I do agree that there is too much abuse on these issues in this forum to the detriment of education. If someone is a die-hard objectivist (or subjectivist) it would be better to ignore topics they disagree with than to pick fights. I wish the mods would enable thread starters to delete such posts or establish some way to curtail these distractions (e.g., a dislike button enabling social reinforcement of good behavior).
Back to the topic: I think the issue of noise vs grain remains open. We know a lot about noise in components. It’s hard for me to make the leap that all grain is caused by noise. Obviously there is a logic to it: it’s anharmonic/orthogonal to the original signal—but *if* noise is the cause, we should try to understand how it becomes audible when it’s not audible as a noise floor. This is basically a topology question. One person said noise in a feedback loop was audible. That makes some sense since a small voltage fluctuations are being amplified and fed through the very same device, causing a resonant growth in noise amplitude at certain frequencies. But even here we don’t really see what we might expect when measuring steady-state signals. We would expect a rise in the noise floor to audible levels with the signal. That usually doesn’t happen. Note: I am not denying the phenomenon exists. I’m saying the measurement is either not a measure of the phenomenon (eg., if it were inherently a transient phenomenon not captured by steady-state FFT) or there is an erroneous assumption about inaudibility of -100dB noise floors, or grain is actually something other than noise. For example what if grain were a tone’s frequency stability, which would also be below the resolution of the FFT.
I was also curious why in #61 a member suggests that a higher noise carbon resistor produces less grain than a low-noise alternative—because this may be useful evidence against the grain=noise theory.
Back to the topic: I think the issue of noise vs grain remains open. We know a lot about noise in components. It’s hard for me to make the leap that all grain is caused by noise. Obviously there is a logic to it: it’s anharmonic/orthogonal to the original signal—but *if* noise is the cause, we should try to understand how it becomes audible when it’s not audible as a noise floor. This is basically a topology question. One person said noise in a feedback loop was audible. That makes some sense since a small voltage fluctuations are being amplified and fed through the very same device, causing a resonant growth in noise amplitude at certain frequencies. But even here we don’t really see what we might expect when measuring steady-state signals. We would expect a rise in the noise floor to audible levels with the signal. That usually doesn’t happen. Note: I am not denying the phenomenon exists. I’m saying the measurement is either not a measure of the phenomenon (eg., if it were inherently a transient phenomenon not captured by steady-state FFT) or there is an erroneous assumption about inaudibility of -100dB noise floors, or grain is actually something other than noise. For example what if grain were a tone’s frequency stability, which would also be below the resolution of the FFT.
I was also curious why in #61 a member suggests that a higher noise carbon resistor produces less grain than a low-noise alternative—because this may be useful evidence against the grain=noise theory.
Last edited:
Shouldn't the conversation be about all tonal imbalance, rather than a single flaw called "grain" ?
Tonal imbalance is a result of two basic elements: design topology and inherent component performance.
The spec sheet (especially IMD/THD/TID) does not adequately predict sonic performance. Often, it's just the opposite.
There is -no- perfect correlation between perception and spec, at least when it comes to things like distortion, slew, etc..
Anyone who has actually designed audio gear (25+ years here) knows that an ideal topology emerges over time and experimentation by LISTENING.
A comprehensive and well-refined blinded AB listening test methodology is foundational to the work and life of any audio designer worth their title. This would, by default, include a broad range of extremely high quality program material that is intimately well known to the designer.
For example, certain program with complex top end, some with dense middle (like choir), some with solo instruments, some with complex ambience signature, mass strings, etc.. Fast A/B comparison (for instance, using DAW soloing) is required, as proven by psychoacoustic testing on perceptual auditory memory.
Last edited:
20 years ago, "grain" (if I understand the term) was more common in opamps. Today, I rarely encounter it. The sense of graininess seems to occur in ultra-high-slew ICs (video-spec, etc.) used for audio. Beyond that, I can't think of an audio-purposed IC opamp that exhibits "graininess."
Part of this is design. I can make a 5534 sound grainy by futzing the i/o parameters. I think the 5534 got a bad name for this very reason: designers using it in ways it doesn't like. I wouldn't use that part any more, but just sayin'
Some film capacitors have quite a bit of graininess until they are broken in, the Multicap PPMFX in particular. The KEF LS50 was unlistenable from grain and brightness, until broken in by playing continuously for over a month. Hard to say if it was the driver or the crossover parts.
Grainy sound may have more than one possible cause. IME, it is often some kind of power supply related issue. In some cases, adding some good quality film caps between the power rails and ground can help. In other cases, trying that approach can do more harm than good as a result of worsening other existing issues, or maybe by causing some new sound quality symptoms.
