Hi Mark,
Do you think I could use an HP 5372A instead of the Symmetricon? I have a 35665A, but my current audio analyser is better. lol! I have all the other RF stuff.
So you're running parts in an application we don't use in audio. What do you think you might learn?
Hi Logon,
Well, to answer your question .. they are creating music. The way the instrument feels in their hands allows them to play a certain way, and they sound different (you could change pickups and change them even more). How many guitarists have but one guitar, or all the same? Noone I have ever heard of.
However, in our home (or wherever), we are reproducing what was created in the recording studio. If you want to create anything, start with a clean system, and stick an effects unit in the chain. At least you can "return to zero" without having a built -in effects unit to fight.
I worked in recording studios. Watched them create each track, each effect and I serviced that equipment. There is a very basic difference between creating a music track and reproducing it. Confuse the two at your peril. You don't want the same effect applied to each instrument, or every song. Some people try for a while (witness single tube class "A" tube amps at low power). You eventually tire of the same sound character.
As I said, you have got to separate the psychology from what is actually happening. If the reproduction is flawed, you've lost already before your mind becomes use to low distortion and natural sound. It is a known fact that people often prefer the sound they are accustomed to (unless it is really bad). After a while, exposure to more accurate sound becomes preferable. You can always drag out special cases where this isn't true. By in large, it is true.
How many times has a system sounded "it's best" on "natural quartets", etc and totally fail with more complex music? That is because they are adding a ton of distortion. how many people turn the music up, yet given a system that reproduces proper bass at low levels, they don't? When there is a problem with a system, people react by doing different things in an attempt to be happy with the sound. Give them a good system, that changes everything. So taking reports from people using flawed systems isn't going to get you anywhere but confused.
Do you think I could use an HP 5372A instead of the Symmetricon? I have a 35665A, but my current audio analyser is better. lol! I have all the other RF stuff.
So you're running parts in an application we don't use in audio. What do you think you might learn?
Hi Logon,
Well, to answer your question .. they are creating music. The way the instrument feels in their hands allows them to play a certain way, and they sound different (you could change pickups and change them even more). How many guitarists have but one guitar, or all the same? Noone I have ever heard of.
However, in our home (or wherever), we are reproducing what was created in the recording studio. If you want to create anything, start with a clean system, and stick an effects unit in the chain. At least you can "return to zero" without having a built -in effects unit to fight.
I worked in recording studios. Watched them create each track, each effect and I serviced that equipment. There is a very basic difference between creating a music track and reproducing it. Confuse the two at your peril. You don't want the same effect applied to each instrument, or every song. Some people try for a while (witness single tube class "A" tube amps at low power). You eventually tire of the same sound character.
And this is exactly how the "high end" audio market gets your dollars. By bringing in confusing factors to cloud the real question. The more accurate a system is, the more life-like it is.
As I said, you have got to separate the psychology from what is actually happening. If the reproduction is flawed, you've lost already before your mind becomes use to low distortion and natural sound. It is a known fact that people often prefer the sound they are accustomed to (unless it is really bad). After a while, exposure to more accurate sound becomes preferable. You can always drag out special cases where this isn't true. By in large, it is true.
How many times has a system sounded "it's best" on "natural quartets", etc and totally fail with more complex music? That is because they are adding a ton of distortion. how many people turn the music up, yet given a system that reproduces proper bass at low levels, they don't? When there is a problem with a system, people react by doing different things in an attempt to be happy with the sound. Give them a good system, that changes everything. So taking reports from people using flawed systems isn't going to get you anywhere but confused.
Are you sure you're not adding an extra variable into the equation, precisely by playing complex music a lot louder than that string quartet? All else being equal, if you double the number of instruments in an ensemble, the average acoustic power goes up by 3dB, and the peaks go up by 6dB. So the dynamic requirements go up significantly, even though no individual member is playing any louder than before.
A corollary to your complaint about tubes, could be that the Achilles heel of high-powered class-AB systems is the milli-watt range. In that case, turning up the volume could also be an attempt to get away from that low-power region. People could go so far as to deliberately select low-sensitivity speakers, so that the amplifier always runs a bit warmer. Cherry-picking 'loud' albums is also likely.
When it comes to actual power limits, I recall doing many tests with oscilloscopes, and frankly, 20-25V peaks (25-39W @ 8ohm nominal) were encroaching on noise complaint levels. And that's with plain average 88dB @ 2-pi space sensitivity, nothing special at all. After a few student parties, back in the day, I opened up a pair of DIY speakers, and found that the speaker cables had gotten hot enough to cut through styrofoam padding! It was a miracle the speakers themselves didn't fail. Or maybe the thin wiring saved them... So when people say they need 100W for their small living room, and that 20dB of unused "headroom" necessitates 100W instead of 1W...
