https://en.wikipedia.org/wiki/Relativistic_heat_conduction
RHC - depends on the material. It has been compared to the speed of sound
but works by a very different mechanism.
Among physicist's , there seems to be a lot of disagreement.
The speed of a measurable change is very fast , but attenuated.
relative change is much slower.
Edit - at an atomic level , it could be as fast as light.
RHC - depends on the material. It has been compared to the speed of sound
but works by a very different mechanism.
Among physicist's , there seems to be a lot of disagreement.
The speed of a measurable change is very fast , but attenuated.
relative change is much slower.
- temperature difference
- length
- cross-sectional area
- material
Edit - at an atomic level , it could be as fast as light.
Last edited:
Yes, that is how I measured delta-Ibias due to thermal transients - measure at current zero crossing. But it's pretty involved.
Linear Audio | your tech audio resource
Jan
I know this is related to thermals vs bias........ But I would work it from the best bias level and then work on the thermals.
We know that under-bias causes thd to rise sharply so that must be avoided. And, optimum bias for lowest thd is difficult to hold there (without dsp monitor/correction). So, the best practical bias level is slightly above the minimum for thd. Then work on the thermals to avoid run-away for bjt etc.
View attachment THD vs Ibias.pdf
THx-RNMarsh
We know that under-bias causes thd to rise sharply so that must be avoided. And, optimum bias for lowest thd is difficult to hold there (without dsp monitor/correction). So, the best practical bias level is slightly above the minimum for thd. Then work on the thermals to avoid run-away for bjt etc.
View attachment THD vs Ibias.pdf
THx-RNMarsh
Doug, this is really quite crude. I suppose you also do not like heat spreaders.
Cheers,
Bob
I don't think your distortion measurements vs time demonstrate this at all.
Anyone who looks at your fig 22.44 (sensor on top of power transistor) and Fig. 22.45 (ThermalTrak) will realize that the ThermalTrak sensing is more than twice as fast - and you even note it. Your statement "the position of the sense diode on the lead frame means it is much closer in temperature to the heat sink than the junction" appears to be speculation on your part based on distortion vs time of a wrongly-applied ThermalTrak bias scheme.
I used good device insulators, but they were certainly not unobtanium.
I also find it very troubling that you made no reference to my book, wherein there was a great deal of prior work on ThermalTrak transistors, including biasing schemes that actually work right.
Cheers,
Bob
The slug in the transistor is copper , insulated (thermal resistance) from a aluminum extrusion. I would think the copper's conductivity would "win out".
The extrusion's thermal equilibrium would have a minor effect on what happens on the copper slug (semi).
Consider a CPU cooler with the center copper bar surrounded by the aluminum. Why ? Just an analogy .
OS
The extrusion's thermal equilibrium would have a minor effect on what happens on the copper slug (semi).
Consider a CPU cooler with the center copper bar surrounded by the aluminum. Why ? Just an analogy .
OS
Ha ha, I got that cover!!! If you look at the picture I posted in #6133, I have all the 0.22 resistors soldered at the bottom of the pcb with leads sticking out on top. I just ordered 100 of 0.33ohm. I plan to bring the OPS up, and check first, then I am going to parallel the 0.33 with the 0.22 on the board. Guess what is the value.....0.13!!!! I have every intention to try that.
I matched beta and Vbe, I even choose resistors so the outer two is 2.5% lower in value and the two outer transistors are 1% higher in beta to hog a little more current than the middle to compensate slightly lower temperature of the outer transistors. I am crossing my finger I can make it to work with 0.13ohm.
I'll play it safe with 0.22 to get my feet wet, then I am going to push it. I also have plenty of 0.22 also........0.11!!!! Now, I received the full assortment of 0.12, 0.15, 0.18, 0.2, 0.22 and will be 0.33.
the attachment looks like a Blameless amp.
