2

I am currently evaluating several power MOSFET’s looking at their switch on and off times.

My testing has employed a few different gate drivers, however the bulk of my testing is with the ST Micro L6491 gate driver IC.

The MOSFET I am hoping to proceed with in my project is the Nexperia PSMNR90-40SSH.

Referencing Vishay document AN608A and the following parameters from the PSMNR90-40SSH datasheet, I calculated the following as the turn on and turn off times.

enter image description here

enter image description here

Having set up a test circuit using a 24 Ohm gate resistor, the following are the scope images for the total turn on and turn off of this MOSFET. These measurements assume turn on is 10% Vgs to 10% Vds, and turn off is 90% Vgs to 90% Vds. The load in the test circuit is a 3 Ohm thick film resistor.

enter image description here

enter image description here

Turn-on = 452ns Turn-off = 636ns

I have made about a dozen sets of measurements using various gate resistor values and with three other MOSFETs; all of my measurements are values that far exceed the calculated estimates.

I did find a SPICE model for the PSMNR90-40SSH and modeled the test circuit in NI Multisim. The measurements for turn on and turn off were:

Turn on = 180ns Turn off = 438ns

How do I account for the discrepancy between calculated times and measured values? How are other people estimating accurate switching times? Are there any considerations I should be making that I may have overseen?

mike
  • 107
  • 9
  • 3
    Layout parasitics like source and gate inductance for example can have a big impact on switching times. – John D Jul 16 '21 at 00:28
  • 2
    You should not be using an x1 probe for this BW.USe a Buffered FET probe or use a 10x passive probe with coil spring and probe tip and gnd clip removed. then calibrate it. Also the layout matters. – Tony Stewart EE75 Jul 16 '21 at 01:18
  • Could you elaborate on why the probe tip matters? I agree layout makes a difference, but by how much? – mike Jul 16 '21 at 01:32
  • 1
    It is the ground lead inductance that can cause issues. 8 nH/cm or so with coax capacitance can resonate. Did you use short coax 1:1 then 30 pF /ft – Tony Stewart EE75 Jul 16 '21 at 01:39
  • the long plateau in Turn ON tells me 24 Ohms is less optimal than 5 Ohms yet that's what they used. Ground coupling also adds capacitance but reduces impedance – Tony Stewart EE75 Jul 16 '21 at 01:41
  • I'm using the probes that came with the scope. I will switch to 10x and see what difference it makes... I suspect it won't load the circuit as badly as the 1x. – mike Jul 16 '21 at 01:46
  • Make 2 test pins with a resistor wire, then lean exposed tip and ring from 10:1 probe for low inductance accurate high speed measurements, then buy the gnd loop spring probe – Tony Stewart EE75 Jul 16 '21 at 01:56
  • Also plot the Vgs to verify risetime – Tony Stewart EE75 Jul 16 '21 at 02:39
  • Once you get the probe stuff sorted out, it would be cool to add in the probe and layout parasitics to your SPICE/Multisim simulation and see how close you can get to your measured values. – Ste Kulov Jul 16 '21 at 04:41
  • Please show circuit diagram and photo of the wiring. – Bruce Abbott Jul 16 '21 at 05:48
  • @TonyStewartEE75.... I just noticed you are in the neighbourhood! Could you expand on/clarify what you meant by "Make 2 test pins with a resistor wire..."? – mike Jul 16 '21 at 07:43
  • Sure mike , invite me over to help test. It was a way to make an easy test points with tip and gnd strap removed for low ESL probing. Resistor wire is steel so its stiff and easy to add on boards for debug say 1/2 ~ 1cm apart for sig - gnd. Or use the probe spring on pad pairs https://electronics.stackexchange.com/questions/431844/does-polarity-of-scope-probe-matter-when-measuring-ripple-of-the-power-supply-mo/431857#431857 – Tony Stewart EE75 Jul 16 '21 at 08:06

1 Answers1

2

Having shortened and used larger gauge wire from driver to gate, and most importantly changing my probing technique (x10, shortened ground path) I was able to get a time that was more in line with my calculated estimate (~160ns with a gate resistor of 24Ohms). Probe ground was way too long; I attached my ground to a more suitable location.

Below is the waveform showing the on time for this MOSFET using 24 Ohms.

oscilloscope screenshot

In summary, layout and proper probing technique makes a huge difference.

SamGibson
  • 17,231
  • 5
  • 37
  • 58
mike
  • 107
  • 9