Hi Ajt
Since you are into tube amp. Tube amp mostly are in the 40 to 50W range ( if that). Can you comment on the need of high power to cover the large transient signal in the program? Ostripper and a few others said they can hear a clear improvement after raising the rail voltage. I have no experience on this at all.
As you can see, I am confused between a large Class A region and sacrificing the peak power vs high peak power with high rail voltage and only have a small Class A region.
I have been on the tube side here before. I decided to go with the SS as my first amp. People on the tube side pretty much said 10W of true class A power is plenty. I have been holding onto this and try to design an amp that was only about 60 to 100W, but having 8 to 10W of pure class A region so in normal listening level, I stay in Class A 99.9% of the time.
What is your opinion? I really like to hear from you that have experience on both sides.
Last edited:
My measurements confirm that ThermalTrack diodes still have a delay in responding to temp changes of a few 100 mS. Long term stability may be good, but transient thermal changes are still there.
See my article in Linear Audio Vol 9.
Jan
I was thinking of your article during the discussion.
I think that it has not been queted yet.
Your findings should certainly not be ignored.
Thermal transient variation of power amp quiescent
Instrumentation and findings - Jan Didden
More information from
volumes | Linear Audio --> preview
Thermal and electrical mechanisms (conduction, resistance, capacitance etc) are actually very much the same!
The reason that they don't track in this setup is that the time constants are very different.
As my article shows, although the Ibias does vary with output power, the variations are most probably not large enough to have an audible effect (3-6% variation in Ibias).
Jan
Sorry Mike, you make the same mistake as people saying feedback can't work because it becomes after the fact.
There is no theoretical reason why it would not work, only the current hardware has too much lag and phase shift. It is EXACTLY like a feedback loop in an amplifier that has too much phase shift or that is asked to operate beyond its bandwidth.
Jan
Is it dead time or phase shift? Is heat conduction constant velocity or variable depending on the gradient?
Last edited:
Glue the To92, or sot23, temp sensor to the Collector/Drain lead where it exits the output package.
TT is faster. Sanken's version will also be faster.
D.Self,
could you do the thermal estimation for a TT similar to the model you have written up in your books/papers?
I don't think there is any dead time. Cannot think of a reason why there would be any (but that doesn't mean it isn't there of course).
Jan
your chassis and transformer are perfect for a very good ClassAB 4ohms capable amplifier.
Just adjust the bias current, using your Oliver criteria, to match the dissipation capability of the sink/s.
5pr biased to 22mVre (100mA bias/pr through 0r22) takes account of some "hidden" internal resistances to give you a dissipation of 10*40V*0.1A + (drivers & predrivers) of around 44W
Assuming the sink has Rth s-a of 0.4C/W, 0.4*44*(DF=1.4) gives a deltaT s-a of around 25Cdegrees. With a Ta=30°C you end up @ Ts~55°C and Tc~60°C which requires the outputs to be derated to ~70% of their cold rating. Even a Tc=70°C (when listening very loud) gives a DF= 0.64
Not a concern since you have a 5pr output stage for a 150W into 4ohms amplifier.
I strongly suggest you build a very current capable PSU to feed the 4ohms speakers.
I suggest around +-40mF for each channel, 50V 15mF, 3 to each rail, which for a stereo amp comes to 180mF.
Last edited:
I have reported this in the past.
It happened on a Roender (BJT) and on a Sugden (LFET), both with closely matched components.
The LFET in the Sugden cannot be due to hFE or stopper resistance. It was due to a small Vgs difference. The same happens in a BJT where a small Vbe difference will also create the conditions to start hogging.
Last edited:
the Leach super amps 250wpc, are effortless when it comes to transients and bass slam....
but then my speakers then are not what you will typically use in a tube amp.....
to compare on the same set of speakers i am building a higher powered tube amp...
I believe they're filming Rimmer at it as we speak.
(he'd probably drill holes in the top of power devices, for his own diodes of choice)
Last edited:
It is dead time (or delay). It is modeled in the s plane as exp(-Ts). See the definition and implications here.
Quote: "delay is a [feedback loop] destabilizing factor, as it adds phase lag".
Every function whether it is chemical, electrical or even mental has a time component to it. The real question you need to ask is if the time factor is enough to cause a problem or if the margin is great enough to allow for these delays without causing a problem or changing the expected end result.
I beg to differ. Having the trajectory f(X1, X2,...,Xn, t) in the phase space time dependent is one thing. Having a delay (specifically modelled as f(X1,...Xn)*exp(-Ts), with the trajectory a spiral, perhaps (and ideally) converging to a limit trajectory) is a very particular case, with specific implications. I don't think there's any doubt that in the application under discussion, the delay plays a very significant role.
From a control theory perspective, the thermal runaway because of a long delay in sensing the temperature is a clear case of an unstable feedback loop. Current hogging maps to what are called "local instabilities", or local cusp points in the phase space. The destructive effect of current hogging (actually destroying the entire plant) is not something that the control theory is really concerned with. Unfortunately, identifying the "thermal" plant is usually very difficult, so we have to rely on recipes, rules of thumb, experience and common sense. In a nutshell, what is called "good engineering practices". It is usually much less expensive to slightly over specify the thermal system, than spending so much time simulating and approximating the system thermal response. For military and space applications, system thermal CAD is almost always the name of the game.
Last edited:
Waly,
I agree with what you are saying, and at the same time I don't think that counters what I was saying. As long as you take the time constants into consideration you would do just what you are saying. As in any engineering we usually deal with SOA or margins that often are much higher than absolutely necessary but that is just expedient and many times a cover your behind approach to engineering.
I agree with what you are saying, and at the same time I don't think that counters what I was saying. As long as you take the time constants into consideration you would do just what you are saying. As in any engineering we usually deal with SOA or margins that often are much higher than absolutely necessary but that is just expedient and many times a cover your behind approach to engineering.
When you dimension fatigue in a mechanical konstruction, good engineering practice was to know your Tochterman & Bodenstein, all was empirically driven. Now FEA rules. Results are visually impressive, but no better than the skills of the engineer feeding border conditions to the number crusher.