers. In this application the faster forward recovery is important, and reverse recovery is less important as flyback pulse current naturally falls to zero during each cycle.

Forward recovery time is highly dependent on stray inductance in the clamping loop. Therefore for things like active PFC boost converters it is critical to keep the switch, diode and output capacitor all as close together as possible. Still in high power applications it is common to see the MOSFET drain voltage peaking 100 volts above the voltage on the output reservoir capacitor! With slack layout even a 500V device is not enough for use with 390VDC output.

Finally, the voltage on the gate of a MOSFET (or IGBT) cannot change instantaneously no matter how much you might try because of the device's intrinsic capacitance. (This capacitance being at it's largest with the bridge DC bus voltage not applied!) Therefore when you attempt to drive the gates with a fast squarewave you will always develope that train on alternate spikes across whatever impedance is driving the gate. If you look at the detail of this voltage waveform it will have a falling slope where the gate voltage has a rising slope, the two waveforms adding up to give the original fast drive waveform. Even silicon power diodes have a an ohmic part and a voltage-drop part to their impedance, so if you put enough current through them in short spikes they will appear ohmic.

-Richie,

doctor electrons

Sun Dec 27 2009, 04:53PM

Registered Member #2390 Joined: Sat Sept 26 2009, 02:04PM
Location: Milwaukee Wisconsin
Posts: 381

Excellent answer Richie! Thank you very much! I would have to say that makes perfect sense
to me now! Answers like that are what make 4hv such a valuable learning tool !!
Thanks again! Mike...

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob