13
\$\begingroup\$

I'm trying to debug a pair of cascaded DC/DC converters and have ran into a brick wall. The local FAE said that it is probably something to do with "Negative Input Inductance" on the second converter messing up the stability of the first converter (but the FAE isn't around to "help" more). The problem is that I can't find any app notes, papers, books, etc. on this issue.

My question is: Do you know of any literature on such issues? Or better yet, some ideas on things to try or look at?

Here's my setup...

Converter 1: +4v to +12v@1 amp out boost converter. Switching frequency is about 350 KHz. Converter 2: This is actually a 10-ish watt Class-D audio amp (which is basically a switching buck converter). Switching frequency is about 310 KHz.

And the problem...

Converter 1 works fine with a resistive load instead of converter 2. It even works if the resistor is switched in/out at audio frequencies.

Converter 2 works fine when powered from a benchtop power supply.

When Converter 1 is feeding Converter 2, C1 will shut down due to over-current through the MOSFET. It shuts down easier if the audio frequency is lower. Above a 1 KHz sine wave it seems to work fine. When it shuts down, the power output is only about 50% of what the converters are able to do separately.

Ideas? Pointers?

Update: I found the problem.

There were two bugs...

  1. Basically, Olin was correct. I did a miscalculation. The first converter should have been able to supply twice the current that it was providing. Instead of +12v at 1A, we needed 2 amps.

  2. Converter 1 is a current mode converter-- meaning that it has a current sense resistor between the MOSFET and GND. It appears that the PCB traces and vias for this signal path was not up to the task. I tried several resistors in the 4 to 24 milli-ohm range, but suspect that the traces/vias were adding another 5 or 10 mOhm. The end result is that we were overcurrenting sooner than we wanted to.

In the debug process, I isolated Converter 1 from the rest of the circuit and tweaked it to provide a solid 2 amps into a resistor load. Once it was solid, I connected it back up to the audio amp and it worked fine under all expected loads and audio frequencies.

So, apparently, it had nothing to do with negative inductance or whatever.

For being a mostly-digital guy, I sure am getting a lot better at analog stuff! :)

\$\endgroup\$
  • 1
    \$\begingroup\$ At least it shuts down. Ours explode. :( The resistive load that works on the first converter is drawing equal or greater power than the amplifier test, right? Including inefficiency of the amp? You're not rating the resistor equivalent to a full-scale sine wave and then testing the amp with a full-scale square wave or something like that? \$\endgroup\$ – endolith Oct 20 '11 at 15:58
  • 1
    \$\begingroup\$ Have you tried a seriously large capacitor on output 1? So big you MAY have to resistor charge it to stop it stopping the 1st converter from starting. \$\endgroup\$ – Russell McMahon Oct 20 '11 at 16:06
  • 1
    \$\begingroup\$ @endolith Yes, when Converter1 shuts down it is running at 50% of what it could do with the resistive load-- based on the RMS measurements. \$\endgroup\$ – user3624 Oct 20 '11 at 16:23
  • \$\begingroup\$ @Russell We've tried putting caps and/or inductors between them with little success. But we haven't tried super huge scary caps yet. \$\endgroup\$ – user3624 Oct 20 '11 at 16:23
  • 1
    \$\begingroup\$ If you've solved the problem please post as an answer to your own question don't edit the question. \$\endgroup\$ – mjh2007 Oct 21 '11 at 18:42
6
\$\begingroup\$

From your numbers it seems like the instantaneous draw when the audio output is at the maximum peak of the waveform is too much for converter 1. This would explain why it works at higher frequencies, since converter 1 then sees more the average rather than peak instantaneous current draw.

You say converter 1 puts out 12V at 1A, which is 12W. To put out 10W RMS audio power would mean the peaks of the audio waveforms draw twice that instantaneously. This would overload converter 1 by almost 2x according to your description. At 1kHz audio output, the overload happens for only about 250µs at a time. Since the average draw is OK, converter 1 doesn't fault.

This is just a guess, of course, but it is consistant with the information you provided.

\$\endgroup\$
4
\$\begingroup\$

I think your FAE is confused.

The amplifier is a constant-power load to the first DC-DC stage. If the first stage output voltage goes down, the amplifier will draw more current to maintain the same output power. The amplifier is therefore presenting a negative impedance characteristic, not a negative 'inductance' (whatever that means).

A negative impedance load does have implications for stability. If the magnitude of the negative impedance cancels out the power supply's output LC filter damping network, the power supply can oscillate.

Venable Industries has a series of papers on the topic of stability (H. Dean Venable = control system guru) - check out TP-12 (free registration required).

If you don't see signs of instability on the DC/DC stage (erratic pulses, sinusoidal ripple) then the premature shutdown isn't due to loop interactions.

\$\endgroup\$
-1
\$\begingroup\$

If you are still looking for some literature about negative impedance (not inductance) of DC-DC converter here are some good links:

\$\endgroup\$

Your Answer

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