I think what you have is the forward path frequency response, and, in any case, looking at the zero dB crossing you have no phase margin or gain margin at all.
That was Opel Loop Gain, you can't just look at the zero crossing, Close Loop Gain should be detracted first.
Re Thermal Trak
The diode choice certainly does not provide a convenient match with the transistor, so there is extra work.
But they still better coupled with vastly superior response speed to an add on temperature sensor, so why do you hate them?
Best wishes
David.
The diode choice certainly does not provide a convenient match with the transistor, so there is extra work.
But they still better coupled with vastly superior response speed to an add on temperature sensor, so why do you hate them?
Best wishes
David.
Well these devices are discontinued. However the focus on thd is imho of some academical interest it doesn't have so much value for the connoisseur of music.
If the spectrum of harmonics matches the spectrum of acoustical musical instruments not even trained musicians can hear a difference between 0.01% thd ( very few if any amp at all will accomplish that with real reactive loads) and 2% thd. What is equally important the harmonics spectrum must not change versus frequency and must be constant over a wide range of power.
These are really important design objectives. Whether or not speakers qualify
is another question.
Above this a third objective the "dynamics resolution" .
This term is not to be confused with dynamic range or dynamic headroom.
Especially in this class of descriptive sound quality tube amps, the QUAD 303,
and amps with Thermal Track excel but you don't see it in any measurement,
but you hear it, and that is exactly what only matters.
(post #500)
the low gain margin at about 6 to 7 dB looks a bit suspicious to me.
I took a look onto your Spice model in a later post. Maybe it is wise to also consider different load capacities behind the output coil, leaving the speaker resistance high (speakers often have a high impedance at high frequencies). The series resonance between the output coil and the load capacitance will influence your gain and phase margins.
Furthermore, building the beast up you may have additional loss of stability margins.
The difference between theory and practice is larger in practice than in theory.
Very interesting is the phase behaviour below the unity gain crossover. At around 50 kHz, the negative feedback seems almost to turn into positive feedback. Here, the loop gain amounts to 60 dB. I know that theory completely allows this and so the circuit really should be stable. Nevertheless it looks unusual.
Good luck,
Matze
(If you are interested in compensation techniques, this thread might also be interesting for you:
http://www.diyaudio.com/forums/solid-state/233122-amplifier-nested-miller-compensation.html
)
Hi Damir,
the low gain margin at about 6 to 7 dB looks a bit suspicious to me.
I took a look onto your Spice model in a later post. Maybe it is wise to also consider different load capacities behind the output coil, leaving the speaker resistance high (speakers often have a high impedance at high frequencies). The series resonance between the output coil and the load capacitance will influence your gain and phase margins.
Furthermore, building the beast up you may have additional loss of stability margins.
The difference between theory and practice is larger in practice than in theory. Very interesting is the phase behaviour below the unity gain crossover. At around 50 kHz, the negative feedback seems almost to turn into positive feedback. Here, the loop gain amounts to 60 dB. I know that theory completely allows this and so the circuit really should be stable. Nevertheless it looks unusual.
Good luck,
Matze
(If you are interested in compensation techniques, this thread might also be interesting for you:
http://www.diyaudio.com/forums/solid-state/233122-amplifier-nested-miller-compensation.html
)
(post #500)
Hi Damir,
the low gain margin at about 6 to 7 dB looks a bit suspicious to me.
I took a look onto your Spice model in a later post. Maybe it is wise to also consider different load capacities behind the output coil, leaving the speaker resistance high (speakers often have a high impedance at high frequencies). The series resonance between the output coil and the load capacitance will influence your gain and phase margins.
Furthermore, building the beast up you may have additional loss of stability margins.
Very interesting is the phase behaviour below the unity gain crossover. At around 50 kHz, the negative feedback seems almost to turn into positive feedback. Here, the loop gain amounts to 60 dB. I know that theory completely allows this and so the circuit really should be stable. Nevertheless it looks unusual.
Good luck,
Matze
(If you are interested in compensation techniques, this thread might also be interesting for you:
http://www.diyaudio.com/forums/solid-state/233122-amplifier-nested-miller-compensation.html
)
Hi Matze,
Thank you for your interest and suggestion. I know that 6-7dB of the gain margin is quite low, but that simulation was just my exercise to see what is possible with the combination of TPC and TMC. Original TT amp, I used for this simulation, uses TMC only and it is playing music for more then year and it’s very stable and nice sounding. You can look at this thread for more info.http://www.diyaudio.com/forums/solid-state/182554-thermaltrak-tmc-amp.html I am not sure if I ever try this TPC-TMC staff in real amp, but who knows.
