John Curl's Blowtorch preamplifier part II

[BV]

Thermal transient effects
Try multitone signal 50+51Hz (beat 1Hz) with 0dB and 10k+11k with -10-20dB relative to 50+51Hz. Observe only IMD products 10k+11kHz(1kHz...), When driving amp with this signal just under limit, power is cycling from 0-100% with 1Hz, so IMD products cycling should be visible..You can try e.g 50+50,2 Hz as "base", so thermal cycling is with 5sec period.

[1audio]

Originally posted by h_a
Well, maybe I'm missing something here, but thermal effects on THD should be quite easily investigated.

Just run a 2 tone test into the DUT, say 1kHz plus a 40kHz sine. Apply bandpass to get rid of the 40kHz. Estimate from standard 1kHz THD the additional intermodulation products and subtract. Certainly a little matlab or similar is usefull for better precision.

Have fun, Hannes

EDIT: or mount transistors on high power Peltier-elements.

THD and IM stuff is interesting but my interest was time based modulations. They may show as sidebands but may be more audible as time shifts. Much harder to measure.

[BV]

They may show as sidebands but may be more audible as time shifts. Much harder to measure.

Sidebands "pumping" of IMD products 10+11kHz with period of 50+51Hz beats is clearly visible with " bad enough" amplifier on spectra in real time.

[PMA]

Have the signal, but see nothing special in a FFT.

[BV]

Have the signal, but see nothing special in a FFT.

So where is impact of thermal effects on linearity? Especialy at hf (like 10+11kHz) it should be visible, or no?

[PMA]

This would depend on a circuit used. BTW, the test signal is periodic ....

[Wavebourn]

Hmmm.... I have some quartz resonators about 7 KHz and pair of 500 Hz 60 db/Oct filters. I'll see what differences in frequencies of that crystals I have.
One double triode with JFets in cathodes would make nice clean 2-tone oscillator.

[scott wurcer]

Originally posted by 1audio


THD and IM stuff is interesting but my interest was time based modulations. They may show as sidebands but may be more audible as time shifts. Much harder to measure.

I'm out of my element but I don't think the diffusion equations support waves as a solution. Analogies between electrical effects and thermal ones will fail badly at some point. jcx probably has better input on this.

[BV]

BTW, the test signal is periodic ....

If you use music with 60 BPM, it is also periodic cycled in power..But serious, if 1 or 2 seconds period between 10% and 100% power (with 10+11kHz present)have no (visible)impact on sidebands (9,11kHz) and 1kHz IMD (with properly constructed and biased amp), how can it be hearable?
BTW, you should use no averaging and FFT e.g. 4096 or 8192.
With underbiased amp i can see so with sugested signal sidebands "pumping" at levels between -120 to -110dB , peaks 10+11kHz at -20dB, 50Hz components peks at -6dB on spectra. Power is cycled from 1W to 100W..But is it from termal effects? Sure not.

[1audio]

Originally posted by Wavebourn


I have a pair of LMI-500 filters by UTC, I can bring them when I come.

Edit: I've found they are LPF actually, 60 dB/Oct, so the task is reversed: like 200 Hz to observe, 20 KHz to disturb. Not so good for phase angle observations, but who knows...

Sometimes you make do with what you have-

Start out with two oscillators, a very stable low distortion low frequency oscillator and a modulatable high frequency oscillator.

Run the lf oscillator a low level through the chain into a sensitive phase meter. Mix into the input the HF oscillator pushing the system to close to its power limits. Cycle the HF oscillator at various rates consistent with the thermal time constants in question. Pass the output through the LPF and see the effect on the phase meter comparing the input phase to the output phase.

I have read but can't cite references that the relative phase of a composite signal are not audible, however if the phase relationship is changed the change is audible. I don't know if there is any data on this.