If we're talking about designs with plenty of feedback, that would depend upon whether you wish to care about what I imagine would be 1 to 10 ppm variations in characteristics.
An amplifier is a time invariant system whose output depends only on its present input.
If a linear time invariant circuit ( quadripole) contains only résistors and current or voltage sources it is a memoryless circuit.
If you have reactive elements, it becomes a memory circuit. The output dépends on the past also by the integrating aspect of reactive elements
Time invariance is related to the elements ( R, C ,L) or the coef of the equivalent differential system équations beeing constant over time. It can be with memory and time invariant aswell
Therefore, thermal so called memory is more related to time (non)invariance and makes the analytical analysis very difficult (IMHO).
It also involves strange frequency slopes. Looking at thermal resistance charts for the BC3x7 I see two slope regions, one with a 1/F**.5, right at the junction plane, and 1/F**.333, probably as the thermal impulse becomes spherical. Both of these have phase/slope characteristics you just couldn't reproduce exactly with passive components. So I wonder what they "sound" like.
Hi jcx,
I do in fact advocate the use of two Zobels in my book - one on each side of the decoupling inductor. With fast output transistors, it is desirable to have the Zobel physically close to the output stage, right on the board with the power transistors. In fact, I like distributed paralleled smaller Zobels there. Having the first Zobel at the output stage mitigates the reality of wiring inductance between the output stage and the amplifier output terminals.
The second Zobel ideally is right at the output terminals, so that RF and EMI from the speaker cable is damped out right there. The decoupling inductor can be placed somewhere in between where any magnetic field it radiates will do the least harm. Some will call this Pi approach to the output network overkill, and it may be in some amplifier designs.
Cheers,
Bob
Hi Mike,
I disagree, a bit. To first order, I agree. However, my favorite example of memory distortion is the program-dependent temperature variations in the class AB output stage that can upset the output stage bias current, and at times drive it well away from the optimum. Crossover distortion will then be dependent on the history of the signal.
BTW, I am not a big fan of the term memory distortion and many have abused it with a bunch of hocus pokus that is all too common in our industry. Many obfuscate their memory distortion by failing to mention that the mechanism of history is temperature. Unless there is some other significant mechanism at work that I am unaware of.
A test might be an adaptation of the SMPTE IM test where 60Hz and 7000Hz are applied in a 4:1 ratio. Replace the 60Hz with 19Hz plus 20Hz, which produces a beat note at 1Hz that modulates the operating temperatures at this slow rate. Make the beat frequency 0.1Hz if you wish. Then look for changes induced in the 7kHz carrier. You could even look at the time-dependent THD.
We need to bear in mind that heat sinks have pretty long time constants, but power transistor cases have much shorter time constants. Finally, transistor junction temperature has a very short time constant. I touch on these issues in my book in Chapter 14. See figures 14.2, 14.3 and 14.4.
Finally, I'm guessing that circuit designs that mitigate sources of thermal modulation may lead to superior performance and sound quality anyway. With regard to crossover distortion, the greater temperature tolerance of output stage bias with power MOSFETs may lead to less temperature dependence in output stage performance. This is speculation, of course, and in the end I'm not sure memory distortion is an issue in reasonably well-designed amplifiers.
Cheers,
Bob
Cordell approved PCB for final Cordell Amp circuit?
I did a quick search and I assume this has been asked a lot ...
Is there a Cordell approved PCB for final Cordell Amp circuit in the book?
It would be great fun to start out with the first basic amp circuit and then listen for a while.
Then, add in each performance-improving new circuit and then listen to see if you can hear the difference. Slowly building up to the final amp and having a better understanding what each addition does along the way ...
This would be a blast!
Is there an approved PCB?
Cheers,
Jeff
PS You could do it with the fancy chipamp too. Approved PCB for the final amp, start with a basic circuit and add in the improvements one-at-a-time... can you hear it?
I did a quick search and I assume this has been asked a lot ...
Is there a Cordell approved PCB for final Cordell Amp circuit in the book?
It would be great fun to start out with the first basic amp circuit and then listen for a while.
Then, add in each performance-improving new circuit and then listen to see if you can hear the difference. Slowly building up to the final amp and having a better understanding what each addition does along the way ...
This would be a blast!
Is there an approved PCB?
Cheers,
Jeff
PS You could do it with the fancy chipamp too. Approved PCB for the final amp, start with a basic circuit and add in the improvements one-at-a-time... can you hear it?
Last edited:
I agree it would be a very interesting project if Bob did design a complete amplifier circuit and board for the Diy audience. We already have the Doug Self designed Pre-amp that we can purchase and build if we want to do that. How about it Mr. Cordell is this something that you could do for all your fans here? I would expect that what you would design would differ in many ways to what Mr. Self would create as you two seem to have divergent design philosophies in how you design active circuitry. Sort of a Bob Cordell Arduino board for the Diy crowd.
Thermal Memory THD?
An easy way to check this is to do
- low level THD20k while the amp is relatively cool
- then do 10min of 1/3 power as FTC to get it nice & hot
- immediately after, repeat the low level THD20k
With even Jurassic gear, you can leave the THD20k setup untouched during the FTC warm up and get the last THD20k within seconds.
Particularly obvious with CFP outputs running low Iq.
Don't forget to check Iq before & after to find the cause.
I second that.
An easy way to check this is to do
- low level THD20k while the amp is relatively cool
- then do 10min of 1/3 power as FTC to get it nice & hot
- immediately after, repeat the low level THD20k
With even Jurassic gear, you can leave the THD20k setup untouched during the FTC warm up and get the last THD20k within seconds.
Particularly obvious with CFP outputs running low Iq.
Don't forget to check Iq before & after to find the cause.
I did a quick search and I assume this has been asked a lot ...
Is there a Cordell approved PCB for final Cordell Amp circuit in the book?
It would be great fun to start out with the first basic amp circuit and then listen for a while.
Then, add in each performance-improving new circuit and then listen to see if you can hear the difference. Slowly building up to the final amp and having a better understanding what each addition does along the way ...
This would be a blast!
Is there an approved PCB?
Cheers,
Jeff
PS You could do it with the fancy chipamp too. Approved PCB for the final amp, start with a basic circuit and add in the improvements one-at-a-time... can you hear it?
Hi AudioLapDance,
Sounds like you are referring the the basic amplifier that is evolved in Chapter 3 (the Evolution Amplifier?). Is that right?
The answer is that there is not. I have not had the time thusfar to do something like that. Darned day job.
But hmmmm... That is an interesting idea. Tempting...
Cheers,
Bob
Yes, exactly. This is a good test, even if it is not thought of in the context of memory distortion.
And you are right about the CFP.
Cheers,
Bob
The thermal stability benefit of the CFP comes at a high price due to the need to operate most versions of it at scary low bias current, resulting in a microscopic class A region. If you operate most of these at higher bias, you get significant gm doubling.
Cheers,
Bob
Could use a smaller RE, 0,1 Ohm would +- 25mA.
Forgot to mention in the previous topic, CFP output decreases NFB in 40% with load, surely this is problem, had an amp released here in the 70s that had this problem, DF decreases with load, when the volume increases, the bass were "floppy".
VBE multiplier is a big problem, even if you use ThermalTrak (are more costly) still have the problem of having "two bias" one for winter where the amplifier operates more cold and other for the summer, dynamic bias control is necessary.
I like the BJTs by having good symmetry PNP/NPN, but I believe that a Mosfet output would be the better choice, I will use in my next amp.