|
Post by ashcatlt on Nov 5, 2020 13:54:11 GMT -5
A couple days late for this I guess, but as Jello Biafra said “Why not every day?” Was just watching a video about Tony Iommi’s sound, and they got to talking about “ghosting” or “ghost notes”, and I kind of want to know more about how it happens and where it comes from. Quick googling brings some results on how to try to fix it, but not enough about the the actual mechanism that causes it, and I figured if anybody has the experience and knowledge to help me, they probably hang out around here. So it seems to have something to do with poorly filtered power supply somehow impacting the tone, but it also depends on the overall volume of the amp ( ), and doesn’t seem like it’s just mains hum mixed into the signal. This leads me to the idea that it’s something like what we call “sag” where the amp wants more current than the filter cap can source, so the entire supply voltage drops. This has the effect of lowering the limit of voltage the amp can provide - the distortion ceiling. If that limit just goes up and down smoothly, it’s just what we call sag. If, however, that change in the limit happens cyclically at an audio rate, it starts to sound more like a ring modulator and introduces side bands that would be the sum and difference of whatever frequency we’re playing and the frequency of that limit change. So ok, cool, but... 1) How and why is it dependent on volume? Just in general, how much we actually hear it depends (obviously?) on how far past the limit we’re trying to go. If we never hit the limit, it doesn’t matter that it’s wiggling, so a truly clean tone (not that there’s actually such a thing in a tube amp, but...) shouldn’t be affected at all. But is that all it is? Or does the filter itself fail worse the more we overdrive the amp? Does the limit actually wiggle more when the PSU is struggling harder? If so, why? I’m trying to think it in terms of resistances where they “shouldn’t” be, and kind of can’t come up with a scenario where it would have a dynamic effect, but I’m quite likely missing several things. B) How much does this affect the bias, and where, and how, and why? And I guess how much/what impact does this have on the overall effect? Does it cause (more) asymmetry in the preamp section? Does it affect crossover distortion in the power amp? Other things I’m not thinking of right now? III) Am I just altogether way off and completely misunderstanding the phenomenon? This is mostly just curiosity, but I also might try my hand at sort of emulating it in code, so if anybody’s got ideas on the sorts of amplitudes these wiggles would have, and like if there is a dynamic component, what kind of time constants are involved in the “attack/release” (or honestly for “sag” in general), or any other thoughts at all, I’d love to read them.
|
|
|
Post by thetragichero on Nov 5, 2020 22:10:40 GMT -5
got a link to the video? as mentioned, I've only read about it in threads attempting to troubleshoot it i can tell you from experience that if you overtighten the bolts holding in the speakers you can get "cone cry" which produces an almost out of tune octave effect, generally above the 12th fret on the b and e strings
|
|
|
Post by ashcatlt on Nov 5, 2020 22:24:43 GMT -5
I didn’t really want to post the video because I don’t really like the dude so I’m a little embarrassed to have even watched it and don’t super want to promote his channel and they only kind of mention the phenomenon in passing at the end of a long sort of wankfest, but...
|
|
|
Post by thetragichero on Nov 6, 2020 14:28:28 GMT -5
okay did some reading last night and just listened to the video
let's start at your wall outlet. you plug in the amp there and you get roughly 120VAC (this can vary a lot. power strip screwed to my bench consistently gets 122v, in the studio it's 119v unless the ac kicks in and then can dip to 117v) that goes into the amp and connects directly to a power transformer(we'll leave out fuse and power switch for simplicity). let's use a 100w Marshall plexi power transformer because that's the first i thought to look up a datasheet for. that converts our (nominal) 120VAC to 350VAC. but ac voltage isn't directly useful for our purposes inside an amplifier, so we run that through a rectifier (in this case a full wave bridge) to get 500VDC (350V/0.7 which is just a standard equation for full wave bridge). the only problem is this isn't a good quality DC like you would get from a battery, this is pulsed DC (DC with AC ripple) so it is then run through filter networks (capacitors, resistors and capacitors, and/or inductors and capacitors) to remove this AC ripple (remember that the capacitor presents a short to ground for AC voltage). the capacitors also act as a reservoir for DC voltage, sort of like a battery. the bigger the capacitor (in farads), the more it is capable of providing clean(er) DC voltage in times of increased current draw (say when the thing is cranked). if either the capacitors are old and dried up, or undersized (the Plexi uses 32uF on the first node compared to a jcm800 with 50uF) there is not enough stored DC so it's 'pulling' more pulsed DC with 60 and/120Hz AC ripple (this doesn't depend on us(60Hz)/uk(50Hz) mains frequency as erroneously stated in the video. it's based on full wave (60Hz or 50Hz depending on mains frequency) vs full wave bridge (120Hz vs 100Hz depending on mains frequency)
if we understand an amplifier as basically taking a small AC input (from a guitar) and 'wiggling' the DC power supply to produce a larger AC voltage, what this ripple does is modulate the wiggle.
