Yes, certainly no yawn and very welcome posts. As my "signature" says theory is for other people..you are one of those "other people"..and if you keep long enough at what you are doing and can explain what it does...I will steal it and use it for my own relentless persuit of happiness in diyaudio..
Thanks,
Bas
Yes, you are right. When the circuit leaves class A mode many things change. Even so, the class A analysis can be helpful.
For example, the AC resistive behavior of the regulator will more or less apply to situations where current demands are asymmectrical. In a screen regulator for AB mode, etc. where the PS has to supply current pulses, the resistance of the PS still applies (assuming that the regulator is fast enough). And so you can know partly what it will do even in non-class A modes. I'll try to investigate this some (but things are starting to pile up
).
One other thing to be sure to say is that the resistor/capacitor models in the simulator are perfect, whereas real resistors and capacitors (especially electrolytics) are not perfect and do have voltage/frequency non-linearities. If anyone can supply me with a good electrolytic capacitor model, I'd really like to have one.
For example, the AC resistive behavior of the regulator will more or less apply to situations where current demands are asymmectrical. In a screen regulator for AB mode, etc. where the PS has to supply current pulses, the resistance of the PS still applies (assuming that the regulator is fast enough). And so you can know partly what it will do even in non-class A modes. I'll try to investigate this some (but things are starting to pile up
).One other thing to be sure to say is that the resistor/capacitor models in the simulator are perfect, whereas real resistors and capacitors (especially electrolytics) are not perfect and do have voltage/frequency non-linearities. If anyone can supply me with a good electrolytic capacitor model, I'd really like to have one.
Bas, theory may simply be for weird folks. Which, for most of us is also "other people."
I do, however, build stuff too. At the moment I'm building a pair of circlotron monoblocks and and preamp to go with them. But I figured that if I posted the schematics for these as my first real post, I might not survive it.
I do, however, build stuff too. At the moment I'm building a pair of circlotron monoblocks and and preamp to go with them. But I figured that if I posted the schematics for these as my first real post, I might not survive it.
Reactive Effects
OK, so here is part two. Do the reactive components of the PS/Regulator affect the circuit? I’ll use the same 6dj8 grounded cathode amp. But first, we have to talk about phase relationships, because this is what the reactive effects are all about. I'm going to do this in two parts.
Everyone on the forum knows what phase is. It is actually very important to audio design, but often left out of consideration. Why is it important? Here is an example.
Everyone also knows that a square wave can be represented by an infinite series of sine waves (Fourier analysis). The basic idea is this:
Square Wave = A1 * sin ( a1 * f) + A2 * sin (a2 * f) + . . . An * (an * f)
Where n typically goes to infinity, A is an amplitude value, a is a multiple of pi, and f is the fundamental frequency of the square wave. Now there are three ways that a circuit can mess up this square wave trying to amplify it:
OK, so here is part two. Do the reactive components of the PS/Regulator affect the circuit? I’ll use the same 6dj8 grounded cathode amp. But first, we have to talk about phase relationships, because this is what the reactive effects are all about. I'm going to do this in two parts.
Everyone on the forum knows what phase is. It is actually very important to audio design, but often left out of consideration. Why is it important? Here is an example.
Everyone also knows that a square wave can be represented by an infinite series of sine waves (Fourier analysis). The basic idea is this:
Square Wave = A1 * sin ( a1 * f) + A2 * sin (a2 * f) + . . . An * (an * f)
Where n typically goes to infinity, A is an amplitude value, a is a multiple of pi, and f is the fundamental frequency of the square wave. Now there are three ways that a circuit can mess up this square wave trying to amplify it:
- Altering the value of any of the As (harmonic distortion)
- Altering the phase relationship among the different frequency components (phase distortion)
- Generating new frequencies (intermodulation distortion, which basically leads back to #1)
[/list=1]
So, if the amplifier causes frequency dependent phase shifts, then the square wave will not be amplified perfectly because the sine waves won’t add up right. You’ll get rounded corners and ringing. This is one reason why good amps must have a near flat response and minimal phase distortion beyond the audio range, so that the transients can be amplified faithfully (or close to it anyway).
OK, so do tube amplifiers have frequency dependent phase? You bet they do. And depending on the number of stages, lots of it. Generally, the more stages, the more the phase shift. Not only that, but at as the frequency gets higher, the phase relationships that hold in the audio spectrum (in good designs) can be completely reversed above (below) the audio spectrum. No big deal you say. Maybe, maybe not. But if you’re applying global NFB, very big deal. Because if you’re feeding a signal back to the input that is supposed to be 180 degrees out of phase, but at 100KHz it gets to be 360 degrees in phase, then you have a positive feedback amplifier, or oscillation. I know that you guys all know this, but I figured it was worth the trip. Besides, you guys are better at this than I am.
