It's not even monotonic. Fit between calculation and experimental is impressive as well. If I can get a chance to scan it tomorrow, I'll post it. Morgan did some careful analysis and found that if the O/L distortion is less than 1-2% and the device is 3/2 law, feedback reduces all the harmonics.
Interestingly, since Ralph sells balanced amps, the Baxandall figure is even less appropriate for his equipment. I haven't seen any good published distortion spectra from the Atmasphere amps, but it would be quite interesting to see. It's a circuit I'd like to try out sometime since I'm loaded up with high perveance power triodes and the midrange/tweeter portion of my speakers has a flat impedance curve.
The only distortion figures published are on the Soundstage! site, unfortunately Bascomb King had one speaker terminal at ground which created an imbalance in the drive of the output section, so distortion was a lot higher and power lower than we actually see on the bench.
That report is from about 14 years ago.
Suitable power tubes are 6AS7G and similar (avoid GA variants), 6336, 6C33 and EL509 and variants. A basic circuit can be found in the thread http://www.diyaudio.com/forums/tubes-valves/161112-what-tubes-tube-amp.html
That report is from about 14 years ago.
Suitable power tubes are 6AS7G and similar (avoid GA variants), 6336, 6C33 and EL509 and variants. A basic circuit can be found in the thread http://www.diyaudio.com/forums/tubes-valves/161112-what-tubes-tube-amp.html
Do you have any distortion spectra that you could share? If you want to email me privately, that's fine, I'll not post anything without permission.
Yes, the amplifier's inherent output impedance is included inside the feedback loop, and decreases along with the gain.
However, the internal operation of the circuit IS different because of the feedback. Some of the internal voltages and
currents will differ significantly from their open loop values, when the input is scaled for the same output level.
This can cause a variety of problems if not properly considered in the overall design.
For example, feedback around a circuit having coupling capacitors can cause a rather large low frequency peak before
one of the capacitors (usually the first one in the circuit). The feedback tries to compensate for the LF rolloff of the capacitor.
Last edited:
Can you give a reference to somewhere where one may read about this calculation? Typically, one views an amplifier as an ideal (zero-impedance) voltage source in series with a resistance Rint, and Rint would be the "output impedance." (Measured by one of the obvious ways, like calculating from the change in output voltage under a change in output load.)
Are you saying that this model for the amplifier is not a good one in the case of the power paradigm? Is that because the amplifier "knows" that it is a power-paradigm amplifier and it actually behaves differently from a typical amplifier, or is it that the usual model is still good, but for some reason not relevant here?
Maybe these points are addressed somewhere?
Thanks,
Chris
No it isn't. The amp doesn't change whether it is inside a feedback loop or not. All the non-linearities with which the amp processes the input signal are exactly the same. How could it be different? All the feedback loop does is 'fudge' the input signal so that what comes out is what we want - a good replica of Vin.
If an amp with an open loop gain of 1000 has Vout of 10V, it has a +Vin of 10mV and a -Vin of zero.
If you wrap it in a 40dB feedback loop with the same 10V out, the +Vin will be 1V and the -Vin will be 0.99V for the SAME 10mV input to the amp itself. No change with any of the internal voltages of currents.
Jan
Chris, see the White Powrtron as an early example:
100 amplifiers – part 1, 1916 -1954. | Lilienthal Engineering (scroll down about 4/5)
It uses a combination of series and parallel derived feedback to give a constant output power for a constant input voltage. Used with 99.9% of speakers (those ones that don't sound like music 😀), the frequency response will be decidedly non-flat.
100 amplifiers – part 1, 1916 -1954. | Lilienthal Engineering (scroll down about 4/5)
It uses a combination of series and parallel derived feedback to give a constant output power for a constant input voltage. Used with 99.9% of speakers (those ones that don't sound like music 😀), the frequency response will be decidedly non-flat.
This is not going to be possible without a non-linear circuit, which has serious inconvenients as I have shown.
A linear mix of series and parallel feedback will result in an output resistance, and this will provide a very poor and limited approximation of constant power around Zload=Zout
Does the 'Power Paradigm' have its own private set of facts, such as an alternative set of Maxwell equations? If not, all you can claim is that an appropriate mixture of voltage and current-sensing NFB can be carefully designed in order to provide a chosen output impedance. If this is equal to the nominal load impedance then moderate changes in load impedance will result in only small changes in output power. This is quite different from claiming that NFB does not affect output impedance.atmasphere said:
See post #100, including the quote.
*******
I've been thinking about that Baxandall example of negative feedback from post #87 and #90. I should have thought of this earlier; had a lot on my mind yesterday...
Anyway, the example given in the two posts is for a simple gain stage. So the feedback is applied around that stage of gain. In this case, the propagation delay is so short that in essence, it works.
The problem comes in when you have a complete amplifier circuit with several gain stages (both voltage and current) nested within the loop. At this point the propagation delay becomes significant (easily measured with very basic test equipment), and in fact its well-known to any designer that you have to limit the bandwidth of the feedback itself, as above a certain frequency it becomes positive instead of negative feedback, causing oscillation.
The example given would work in a preamp, but in an amplifier is probably not a practical application. So if one is to make an argument for GNF in an amplifier, the example of a single gain stage should probably not be used.
You are confusing phase delay with propagation delay. The first is well-known and easily dealt with (as any student who has passed Control Theory 101 will attest). The second is well-known, difficult to deal with, and completely irrelevant to audio; it's something that microwave designers have to consider.
It is well known to any audio designer that this problem is not caused by propagation delay but by low pass filters or (possibly) integrators. Propagation delay is only a problem for people designing feedback into RF amplifiers. The problem you describe is simply one of loop stability; the mathematics of this is well known to audio and servo system engineers but it does require a little knowledge of complex numbers.atmasphere said:
So when I can show that a certain amount of time as passed between when the signal arrived at the input, as opposed to when it arrives at the output, you call that 'phase delay'. Do I have that right?
In the example that I gave, the LF peaking at an internal circuit node (due to the attempt to extend the LF bandwidth by the feedback)
is NOT present in the open loop case.
Yes of course (ideally) the amplifier circuit works the same either way, and the input is combined with the fed back signal from the output,
but this is the point. The error signal is larger at LF, causing the peak. The response open loop does not have that peak. Overload could occur
because of the peak with feedback that would not occur without the feedback.
Last edited:
A few cascaded low pass filters with high corner frequencies can give the superficial appearance of a genuine propagation delay if you look quickly. The difference is that filters can be undone by inverse filters; a true delay requires a Tardis to undo. Only 'high-end' manufacturers can include Time Lord technology in their circuits.
Sure I get that- and I think we are on the same page.
So let's not think about the timing constants near cutoff (IOW this isn't a filter issue), what I am talking about is the fact that the time it takes for a signal to propagate through the circuit is time t regardless of frequency. IOW this value will be invariant at 20Hz and at 20KHz or 100KHz. Its easy to measure- just put a square wave though the amp within its passband and compare input to output in phase on the same time base. What you will see is the output is delayed by the same amount (time t) at all frequencies for a given amp or preamp.
So do you call that propagation delay or phase delay?? Where I come from it is propagation delay, which is different from phase delay as timing constants and cutoff frequencies have nothing to do with it.
Propagation delay through an amp (i.e., the constant delay vs frequency) is on the order of nanoseconds. How does (say) 10ns translate to phase change at (say) 100kHz?
SY, I disagree. It seems that the times I have seen are longer than that.
The concern has nothing to do with phase change BTW.
The concern has nothing to do with phase change BTW.
- Status
- Not open for further replies.