so yes it is like a ring mod at 50, 60, 100, 120Hz. their explanation was close. those are the values I'd mess around with to try to emulate
I'd welcome other input/clarification/corrections from others
|
|
|
Post by ashcatlt on Nov 6, 2020 17:54:40 GMT -5
I do appreciate your input, but if I'm honest you've kind of just filled in the information that I left as given/understood. Probably good exposition for others who maybe aren't quite so familiar with the concepts, though. Real quick re: the video, when they played the comparisons, the first track was just way off from the original tone, and to my ears it's completely the Rat's fault. I mean, I love a good Rat, but cannot understand why they went into great detail on Tony's signal chain just to completely not do that, and throw a Rat on it. The second track, while not as close in the playing, definitely did sound a lot closer to the original. I also found it super amusing that he specifically said how Geezer's bass sound was essential to the whole thing, but then totally doesn't play or sound anything like him. the bigger the capacitor (in farads), the more it is capable of providing clean(er) DC voltage in times of increased current draw (say when the thing is cranked). One of my questions is pretty much exactly why this would be the case. A bigger cap means a lower corner frequency to the lowpass filter, and since it's a rather gradual slope, it will usually mean less ripple overall, but at least in theory with ideal components that should be dynamically static regardless of current draw. I'm trying to understand if/why there would proportionately more ripple when the amp is asking for more current. Can we model this action as a capacitor in series or parallel with some other component? What actually is happening?!? I noticed that, too. Not sure if they actually didn't know what they were talking about or just misspoke in the moment and then failed to catch it in editing, or... I guess another part of my question is why isn't it just hum, though. In most circuits, including guitar amps, if the power supply filtering is insufficient, it usually presents as a constant buzzing, but this doesn't seem to be that. It seems to be dynamic, which brings us back to that first question of why. Or maybe it IS just a tiny bit of hum that's not bad enough to really hear on its own until you really cram the signal past the rail hard enough to hear it like amplitude modulation? That would sort of explain the dynamic nature of it, but IME, it seems like the hum would need to be a lot more significant to cause much of a noticeable effect. Granted this is pretty subtle, and in fact I didn't really hear it on the video, but then I listened on phone, so probably wouldn't have gotten that sub information anyway. Sure, but like at what amplitude? If it IS dynamic, then does it sort of fade in and out the way sag kind of takes a (split) second to suck down, and if so like how long and how far?
|
|
|
Post by thetragichero on Nov 6, 2020 19:20:44 GMT -5
maybe i didn't read the original post clearly
the reason why it's occurring only when there is more current being drawn because instead of drawing current just from the filter capacitors it's having to draw directly from the rectifier (hence the wiggle)
in terms of amplitude in a modeling situation I'd imagine empirical testing would be in order. i don't have any amps on hand with undersized first supply nodes to throw on the scope
and yes i thought the decision to use the rat was an odd one. "he used a simple treble booster but we're going to use something completely different"
|
|
|
Post by ashcatlt on Nov 7, 2020 13:24:40 GMT -5
maybe i didn't read the original post clearly Naw, I didn’t really lay out my foundational knowledge. (just enough to be dangerous) That’s the part I can’t get my head around. Intuitively it sort of makes sense, but my understanding of electronics in theory just doesn’t bear it out. A filter is a filter and a “short for AC signals” is just that. It shouldn’t matter how much current you’re trying to pull out of it. The way I see it, it should just sag all together. Maybe I’m just being dense, but if I was to make a spice model using just idealized versions of exactly the components in the circuit, I don’t think it’ll show that behavior. I feel like there needs to be some “hidden” components involved to represent the ways that real components are not ideal, but I’m having trouble figuring out what I could put where to make it happen.
|
|
|
Post by reTrEaD on Nov 7, 2020 14:04:38 GMT -5
I didn’t really want to post the video because I don’t really like the dude To say that 'I really don't like the dude' would be putting it mildly. I could have more harsh words about him, but I'll just agree with your assessment and leave it at that. To begin with, in this case he's perverting an existing musical term and applying it to an entirely different phenomenon. I find that heinous. A ghost note is a note which has rhythmic value but no discernible pitch. For instance, when you mute a string but pluck it. Now that we've set the record straight on accepted terminology, lets move on to the phenomenon he noted.
