|
Post by ashcatlt on Oct 4, 2017 19:28:25 GMT -5
We've all heard the term. Some of us even know how and why it happens. I guess there are several mechanism that can cause or contribute to the phenomenon. I'm not really worried about any of that. What I want to know is the numbers. The actual hard figures: How "far" does a relatively squishy amp sag. Like, if the supply voltage is V when "unstressed", and sags down to V * x (where 0 < x < 1) at full bore, then what is a reasonable value of x? Then, what kind of drive causes that much sag. Seems to me that it has something to do with input * gain = how much voltage it wants to put out, and how much current that wants to draw, but how does that map? How close to the rails can it get without sagging, how far past before it doesn't get any worse? Is it more or less a linear droop, or kind of curvy? Then, what about timing? It takes some time after a loud event for the thing to sag significantly, right? How long does the (attempted) output have to stay loud before it gets as sagged as it's going to get? And how long does it take to recover after it stops trying to be loud? I've Googled around a bit, and while everybody and there brother has a useless fluff piece about how important this phenomenon is to the sound and feel of a tube amp, and like I say some folks even kind of explain the mechanism behind it, but I can't find anywhere anything even close to real numbers. Anybody know anything about this, or where I might find this kind of info, or willing to help figure out?
|
|
|
Post by newey on Oct 4, 2017 21:13:51 GMT -5
You're way beyond my limited technical knowledge. SG would be the one to know how/where to quantify it, but he's off in the RV at the moment. I had always thought the term "sag" was just some subjective perception of tube amp sounds, not an actual quantity. Either that, or it stood for the Screen Actor's Guild.
|
|
|
Post by reTrEaD on Oct 4, 2017 23:27:16 GMT -5
There's really only one mechanism. The relationship of source resistance in the power supply relative to the amount of current being drawn. Capacitance in the primary filter also plays a role. Not so much in the amount of sag but in how long it takes for the B+ voltage to sag under a given load. DC resistance in the power transformer factors in. On the primary side we probably need to multiply by the voltage ratio or something. On the secondary side it depends on the style of rectification. For a full-wave bridge, the DC resistance is all we need. For a full wave CT, 1/4 of the total resistance, end to end or 1/2 the resistance from CT to end. A power supply with solid state rectifiers will inherently have much less sag than one with a tube rectifier. Also adding to sag would be the internal resistance of a filter choke, if one exists before the capacitors feeding the CT of the output transformer. But in power supplies with a tube rectifier the resistance of the rectifier is accounts for the lion's share. Tube data sheets would be a good resource. Most of the amp guys reserve the term 'squish' for effects in the preamp stages and sag for those in the power amp section. In a push-pull output stage, sag is very non-linear and rather abrupt. A class AB push-pull amplifier draws a significant amount of current at idle. So the B+ is lower than it would be if there was no load. But that really doesn't change much as we apply signal until we get the output stage into a non-linear region. When one side is drawing more current, the other side draws less. So the load on the power supply remains fairly constant until the signal becomes so large that one side is past the point of not conducting and the other side is drawing more than the total idle current. Yes, this is where the size of the capacitors in the primary filter becomes important. Larger capacitors require more time to discharge and to charge. I'm not sure if anyone has ever boiled the effect down to a formula but if anyone ever did, I imagine one of the guys at The Amp Garage would have been the one to do so. ampgarage.com/forum/
|
|
|
Post by ashcatlt on Oct 6, 2017 12:33:23 GMT -5
I suppose what I should do is model some thevenin equivalent approximation and see what happens....
Really, I could could just try it and see, turn the knobs till it sounds good. I'm not trying to emulate anything, really, but some real world perspective is sometimes good.
I'm playing around with different saturation schemes in Reaper's JS plugin environment. Got some pretty cool stuff happening - "diode" clipping, slew rate limiting, something like transformer saturation, an RMS limiter - and adding "sag" to any or all of them would be pretty easy, and might be fun.
|
|
|
Post by sumgai on Oct 6, 2017 23:28:59 GMT -5
Sag is a phenomenon that we can hear acoustically, and observe electrically (I mean, with measuring instruments).
Acoustically, it's a result of the power-amp stage temporarily being unable to keep up with the demands of the input signal, i.e. it's overdriven. Conversely, because that result is attractive to some guitarists, it has become a design 'feature' instead of what it should be called; a poorly designed amp stage. Of course, I intentionally leave out the fact that some players (you pick a percentage...) abuse an amplifier in order to take advantage of this phenomenon. For me, if you abuse your amp for any reason, that's OK - it's your amp; have at it!
reTrEaD hits all the high spots in how this occurs, possibly excepting the omission of the fact that overloading a transformer usually causes eddy currents that will play hob with a power supply's delivery of needed current. (At this point in the discussion, it's important to remember that a power-amp stage delivers current to a load as demanded - voltage is present simply as a by-product of Ohm's Law. Because of this distinction, it becomes obvious that the term 'sag' came about because as the current draw went up, the measured voltage went down.) This is exactly why sag is unpredicable, and can't be (easily) modeled - it will happen if an amp stage is forcefully overdriven on purpose, but there are times when the signal is just a tiny bit under the threshold of saturation, with no acoustically apparent 'sag'. Then, the player need only 'touch' the strings a wee bit harder, and Voila!, the amp goes gunnysack in the electrical department, which gets translated to the acoustical department - in a hurry.
Sadly, numbers cannot be spewed across the screen so easily as the theory. The fact is, the observable phenomenon is very non-linear. It happens fast, and it recovers almost as fast - the duration is quite short, due to either one of two conditions - the overdriving incoming signal ceases to overdrive, or the main fuse blows. Since fuses seldom blow under these conditions, one expects that it is the percussive attack of a guitar string that is the essential culprit in causing sag.
I could go on, I've conducted more than my share of investigative experiments, but in the end, it won't matter. As I noted above, the desired acoustical effect is a direct translation from an electrical effect. Even that very translation is subject to so many factors that I'd be hard-pressed to name half of them, I'm sure. For me, it was enough to see the voltage drop about 30 percent for a second or so, and rise back up to normal at about the same rate. (Using my eyeballs for drop-and-rise, I had no digital storage scope at the time.)
I do have one bit of advice for players though.... If this is your cup of tea, then emulate EVH. He wanted a sustained sag, and he obtained it with a Variac. Thus was the beginnings of 'the brown sound'. Look it up.
HTH
sumgai
* ash - that last bit about EVH was not for you, I know you're taking an entirely different path.
|
|
|
Post by ashcatlt on Oct 7, 2017 13:20:13 GMT -5
My plan is to modulate the limit of a clipping function based on the RMS average over some length of time, possibly with an attack/release filter, and I'm really just looking for any kind of eyeball ballpark for starting parameters.
"30% for a second or so" actually is a lot more info than I've found so far.
So is the fact that it's some kind of nonlinear. The current draw is constant until it suddenly breaks over to very large. Gonna take some kerthunkin to get that happening, but have some clues...
|
|
|
Post by sumgai on Oct 7, 2017 21:30:50 GMT -5
ash,
This is all from memory, my books and such are in a storage unit... but at least they aren't in another city!
I had a well-used FSR-SF, I think it was a '74 or '75, something like that. IIRC, it was supposed to have 480vDC on the plates, and I was getting 455 on one side, and 465 on the other - pretty well balanced. The scope was looking at the dummy load, and the meter was looking at the plates. The meter (a Fluke 187) had a min/max hold function that captured the data, but the scope (an Hitachi V220) was too old for that.
It seems to me that the plates went down to just a tick over 300vDC, perhaps a bit more. On the output, I saw a drop from 11.25vAC to just something like 8vAC. Those two don't compare exactly, but then again, the dummy load is now 'in the equation' as well, so when considering non-basic conditions like this, it often pays to take measurements at more than one point.
Sadly, I had only the one DMM, so I couldn't also do a current reading at the same time. Besides, even though the meter was fused at 10aAC or DC, I was still leery of testing that much current draw - I had already fried one meter, earlier in my career, attempting such on a Marshall (that was smoking the output transformer). Better a little too cautious, and have some money remain in my wallet, than repeat a costly mistake, eh?
Oh, and the time all this occured was subjective - I'm pretty sure that it dropped like a stone, but it took no more than a second or so to come back up to nearly full steam ahead. I had no way to capture any clock readings at the time, so my memory may well be fully faulty on this part.
HTH
This was all in 1990 or thereboots, and even then I was fully into solid-state amplification. Fender was still the top sounding amp in my book, but for some reason, I just had a thing for building huge, honkin' powerful, banks of transistors, sissies-need-not-apply, tube-lover-destroying power amps. (For the record, Fender is still my favorite. But I've "trimmed back" (see gumbo's tagline) to only one FTR-SF and one Bassman 200. A nice PA system pretty much does what I need now, vis-a-vis a synth setup.
sumgai
|
|
|
Post by b4nj0 on Oct 8, 2017 3:18:32 GMT -5
Boy I love it when I read "stuff" that isn't just cut'n'pasted from Wikipedia. I would just add that we remember that a variac is an auto transformer and as such provides no DC isolation. Great for reforming capacitors but to be used with caution.
e&oe...
|
|
|
Post by reTrEaD on Oct 8, 2017 11:51:22 GMT -5
Sag can be quantified but only in coarse terms.