Last edited:
I have no intimate knowledge of the TT, but common sense tell me that if one goes through the trouble to put the diode in with the transistor, it would be insane to not put them on the same metal back plate. You cannot get closer and faster response than that.
The main reason I did not choose this is because I want the standard footprint so I can change to any transistor I want and not having to modify the pcb. I am sure I can find a whole bunch of transistors to replace the MJW.
Again, people been producing reliable amps like forever before the TT. So there is a way to make it reliable.......without uP or servo.
Last edited:
Hi Mr. Cordell
I was looking at quite a few schematics of the commercial amps, Ostripper and Bonsai schematics. Commercial brands including Aragon, Acurus, Bryston, Krell, Mark Levinston, Pass Lab and YBA. These are serious and high end amps. Other than Bonsai's schematic using exactly the same way I did using the diode from collector back to the base of the dalington to prevent the transistor from turning on. No body else even have anti clipping circuit.
Is it that important to have that?
I did simulation on 32V peak swing on the IPS/VAS both in 1K and 20KHz, the difference in distortion is minimal with the diode I have in my circuit.
Is this LTspice?
If so, can you post the *.ASC and a pic of the circuit?
If not, please post a pic of the circuit, results and what diode you used?
_______________________
Certain topologies are less prone to saturation so don't need these diodes.
Last edited:
My measurements seem to indicate it is NOT on the die - there's appreciable delay of a few 100 msecs before the diode reacts to changes in die dissipation.
Jan
Hi Jan,
I don't think that's in dispute. Someone posted earlier a picture of an NJL cracked open that clearly shows the diode as a separate die. jacco was just pointing out that whilst the thermal response is fast, it could be even faster if the diode were on the same die as the transistor.
This raises the question, would that actually be a good idea or is it theoretically possible for thermal response be too fast? Could it get to the point where bias is being altered in realtime with low frequency signals and creating distortion?
Completely agreed. It is critical to avoid a significantly under-biased condition. This can occur when there have been loud passages and the heat sink is hot, and the program suddenly goes to a low-level. The bias spreader will remain in a lower-voltage condition for some amount of time, while the temperature of the transistor junction with respect to the heat sink will cool down quickly (within seconds), potentially leaving the output stage seriously under-biased. Faster-reacting bias spreader sensing can help this (like mounting the sensor on top of the power transistor or using a ThermalTrak transistor), but speed of response is important.
Of course, it can go the other way, depending on whether the thermal is over-compensated or under-compensated - bias may be quite higher after the long hot interval, but then may under-shoot.
Run your amplifier at 1/3 power for a half hour, then kill the signal and measure the voltage across RE and compare it to the voltage that was across RE in the beginning before the signal was applied (but after the amplifier has had perhaps a half hour to stabilize at idle after turn-on.
Cheers,
Bob
Yes, that is right. However, if the diode is integrated on the die, it cannot float with respect to the transistor voltage, so biasing circuit options are more limited.
Cheers,
Bob
Actually Harry I did worry about this and I think I mentioned in my book that is is preferable to use two ThermalTrak diodes, one from the NPN device and one from the PNP device to mitigate this possibility somewhat.
But we can also note that most amplifiers have quite a bit more negative feedback at the lower frequencies, so that would likely mitigate this concern.
It would be very interesting to operate an amplifier at a high level at 20Hz and look at the amplitude and waveform of the Vbe of an unused ThermalTrak diode. This would also give some indication of the sensing phase delay.
Cheers,
Bob
We've hit on this before but needs recycling every once in awhile;
Let's contrast that bias spreader/thermal affects with FET OPS bias needs and thermal affects on bias and distortion, pls.
THx-RNMarsh
Absolutely right. And it works very well- see the Class-A amplifier design in my book where the Iq is controlled very closely by negative feedback. However, this is of course the case where we don't need to control it very precisely.
A cruel irony is a fickle mistress.
Spare my blushes.
What I do find unhelpful is a tendency to make up new terms when perfectly good ones already exist. This just sows confusion.