Thanks
Damir
Where did you get this information from? I am looking at the OnSemi online catalogue and the NJLs are still marked as "Active".
Audio Transistors
As you mentioned, there is a significant tempco (and VF) mismatch between the diode and the transistor. If I remember correctly, the diode and the transistor tempcos become equal at a pretty high current through the diode (some 25 mA?), definitely not practical for a VAS. Moreover, at the VF where the tempcos match, the power transistors bias current is already uselessly high. There are, of course, solutions ( and Mr. Cordell's book illustrates quite some) but obviously this situation always leads to a more complicated bias circuit (compared e.g. with the functionally equivalent Sankens).
Also, implementing a fast thermal compensation loop has obvious advantages, but it's difficult to design up front (I don't call tuning on a prototype "design") and there's virtually no good references about (Mr. Cordell's book is also lacking on the thermaltrak circuits design methodology). When using multiple output pairs (which is almost always the case), it gets even more complicated and difficult to design (or you must give up most of the advantages of the fast thermal loop).
All the mess for some theoretical benefits (like not starving the output devices during fast power transients) which probably nobody really cares about.
TPC + TMC
Indeed, it can work.
But first, read this about the optimal capacitor ratio for TPC.
It is Harry Dymond who deserves all the credits for pointing this out.
Next, look here how TPC + TMC should not be implemented.
And finally look here for one of the correct possible implementations.
Bottom line: all these Nth order compensation schemes only make sense if the stage (i.e. the TIS) to which it is applied, has sufficient gain.
Also notice the shunt compensation (C8 & R21) Do you hear me Mr Self? I repeat SHUNT compensation. Also Bob Cordell calls this shunt compensation.
Cheers,
E.
Indeed, it can work.
But first, read this about the optimal capacitor ratio for TPC.
It is Harry Dymond who deserves all the credits for pointing this out.
Next, look here how TPC + TMC should not be implemented.
And finally look here for one of the correct possible implementations.
Bottom line: all these Nth order compensation schemes only make sense if the stage (i.e. the TIS) to which it is applied, has sufficient gain.
Also notice the shunt compensation (C8 & R21) Do you hear me Mr Self? I repeat SHUNT compensation. Also Bob Cordell calls this shunt compensation.
Cheers,
E.
Last edited:
Actually I first pointed this out in a paper I wrote back in 2004, as Jan Didden can confirm-many years before Harry Dymond's AES paper was published.
Note, however, that one need not adhere to this "optimal" capacitor ratio for TPC if one's TIS can supply enough current to the compensation network when the capacitor connected to the TIS is made larger than that connected to the output of the transadmittance stage (TAS).
Steady on, old chap, D. Self does not really need to take your condescending attitude.
Hi, Edmond,
Thank you for the good information you provide.
I just took several hours to dive into the math mechanism behind the TPC.
I find an improper implemented TPC will increase THD, as the link you provided.
There are two capacitor involved in TPC, the problem is the one near input stage of VAS should be small value. If that capacitor is too big, it is attempt to behave more like shunt compensation before VAS. That will reduce global negative feedback, it will result larger THD figure.
David,
Now I am somewhat confused. For the near ideal case with the injection ac voltage source referenced to a low impedance doesn't it represent the loop gain and not loop gain + 1.
Thanks
-Antonio
Hi Waly,
The ThermalTraks are surely not perfect devices, and many approaches to using them have failed. Moreover, there necssarily is some tweaking of the relative temperature compensation of the drives and the output devices for any given amplifier with any given approach to heatsinking (for example, how are the drivers heatsinked? Are they on the main heatsink or do they have their own? Where are the pre-drivers heatsinked in the case of a Triple? As mentioned in my book, I try to use bias spreader circuits that allow for some thermal feedback from the driver and the output stage to allow establishing the best balance. I'll be the first to admit that it takes some experimentation on the specific amplifier, but I have been able to achieve very good results, as shown by some of the figures in my book. Their faster response to the actual conditions on the power transistor die is very helpful. Nevertheless, they are certainly not perfect.
The output stage bias is important, and I have always been troubled by the way it moves all over the place with traditional approaches to bias temperature compensation.
Until ThermalTraks came along, I was much more comforatble with power MOSFETs in this regard. Not only are they less fussy about bias, since with them the more the better, but they are at the end of the day less variable in bias with changing conditions of warm-up and program material.
Cheers,
Bob