What gives with this square wave?

[martyh]

Here are two square wave traces from an output transformer that is 6K6 to 8/16

Both are at 10K, the one with the visible ringing is from the 8 Ohm tap unloaded the second is with an 8 ohm resistor across the 8 ohm secondary. I’ve seen the rolled off front edge before but what is the spike at the back side? While I am on the subject, can anyone point me to a good tutorial on interpreting square wave response? Please excuse the picture quality

Thanks,
Marty

[TheGimp]

The unloaded tap is the classic under-dampened square wave. It will ring, and is not good for tube outputs.

Loaded waveform shows over-dampened response, loss in high freq band.

Here is an explanation on how to interpret the results.

http://www.kennethkuhn.com/students/...ve_testing.pdf

[HollowState]

A fundamental square wave test frequency of 10KHz is improper for analyzing output transformers. To reproduce a square wave without distortion, the transformer must be capable of passing frequencies of F/10 and Fx10. This means if a fundamental square wave of 10KHz is applied to a transformer, the transformer must have a bandwidth out to 100KHz. Something virtually no standard power output transformer really has.

Testing with a 2KHz fundamental will equate out to 20KHz. One could even push this fundamental up a little more to perhaps 3KHz, but not 10KHz. For a low end of 30Hz, a fundamental of 300Hz should be used. 20Hz would require 200Hz and so on.

[Geek]

**NEVER** run a tube amp without a load.

Those spikes can be cured by parallelling a correct capacitor across the NFB resistor. Or better yet, getting the OPT out of the NFB loop all together with a little redesigning

Cheers!

[tubelab.com]

A fundamental square wave test frequency of 10KHz is improper for analyzing output transformers.

True, to see a perfect 10KHz square wave you need frequency response from 1KHz to 100KHz. But we can look at the "imperfectness" of a 10KHz square wave to judge the RELATIVE merits of a few OPT's under the same conditions, and we can use that 10KHz square wave to tweak the components in the NFB loop for best response. Achieving the best looking square wave for a given OPT's correlates well with good sounding high frequencies. Too much leading edge roundness sounds dull, too much ringing sounds brash and annoying.

The photo shows a 10KHz square wave at 1 watt through a Simple P-P using some OPT's that I paid $16 for. The OPT's were designed for guitar amps and use no interleaving whatsoever. The transformers were made by Schumaker using the same technology applied to their battery chargers.

Spikes like those are an indication of some really high frequency energy. That didn't come through the OPT. I would suspect some interaction between the scope and the circuit. Do you have one end of the secondary grounded? Is this measured in an amplifier? If so, do not run the amp unloaded. As mentioned it can lead to damage in most tube amps.

[TheGimp]

Here is another site with interesting info on testing with simple test equipment.

Waveform Analysis

[gootee]

Originally Posted by HollowState

A fundamental square wave test frequency of 10KHz is improper for analyzing output transformers. To reproduce a square wave without distortion, the transformer must be capable of passing frequencies of F/10 and Fx10. This means if a fundamental square wave of 10KHz is applied to a transformer, the transformer must have a bandwidth out to 100KHz. Something virtually no standard power output transformer really has.

Testing with a 2KHz fundamental will equate out to 20KHz. One could even push this fundamental up a little more to perhaps 3KHz, but not 10KHz. For a low end of 30Hz, a fundamental of 300Hz should be used. 20Hz would require 200Hz and so on.

Another way to think of it is in terms of the input square wave's amplitude and the gain and frequency-response of the amplifier, and band-limiting a test-input square wave so that it doesn't try to exceed the maximum slew-rate that can be achieved by all parts of the amplifier (unless, of course, you're hoping to excite any high-frequency instabilities that might be lurking in the amp).

After all, even a low-frequency square wave could have rise and fall times that might try to induce output slew-rates that far exceed an amplifier's capabilities. i.e. The frequency of the square wave, which is just its "repetition rate", is not nearly as important for this purpose as, and is not necessarily even related to, its rise and fall times.

I did a little work on this, when trying to simulate amplifier responses to square wave inputs, with capacitive loads. Here is one of the relevant threads, where I have included the necessary calculations in one of the posts:

square wave testing and slew rate

Tom

[martyh]

Thanks for all of the replies

Just for clarification, I was testing the transformer itself not a whole amp. There was only 10vrms on the primary from a 600 ohm source. My main goal was to find the turns ratio with a sine wave which was easy enough. I was more or less messing around when I switched to square waves and saw the weird spikes at high frequencies.

[rascal101]

I believe that the primary cause is the leakage inductance for this transformer which maybe be a bit high. This means that the coupling in between the windings is not that good or the turns distribution in between layers has not been maximized. To measure the leakage inductance, you will need to short all other windings then with an LCR meter, measure the inductance of the concerned winding.

[richwalters]

Originally Posted by HollowState

A fundamental square wave test frequency of 10KHz is improper for analyzing output transformers. To reproduce a square wave without distortion, the transformer must be capable of passing frequencies of F/10 and Fx10.

Where did you find this from ? That's only possible from a theoretically perfect transformer. In the real world equals compromises.. Especially when going from class AB, mismatches,overshoots, misbalanced wiring, Miller effects and so on. Also not forgetting the probe and ground compensation. We have to accept much less.
It doesn't correlate with Fourier analysis. With global NFB, the old rule was a transformer reproducing a 10Khz square wave should see
minimal 3rd harmonic attenuation at 30Khz. So a transformer with good response at 20Khz should imply a droop not less than -3dB down at 60Khz.

So in p-p tube amps using global nfb, the 5th harmonic is already very low and I ignore this. The result is then an amplifier design based on 3rd harmonic and the enclosed is a 10Khz square wave pic of high power p-p 100W amp at rated o/p. The response of transformer (E&I) was specified by the maker as -3dB at 70Khz which for it's size is a very well designed balanced component and the resonant frequency at approx 100Khz. The final amp response was trimmed at 55Khz by both combination of trimming the Bode plots of the 1st stage and the capacitor on the 20dB global nfb loop.
The result is fully optimised.