Two were just mentioned: SMPS power causing correlated noise which doesn't show up well on standard FFT measurements. Another is asymmetrical crosstalk which can happen for reasons like putting the input to one channel too close to the output of the other channel, but not vice versa.
Phono is different because at least some spatial effects are believed to come from tracking issues when using cartridges and tone arms that don't track the groove exactly the same way the cutting lathe did.
DACs are different yet again because they can be subject to correlated noise effects from random timing errors (jitter) which does not show up well on using typical J-Test measurements. They are also subject to random noise effects from Vref noise, which again may not show up well in standard spectral measurements. The mechanisms involved can be complex and take some time and study to understand but they are quite real.
IOW, the reason people can't think of possible equipment problems which would affect sound stage is because they are things not easily seen in standard measurements, and almost nobody talks about those types of problems.
Last edited:
I was thinking about AMP circuit performance. If you think about an amp like a "black box" with everyhing inside poorly integrated, I understand.
These items are very easy to fix by using linear power supply and carefully route the cables.
There can also be problems in amplifiers with ground current paths causing asymmetrical shared-impedance coupling between channels. IOW, its doesn't always have to be something visually obvious.
As mentioned before, I have an amplifier that measures superbly in terms of nonlinear distortion which was designed by pretty well respected designer of such things. Yet it ended up having a (presumably asymmetrical crosstalk caused) lopsided soundstage. Standard measurements with the AP555x didn't find that problem.
As mentioned before, I have an amplifier that measures superbly in terms of nonlinear distortion which was designed by pretty well respected designer of such things. Yet it ended up having a (presumably asymmetrical crosstalk caused) lopsided soundstage. Standard measurements with the AP555x didn't find that problem.
Its comes down to the 'noise' level of your loudspeakers driven by said amplifier.
There is THD, which is important, then, IMD, then RISE/Fall time, then output Z, then how the amp deals with feedback, is is perfectly 180o at all frequencies at all conditions, then there is cross talk, then power supply , then many other internal components interaction etc,
But one very important is the noise which is under the THD and on the sides of each signal.
There is THD, which is important, then, IMD, then RISE/Fall time, then output Z, then how the amp deals with feedback, is is perfectly 180o at all frequencies at all conditions, then there is cross talk, then power supply , then many other internal components interaction etc,
But one very important is the noise which is under the THD and on the sides of each signal.
Not for correlated noise. Its very important to understand what "correlated noise" means. Its only there when the audio signal is there. It can stay lurking just below the surface of whatever you hear. It doesn't necessarily sound like noise, yet you may perceive it as a gloss, veil, or some other vague character to the sound.
It may not show up in typical spectral measurements, so its easy to miss. If you know what to look for then you might see some evidence of it, but spectral measurements can be deceiving. What looks like a very low level effect in terms of average frequency magnitude may not be nearly so low when phases of noise components line up in time. Since phase information is discarded in typical spectral analysis, you probably won't have much a clue as to how audible it might be in the time domain.
Here you may be referring to correlated noise that has become intermodulated with the audio signal, which then produces additional noise sidebands. Using THD as a metric will tell nothing about the degree of objectionalness of correlated noise as compared to a fixed level noise floor. The latter tends to be less audible because its easier for the brain to filter out as an uncorrelated noise signal.
Last edited:
Very interesting - I'm curious to know:
1. when you say "lopsided soundstage" ... do you mean the soundstage height, each side?
2. what in the circuit would cause asymmetrical crosstalk?
Thanks.
1. Lopsided from left to right.
2. As mentioned before, there has to be some kind of crosstalk which always involves some coupling between two different channels. There is nothing that says its impossible for something like, the left channel may pick up some of the right channel, but the right channel may pick up a different amount the left channel. It would depend on the symmetry of the particular electromagnetic coupling mechanism in each case.
Also, there is no way to make two channels exact mirror images of each other so they are perfectly symmetrical with respect to each other. If for no other reason, components with more than two leads can't always be flipped around in a perfect mirror image way (unless maybe one channel circuit is on top, say, maybe on the top-right side of the board, and the other channel is a perfect rotation of the first channel which then appears on the bottom-left side of the board). That's just one example that isn't too hard to visualize to illustrate the difficultly of providing perfect EM isolation and or symmetric coupling between channels.
2. As mentioned before, there has to be some kind of crosstalk which always involves some coupling between two different channels. There is nothing that says its impossible for something like, the left channel may pick up some of the right channel, but the right channel may pick up a different amount the left channel. It would depend on the symmetry of the particular electromagnetic coupling mechanism in each case.