Reactive Effects Cont'd
OK, so the main effect of the reactive components of the PS (capacitors and chokes) will be in altering the phase behavior of the amplifier. Chokes and capacitors store energy in magnetic fields or in electrostatic charges and this stored energy is returned back to the circuit in a frequency dependent way causing phase shifts in V and I.
So, we start here with our basic amp but with a perfect PS. Like this:
Looking at the voltage phase at the point marked “Phase” (the output of the amp) we see this:
Since this amplifier inverts, the primary phase relationship is -180 degrees. However, at low frequencies the phase is slightly less than -180 and at higher frequencies it is slightly more. What this means is that the high frequencies effectively get to the output “faster” than the low frequencies. Practically speaking, however, these phase shifts are probably audibly unnoticeable. But, I don’t have data to back up this claim.
But, no matter what we do with the regulator we can’t get any better than this, unless we can actually improve the phase shifts somehow.
Starting as before, if we insert our basic RC regulator and re-plot the phase shifts caused by the amp we get this:
In other words, the RC regulator has absolutely no effect on the phase behavior of the amplifier.
So, how about the series regulator? Taking the “raw” regulator with the 470n damping cap we get:
Absolutely no effect. In fact, there is no discernable effect for any of the regulators that were used previously. So, for this very simple analysis, we can see that, from a phase point of view, all of these regulators are the same. Meaning that the reactive behavior of the regulator on the circuit is irrelevant. This is primarily due to the fact that the resistive component is also irrelevant and, hence, the phase of the delta-V just doesn’t show up.
I should also add that the phase of the current in all of these examples is identical to the voltage phase.
And, I should give the caveats that this is simulation, these are perfect Rs and Cs, the circuit is pure class A, and the line side of the PS is not being taken into account.
You may ask, is there every any change to the phase relationship? There can be, depending on the exact time constants of the PS and the circuit. For example, here is a result for the RC regulator where the cap is set to 1u instead of 330u.
The voltage phase shift is now 180 degrees advanced and the difference is almost 15 degrees at 20Hz. In the case where C=330u, the phase shift at 100Hz is almost exactly -180d. I’ve also plotted the current phase shift in red. Even so, these phase differences are unlikely to be audibly noticeable.
Now, when chokes and big caps are involved in high current situations (for example, OTL output PS), YMMV. Maybe we can look at this later.
In the meantime, I think I've figured out how to account for most of the pre-regulator PS. If this works, I'll be back with more. But, please let me know how we're doing.
OK, so the main effect of the reactive components of the PS (capacitors and chokes) will be in altering the phase behavior of the amplifier. Chokes and capacitors store energy in magnetic fields or in electrostatic charges and this stored energy is returned back to the circuit in a frequency dependent way causing phase shifts in V and I.
So, we start here with our basic amp but with a perfect PS. Like this:
An externally hosted image should be here but it was not working when we last tested it.
Looking at the voltage phase at the point marked “Phase” (the output of the amp) we see this:
An externally hosted image should be here but it was not working when we last tested it.
Since this amplifier inverts, the primary phase relationship is -180 degrees. However, at low frequencies the phase is slightly less than -180 and at higher frequencies it is slightly more. What this means is that the high frequencies effectively get to the output “faster” than the low frequencies. Practically speaking, however, these phase shifts are probably audibly unnoticeable. But, I don’t have data to back up this claim.
But, no matter what we do with the regulator we can’t get any better than this, unless we can actually improve the phase shifts somehow.
Starting as before, if we insert our basic RC regulator and re-plot the phase shifts caused by the amp we get this:
An externally hosted image should be here but it was not working when we last tested it.
In other words, the RC regulator has absolutely no effect on the phase behavior of the amplifier.
So, how about the series regulator? Taking the “raw” regulator with the 470n damping cap we get:
An externally hosted image should be here but it was not working when we last tested it.
Absolutely no effect. In fact, there is no discernable effect for any of the regulators that were used previously. So, for this very simple analysis, we can see that, from a phase point of view, all of these regulators are the same. Meaning that the reactive behavior of the regulator on the circuit is irrelevant. This is primarily due to the fact that the resistive component is also irrelevant and, hence, the phase of the delta-V just doesn’t show up.
I should also add that the phase of the current in all of these examples is identical to the voltage phase.
And, I should give the caveats that this is simulation, these are perfect Rs and Cs, the circuit is pure class A, and the line side of the PS is not being taken into account.
You may ask, is there every any change to the phase relationship? There can be, depending on the exact time constants of the PS and the circuit. For example, here is a result for the RC regulator where the cap is set to 1u instead of 330u.
An externally hosted image should be here but it was not working when we last tested it.
