Yes, I am quoting RMS of the error divided by RMS of the original. I am counting all sources of error other than a constant time delay. I wanted to see how bad MP3s really are.
Ed
That's the case with global feedback, too, and thus it can't work - we've come full circle 😛 !
(SCNR)
True. But unless its pointed out explicitly as you just did, some readers of threads like this may take it for granted that distortion is distortion, its all like THD. Why? Because in forums like this the word "distortion" tends to be used loosely. Usually it means HD/THD (i.e. distortion produced by a curved transfer function), but not always.
Apparently not as bad, there are plenty of articles that can be found on AES or IEEE , where even a group of trained professionals can't hear the difference between anything above 256kbps vs uncompressed audio in a VERY controlled environment with blind AB and ABX tests.
So yaay, that you can put numbers to something, but it doesn't mean much without any context.
It doesn't even say where most of that distortion is for example.
I am subtracting the MP3 output from the original and computing RMS of the error divided by RMS of the original. I don't think the algorithm can be more clearly described.
I am not bucketizing the error into frequency response, noise, harmonic distortion, and phase shift (each of which is perceived differently).
Ed
I am not bucketizing the error into frequency response, noise, harmonic distortion, and phase shift (each of which is perceived differently).
Ed
Is the uncompressed audio equivalent to CD, SACD, 24/192 PCM, or something else? And what are the listening conditions? Nearfield recording monitors, headphones, middle of a listening room with speakers behind a curtain, or something else?
Reason I ask is those are all different experimental conditions, and results may be different under different conditions.
Also its not clear if a group of professionals could not hear a difference on average, or if none of them could. IOW, how results are reported can matter too.
The point I am trying to make here is that another problem in forums like this is the way research is reported/summarized. Some readers may tend to interpret the quote above in terms of absolutes, such as "no person on Earth can hear a difference between CD and 320kpbs MP3 under any conditions whatsoever." There have been many arguments in forums over the years that hinge on misinterpretations of published research.
Last edited:
Still though, negative feedback still works in spite of that. Frequency compensation components in the feedback circuit can correct for the delay if it's not excessive.
I think you don't really understand the bigger picture of such findings if such a group of people can't hear any differences in a very controlled environment with blind tests?
Because with deductive reasoning we can conclude that in actual practical environments this would only become orders of magnitude more difficult.
Most of these tests are done with headphones btw.
Since they amplify these kind of issues and problems, like distortion problems, much better than loudspeakers in a room.
And yes, statistically speaking, there are by definition always people who might be able to hear something.
I wish them all the best in life, because if your ears are that sensitive you have a lot more serious (mental) problems to worry about.
In fact, I doubt that these people even enjoy listening to many things at all.
There are a couple of measures I could think of to make a negative-feedback amplifier work well in a subtractive test without phase compensation, but most of them would be undesirable for normal use.
1. Make sure you have lots of loop gain over the audio band. This is actually useful to lower distortion. You probably need some high-order compensation scheme with zeros.
2. Don't put zeros (lead compensation) in the feedback network (phantom zeros), but only in the forward path. This constraint leads to a bump in the far ultrasonic region.
3.Don't use a series inductor or filter at the output. Not using one will make the amplifier more susceptible to weird load impedances and RF picked up by the loudspeaker cable.
4. DC couple everything without any DC servo. If you must block DC, use a DC servo with a very low bandwidth of which the loop gain drops of with some high order, or find some other way to realize the kind of transfer that gives (1 - H(s) shaped transfer where H(s) is a unity-gain low-pass transfer).
DC coupling may be dangerous for your loudspeakers if the preamplifier has much too much offset, the other options will give a bump in the subsonic region.
5. Don't use an input low-pass filter, or one with a 1 - H(s) shaped transfer where H(s) is some unity-gain high-order high-pass transfer at a far ultrasonic frequency. Using no filter makes the amplifier more susceptible to RF at the input, the other choice causes some far ultrasonic bump.
These are all tricks to keep the vectorial difference between input and output signal as small as possible, and most make no sense at all if you just want a good audio amplifier.
Last edited:
With deductive reasoning we can conclude many things that are not scientifically true. Its the difference between trying to use philosophy and debate to find truth, as opposed to using the scientific method.
That's not an argument, nor does it go into the subject.
Btw, deductive reasoning is used in all kinds of fields of science.
Btw, deductive reasoning is used in all kinds of fields of science.
I have made recordings and made high bit-rate MP3s from the recordings (320kbps). The recordings and the MP3s sound a little different in places. Interestingly the MP3s improved some little problems such as finger squeaks on guitar strings or tongue clicks from a vocalist. In other ways the MP3s sounded a little worse. Overall they sounded about equally good. At the time I was listening nearfield using NS-10 speakers, maybe a Bryston 4-B amp, and a Lynx sound card.
So, when someone comes along and said what I did is impossible, to my way of looking at it the person may be misinterpreting research, or maybe there was bad research done (more replication crisis stuff). I would have to see the exact research papers to see if I have any more to say about it.
So, when someone comes along and said what I did is impossible, to my way of looking at it the person may be misinterpreting research, or maybe there was bad research done (more replication crisis stuff). I would have to see the exact research papers to see if I have any more to say about it.
I think when you talk to B-force, you are talking to a wall, all components and software have an influence in the sound. Theory is good, practice is better. Do we have DNA measuring equipment to unravel that.
regards wollie
regards wollie
Without deductive reasoning we would have no science to speak off. You really are out of your depth here.
Jan
Deductive reasoning can be used or misused in science. That should be obvious. One can deduce that feedback goes around and around. One can deduce that the earth is flat.
When deductive reasoning is used in science it is usually to aid hypothesis formulation. Hypotheses then need to be tested and the results analyzed.
To claim nobody on earth can hear a difference between CD audio and 320kbps MP3 is non-scientific nonsense.
EDIT: To further clarify, deduction used alone to directly reach final conclusions is only valid is mathematics, such as in formal logic, and or mathematical proofs. That's not the same as the role of deduction in science.
When deductive reasoning is used in science it is usually to aid hypothesis formulation. Hypotheses then need to be tested and the results analyzed.
To claim nobody on earth can hear a difference between CD audio and 320kbps MP3 is non-scientific nonsense.
EDIT: To further clarify, deduction used alone to directly reach final conclusions is only valid is mathematics, such as in formal logic, and or mathematical proofs. That's not the same as the role of deduction in science.
Last edited:
The propagation delay is in nanoseconds i.e. relativistic velocities and of no consequence in a null test. Adjusting for phase shifts is the difficult bit.
Yes ,adjusting amplifier for minimal phase shift is the difficult bit , hard example is on many conventional tube amps with output power transformer , where those output transformers introduce major phase shift which also limits max. level of inserted global negative feedback , only exceptions are OTL tube amps (output transformer less ) , but they are very rare beast today.
Last edited: