Diode voltage drop stability/accuracy

[Infraviolet]

Ok, so I did a test with the basic single diode (or single b-e junction of an NPN transistor) detector. It LOOKS like temperature barely affects the result at all, I chilled some transistors and diodes over night in a freezer, got them out during the day handling with tweezers so my wrm fingers wouldn't heat them too fast, plugged them in to the breadboard layout and watched whether the voltage drop between a fixed amplitude input sine wave an he detected DC output changed over time after they were plugged in as the parts were warming up to about the 25 degrees C summer room temperature. No noticable change over time from plugging them in for many tens of minutes, nothing more than the 10mV or so of range which the signal flucuated over anyway (probably from not-quite perfect conatct on a breadboard wire somewhere).

...either that or the heat capacity of the through hole components I breadboard tested with is so tiny they reach room temperature within seconds of being lifted from a -10C freezer and popped in to a breadboard.

[Vovk_Z]

1. Yes, small components change temperature quite fast.
2. It is often easier to test temperature dependence by heating component with  available instrument. E.g. heat-gun.

[Infraviolet]

I've had some more thoughts, can anyone suggest if they make sense:

I need 3 channels of signal, I need to accurately know the ratio of the peak-to-centre height of all of them.

I'm thinking with 3 channels plus one dummy channel solely doing the peak detection section I should have enough stability?

All sections of the 3 channels' signal processing before the peak detection can be done within a quad transistor IC for each section (MMPQ3904 looks to be about the only quad NPN made any more, but seems to be from multiple manufacturers so fairly generic and not too likelyto stop being available, CA3046 was popular in things I've read but looks to be out-of-production, MMPQ2222 looks to be nearing end-of-life, many other transistor array ICs aren't independent transistors at all, but rather arrays of darlingtons with certain pins commoned) of processing. With 3 signals undergoing any give process within an IC containing the relevant transistor for each channel then any temperature variations or manufacuring differences between ICs ought to affect all channels equally. So the ratio between signal amplitudes ought to remain consistent regardless of environemtnal or manufactruing variations?

It is only at the peak detection stage that my signals get coupled to a fixed DC reference level, until then they could all be centred about different reference levels and it wouldn't change anything. So at this point if I use a fourth channel in the IC* as a dummy channel with a peak detection circuit around it, but just feed it the fixed level to which other channels get their signals coupled, then in theory any variations will affect all 4 channels of peak detection equally. If the "diode" drop of the peak detector changes then the dummy channel will drop by the same amount as the other 3, meaning the difference between each of them and the dummy would stay constant. 

Then as the difference between the real channels and the dummy should stay constant, and the ratio of signal strengths given in earlier stages ought to be kept constant by having shared ICs, I would get stable output values when comparing the ratios of (signalXpeakdetected-dummyChannelVoltage) vs (signalYpeakdetected-dummyChannelVoltage) vz (signalZpeakdetected-dummyChannelVoltage)?

And my guess is this would even make thos ratios stable against slight variation in supply voltage too?

Thanks

*as NPN's can be used like diodes, and I couldn't find a quad diode array which didn't have commoning of one pin or the other