The voltage phase shift is now 180 degrees advanced and the difference is almost 15 degrees at 20Hz. In the case where C=330u, the phase shift at 100Hz is almost exactly -180d. I’ve also plotted the current phase shift in red. Even so, these phase differences are unlikely to be audibly noticeable.
Now, when chokes and big caps are involved in high current situations (for example, OTL output PS), YMMV. Maybe we can look at this later.
In the meantime, I think I've figured out how to account for most of the pre-regulator PS. If this works, I'll be back with more. But, please let me know how we're doing.
first RC network
Runeight, sorry I am late in seeing your first post.
The first analysis is way off at higher frequencies. The reason being that a typical electrolytic capacitor of that size will series resonate at some point, probably between, say, 7khz and 30khz. Then the cap will become an inductor (simplification) as the frequency rises. So the impedance will start rising above the series resonant point. The same with the cathode bypass electrolytic capacitor.
The frequency response will also change at both low and high frequencies as the reactance of the cap rises.
Just a thought.
ps. As you study the subject at hand, you will see other implications that need addressing.
Runeight, sorry I am late in seeing your first post.
The first analysis is way off at higher frequencies. The reason being that a typical electrolytic capacitor of that size will series resonate at some point, probably between, say, 7khz and 30khz. Then the cap will become an inductor (simplification) as the frequency rises. So the impedance will start rising above the series resonant point. The same with the cathode bypass electrolytic capacitor.
The frequency response will also change at both low and high frequencies as the reactance of the cap rises.
Just a thought.
ps. As you study the subject at hand, you will see other implications that need addressing.
reasons
By passing electrolytic caps must be done carefully so as not to set up a parallel resonant circuit. Even a little tweek in the higher registers can affect the sonics.
I did some testing on a phono RIAA section some years back by changing the -1db response from 200khz to 150khz, which made a noticeable sonic difference.
If electrolytic caps are that bad, do you really want one providing an AC path to ground so close to the signal tube? There are others too, but I will let you figure them out.
By passing electrolytic caps must be done carefully so as not to set up a parallel resonant circuit. Even a little tweek in the higher registers can affect the sonics.
I did some testing on a phono RIAA section some years back by changing the -1db response from 200khz to 150khz, which made a noticeable sonic difference.
If electrolytic caps are that bad, do you really want one providing an AC path to ground so close to the signal tube? There are others too, but I will let you figure them out.
I can add an ESR to the caps.
Here is one good link that I found:
Capacitor Model
But, I haven't found any means of computing the various lumped circuit elements here. If anyone can give me some decent values, I can make this model work.
In the meantime, I'll keep looking.
Next, however, I was planning to try to include some line side effects. I'm still noodling on this.
Here is one good link that I found:
Capacitor Model
But, I haven't found any means of computing the various lumped circuit elements here. If anyone can give me some decent values, I can make this model work.
In the meantime, I'll keep looking.
Next, however, I was planning to try to include some line side effects. I'm still noodling on this.
The only measured values I could find for an electrolytic were for an Elna RSH 1000u 10V:
C = 765uF
ESR = 0.158R
L = 41.3n
These values were measured by a friend on a rather swanky HP component analyser. I have rather more data for 100n plastic capacitors. When I've been able to find data for electrolytics, the ESR has always been higher than expected. For recently made components intended for switch-mode PSUs, L has been lower.
C = 765uF
ESR = 0.158R
L = 41.3n
These values were measured by a friend on a rather swanky HP component analyser. I have rather more data for 100n plastic capacitors. When I've been able to find data for electrolytics, the ESR has always been higher than expected. For recently made components intended for switch-mode PSUs, L has been lower.
Hi,
Hmmm...I guess that just like myself you know the answer to that one already....
Still, lets have a third party confirm it and have it out of harms way.
In that way we could both fall back on an independent measurement....
Sorry if it sounds like abuse to any of you, it really isn't.
Cheers,
Hmmm...I guess that just like myself you know the answer to that one already....
Still, lets have a third party confirm it and have it out of harms way.
In that way we could both fall back on an independent measurement....
Sorry if it sounds like abuse to any of you, it really isn't.
Cheers,
Electrolytics Models
Hey guys, I found that modern electrolytics are modeled using a series inductance/resistance and a ladder of RC pairs. Kemet actually has a nice program that you can download to see the frequency dependent ESR of some of their caps. Here is what one of the models looks like:
I've simulated this and I don't find any resonances until up past 1GHz. This is a 330u low voltage cap. So I'm figuring that the high voltage electrolytics have different parameters. If anybody can provide me with more help here, I'd appreciate it.
More regulator analysis coming. . . . . . . .