Looks like the instruments you have are rather different from the Symmetricon. Some info on that type of instrument at: https://www.testequipmentconnection.com/specs/Symmetricom_5125A_.PDF
Also, from looking at some of how dac AN and PN are being measured, its just noise of the dac itself (including Vref), but not the combined dac and crystal clock system as we use for audio reproduction. I say that because it looks like they are using a synth and NCOs (Numerically Controlled Oscillators) to clock the dacs in order to make nulling adjustments.
So, it might still take some more work to combine a crystal clock phase noise measurement with a dac phase noise measurement so as to calculate an estimated total phase noise spectrum. Something like that.
Last edited:
Hi Mark,
I would argue that tiny phase noise in a single DAC is not audible. It also affects both channels equally and at the same instant. This will have no impact on imaging even if it did occur in other words. If they use a standard quartz crystal, short term stability is the best of any solution save an ovenized quartz oscillator. Other sources ae not as stable short term, so when I measure phase noise of a clock, I pull the GPS reference to put it in holdover mode.
When playing music, with the amount of long term frequency drift in a normal oscillator (quartz) absolutely doesn't matter. What matters is short term stability. Now we are talking expressly about the execution of a design. So things like trace routing, crosstalk, power supply noise and all that. That is why I have the 5372A.
Anyway, unless the DAC chip or chip set is defective in design, it shouldn't have any more jitter or phase noise than the master clock. Of course physical execution takes over as the primary problem and what we are discussing has no bearing. You also need to make sure your experimental setup and instruments don't have near the same phase noise as what you are trying to measure.
Right now I suspect we are talking about things that do not have any practical impact on audio. Only the execution of a decent design does.
I would argue that tiny phase noise in a single DAC is not audible. It also affects both channels equally and at the same instant. This will have no impact on imaging even if it did occur in other words. If they use a standard quartz crystal, short term stability is the best of any solution save an ovenized quartz oscillator. Other sources ae not as stable short term, so when I measure phase noise of a clock, I pull the GPS reference to put it in holdover mode.
When playing music, with the amount of long term frequency drift in a normal oscillator (quartz) absolutely doesn't matter. What matters is short term stability. Now we are talking expressly about the execution of a design. So things like trace routing, crosstalk, power supply noise and all that. That is why I have the 5372A.
Anyway, unless the DAC chip or chip set is defective in design, it shouldn't have any more jitter or phase noise than the master clock. Of course physical execution takes over as the primary problem and what we are discussing has no bearing. You also need to make sure your experimental setup and instruments don't have near the same phase noise as what you are trying to measure.
Right now I suspect we are talking about things that do not have any practical impact on audio. Only the execution of a decent design does.
I would disagree on that. For example, if you look at discrete dac designs its easy to see where there can be more sources of jitter. Aperture noise at the clock input of a shift register can be one source that should be easy enough to understand. Also, some dacs have internal ASRC which may add some phase noise. There is also substrate coupled noise. Mode conversion of amplitude noise to phase noise is a well known mechanism. Once an MCLK signal enters a dac chip, of course it is subject to pollution by other internal noise sources. Etc.
Last edited:
There's another source of phase shift that may affect stereo imaging, and that's Global Negative Feedback.
Assuming the amplifier is well designed, it's stable.
I define 'stable' here as having less that 180 degrees of phase shift in the extreme treble the amplifier ever experiences.
Why does the phase shift matter only for the treble? Because it's a short delay. But 'stable' can accomodate some phase errors..
When a treble, stereo signal is amplified, it's possible that the actual phase shift on one channel is slightly different to the other, due perhaps to a slightly different waveform / frequency on one of the channels, or the work the GNFB is doing in the background,
Why do I think GNFB affects this? Well, if one looks at the gymnastics the NFB vs input signal do on a scope, with music, it's most interesting, and seems to me that it has the ability to affect the phase of some of the treble.
Therefore, with local feedback only, I would expect the stereo image to be better...
Anyway, this is my thought, and may help explain some of the differences heard 🙂
Assuming the amplifier is well designed, it's stable.
I define 'stable' here as having less that 180 degrees of phase shift in the extreme treble the amplifier ever experiences.
Why does the phase shift matter only for the treble? Because it's a short delay. But 'stable' can accomodate some phase errors..
When a treble, stereo signal is amplified, it's possible that the actual phase shift on one channel is slightly different to the other, due perhaps to a slightly different waveform / frequency on one of the channels, or the work the GNFB is doing in the background,
Why do I think GNFB affects this? Well, if one looks at the gymnastics the NFB vs input signal do on a scope, with music, it's most interesting, and seems to me that it has the ability to affect the phase of some of the treble.