Also, there is no way to make two channels exact mirror images of each other so they are perfectly symmetrical with respect to each other. If for no other reason, components with more than two leads can't always be flipped around in a perfect mirror image way (unless maybe one channel circuit is on top, say, maybe on the top-right side of the board, and the other channel is a perfect rotation of the first channel which then appears on the bottom-left side of the board). That's just one example that isn't too hard to visualize to illustrate the difficultly of providing perfect EM isolation and or symmetric coupling between channels.
Last edited:
Ill post in other thread 2dacs/2amps/through ADC with 2 samples, for fun.
Link: https://www.diyaudio.com/community/...roskie-unbalancer-audible-differences.425505/
Link: https://www.diyaudio.com/community/...roskie-unbalancer-audible-differences.425505/
Last edited:
Ok, but if this coupling exists, and some amount, even minimum, really does, why it cannot be measured?
If we drive one channel and left the other with no signal, we can observe how much output signal we have in the channel without input signal.
In fact this is the basic crosstalk test.
I'm just trying to understand the fenomenum that affect the so called "image problem" and finding a way of measuring it.
I'm interested in this point. It's been explained to me that one source of IMD is too much/improperly applied negative feedback. Which is to say that low THD isn't an indicator of good sound if your effort to achieve it caused any of the many other kinds of distortion.
The OP's two questions have been rolling around in my head all afternoon (much to the detriment of my home-stretch-o-the-work-week productivity).
I have answers, but they're not ultra-experienced DIYer or engineer answers...they're music lover answers. So if anyone has a technical critique about these thoughts, I'm interested.
Forget about amps for a second. Instead, think about striking a piano string. Let's say it's B6 - closest thing on a piano to 1,000 Hz. Strike it hard enough to hear the overtones... well, there's your "harmonic distortion." I don't have the data, but I feel ok arguing that by amplitude, the "THD" percentage of the sound produced by that piano string is MUCH higher than any (properly working) amplifier you've ever heard. The relative amplitudes and phases of those overtones are what determines whether the note is bright, cool, warm, fat, hollow, and/or the umpteen other qualities that add up to what we call timbre – which is obviously affected by the technique/artistic intent of the pianist.
So... Class D Guy just read that and maybe he's thinking "Exactly! Which is why I want an amp with 0.0001% THD, because anything else gets in the way of the pianist's artistic intent." Meanwhile, Single Ended Triode Guy is thinking "Exactly! That's why I don't care that my amp makes 2% THD, long as it's low order, because the tonal boost makes the pianist's performance sound more 'present'."
I think they're both right.
In my admittedly limited understanding of music theory, the conventional idea is that certain frequency ratios just "sound good" to the human ear. And of course, what's "conventional" to Western tradition isn't necessarily "conventional" to Eastern tradition.
Are the harmonic overtones generated by a warm-n-fuzzy major third chord "lower order" than those of a dissonant minor second?
Are the differences between amp topologies akin to the differences between the chord structures used in Country Rock vs Free Jazz?
Correct me, but I think what Markw4 is saying is that we can hear things our tools can't measure everything we can hear. I'd add that when it comes to harmonics, our tools are measuring through the lens of physics, and our brains are measuring through the lens of timbral perception - and how it makes us feel.
TL,DR:
1. THD is a useful predictor of the quality of sound, but not "sound quality."
2. "Good" sound quality can ONLY be measured subjectively.
The first things to come to mind is room treatment.
A flat surface on the sides and behind and unequal placement of speakers breaks the imaging through sound reflections.
high frequency is the image, under 100hz nothing much brains loses capacity to find from where the sound comes.
At higher frequency the polar dispersion of the woofer and tweeter is what comes in mind.
Probably the noise floor (dynamic) is also a great culprit, so regulated power supply helps a lot, mono block too. Hifi sources.
Hifi loudspeakers are designed with this objective to image well : they match the woofers, tweeters, XO, then match each box to its closest one to make the best matched pairs possible. high frequency sum at the listener position at some 20o angle is also calculated in the XO to make the best out of it.
Hi, Your experiment and your conclusion is correct for a piano. The harder you hit the keyboard, more the harmonic levels increase as well as the harmonic profie changes. Due to he piano hammer position, close to the one of the ends of the string, a set of harmonics will always be generated, so creating what we know as piano sound - far from a sinewave. If there weren't harmonics, just the fundamental (sinewave), all instruments would sound the same with the exception of attack and decay.
But note this effect in the piano is not harmonic distortion, it's just the creation of the sound.