Last edited:
Many, many years ago i experienced a revelation trying the old Signetics version in the inverted configuration. Equalising the impedances seen at both inputs and making sure it only worked into high impedance loads got rid of the last traces of grain. Not sure if this is what you had in mind, but if it is, yes, a great example.
I've never experienced capacitor "break in". Did you do a blind AB of new vs. old identical film caps?
We know that certain ICs can sound poor in the first few minutes after power-up. Tubes, of course, same thing.
There are distinct tonal colorations among electrolytic capacitors used in microphone preamp front-ends (phantom block) -- even among those with identical specs (ESR, etc.). A routine blinded AB test immediately reveals such differences. I wouldn't call it "grain" but rather tonal shading. Adding a small film cap in parallel with elec often reduces the problem.
The first suspects I would look at would be electrolytic caps. They're notorious for drying out, drifting and leakage. A big problem I ran into servicing audio equipment is customers not removing dust on a regular basis. A thick coat of dust holds heat in. Second would be the supply rails. Third are volume and tone controls. If the quality wire-wounds, you can use a good electronics cleaner to break up the oxides that form. Mechanical interfaces like RCA jacks and speaker connections are always worth a look. I found over the years that mechanical problems account for 60% problems.
In this case the “grain” would correspond to odd-order distortion, which would include harmonic and IM components. If you run a push pull stage into a high enough impedance, it runs class A which minimizes crossover distortion. Unbalanced source impedances unbalances the diff pair, and you may as well have mismatched input transistors. A mismatched difff pair has a lot more distortion than a properly matched pair, since the feedback summation is less perfect.
In the RF world, noise and distortion are often indistingushabld from one another. Especially when the signal is a complex, high-order modulated digital signal with a high peak to average ratio, such as an LTE cell phone signal. These signals have a lot in common with audio, actually. So which produces “grain”, noise or distortion? Both. Because in.the limit when the number of tones (carriers) aporiach infinity, both contribute to “noise power”.
If pure static white noise is added to the source, it would not make sound grainy, but it just raise the noise floor. Every electrical component should produce some kind of noise, and I suspect that the noise signature and its dynamic behavior of the each electrical component would be very different each other. I guess synergistic effect of them would make a difference, and so that a very simple amp with little components sounds less grainy in general even it has higher noise floor.
Please clarify: how does odd order distortion have IM components? As for the harmonics, once you attach numbers to them you can only start scratching your head. Little doubt that it is distortion, but perhaps not a type we have a definition for.
In my, third harmonic rich valve amp, the last source of grain was associated with the fixed bias voltage source. Topology and especially parts quality was very important to solve this.
Perhaps post #206 A Tube amp without coupling capacitors? Possible? has some relevance.
Odd order IM is 2*f1-f2. For closely spaced frequency components, the IM product fallls back into the same frequency range. Tends to raise the noise floor. Audio-wise, you don’t hear it as a discrete tone. With complex signals (like music) you just lose resolution. The higher the order components are (ie, 5f1-4f2 for example) the less “correlated” to the original signal it “sounds”. It could be interpreted as “grain”. Crossover and switching distortion is the biggest source of high-order IM - something actually has to generate the harmonics which produce it.
I think this is the best theory I’ve heard thus far. It’s consistent with the observables. Thanks wg_ski
For audio, IMD (at least typically) sounds different from graininess. Graininess is perhaps closer to quantizing distortion, but not exactly the same sound as that either.
Of course, any type of nonlinear voltage transfer function will produce IMD when more than one frequency is present. Doesn't mean all types necessarily sound the same.
Of course, any type of nonlinear voltage transfer function will produce IMD when more than one frequency is present. Doesn't mean all types necessarily sound the same.
Actual measurements you've made, or data sheet measurements - if the latter I'd suspect the bias has drifted on the actual amps.
Mark Tillotson: Those are actual measurements. -118dB was the noise floor on the Pioneer M22. And since others have asked, the M22 has been fully re-capped with carefully paired (and generally low-ESR, low-thermal drift) caps. As for the Pass X25 and Pass/FW F5, those are relatively new amps only a few years old. I would say both the F5 and M22 have noticeable grain. The XA25 is serving as my reference for clean sound.
The reason for this thread was more to seek an understanding of the phenomenon generally—but of course, if I can figure out how to eliminate the grain in my amps, I will.
The reason for this thread was more to seek an understanding of the phenomenon generally—but of course, if I can figure out how to eliminate the grain in my amps, I will.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.