we do not disagree.Bob Cordell said:
It depends very much on the program material and how we use our amplifiers.
Whether it's a 12W amplifier driving 104dB/W/m speakers or JC's example 800W into 4ohms driving 83dB speakers.
If we listen at levels ~ 20 to 30dB below the maximum capability of the amplifier then PMAs plots are very much a worst case scenario.
There will be periods when we stress the transistors more, but these should be rare maybe very rare, what about 20ms during an evening's listening? Or would be happier if I suggested we tolerate 300ms per listening session?
How does the transient thermal change affect the signal passing through the amp? This is clearly not easy to measure with steady state tests and it may be significant. I have been reading about delay vs. temperature in very precision systems and the numbers are in the 10 pS/degree C in a carefully compensated system. I seriously doubt anyone would hear something that small as a steady state change but an abrupt change could modulate a signal enough to be audible (if we knew what was audible).
scott wurcer said:
I agree Scott and if someone likes to simulate it with Spice it’s important to use a model that is as good as possible and not a very simple model like I have seen someone used on this forum.
Attached is an example of a slightly better model.
BTW: the thing I showed was not made by SPICE, but by COMSOL
OT: I really hate the limitations in file size that is allowed to attach.
I really like your heat sinks, Nelson. I wish that I had the same. I have close, but not the same. Of course, it is the thermal capacitance of the sink that will smooth out much. Then, a real engineer, looks at the thermal capacitance of the individual devices, which is usually available from a good data sheet.
So, there are 3 approaches to smooth thermal performance, just from a thermal point of view.
First, there is the use of rich class AB-1 to smooth the transition in average power, from low levels to higher levels. Two, spreading out the power dissipation, lowers the overall thermal resistance of the output devices, by using a number of output devices. The electrical insulator matters too.
Third, with a huge thermal capacitance and a good thermal resistance, it is difficult to rock the temperature of the output devices too much.
So, there are 3 approaches to smooth thermal performance, just from a thermal point of view.
First, there is the use of rich class AB-1 to smooth the transition in average power, from low levels to higher levels. Two, spreading out the power dissipation, lowers the overall thermal resistance of the output devices, by using a number of output devices. The electrical insulator matters too.
Third, with a huge thermal capacitance and a good thermal resistance, it is difficult to rock the temperature of the output devices too much.
...Fourth approach exists that I use, you don't: an asymmetrical class A amp, with a source follower powered by one more bootstrapping source follower, and loaded on a counter-modulated current source. In such case power variations will be minimal due to phase shifts between voltage and current when loaded on a real world speaker. 
The concern I expressed above is that the time constant at the sensitive junction for the temperature shift can be near the audio range or at least the range of modulation of audio tones (attack and decay of a note). Even if the heatsink is infinite the resistance to it is not zero and the can be a short term chnage in the devices. Thgis is well understood in IC's and efforts are made to control the modulation of sensitive parts by output devices. However the issue may still be significant in an amp. Is there a way to measure it? Perhaps a change in a low level HF tone while a large change at nearly DC is applied?
KSTR said:
This good idea was implemented thirty years ago, about month ago I described it in details on this forum. I used a complex HPF to separate a high tone from a low disturbing one so it won't swing a high pitch tone from an oscilloscope screen. The problem was, first filters I used had non-linear cores so could not be trusted.
) . Mind-boggling (18" wafers that is).
KSTR said:
No Klaus, it was made for observations of linearity change with voltage swing, I used it to experiment with everything in the amp to see how different changes of topology and regimes were reflected on linearity. I never measured THD, I always used to observe waveforms and to listen to sounds the same time. It was more revealing and useful, like giving thermometers to each and every patient VS averaging of a temperature in the clinic for some kind of a report.
I think the most interesting to me right now would be the transient phase change to the HF signal. I'm not sure how the filter will effectively isolate to tone or if having the filter will cause an effect in the process. Making a good isolation filter that won't distort the tone is the hard part.
- Status
- Not open for further replies.