1 - Lift one up with your left hand. 2 - Place a pencil parallel to the ribs with your right hand. 3 - Remove your left hand first. 4 - Then release the pencil from your right hand.
If the pencil falls to the ground there is no sag. If it doesn't, there is sag. If you had to reach below the navel with your left hand, there is way too much sag.
|
|
|
Post by ashcatlt on Oct 8, 2017 14:20:06 GMT -5
sg - do you have any clue how big the voltage at the amp input was, how much gain there was in the circuit, or basically what the power amp wanted to put out? It sort of sounds like this happens all of a sudden when the attempted output comes close to the maximum possible output. Does that sound about right? Or do you have to try to go WAY past? Or does it start happening well before we start bangin the rails? rT - I think I'm gonna need a diagram (better yet a vid) on that one.
|
|
|
Post by reTrEaD on Oct 9, 2017 4:34:32 GMT -5
I would draw you a diagram but I seem to have lost my pencil.
|
|
|
Post by sumgai on Oct 9, 2017 13:07:54 GMT -5
sg - do you have any clue how big the voltage at the amp input was, how much gain there was in the circuit, or basically what the power amp wanted to put out?
Not really, but in most cases, a stock single-coil (not an aftermarket hot-rod) can deliver between 0.6 and 1.5vAC to the input jack, straight out the axe.
Now much gain? That's a seque for a discussion for another time, but the bottom line here is: no, sorry but I never worried about 'gain' between stages or between any other particular points in the circuit. While I'm an Engineer, and I'm (almost) morbidly curious about the inner workings of various esoterica, I truly don't think it's necessary to worry about the fine details in these cases. The amp's manufacturer gave us one volume level control for the entire amp, and that was it. (Although now that I think on it, you're correct - all of the bigger Fender tube amps of that day and age had Master Level controls - but I never adjusted them for any reason. To me, they were fluff, put there by Marketing people, and not by real engineering designers.)
The basic idea of dialing in a tube amp is to set the power-stage bias such that the output at the speaker for the least discernible distortion at the highest possible volume level. The details of how to do this are rampant across the web, so I won't get into it here. But during this setup is when we use the o'scope, to look at the waveform. Sadly, the only way to see anything on the scope is use a steady state signal, from a signal (tone) generator. This has to be very clean, to say the least - no harmonic distortion, etc.
That said, I was curious about something - I could drive the amp into massive distortion (more tha 70 watts across the load! And the OT got hot, reall hot.) and it sounded like crap when a guitar was plugged in. But it sustained something fierce, like a compressor and a fuzz-tone together. Hmmmm....
Using a guitar, I then played both single-note lines and chords, and watched the scope for least distortion. Yeah, the bias turned it all down to less than 30 watts or so before it cleaned up nicely. Say what? Wasn't I supposed to see 45 watts? This was shaping up to be an insanely heavy FDR with 4-10s instead of 1-12.
Well, yes, that was the stated rating. (Remember, in "professional" audio, versus any grade of audio reproduction, the acceptable limit of distortion is an order of magnitude higher. If you looked closely, you cold find that Fender rated their amps for a given wattage at 4% distortion. Sounds bad at first, but all you need do is recall about 7 decades of history - was everybody out of their minds, or are you assigning a good/bad value judgment to something out of misplaced familiarity?) So I went back and re-biased the amp properly with a steady signal, and then watched the scope whilst playing. That was when I found a large drop in the scope's display for a moment, as I hit the strings. That lead me to put a DMM on the output as well, which confirmed my observation. Moving the meter to the plates showed the same thing happening, only on a larger scale. Interestingly, the scope also showed the waveform becoming quite flattened at the peaks, for those short periods of time. Could this be the 'compression' that folks speak of? I dunno, but....
Finally, recall that in a tube pre-amp stage, self-bias is usually the case. This means that stage gain is variable, though usually linearly so. Recall too that intentionally overdriving the grid will affect the observed gain. From my (very) early days of amp modding: want more "bite" in your tone? Simply remove the grid drive-limiting resistor... that'll make a noticible difference!
I trust I've answered your questions?
HTH
sumgai
|
|