Analysing and changing a resonant circuit

[F._Schubert]

Ben,

notch filters won't help, as all parameters change with the DUT and therefore I don't have a fixed resonance frequency. I have configurations of DUT and measurement resistor, where I don't have the resonance in the amplified range. In this situation I can compare the signals to a custom made load which gives me a correct value. This custom made load just doesn't do what I need for some new specs, it is tailored to the standard configuration.

Felix

[johnsold · Accepted solution]

Felix,

Ben's idea may be helpful if the frequency is constant or readily predictable.

A scope will probably be needed to figure out what is really going on. The bandwidth of your DAQ device is too low, even accounting for aliasing to really see what is happening.  I have seen power devices oscillate at hundreds of kHz or even MHz in systems with nominal bandwidths of a few hundred Hz.  Looking at various points in the system with a scope may give clues as to which part is the actual oscillator and which parts are just amplifying or spreading the oscillation around.

Force to DUT is where the power flows.  Sense to DUT is just a voltage signal, hence easier to filter because very little power will flow in the filter. 

It may be worth the time it takes to understand what is causing the oscillation and design a proper remedy before going to those very low impedance systems.  If those oscillate, DUTs or test equipment may be destroyed before the safety circuits can shut things down.

Lynn 

[Henrik_Volkers]

---

Felix,

I know scientist can build funny stuff ....  (I work at PTB)   but more and more I have the impression that your load goes creasy. Maybe some sort of missleaded offset compensation? Hard to tell .. and your observations lead me more to higher frequencies.       

At this point I would insist to get hands on a (as fast as possible) scope. Debugging analog circuitry without is like trying to do a operation blind.( Und blinde Chirurgen sind wie Maulwürfe, ihre Hände wühlen im Dunklen und ihr Tagwerk sind Erdhügel! Ich leih Dir mein privates, hat nur 20MHz BW ... aber besser als nix)

You can try your 'poor mans DAQ' the soundcard as a ac input , depending on your board you might have 192kS/s 😉  and (a little bit 'MacGuyver') a Radio (AM, SW) make a simple RF detector.. 

but seriously: you need the right tools!!

Message Edited by Henrik Volkers on 04-30-2010 10:45 AM

[Henrik_Volkers]

here are some links to appnotes from linear I mentioned in an earlier post

AN118 High Voltage, Low Noise, DC/DC Converters

AN104 Load Transient Response Testing for Voltage Regulators

AN101 Minimizing Switching Regulator Residue in Linear Regulator Outputs

Last page in AN113 🙂

[F._Schubert]

Henrik, Lynn,

you were so right! I checked with an oscilloscope. It's at about 140 kHz. I wasn't expecting anything that high, as the Load is specified only up to 100 kHz.

Felix

[johnsold]

Felix,

For the load to work as  specified to 100 kHz it clearly will have many components with much wider bandwidths.  Some of the small signal circuits within the device probably have MHz bandwidths.

Taming the oscillation amy still be a challenge, but now you can see more of what is happening.

Good luck!

Lynn 

[F._Schubert]

So a short final update to you all (and in case someone is reading it for solving his own problems).

Now that I know the frequency, it is really clear. My first thought of a resonance circuit were wrong, the detailed analysis of noise gain versus amplifier gain explains it. The amplifier is specified up to 100 kHz, this means a gain of above 1 is available much higher. Because the gain plots are log-log plots,  the final cut will be more likly at 200 kHz or even some MHz. The noise gain ramps up due to the capacitance (I wasn't able to measure it) and the resistor and intersects the amplifier gain at that frequency of about 140 kHz where I see the oscillation. Changes in C and R result in changes of the frequency. Propably there is a relation of 1/Sqrt(R C), I'm not sure.

I was already playing with 400 uF bipolar capacitors and I knew where I have to connect them to cut the oscillation. Once I knew the oscillating frequency (it was the highest peak, others where caused by harmonic distortion due to some diodes, and some of those I saw with the DAQ card), I could calculate a capacitance which forms a low-pass with the resistor. Some experiments with different values showed that I found a good value (higher R*C values introduce a phase-shift, lower allow for oscillations).

And even better: I can send a report to the developers of the electronic load. They can now give me feedback based on their knownledge of the inner workings. And they can hopefully even provide me with a high-quality product that is working without external capacitors in the near future.

To all a big thank you.

Felix

[johnsold]

Felix,

Thanks for the update and the nice explanation of the behavior.

Lynn