This is explained by Fourier theory, where any sound can be decomposed/composed by a sum of sinewaves (fundamental + harmonics).
Take a Monet artwork. It is not highly defined picture on purpose, so it is not a distorted image, but just an art style.
If you are an old person that need glasses, you see highly defined images, like your wife image, out of focus if not wearing glasses. This is distortion, cause you are adding things (blur) to the image that don't exist and should not exist at all.
The idea of an amplifier is to amplify the signal and not create harmonics. So an amp should amplify a signal and keep the harmonic profile identical.
If it changes the harmonic profile, it changes the original sound. So a piano might not sound like a piano, violin not as a violin etc.
On the other hand, music is entertainement, you can do whatever you want. If you want to use an amp to modify the harmonic profile, no problem.
And just remember that hi-fi tube amp designers didn't want to modify the signal, it just happened due to the limitations of the tecnology.
On the other hand. it's totally different from a guitar tube amp and speaker. In this application, the guitar sound is composed by the guitar strings, pickups, amplifier and speakers. You want the amp and speaker "distortion" to create the guitar sound.
The whole set will give you the sound signature of, for example, Eric Clapton, Steve Vai, Hendrix etc.
But for a hi-fi (high fidelity) system once you reproduce this guitar sound, you don't want the hi-fi amplifier to modify it, other wise Eric Clapton sound will not sound like Eric Clapton. High Fidelity sytem means as close to the original as possible.
Here, as an electrical engineer (yes, a nerd... lol), my instinct is to try to measure everything.
This is what moves me forward. I see a fenomenum and I want to measure it, so I can understand, control, fix, modify, build etc.
I am on the side that thinks that we can measure everyhing - we just need to find how.
First step is to understand the problem, then find the way - that's why I was making questions.
Good discussion! 🙂
Sure, room makes a lot of changes.
But, my question was focussed on the amplifier.
I want to understand why an amplifier changes the image of a system and, if it does, how I can measure it.
Initially, I think amplifier crosstalk is something mininum, but if other people say that it isn't, I'm trying to understand other's people point of view.
Maybe I learn something I'm overlooking and fix that.
To correct that.... we measure through the lens of Electrical Engineering, not Physics.
Physics is a model, like my advisor told me in my verbal last final:
" Tony, if someone came up with a model that Green Men from Mars were behind it all and the theory explained what we see and correctly predicted criteria we'd have to accept it"
He was very much correct. Dr. Brown's words really stuck to me.
My question thus is... what do we measure to quantify/qualify soundstage?
The claim was made than interchannel crosstalk affects this. How about full dual mono designs, where the L and R channels are on separate boards, and in case of amps, using separate power supplies and chassis?
Yet, we still still have soundstage in a case of minimal electronic interaction between the channels..
I think it's likely much more complicated ( or fundamental ) than that as it involves the interaction between amplifier and speaker and room.
Also, "good sound quality" is capable of being measured objectively... once we develop the model that describes our hearing cognition.
Hey Ron, I think you can get good stero image on any amplifier, the most impressive image system I heard was a top $ turntable system in a treated room at planet of sound we played some excellent jazz and classical it was a great afternoon but the whole set up was 50k
When I place a sound damper between my harbets and listen close It image very well,I think it is more recording dependent then room accoustics
Making abass trap / a sound diffuser between loudspeakers on a back wall at least 1 meter away back an some carpets / side walls sound absorbers and it should make a big difference
When I place a sound damper between my harbets and listen close It image very well,I think it is more recording dependent then room accoustics
Making abass trap / a sound diffuser between loudspeakers on a back wall at least 1 meter away back an some carpets / side walls sound absorbers and it should make a big difference
I guess that supports the argument that a pair of amplifiers arranged as monoblocs ... will always sound better than when they are housed in the one box - even in a dual mono configuration (where each amp has its own PS)?
Nobody said that particular thing can't be measured. It just isn't usually a part of standard measurements.
Yeah, I wish we could.
My experience tells me that every time I arrived at the power supply with lower noise and more uniform/wide & very low output impedance, the soundstage improved. But so did other aspects of the sound... so, yeah - not easy to isolate the soundstage as such.
IME it involves everything. Small changes at certain spots in room treatment can have substantial effects. Also, IME the dac can put some limits on soundstage imaging precision. Unfortunately, it seems to involve some cost at every stage of reproduction starting from the dac and ending with the speakers and room. Anything wrong anywhere in the chain can have some deleterious effect. Measurement of some aspects of performance may take a very good, low random noise jitter ADC.
Regarding measurements, there is the attached, and maybe some of the papers over at the Linkwitz site.
Last edited: