Sound Quality Vs. Measurements

[dvv]

Originally Posted by Audionutz

I am lucky to own an example of the XX, along with the matching preamplifier. Some small repairs mean that I have not had a chance to listen to it yet, but I look forward to hearing it. I also have the smaller X-1 amplifier and matching X-11 preamplifier, both of which sound very good indeed.

What can I say? Lucky you, and then some.

[DF96]

Originally Posted by magnoman

. . . these next generations are smarter and will face different challenges than we did.

Smarter? No. Different challenges? Yes.

My experience is that the cream of new talent are every bit as smart as the best of the older generation, but they are now less well educated. This is because the average has declined, so courses have had to be simplified so they don't fail too many students.

I remember chatting to someone a few years ago. He was bright, and had a good EE degree from a good university. When he started work, doing real engineering, he and his colleagues were surprised to discover that he knew almost nothing about noise because his course had taught him almost nothing about noise. I was surprised too, because Johnson noise (and its derivation using the transmission line model) is first-year stuff for physicists (or it was 30 years ago - I don't know about nowadays).

Before using simulators for real, all newbies should be forced to simulate an LTP with a perfect CCS tail. Then build one and measure it, and explain the huge differences in CMRR and even-order distortion performance.

[dvv]

Originally Posted by Wavebourn

If you have hands-on experience and well trained imagination no simulation is needed. But in order to obtain this skills you need to work with breadboards and measurement tools. Like pocket calculators deprived people from ability to calculate numbers, simulators deprived people from ability to see, feel, and hear circuit diagrams.

But it's people, Wave, not tools.

Look at a simple hammer. You can waste time with it by cracking walnuts, you can build a house with it, or you can kill a man with it. All with the same hammer. So it's not the hammer, but the man using it.

All the tools we have now are all wrong if taken as absolute. They are not an end unto themselves, they are tools, and as any tool, they are only as good as the man using them.

In other words, if the PC is understood as a tool which is here to help us by saving us from menial work, that's just fine beacuse the thought process is still our own, and has not been relegated to a machine "because it's faster".

It's a simple fact that my simulator will allow me to make a fault free electrical circuit, but that says nothing about how that circuit will sound. I have had it sound absolutely awful, so it was back to the drawing board. On a few occasions, using the same tools, I have hed it produce great sounding circuits, but I never lost sight of the simple (at least simple to me) fact that it only does what I tell it to do far better than I could protoboard the sam circuit.

But I must clean here. The best aspect of it is that allows me to create SPICE models of my own choosing. I test drove that point once, and indeed got different results that from the simulator model database. After that, I downloaded an unholy collection of transistors I tend to use plus some new ones and created my own database, as per the manufacturers' data sheets.

Ever since then, over the last 7 years, my suimulator hos not been wrong once, not even once. It it says the circuit will work, it has worked each and every time. If anything, it's a bit conservative, and I just love that - if it says the circuit will hit say 300 kHz, you can bet your bottom dollar that as a model on the testbench it will hit at least 330 kHz, and more often 350 kHz.

I much prefer that to overoptimistic results, always have. If nothing else, it keeps building of models and test bench measuring them still a lot of fun, there you are investigating by how much it shortchanged you this time.

Hoqwever, in the end, it's all up to how I hear it under real world conditions, and I have never in my life had a circuit made which got it just right the first time, all to the one had to be reworked, some more, some less. The ultimate judge are my ears, such as they are.

[nigel pearson]

The typical LTP of old was 1 mA tail current with no extra emitter degeneration . I have played with various currents and found 1 mA not such a bad choice . The choice then to run more current or make the VAS a Darlington or whatever needs to be considered . I suspect Douglas Self struggled with this and came to no absolute conclusion . Anyone brave enough to suggest an optimum LTP current ? I will start the ball rolling and say 2 mA . Is 10 mA excessive ?

I know what I have just said is a bit daft . However in motorcar engineering it was "suggested " 330 cc to be an optimum capacity per cylinder if petrol . I believe this was said in the 1920's and might have been Citroen . There is evidence to say it is a good rule of thumb today . I forget how square the engine is at 330 cc ( bore and stroke ) . The parameters being for example , piston velocity , economy of fuel and parts , power and torque , vibration . For a simple hybrid car 990cc would be practical and fast .

John suggested 50 V / uS @ 100 W as desirable . We arrived at 22 000 uF x 2 as the PSU for the same . Optimum tail current and perhaps bandwidth would be useful . I suspect 100 W @ 8R is also a good universal power level . I will suggest 5Hz - 3 dB and 100 kHz - 3db as a good minimum bandwidth . To ice the cake I see distortion below 0.1% as still the hi fi standard with as littler high order distortion as possible . 0.1% THD @ 100 W . Preferably the amplifier should behave as a class A design ( op amps more or less do these days ) .

[DF96]

Originally Posted by nigel pearson

Anyone brave enough to suggest an optimum LTP current ?

Almost meaningless question, as you may be aware.

"How long a piece of string do I need?" is a similar question; the correct answer is "It depends; what do you want to use the string for: measuring a distance, lifting a weight, communicating via two tin cans?".

[jan.didden]

Originally Posted by dvv

The key sentence for me here is: "Otala got the ball rolling". Exactly.[snipped to preserve bandwidth]In my general view, as it somehow always turns out, it's more of a question of overall balance. Global NFB is GOOD if it's relatively moderate, say no more than 26 dB (a loose pointer, no more), and if your OLFPB is at least say 30 kHz, not to be satisfied with the barest of minimums. Finding the right sonic balance between local and global NFB, which will obviously vary with circuit topology and other aspects, so there is no simple one-fit-all formula.

I agree with your post here - well said. I'm just wondering why you feel that nfb should be limited to 26dB (or whatever number).
If you can maintain stability, what's against the highest nfb factor you can get away with? Is there some point where nfb turns 'bad'?

jan

[nigel pearson]

Originally Posted by DF96

Almost meaningless question, as you may be aware.

"How long a piece of string do I need?" is a similar question; the correct answer is "It depends; what do you want to use the string for: measuring a distance, lifting a weight, communicating via two tin cans?".

I know . I think 2 mA is as far as this will go . I don't think many will go wrong with that . I did see somewhere a very lengthy analysis of the Douglas Self's designs . It argued that noise could be improved . It did come to an optimum current . If I can find it I will post it . It was based on assumptions of typical pre amp output impedance and typical uses . I was very impressed by Mr Self saying he was a little nervous when bootstrapping the input to have lower noise ( 2 x 1K input from memory ) . I assumed all was simulation in his books ! I do think noise is important as does he . I was told that lower noise transistor amplifiers sound warmer . I have no way of knowing if it is true . It seems no bad thing to believe . Noise gating is not an alternate .

[jcx]

actually the 10-20 dB is the range where feedback is "worst" - where "harmonic multiplication" gives the highest peak in the feedback generated new, higher harmonics when looping a low order open loop distortion

much higher feedback is needed to have the new harmonics all fall with increasing order as we are told is necessary

the "harmonic multiplication" is fundamental feedback math - applies to degeneration as well as "local loop" or "global" feedback for those who think there is a distinction - it is not in this property

[nigel pearson]

Originally Posted by jcx

actually the 10-20 dB is the range where feedback is "worst" - where "harmonic multiplication" gives the highest peak in the feedback generated new, higher harmonics when looping a low order open loop distortion

much higher feedback is needed to have the new harmonics all fall with increasing order as we are told is necessary

the "harmonic multiplication" is fundamental feedback math - applies to degeneration as well as "local loop" or "global" feedback for those who think there is a distinction - it is not in this property

Taking an op amp as a mini amplifier I find some curious things . A cheap and nice chip is MC33078/9 . I use it with a gain of 62 and also a gain of 1 . Both sound exceedingly nice like this . A gain of 3 not so much my cup of tea ( two different uses ) . This is not evidence , it just reminds me of your statement . Lets be clear the gain of 1 sounds truly excellent . Gain of 62 to 100 superb considering how it looks on paper ( <250 uV in and 60 dB noise if lucky ) . Gain of 30 was not special either . I am very happy to be shot down in flames for this . If not you then I should do it .

Although troublesome should we expect power amplifiers to be usable as big op amps with a gain of 1 ? The output stage is that . I said before if the output stage is preceded by a big op amp it could have a high degree of feedback and the output stage enough global to the op amp like stage to maintain low output impedance .

[jcx]

just to be clear it is 10-20 dB of feedback factor, not the closed loop gain that we are talking about with the "harmonic multiplication" math



I have seen "unity gain" op amp datasheet open loop curves with a bump in the phase that gives less phase margin at low closed loop gain of 3-5 than at unity gain or much higher closed loop gain