Yup, but that should't be a big problem, as a lot of DIYers can get them by a company order. If so, it can be solved by a group buy.
Both are silicon transistors, Mike. Temperature dependency is the same. Higher uprise of current is probably caused by oscilations...
No oscillations, scope was attached ! I observed different behaviour for thermal co often, not only output-bjts.
Mike
Mike
Maybe they have different Rjc value, so the Toshipa chips run hotter at the same heatsink temperature?
Which BW have your scope, Mike ? Oscilations can be at very high frequency, maybe you can't see it...
Christer said:
Yes, that's good possible, the Toshibas are likely to have smaller die, and have lower power rating, giving higher junctiontemp for same heatsink temp !
Upupa Epops said:
200mhz scope with 2mv resolution should not hide anything !
Bias was slowly rising while temperature was increasing.
Mike
Upupa Epops said:
No, the amp need not necessarily oscillate with Cf compensation. It is a delicate game, but can be solved. Depends on intersection of OLG gain curve and compensated closed loop gain curve, and on phase shift at this intersection. Only that the value must be "tuned" to specified devices, the faster devices, the more game.
Symasym's strategy is based on high OL gain with high frequency of the 1st pole, this results in several poles close each other at HF part of the amplitude characteristics, which brings problem to find compensation scheme, of course.
HF oscillations would result in output device warming, if he would not have measured.
I can bring 400 MHz oscilloscope, but don't see a reason.
Symasym sings
It has been playing music for some time in my room, not only waves 😉 My cups and glasses are trembling 😉 , very good amplifier. Let me test a little bit.
The symasym story confirmed importance of measurements. If I had started with the amp as I had assembled, I would have said flat, boring, etc. The measurements prevented this situation.
It has been playing music for some time in my room, not only waves 😉 My cups and glasses are trembling 😉 , very good amplifier. Let me test a little bit.
The symasym story confirmed importance of measurements. If I had started with the amp as I had assembled, I would have said flat, boring, etc. The measurements prevented this situation.
Carlos,
I am playing here:
Doralice
Desafinado
One Note Samba
Para Mechuchar Meu Coracao
Corcovado
So Danco Samba
Menina Flor
Eu E Voce
.... and many others.
Ever heard some of them? 😉
I am playing here:
Doralice
Desafinado
One Note Samba
Para Mechuchar Meu Coracao
Corcovado
So Danco Samba
Menina Flor
Eu E Voce
.... and many others.
Ever heard some of them? 😉
Hi,
I almost ready with my version. The first measuring results are good, and when I finish the power supply (300W transformer+2*22000uF BC capacitors/channel), I can start the first listening test. Due the increased local feedbacks (in tha VAS too), I found that my version is rock solid. No oscillation, even with pure 2.2uF capacitive load! I use onsemis (MJL21193/94), but two pairs, to drive low impedance loads.
sajti
Ps.: Pavel! I spent the long weekend in Prague. Nice city, and very very good pubs! You are lucky man to live in Prague!
I almost ready with my version. The first measuring results are good, and when I finish the power supply (300W transformer+2*22000uF BC capacitors/channel), I can start the first listening test. Due the increased local feedbacks (in tha VAS too), I found that my version is rock solid. No oscillation, even with pure 2.2uF capacitive load! I use onsemis (MJL21193/94), but two pairs, to drive low impedance loads.
sajti
Ps.: Pavel! I spent the long weekend in Prague. Nice city, and very very good pubs! You are lucky man to live in Prague!
sajti said:
sajti,
I remember that Budapest is also a very nice town. Hope to visit again in the future.
It can be some thermal ringing. It happens somtimes, depends by the size of the heatsink, type of the transistors etc...
The thermal compensation is a kind of feedback, so there can be some instability.
sajti
The thermal compensation is a kind of feedback, so there can be some instability.
sajti
It is exactly that. It can be thought of as a servo that hunts. Whether the loop is in control or not (damping) is up the the designer.
-Chris
-Chris
It did only go up, never down, Pavel has a much bigger heatsink, maybe that's the reason why it did not happen ?
The MJLs showed in the exact same configuration no change.
Mike
The MJLs showed in the exact same configuration no change.
Mike
It can be a game of all of those thermal resistances inserted (used - like pads, grease, device case, heatsink etc.), time constants, component positioning. I would not mind.
Actually, Chris, calling it a 'servo' is a misnomer - it's more akin to an error correction scheme, where an error is corrected by introducing an oposite sign but equal magnitude error elsewhere in the circuit.
Bias current fluctuations are possible for many reasons. The best way to observe them is to track them with a storage scope or even better, an old style X/Y plotter with a built-in time base (in fact I have an old HP XY plotter that I use for this amongst other things).
It should be noted that the compensation element is not really in close thermal proximity with what it is sensing, namely the output transistor. The system is really a series of 'thermal delay lines' which can be modelled as lumped components, namely time constants. For every connection of two Rth (for instance, junction to case, case to heatsink etc), there is a Cth connected to tha node - a thermal inertia component, in other words, mass. Further, our output BJTs are really arrays of degenerated smaller transistors (in order to insure good secondary breakdown characteristics and speed at the same time), which effectively produces many units with a lower thermal coefficient, coupled to each other in not so easily describable ways. What you get as a result is that the temperature vs 'effective Vbe' curves do not cancel exactly but may have several points of intersection. What is worse, in some designs, idle current is dependent on itself through a different mechanism, namely change of power supply voltage under load changes the current through the bias generator, which in turn changes idle currrent, which in turn changes the load and that in turn changes the power supply voltage. Couple all that with the thermal 'capacitnce' delays, and you can get all sorts of initial transients, until bias stabilizes.
Bias current fluctuations are possible for many reasons. The best way to observe them is to track them with a storage scope or even better, an old style X/Y plotter with a built-in time base (in fact I have an old HP XY plotter that I use for this amongst other things).
It should be noted that the compensation element is not really in close thermal proximity with what it is sensing, namely the output transistor. The system is really a series of 'thermal delay lines' which can be modelled as lumped components, namely time constants. For every connection of two Rth (for instance, junction to case, case to heatsink etc), there is a Cth connected to tha node - a thermal inertia component, in other words, mass. Further, our output BJTs are really arrays of degenerated smaller transistors (in order to insure good secondary breakdown characteristics and speed at the same time), which effectively produces many units with a lower thermal coefficient, coupled to each other in not so easily describable ways. What you get as a result is that the temperature vs 'effective Vbe' curves do not cancel exactly but may have several points of intersection. What is worse, in some designs, idle current is dependent on itself through a different mechanism, namely change of power supply voltage under load changes the current through the bias generator, which in turn changes idle currrent, which in turn changes the load and that in turn changes the power supply voltage. Couple all that with the thermal 'capacitnce' delays, and you can get all sorts of initial transients, until bias stabilizes.