Hey guys, I found that modern electrolytics are modeled using a series inductance/resistance and a ladder of RC pairs. Kemet actually has a nice program that you can download to see the frequency dependent ESR of some of their caps. Here is what one of the models looks like:
An externally hosted image should be here but it was not working when we last tested it.
I've simulated this and I don't find any resonances until up past 1GHz. This is a 330u low voltage cap. So I'm figuring that the high voltage electrolytics have different parameters. If anybody can provide me with more help here, I'd appreciate it.
More regulator analysis coming. . . . . . . .
That's a remarkably low inductance, and much lower than my measured value. Rule of thumb is that the inductance of a straight piece of wire is 0.5nH/cm. I also find that ESR hard to believe.
In isolation, individual components look reasonably harmless. Due to falling open loop gain, regulators have a rising output impedance (as you've shown). Another way of viewing this is to say that they look like perfect regulators in series with an inductor. Add a capacitor across the output, and you have a resonance.
In isolation, individual components look reasonably harmless. Due to falling open loop gain, regulators have a rising output impedance (as you've shown). Another way of viewing this is to say that they look like perfect regulators in series with an inductor. Add a capacitor across the output, and you have a resonance.
The inductance seems low to me too. I'm assuming that this is a small capacitor.
However, if I simply use a series LRC model for the cap, here's what I have to do.
Assuming C=330u and R=.01 (low, I know), then L=3u to get a series resonance around 7K (where Steve put the earliest onset). This seems like an impossibly large inductance.
Increasing R, of course, will slowly turn this into a high pass filter.
If I use the parameters you provided above, I still don't get any visible resonances between 10Hz and 1MHz. I do get a very nice high pass filter that turns over at about 300KHz, but hardly effects the audio spectrum at all. Is this the heart of the issue?
I'm assuming that other folks have measured the series resonance features of aluminum electrolytics. Perhaps you all could point me to where to look.
However, if I simply use a series LRC model for the cap, here's what I have to do.
Assuming C=330u and R=.01 (low, I know), then L=3u to get a series resonance around 7K (where Steve put the earliest onset). This seems like an impossibly large inductance.
Increasing R, of course, will slowly turn this into a high pass filter.
If I use the parameters you provided above, I still don't get any visible resonances between 10Hz and 1MHz. I do get a very nice high pass filter that turns over at about 300KHz, but hardly effects the audio spectrum at all. Is this the heart of the issue?
I'm assuming that other folks have measured the series resonance features of aluminum electrolytics. Perhaps you all could point me to where to look.
I find it oversimplified
Different capacitors and their sonic signatures, seems to me, to be different because the models are all different.
For instance, my source informs me that Hovlands (film cap) is complicated. The lead attachment is at one end of the foil and then the foil is rolled. Thus one has an infinite number of resistors and inductors and capacitors instead of one inductor and an infinite number of RCs. Electrolytics, I think are similar.
Infinis, on the other extreme, have the "metalized foil" deposited and the lead is attachmented on the entire edge of the whole roll. Thus the inductance is minimized and the resistance is in the the "metalized foil" itself from one edge to the other, which is where the wire is connected. I hope I explained it clearly.
"The Multicap" has ten sections of similar capacitance, I believe, yet it sounds different than the other two. I think each section is terminated at one point on the foil, so you have ten little "hovlands" so to speak paralleled in one package.
They sure do sound different, yet the specs seem to be similar.
ps. If you like, you might read "Picking Capacitors" by Walter Jung and Richard Marsh. Provides some interesting insights. I also have an article by a chemist on capacitor dielectrics.
pss. Remember, some electrolytics only terminated the foils at one or a few points, so you have alot turns.
Different capacitors and their sonic signatures, seems to me, to be different because the models are all different.
For instance, my source informs me that Hovlands (film cap) is complicated. The lead attachment is at one end of the foil and then the foil is rolled. Thus one has an infinite number of resistors and inductors and capacitors instead of one inductor and an infinite number of RCs. Electrolytics, I think are similar.
Infinis, on the other extreme, have the "metalized foil" deposited and the lead is attachmented on the entire edge of the whole roll. Thus the inductance is minimized and the resistance is in the the "metalized foil" itself from one edge to the other, which is where the wire is connected. I hope I explained it clearly.
"The Multicap" has ten sections of similar capacitance, I believe, yet it sounds different than the other two. I think each section is terminated at one point on the foil, so you have ten little "hovlands" so to speak paralleled in one package.
They sure do sound different, yet the specs seem to be similar.
ps. If you like, you might read "Picking Capacitors" by Walter Jung and Richard Marsh. Provides some interesting insights. I also have an article by a chemist on capacitor dielectrics.
pss. Remember, some electrolytics only terminated the foils at one or a few points, so you have alot turns.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.