5
\$\begingroup\$

I have designed a full bridge that handles +/- 2000V. It seems to work fine, however I am getting a voltage dip after the switch. The voltage reaches the setpoint fine, but then dips a couple hundred volts, back to setpoint in about 150ms. I can't seem to find the cause. Feeding into a 1000:1 divider to look at it.

I hopefully attached the dip, and the schematic. This runs at 50% duty cycle, feeds a 6pF capacitive load (Pockels Cell) at <= 1KHz. Switches very little current. I used drivers that worked well in the past, set up with lots of dead time. I bootstrapped the high side. Used standard 10 ohm gate resistors, and 10 on output to eliminate spikes (schem shows 200R, I lowered).

I'm wondering since the low side is referenced to a solid negative voltage, and the high side is swinging around maybe the issue is there somewhere. All my bridges switch to GND until this design!

EDIT: MOSFETs are: IXTF1N450 (4500V) RDSon: 80Ohm GateCharge: 46nC

Ideas?

HV Dip

Schematic Schematic

\$\endgroup\$
14
  • \$\begingroup\$ "...the high side is swinging around ..." -- can you add that to your scope trace? That's the obvious first place to look. \$\endgroup\$ Commented Sep 11 at 20:29
  • 1
    \$\begingroup\$ What probe are you using, and how is it connected? \$\endgroup\$ Commented Sep 11 at 20:41
  • \$\begingroup\$ Where does your high side driver get its isolated 12 volts from? Please show a complete circuit. \$\endgroup\$ Commented Sep 11 at 20:42
  • 1
    \$\begingroup\$ My point was that you don't show where Q1_DRV, Q2_DRV, etc. come from. Are you absolutely certain that those voltages are stable during the switching event? \$\endgroup\$ Commented Sep 11 at 21:49
  • 1
    \$\begingroup\$ A simple fix would be to use internally compensated high voltage probe, such as GE3421, instead of the resistor divider. \$\endgroup\$ Commented Sep 12 at 6:30

1 Answer 1

Reset to default
6
\$\begingroup\$

So this is your probe(s)?

enter image description here

A scope's input circuit looks like either:

schematic

simulate this circuit – Schematic created using CircuitLab

(C2 probably has some ESR, too, hence R3), or a direct 50Ω termination instead of the 1M. (C2 is probably still present, just swamped by the resistance, hence the some-100s MHz upper bandwidth limit for typical scopes bearing specs around here).

In the terminated case, you can extend the 50Ω with any length of transmission line, and the input impedance remains the same. Downside: that impedance is pretty low, for probing purposes.

The traditional 10x probe acts as a resistor and capacitor divider in parallel, depending on which frequency it's measured at.

Note that C1's value was indeed adjusted to be that particular, and this is the trimmer provided in the probe handle for just this purpose. Try different values in the simulation to see.

The probe cable is a high impedance, with a lot of series resistance, giving it an RC characteristic; this is why, in 1x mode, the cutoff frequency is very low (much to the surprise of many a newbie that skips reading the probe manual), typically 6MHz.

If we consider that the typical resistor (unstated type) is capacitive at high frequencies, for values above a few 100 Ω, we can model the overall circuit like so:

Oh god damnit it overwrote the fucking first schematic again, to hell with Circuitlab

Anyway this doesn't make any sense anymore thanks for that have a waveform ugh enter image description here

\$\endgroup\$
9
  • \$\begingroup\$ Tim Williams, Regarding the annoying overwrite problem: I contacted CircuitLab support at the beginning. They told me the problem is not with CircuitLab. That is obviously the case because this is a problem with saving the edited image and text. For now, only @Chester Gillon's CTRL-Z trick helps; it restores what was done with so much effort. \$\endgroup\$ Commented Sep 12 at 7:09
  • \$\begingroup\$ We have very expensive HV probes rated to 1000V, will switch and try. Also, the bootstrap diode is a Schottky 6kV, I just haven't updated the schematic. If it is an artifact of a resistive divider I'm gonna feel pretty dumb! This area isn't my forte, I just copied reference circuits and bought eval kits, read lots of white papers relating to my application. That pic sure looks like exactly what I am seeing! BTW the scope is on the two unnamed nodes heading up off the schematic (the outputs proper) to external divider box. I am not using the dividers on the schematic, those are for ADC's. \$\endgroup\$ Commented Sep 12 at 12:38
  • \$\begingroup\$ However I never thought of the on board divider being an issue! Will remove them and retry. \$\endgroup\$ Commented Sep 12 at 13:06
  • 1
    \$\begingroup\$ @tim Williams You sir are a genius! I switched to a Tekronix HV probe and it’s perfect. \$\endgroup\$ Commented Sep 12 at 14:18
  • 1
    \$\begingroup\$ @TimWilliams I found that that the wrong section of the answer text (specifically the <!-- Begin schematic:...--!> section) is being overwritten by the schematic editor, which points to a problem on the SE side (javascript). I also found that "undo" (CTRL+Z, immediately following the schematic edit and save) will restore that text, and the associated schematic image and CircuitLab link, and also retain the new schematic. That suggests that answer text is being amended twice by a single schematic edit. Awkward, but thankfully recoverable with CTRL+Z after the edit. \$\endgroup\$ Commented Oct 9 at 3:27

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.