Standard remote control IR runs between 36 and 56 Khz carriers with a few odd-balls running as high as 70K. What the actual diodes can do I have not looked at. Considering that are talking about 10Base Ethernet being piped over the LED overhead lights, I would suggest there is some room there.
Which matters most: chip surface instantaneous temperature, or junction instantaneous temperature? A bit of dynamic thermal modelling would give a connection between them, but a zero result from aiming an IR diode at the chip surface is not sufficient to show that nothing is happening at the junction (in the absence of such a thermal model). At the very least a junction thermal time constant would be useful, giving a first-order thermal low pass filter.
That gets back to my original observation about the lack of distortion rise with lower frequency.
Right. And junction itself is the best tool to measure own temperature directly. Chemist SY knows physics well. But moderator SY wants the show to go on, I suppose. All topics where he participates are usually hot and long.
Very strange observation, SY!
Use one common emitter stage, with stiff voltage bias, no emitter degradation, no feedback, feed signal from zero volt source directly to the base, and measure what you get on collector. Then come back and let's talk about observations.
Proposed apparatus: opamp follower with output connected directly to the base of a common emitter stage. Bias applied to the input of the opamp follower.
I can't. No convenient drawing software for Linux.
1. NPN transistor. Emotter to ground, resistor in collector.
2. Opamp. Output to inverting input (follower). Output of opamp directly to base of transistor.
3. Voltage divider, 1 resistor from + rail and one trimpot to ground. Whipper of the trimpot to the non-inverting input of the opamp.
4. Signal from signal generator to the non-inverting input of the opamp through coupling capacitor.
For measurement first set 1/2 of B+ on collector of the transistor adjusting trimpot.
Measure. Enjoy. The junction itself is the best indicator of it's temperature.
The fundamental problem is, how to tell which distortions are thermally induced on background of huge distortions of the stage itself.
Last edited:
OK, but I'm still interested in brief signals that are 20dB or more greater than the average level.
Does running the output devices and drivers at room temperature or above the boiling point of water make any difference? How could you tell and how could you separate any difference found from other causes (PSU, or other)?
What if the corner frequency for the LF distortion rise is significantly lower than the LF rolloff of the amplifier? You can't then measure it using the normal methods, but an envelope technique might still show it.
Of course, NFB would still reduce it as it is something in the output which is not in the input but NFB second-order effects might create problems.
So, what sort of junction thermal time constant are we talking about for a typical audio output BJT chip? I guess we can assume the case is at a constant temperature, so it is junction-case heat flow which matters. Case to ambient thermal time constants would be in the 10s secs-mins range?
Of course, NFB would still reduce it as it is something in the output which is not in the input but NFB second-order effects might create problems.
So, what sort of junction thermal time constant are we talking about for a typical audio output BJT chip? I guess we can assume the case is at a constant temperature, so it is junction-case heat flow which matters. Case to ambient thermal time constants would be in the 10s secs-mins range?
Have you tried GIMP?. Not drawing software per se, but you can do a lot with a little practice. There isn't a good schematic capture software for Linux? None of the SPICE versions?
Undoubtly well over 10 sec.
Given its size , the heatspreader can be assumed being at constant
temperature in respect of the junction instantaneous thermal dissipation.
At the frequencies of interest , what matters is rather the junction/case
thermal coupling since the variations are too fast to be absorbed in
usefull time by the main heatspreader , but even then ,according to the datasheets,for audio purposes the instantaneous temperature variation is a quite secondary phenomenon among all parameters that have an influence in the device linearity.
Last edited:
According to this (see page 12), the junction temperature of an RF PA device can follow the modulation envelope below about 100kHz and this is a problem for amplifier linearity. Assuming that this 100kHz is a ballpark figure, and assuming that RF devices have much smaller junctions than audio devices (and so much faster thermal time constants), I am still left with the impression that junction temperature could easily follow an audio envelope (at 0.1-1Hz?) and possibly LF audio too.
Indeed, those are the signals in question. if I understand Anatoliy correctly, he's adding a DC background which is totally different than modulating the temperature. Yes, the steady-state temperature matters, that's not in question (Ebers-Moll). The question is modulation of that temperature with signal (which I hope clarifies things for Antonio).
There s nothing to be worried about in an audio app since
NFB wich is maximal at LF will clean out any weak non linearity
that may occur at such frequencies.
The case may be different for non NFB designs , though.
The exemple you re quoting is a little different as RF amplifiers
have very little NFB to start with , so the device intrinsical
linearity and thermal induced variations of its parameters
are way more critical.
Last edited:
There are two problems with assuming high levels of LF NFB will clean up the problem:
1. NFB doesn't just attenuate error, but can also distort it - IMHO this is probably not an issue in this case, though, as we are talking about quite a small effect.
2. The IM could appear at higher frequencies where NFB is less effective. We are talking about IM sidebands appearing at a spacing set by syllabic rates e.g. 0.5Hz sidebands on a 10kHz signal.
RF linear PAs often do have local NFB, but these thermal memory effects make this less effective.
I'm not saying this is all a problem. I am saying I am not yet convinced that it is not a problem.
1. NFB doesn't just attenuate error, but can also distort it - IMHO this is probably not an issue in this case, though, as we are talking about quite a small effect.
2. The IM could appear at higher frequencies where NFB is less effective. We are talking about IM sidebands appearing at a spacing set by syllabic rates e.g. 0.5Hz sidebands on a 10kHz signal.
RF linear PAs often do have local NFB, but these thermal memory effects make this less effective.
I'm not saying this is all a problem. I am saying I am not yet convinced that it is not a problem.
Pretty much were I stand at the moment.- Status
- Not open for further replies.