How about Blackman’s Theorem, do we agree on how to apply that?
R_out of 1 Ohm gives 1 Ohm output impedance to this circuit when all feedback is disabled, yes?
Then with all loops operating do we not show 168 dB of “something” (I like to call it “loop gain”) reducing the output impedance measured with I1 test current source connected to the output?
can’t be bothered to check anyone else’s math Mike?
[a couple of quick discriptive text edits in the math comments]
R_out of 1 Ohm gives 1 Ohm output impedance to this circuit when all feedback is disabled, yes?
Then with all loops operating do we not show 168 dB of “something” (I like to call it “loop gain”) reducing the output impedance measured with I1 test current source connected to the output?
can’t be bothered to check anyone else’s math Mike?
[a couple of quick discriptive text edits in the math comments]
Nixie wrote on the efficacy of THD measurement:
Yes it is. I'm not prepared to elaborate freely on this topic. It's the gold-dust of audio design. Bob gave a good answer. The aim is to measure how closely the output voltage of the amp equals a scalar multiple of its input and nothing else, across all "in range" inputs and all practical loads. The standard THD test is not comprehensive. And that prevents it correlating perfectly with preceived sound quality.
traderbam said:
Actually, leaving aside certain schools of subjective evaluation I suspect biased to some forms of coloration, I believe there are orthogonal specifications I feel comfortable with.
THD 20 KHz
As Bob has remarked frequently, THD20 is a stress test (better yet applied on a representative variety of real world speaker loads) almost certain to catch weak points except phase deviations. This has been shown conclusively by Cabot.
Yet, THD20 is by no means an easy test to perform except with costly lab equipment. Somewhat more affordable is IMD 19/20KHz as Bob also suggests.
THD 1KHz
On the other hand, THD @ 1K, by far a popular benchmark, is fairly easy to perform with good quality sound boards and appropriate software. Is this enough? I dare suggest it depends. If one knows with some certainty the general behavior of distortion reduction available, - 6dB/Oct an optimistic starting point with a lower frequency dominant pole - one can glean at least a ballpark figure for 20KHz. Starting with a very good 1KHz performance and knowing in advance slew rate capabilities are well beyond program requirements, I tend to feel comfortable.
Phase deviations of course can and should be kept close to null within the operating range.
Rodolfo
Nixie said:
Sorry, I should have been more specific.
Linear phase systems, i.e. those who exhibit constant delay are no problem. This is the case for example of digital processors. For normal analog circuits there is no linear phase deviation possible (constant delay / group delay) and the only alternative is zero phase deviation.
Zero deviation is important because the inner ear is highly nonlinear, making actual waveshape preservation important for accurate rendition of the original sound. This is mostly of concern with transient phenomena and irrelevant for example with steady tones.
Rodolfo
ingrast said:
I tend to agree with a lot of what you have said, but there are, I think, too many "IFs" in extrapolating THD-1 results to matters of high frequency nonlinearity that are best picked up by THD-20 or by CCIF IM with full spectral analysis.
THD-1 is popular by manufacturers because it is easy to get very low numbers, even in a relatively crappy amplifier.
With regard to the cost of analysis equipment, I think things have gotten a lot better with PC-based systems involving spectral analysis. Using affordable 24-bit, 96 kHz sound cards, one should be able to get spectral analysis out to well beyond 20 kHz with dynamic range greater than 100 dB. Has anybody here tried to use the Visual Analyzer shareware (www.hacca.altervista.com) for audio spectral analysis with a soundcard yet? I'm about to give it a try, and would like to hear about your experiences.
Along that same line does anybody have a decent-priced 24-bit, 192 kHz sound card to recommend?
Bob
ingrast said:
This is a point of big controversy. I am inclined to believe you, as I am a proponent of transient-accurate loudspeaker systems (minimum phase). 99% of loudspeakers are not even minimum phase, much less linear phase, so if you are concerned very much about the phase response of an amplifier, you should REALLY be concerned about the phase response of your loudspeakers. Taken a look at a step response of a typical multi-way loudspeaker lately?
Bob
Bob Cordell said:
Sure agree, but harmonics get restricted to second and third at best. This is why I find your suggestion of 19/20 KHz IMD very attractive, for products fall well within the acquisition bandwidth (and are fairly representative of linearity or lack thereoff).
Though I have read bad reviews afterwards, I am quite satisfied with the Creativ Labs Audigy II USB which I used in all my measurements. Being external it is easy to hook up and imprevious to the noisy computer innards. When I bought it here in Uruguay it was about U$ 200 or something, don't know now.
Rodolfo
Bob Cordell said:
Agree too, and crossover networks - unless implemented in DSP - add to this and frequently enough smack in the most sensitive part of the spectrum from the hearing standpoint, the 300 - 3000 Hz range.
My point is to not add insult to aggravation, try to make at least the amplifier as good as possible, besides, it is not really difficult nowadays.
Rodolfo
Digital crossovers multiply the DACs and volume controls needed. Not very cost effective if you've spent a lot on the source, which often makes at least as much difference as the amplifier. What about transient-perfect crossovers such as this one? Insertion loss is high, but this approach would be fine as a line-level XO -- still need multiple amplifiers, but at least not multiple DACs.ingrast said:
On the side, I'd like to ask: SPICE has an option to display group delay instead of phase; which view is 'better'?
Nixie said:
Interesting, will have a look.
If you are concerned for example with stability, phase is more relevant. If you are concerned with phase linearity, group delay is more informative.
Rodolfo
PS And Bob certainly is better qualified than I on this issues, he is working on this, let's listen.