Question might be, whether or not standard figure-of-merit steady-state FFT measurements fully describe accuracy?
They seem to come much closer to accurate audio frequency linear amplifiers than maybe they do for sigma-delta dacs. After all, we are not dealing only with a slightly curved and mostly unchanging voltage transfer function. With dacs there can be aliasing, post-transient state-variable settling, noise modulation (such as can measured using digital stepped DC offset signals), clock phase noise effects, etc.
Also, I would question whether measuring crosstalk at 20kHz is sufficient for assessing stereo imaging even for simple linear amplifiers. Dual monoblocks typically don't tend to exhibit as much imaging instability during musical power transients as do stereo power amps (which is more of a dynamic effect not a steady state one).
They seem to come much closer to accurate audio frequency linear amplifiers than maybe they do for sigma-delta dacs. After all, we are not dealing only with a slightly curved and mostly unchanging voltage transfer function. With dacs there can be aliasing, post-transient state-variable settling, noise modulation (such as can measured using digital stepped DC offset signals), clock phase noise effects, etc.
Also, I would question whether measuring crosstalk at 20kHz is sufficient for assessing stereo imaging even for simple linear amplifiers. Dual monoblocks typically don't tend to exhibit as much imaging instability during musical power transients as do stereo power amps (which is more of a dynamic effect not a steady state one).
Last edited:
And i'm sorry Ken, by your statements, I must not enjoy music on an emotional/spiritual level, its just some sort of intellectual diversion?? Sorry mate, I know its very popular to keep this deeply illogical belief. Its very a very strange belief and I wonder what even points that way? That is absurd. I lost half my teens and all my 20's bobbing around concerts and dance parties and most of my long term friends/family relationships and work life were shaped by it and the pursuit of better sound, some would say i'm obsessed by music and music production. I would say it motivates me more than just about anything. A love matched only by my love of good food, art, language and science. I wonder where this illogical need to pigeon-hole people who are driven in my particular way as being somehow bereft of spirit, comes from?.
I'm not offended though @Ken Newton . I know you didnt mean to offend. I guess i'm just disappointed; as its a terrible cliche.
Given our current understanding of how measures actually corelate to what we here, there are many subjectively choosing objective measurements over listening.
Most of thos emeasures are usful for the designer when developing and maintaining quality of their products. But few actaully correlate to what we hear.
So is trusting only mostly meaningless measures (as far as what it actually sound like), is a largely subjective choice. And the ones that hold meaning have to be jugded in the context of the entire system, each bit with its own measurements, how do those interact?
We are in way too complex a situation to clearly assign causality in many cases.
To quote Floyd Toole:
dave
:^)Looks like human interpretation of measurements can be highly subjective
Yes. What do the measures mean? I quickly learned, during the Japanese spec wars of the late 70s’ 80s that a really low THD was all to often caused by excess feedback instead of careful design and they all sounded kinda lifeless. While some stil hang on that measurement i realized early that it was much more complicated, and that required looking at the spectrum, the subjective interpretation of the much larger set of degrres of freedom in this precursor to the THD number makes that much harder to interpret.
If you are going to “listen” to the numbers, there are so far only a few that give solid clues as to sonics, and given the large mix of these things can get quite chaotic quite quickly. One can use measures as clues, but in the end one has to judge the gestalt of th sound coming out of the SYSTEM.
dave
The real question might be, who even said that it does? Not quite sure I take your meaning though. There are many measures we should look at. Perhaps IMD vs level/frequency and FFT are the most complete, yeah I guess. They represent the recording quite well; but do they represent the musical event? Not all that relevant for a lot of music today though I suppose; as much was created in the studio.
Myself, when i'm listening to music; I tend to lose myself in the music; let it take me for a ride. Sometimes I point my browser at my music searching/history algorithm and just let it take me on a journey of music i've never heard before and lose the entire afternoon.. I dont and dont want to have to convince myself that its correct in order for that to happen and suspend disbelief; for the system to disappear and leave just the music. I dont confuse my ears for a measurement and logging system. I know what is more reliable if that is my objective.
Yes, I use active listening too, with familiar music, but it is a separate endeavour and not casual at all. Listening tests are hard work and some of the music I tend to use for the purpose; I know well, but dont enjoy all that much
Last edited:
I really don't get this black-white approach to measurements vs. listening. I do both and actually most my builds have been tube based. However when assessing SOTA dac chips which are supposed to have ultra-high performance skipping on measurements seems rather strange. Making measurements does not break anything or make listening tests invalid. They could be valuable tools for analyzing the listening test results.
This is yet another example of something visible in measurement. But not necessarily audible in listening test.
Yes, that aspect is very confusing. Yes, perhaps a 'bad' measurement that does not correlate with a magical listening session could be disconcerting, but whether that listening session contained some magical element in actuality isnt important; the listening session happened and you had your experience of it. That experience is not rendered moot because the magic was a trick played by your mind. The entire experience is ALWAYS rendered by your mind and is ALWAYS a trick.
It certainly won't show up in the standard performance measurements, like SINAD and frequency response, and it shows that there is something in the D90 that could be improved.
Why would bigger noise skirts not show up in standard performance measurements?
The thing that happens when you introduce a better measuring unit into the chain and think that the sound got worse is that what you hear now is the worst unit in the chain. If it is your choice to switch back the unit in order to get back to what you liked, be sure to not get any further with that chain as long as you keep that returning unit in the chain (phonogram -> ear wax)... ->
//
//
You need a smaller resolution bandwidth/bin size to see these skirts than for SINAD measurements. You also need to look at the spectrum, instead of letting the audio analyser calculate the SINAD and just reporting the number. Besides, any measurement with a notch filter at the fundamental will also suppress close-in sidebands. Joseph K found them because he specifically looked for the noise sidebands.
- Status
- Not open for further replies.