What were the numbers with EC, for comparison?powerbecker said:Nix "error-correction":
To throw some numbers at that moment from my latest simulation with (a lot of) common NFB:
50W 8E max, 20kHz 36W/8E : 0.000022%, bandwidth DC-500kHz, gain 24dB, PSR 100 (!) kHz - 135dB.
I use Andy´s fets.
Who bid better
Regards
Heinz!
First I have to write a "paper"....
Nixie said:
That is a good question, which I have also asked myself. I suppose control theory is mostly bothered with stability and usually do not consider such aspects as distorsion spectra. I wouldn't expect the various 2DoF cases in Horowitz to be equivalent also when looking at distorsion spectra. Actually, it is probably not even something we can analyze within the models, since Horowitz, just like small signal analysis, assumes linear models. The equivalence between Horowitz' cases most likely breaks down if we switch over to non-linear models to allow for a better modelling of distorsion generation.
I wasn't even asking about the differences between the 2DoF forms, but in regards to the 'fuzziness' you claimed between them and 1DoF.Christer said:
I think the specific distortion profile is very important, and indeed more important than non-perceptually-weighted summary measures like THD or IMD (which have been shown to be uncorrelated with perception of distortion; see, for example, "Auditory Perception of Nonlinear Distortion" in the October 2003 AES convention).
Topology-specific distortion mechanisms such as phase intermodulation have been discussed elsewhere in the forum and some suggested to be perceptually important. I wonder how they would also potentially differ in an EC vs classical NFB control system.
powerbecker, what are you smoking? The ear's masking effects make it sensitive to some distortion orders of magnitude more than others. By perception testing I'm talking about blind testings subjects' ability to detect any distortion when compared to a reference. I'm not talking about which people thought sounded better -- it's a completely different thing. Perception testing with blind testing procedures is the most important measure for the same reason that drugs are best evaluated in clinical trials. You should have read the reference before exercising your itchy Submit-button-clicking finger.
Nixie,
Yes, there are most certainly many differences between different topologies that are considered equivalent on the abstract level and also between different implementations of the same topology. I never meant to imply, and I hope I did make that clear, that it does not matter what topology we choose and how we implement it. It most certainly matters, but it is still useful to understand equivalences and differences on the more abstract level too. The "fuzziness" I mentioned is mostly about a fundamental difference between the control theory situation and the audio design situation on the abstract level, namely whether the plant is given to us or is under our own control to design.
Yes, there are most certainly many differences between different topologies that are considered equivalent on the abstract level and also between different implementations of the same topology. I never meant to imply, and I hope I did make that clear, that it does not matter what topology we choose and how we implement it. It most certainly matters, but it is still useful to understand equivalences and differences on the more abstract level too. The "fuzziness" I mentioned is mostly about a fundamental difference between the control theory situation and the audio design situation on the abstract level, namely whether the plant is given to us or is under our own control to design.
Nixie said:
It is not uncommon that control engineers are called in to improve an existing plant. However, I also think that it usually makes sense to make a distinction even when designing a new plant. The plant is most often not an electronic system (although in many cases it is), but rather something like a chemical plant. It probably wouldn't make sense to make extensive modifications to this chemical plant when it is sufficient to add an op amp and an RC filter to its input. Of course it is a matter of taste if we consider this op amp and RC filter as part of the plant or the controller, but in the case of a chemical plant for instance, it seems to make more sense to consider it as part of the controller.
But your question does pinpoint something important, that many abstractions and classifications are not as crystal clear as they might seem, and one easily ends up in philosophical discussions if trying to scrutinize them. Such abstractions and classifications are still often useful, if one just keeps in mind that this problem may exist.
The point is that said hearing threshold varies depending what it is you're checking to be below the hearing threshold. For some types of distortion, the threshold can be a few percent. For others, it can be a fraction of a part per million. Can you build an amplifier that has all possible distortion mechanisms brought below that threshold, including phase intermodulation (discussed elsewhere in the forums), thermal/memory distortion (discussed, say, here), and so on, while driving a real speaker?
And then, you have to consider the system as a whole, and the possibility that the amplifier can, through predistortion or masking, ameliorate or the effects of distortions of other equipment in the chain to result in an overall less perceptible system distortion. I don't think there's yet a transport+DAC that can manage jitter performance below the perceptible threshold which is a few picoseconds, or a speaker withthout a host of problems such as Doppler distortion (see here).
Tell you what, you build a system that, when I listen to it blindfolded, I will not be able to tell you if I'm listening to a reproduction or the live, non-amplified orchestra, and then I'll believe you can build a sufficiently distortionless piece of equipment. Otherwise, the best remains in compromising by minimizing the distortions that are most perceptible.
And then, you have to consider the system as a whole, and the possibility that the amplifier can, through predistortion or masking, ameliorate or the effects of distortions of other equipment in the chain to result in an overall less perceptible system distortion. I don't think there's yet a transport+DAC that can manage jitter performance below the perceptible threshold which is a few picoseconds, or a speaker withthout a host of problems such as Doppler distortion (see here).
Tell you what, you build a system that, when I listen to it blindfolded, I will not be able to tell you if I'm listening to a reproduction or the live, non-amplified orchestra, and then I'll believe you can build a sufficiently distortionless piece of equipment. Otherwise, the best remains in compromising by minimizing the distortions that are most perceptible.
traderbam said:...and listen to it.
Brian,
This is something often repeated that never fails to mistify me.
Probably the electronic part is the one most approaching perfection as perceptive thresholds go. And this is no matter the design approach selected as long as it is not a fashionable fad or a sloppy job.
Amplifiers should not "sound" this or that way, they should simply do their job to a high standard so we can forget their presence and concentrate in those parts of the chain where real problems are far from being corralled at the same level of performance.
Then, if someone exercising his legitimate right to modify reproduction this or that way wants to depart from accuracy, give tone controls or whatever DSP massaging better suits his/her tastes with a far richer menu of possibilities.
Rodolfo