Therefore, with local feedback only, I would expect the stereo image to be better...
Anyway, this is my thought, and may help explain some of the differences heard 🙂
Mark, that's what I was saying. Execution is everything. If you start with a problematic design, what are we even talking about???
Hi Globulator,
We are talking about decent designs. There is no end to problematic designs. Why would you even want to talk about those?
I can tell you from decades of experience and direct observation and measurements. Channels are close enough in response so this isn't an issue. To get phase variations you absolutely will have a frequency response difference between those channels - a big one. So let's not waste time talking about defective equipment, you can prove anything you want by examining stuff that doesn't work properly.
What you are repeating here is the anti-feedback crowd viewpoint. It isn't a valid argument, look at the speed of the circuit, then look at our upper limit of hearing and convert that to time. Compare them. Feedback is never the problem, a bad circuit is - whether or not feedback is used.
So no, your points do not explain anything because they are not valid when you compare any reasonable phase differences between channels.
Hi Globulator,
Absolutely not the case!
We are talking about decent designs. There is no end to problematic designs. Why would you even want to talk about those?
I can tell you from decades of experience and direct observation and measurements. Channels are close enough in response so this isn't an issue. To get phase variations you absolutely will have a frequency response difference between those channels - a big one. So let's not waste time talking about defective equipment, you can prove anything you want by examining stuff that doesn't work properly.
What you are repeating here is the anti-feedback crowd viewpoint. It isn't a valid argument, look at the speed of the circuit, then look at our upper limit of hearing and convert that to time. Compare them. Feedback is never the problem, a bad circuit is - whether or not feedback is used.
So no, your points do not explain anything because they are not valid when you compare any reasonable phase differences between channels.
Of course, this is my view 😀
Not all feedback, but the long, arduous trip across many stages and perhaps an OPT 😀
The FR and THD don't give the picture of the TIM distortion, in my view.
I think Lynn Olsen goes into some detail about this, e.g.:
http://www.nutshellhifi.com/library/FindingCG.html
I realise that many don't agree with me, this phenomena I'm familiar with for much of my world views 😀 😀
How do you know which one's which?
As mentioned earlier, stacking multiple stages in series and connecting a GNFB loop could be seen as 'problematic' already, but writing off the "anti-feedback crowd" as flat-earthers could make one a showcase of the Dunning Kruger effect in action if one is not careful.
Consider: 2 or more gain stages in series. GNFB takes the total gain and divides it down to a more reasonable value, dividing the distortion down as well. That seems simple enough, but what's often missed is that stacking multiple gain stages together also results in more complex types of distortion that didn't exist prior to the decision to optimize.
With one transistor, at worst, multiple signal frequencies passing through simultaneously will be modulated, once. (To the extent the transistor itself is treated as an ideal block with non-linear gain).
With two transistors, HD, IMD, AM, (and in some cases the transistor parameters themselves) (etc.) will be modulated again.
With three transistors, (HD, IMD, AM, (etc.))^2 will be modulated a third time.
What the "GNFB = Panacea" crowd seems to be claiming, is that for any given feedback ratio, the subjective quality of the distortion doesn't matter. It's basically the Douglas Self "blameless" argument. Reduce total distortion to some infinitesimal fraction, and it doesn't matter what the sound of silence sounds like.
The flaw in the above, or one of the flaws, is that the line between inaudible and barely audible is not so clear-cut. And if there's a possibility of at least slight audibility, then the subjective quality of the unwanted distortion starts to matter.
What's more, the design decisions necessarily attempt to reconcile a number of "apples vs oranges" choices.
2 stages could give something like 10 x 20 gain to play with. And a certain "grain" or colour to the distortion.
3 stages could increase that to 10 x 100 x 20.
GNFB = Panacea would argue that 3 stages are therefore better by design, as long as the end-goal is achieved. Success is measured by the perfection of putting the distortion below the threshold of audibility. If it fails, it means that in hindsight the design/execution was bad. Kind-of a strange mode of thinking, if you ask me.*
Another angle to what may be missed, is that the "horribly coloured" distortion that you hear, may be subjectively less coloured than the other "horribly coloured" distortion that you are unable to discern but some others can. What if somebody hears faint IMD like a symphony of Daleks chanting "Exterminate!" on every song, 10dB into the noise floor, and therefore, they would rather listen to a single-stage "distortion box" with a mediocre 0.05% - 1.0% HD, which doesn't have that?