At the core of this is distortion. And in particular, intermodulation distortion. While there are more types of distortion, the two we'll focus on in this discussion are Harmonic distortion and Intermodulation distortion. When a single signal is passed through a non-linear device, Harmonic distortion occurs. The output is the original signal plus the integer multiples of the input frequency. If we call the input frequency f, the output would be f and 2f and 3f, etc. The more non-linear the system, the greater the amplitude of the multiples, but the amplitude of the multiples will nearly always decrease as the integer increases. Any integer that can be expressed as 2 x will be highly 'musical'. It will create perfect octaves. Other integers will be somewhat less musical, particularly in even-temperament. For instance, the third multiple of A 2 (110Hz) will be 330Hz. This lines up with E 4 an octave an a perfect fifth above the original in just-intonation. It's slightly off of the 329.63Hz for that note in even-temperament. But hey, close enough!
Intermodulation distortion occurs when two (or more) signals are simultaneous passed through a non-linear system. And the more non-linear, the greater the amplitude of the distortion products. We get the original signals out plus all the sums and differences of those originals. Even with as few as three signals, thing can get very 'mathy', chaotic, and non-musical in a hurry. Hence the common use of 'power chords', just the root and fifth. Let's examine an instance where we have the root and fifth going through a non-linear system. In this case we'll choose A 2 (sixth string, fifth fret) and E 3 (fifth string, seventh fret). 110Hz and 164.81Hz. Our distortion products will be 54.81Hz and 274.81Hz 54.81Hz is very nearly equal to 55Hz or A 1, which is an octave below our root. 274.81Hz is rather sharp compared to C 4 (261.63Hz) in even-temperament but it falls just about dead-on to that note in just-intonation. That would be an octaves and a major third above the root ... fairly musical. ... but getting back to the octave below, this occurs without the power frequency even being a factor in this equation. So these two 'experts' have it completely wrong.
Will the mains frequency enter into this? I dare say yes, but not so much in the way they suggested. In this particular case, if 60Hz (or 50Hz on the other side of the pond) is also introduced, that would beat at approximately 5Hz with our subharmonic A 1. Like a very rapid tremolo. I doubt we would see 60Hz (50Hz) introduced into the system because of power supply limitations. There I would expect to see 120Hz (110Hz) because nearly all amplifiers use a full-wave rectifier, either full-wave center-tap or full-wave bridge. But with a P-90 or any other single-coil and tons of gain, I could definitely see a path for mains hum entering the system right at the source.
How does sag fit into this? In my estimation, there may be some greater introduction of 2x mains frequency, but I reckon the larger effect will be that when the PS sags and the voltage available to the plates drops, we'll go deeper into clipping (distortion). JMO. As always, Your Mileage May Vary.
|
|
|
Post by ashcatlt on Nov 7, 2020 15:07:22 GMT -5
The more I read about this phenomenon, the more convinced I am that it’s not really a dynamic thing. It’s really just basically static supply ripple. I could very well still be missing something, but what I’m seeing is that the supply just wiggles more or less based on how far the cutoff frequency of the supply filter is from the 100/120Hz rectified supply frequency and probably also because ESR (effective series resistance?) in the cap makes it into a shelving filter, basically mixing in some of the unfiltered signal. When the supply sags, the ripple goes with it. The reason it doesn’t sound so much like hum is that - for various reasons including the nature of the push-pull output stage itself - it’s not wiggling the “middle”, but the “top”. That is, it’s not introducing an actual oscillating signal, but rather it is oscillation in the ultimate limit - the clipping threshold itself. The reason it seems to be dynamic is, like I said in the OP, if you’re not hitting the limit to begin with, then it doesn’t matter that the limit is wiggling, and the further you try to go past the limit, the more effect it will have. Sag plays into it I guess in that, as reTrEaD says, sag puts us further into clipping. The whole limit moves further below the desired output, so the ripple has somewhat more impact. I have done some experiments with this sort of “limit modulation” in the past. I have a plugin that I made which basically has CV-type input to set the clipping threshold. It can be used to emulate sag basically by feeding a short-term RMS of the actual audio signal in, but you can also just feed it an oscillator. It can be an interesting effect, but...well I guess the best I can say is it’s touchy. I will say that the term “ghost note” seems to be pretty commonly used for this kind of IM byproduct. Rick is a dink, but he definitely didn’t coin this usage.
|
|