Blog Post Actions
Group Actions
Engagement
  • Author Author: michaelkellett
  • Date Created: Date Created
  • Views 402 views
  • Likes 2 likes
  • Comments 4 comments
Related
Recommended

Power Supply Dynamic Load testing

michaelkellett

Following on from earlier discussions I thought I'd take a look at how easy (or not) it is to use a big MOSFET in a simple set up to switch a 5A load to test power supply transient response:

 

I used an STW60N10 MOSFET, a 3.3R load resistor, 4 x 4700 uF 100V caps in parallel. The MOSFET is driven from Rigol 1030Z signal generator set to make 20uS pulse at 100Hz rate. I tried it with and without a 100R resistor in the gate.

The power supply was 16.5V via  a100R resistor driving the capacitor bank.

 

Here's a picture:

image

 

 

And these are the results:

Red trace is MOSFET drain voltage, Yellow trace is Sig gen output, Blue trace is MOSFET gate.

image

image

 

It's fairly horrible - the stray inductance in the load makes it ring when the MOSFET switches off and it turns out to be quite tricky to control this (using only passives and a sig gen for the drive pulses.)

 

Time for an LTSPice model:

image

The magenta trace is the MOSFET drain voltage, the blue the current though R1 and the red is the voltage on C1 (which dips due to C1 series resistance)

R3 and C3 slow down the drive to the MOSFET,  R4 and C2 'snub' the high frequency ringing when the load inductance is small.

L1 is the load inductance - this ciruit isn't too bad for values up to about 20uH.

 

If you are interested then run the model in LT SPice - it's free and try some tuning of your own - no prizes image

 

MK

Attachments:
mosfet_load_switch.asc.zip

  • The co-fired MLCCs meant to replace tants work really well.  Their existence slipped my mind the other day.  Really good of the industry to try to replace tants.  Frankly, I was distracted, been working on an involved project.  Good luck and keep us posted.

  • in reply to michaelkellett

    I tried putting small caps across the 18800uF main cap and it made almost no difference.

    So I added a snubbing network (22R and 47nF) and a current sense resistor (0.1R) and it doesn't look too bad now:

    image

     

    image

    The scaling for Is is 10A/V so the on current is very close to 5A.

     

    This is probably about as fast as these MOSFETS are going to switch without using high current drivers. Although I've labelled a trace as Vg it's actually the voltage on the signal generator side of the 100R gate resistor so the gate load of the MOSFET is quite large to distort the sig gen output so much.

     

    MK

  • in reply to D_Hersey

    I'd been pondering how much of the stray inductance is in the caps and how much in the wires - the rig is still set up so I'll experiment later today.

    Won't use a tant - in the olden days when I did a lot of high power mid voltage (50V or so) stuff for audio I liked polyester caps for this sort of thing - they never blow up but I have had ripple current explode tants or small electrolytics.

     

    Modern ceramics might do but I won't have anything big lying around.

     

    Because of the intended use as a power supply test load it needs to work decently even with long wires image

     

    MK

  • Hi, MK.  Perhaps the ringing could be moderated somewhat if you bypassed your bypass capacitor with something smaller and faster, a tantalum* perhaps.  If that doesn't help, it will ring in a more complex way at least.  I doubt it would help much, but bringing the filter cap electrically closer might be of benefit.  Solder braid is lower Z at higher frequencies than wire. 

     

    *Obtained from a conscious supplier, of course.

     

     

    --Nice work, BTW.