*PS : not to pooh-pooh the way anyone else does things, but I'm more inclined to go for the "low gain, minimum number of stages necessary" approach. If there's a grey area in that distortion might be audible, steps can be taken to reduce it at the source. Following on from that, the harmonic content is kept low.
Last edited:
Please can anyone explain me why different devices measures almost the same with absolutely different components, but I hear how sounds differ with changing one capacitor in the same device ?
Hi Anatolii_A,
Either expectation bias, or differences you don't have the resolution to measure. If you can actually really hear something, it is very easily measured.
Hi abstract,
Easy as heck. Performance. The cleaner product will both measure better and be reported as sounding better given an unbiased audience. Done that for decades, it is very old hat. You can't make sweeping generalizations about layout either. You can have identical amplifiers from a schematic point of view, same components. The circuit layout can change everything and I have done that.
Ever notice that the "best" stereo component never stays at the top? Yet, ones that measured very well stay in the running over the years, only held down by the lack of a good story in the audio press.
Any equipment with higher distortion colours the sound. Period, end of story. Just like an effects unit hung in front of a better amplifier. The only problem is that it is there permanently, you can't unplug it once you get tired of the sound. Want to go that route, great. Most of us would rather not.
Either expectation bias, or differences you don't have the resolution to measure. If you can actually really hear something, it is very easily measured.
Hi abstract,
Easy as heck. Performance. The cleaner product will both measure better and be reported as sounding better given an unbiased audience. Done that for decades, it is very old hat. You can't make sweeping generalizations about layout either. You can have identical amplifiers from a schematic point of view, same components. The circuit layout can change everything and I have done that.
Ever notice that the "best" stereo component never stays at the top? Yet, ones that measured very well stay in the running over the years, only held down by the lack of a good story in the audio press.
Any equipment with higher distortion colours the sound. Period, end of story. Just like an effects unit hung in front of a better amplifier. The only problem is that it is there permanently, you can't unplug it once you get tired of the sound. Want to go that route, great. Most of us would rather not.
Hi Globulator,
THD and IMD at various power levels can completely describe any audio equipment. TIM doesn't exist in normal designs, this was proved ages ago. The "T" means transient. The circuit operates too quickly to have transient distortion unless you clip the stage, then the feedback loop is open.
THD and IMD at various power levels can completely describe any audio equipment. TIM doesn't exist in normal designs, this was proved ages ago. The "T" means transient. The circuit operates too quickly to have transient distortion unless you clip the stage, then the feedback loop is open.
That is I'm talking about, how we can see on graphs these changes with swiping capactior for example in rectifier ? Many different devices with different power supplies measures very good - almost the same, but they don't sound the same...
If you have problems in rectification, that is incredibly easily measured and shows up like a sore thumb in tests. Otherwise it is imagined, expectation bias.
You have no idea how many ideas and modifications (and wire) I have been asked to investigate over the decades. That is an old one you just brought up.
You have no idea how many ideas and modifications (and wire) I have been asked to investigate over the decades. That is an old one you just brought up.
It is very audible to hear difference for example between Nichicon KG - Nichicon KA - Panasonic FC - Panasonic FK - and another series/vendors. How this difference measures on graphs ?
Sure, because standard measurements don't measure everything that can be audible. For one thing, they don't measure linear distortions such as phase distortion at low frequencies. They also don't measure some different types of noise. What looks like white noise on an FFT can sound like a little hiss, like frying, like pops, and so on. It can still be white in the frequency domain. Also, ESS would say standard measurements don't measure non-PSS noise and or distortion. In addition, leakage currents in electrolytic caps cause 1/f noise, which may cause audible effects, depending.
Actually, what can be measured depends entirely on the equipment involved, and the setup. It is as simple as that. Our understanding on the subject has grown along with the resolution of the equipment we use and our understanding on what it shows us.
Sadly, those who don't have access to this kind of thing will forever be in the dark doubting what is known.
Sadly, those who don't have access to this kind of thing will forever be in the dark doubting what is known.
If it is actually audible, it is easily shown by putting those components in the same circuit configuration as used. You run the same impedance levels and frequencies / levels. Simple, has been done but not with your unknown examples.
If it can't be seen on measured information, it is your mind at work. Capacitor distortion has been measured very successfully in the past by various people. The basic rule is this, if there isn't signal voltage developed across the capacitor, you can't hear it as it does not generate distortion.
Some capacitors affect the distortion of a circuit greatly. One of my secrets, and they follow rules. The material and type is far more important than the brand or model folks. This is all physics, not very romantic at all. Sorry.
But it can generate 1/f noise due to leakage currents?
IME, it can be a mistake to assume only steady state distortion measurements can correlate with audibility. Correlated noise can be another matter, for example. Many people don't think to measure for that.