|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
65.202.22.70
In Reply to: RE: PSUD Playtime posted by 91derlust on May 27, 2015 at 02:40:51
Step response and recovery are most important for amps near class B, where current jumps with drive. A class A amp draws a nearly constant average current, though more output capacitance will help bass somewhat.
For a first cut at the ripple requirement, just multiply the ripple you can live with at the speaker (maybe 1 mV) times the output transformer ratio (about 20 for a 2A3 output). So try for less than 20 mV. Pentode mode and/or push-pull can live with much more than that.
Follow Ups:
There is a lot of emphasis on this forum about "recovery" in power supplies for class A amplifiers. I think there might be 2-3 people who drone on about this, the rest of us don't particularly agree with what they have to say.
In a class AB push-pull amp, when you move out of class A, regulation becomes pretty important and ripple is not very important. In a class A SET, average current draw is nearly constant (assuming your output transformer has a decent amount of inductance, and you aren't driving your output tube into A2), but power supply ripple becomes audible (again, assuming no feedback is used to mitigate these effects).
I don't seem to be able to reconcile a couple of things:
LcritCLC supplies are often discussed as being suited for SET amps due to their superior regulation. But if the current drawn/ load does not vary, then there should be no change of voltage across the PS resistances... why would regulation be important?
Also, at least one EE and another enthusiast (no, not Jeff or Dennis) have measured the current draw changing during playback. Maybe the amps were running into clipping or some such or maybe they were measuring the wrong things, but I think the guys performing the measurements would be aware of that.
Can someone clarify my misunderstanding?
Regards,
91.
"Also, at least one EE and another enthusiast... have measured the current draw changing during playback."That's kinda funny. I'm pretty sure that tens of thousands of engineers have seen this, probably going back to at least the '30s. Anyway, as Paul pointed out, it can be caused by non-linearities in the output tube. Another potential cause is the fact that source material can contain significant content at frequencies to 20Hz and below. Most mechanical ammeters will respond to those frequencies if installed in series with the anode or cathode of the tube. Meter fluctuations in this regard do not indicate changing average current, only that the average must be calculated over a longer time. A third possible cause for changing meter readings is simply the fact that music isn't sinusoidal. Various patterns can cause ammeters to respond differently.
As an aside, this attempt by the output stage to "modulate" the power supply at the frequency of the source material is the reason the power supply must be designed to maintain the lowest possible output Z over frequency. Lcrit regulation, a longer-term characteristic, is only peripherally related to this. I personally consider Lcrit regulation to be one of the least necessary requirements to designing a high quality SET. In fact, if not done exactly right, it can seriously degrade sound quality. Chokes are essentially current sources, the antithesis of the voltage source that's needed to maintain steady B+ against instantaneous current changes.
--------------------------
Buy Chinese. Bury freedom.
Edits: 05/27/15 05/27/15 05/27/15 05/27/15
"That's kinda funny. I'm pretty sure that tens of thousands of engineers have seen this, probably going back to at least the '30s"Glad you had a chuckle. I'm not an EE - am I expected to know this, or is asking questions okay?
"Another potential cause is the fact that source material can contain significant content at frequencies to 20Hz and below. Most mechanical ammeters will respond to those frequencies if installed in series with the anode or cathode of the tube. Meter fluctuations in this regard do not indicate changing average current, only that the average must be calculated over a longer time. A third possible cause for changing meter readings is simply the fact that music isn't sinusoidal. Various patterns can cause ammeters to respond differently."
So, the fluctuations do occur in SET amps playing music, but should average over time. Are the fluctuations relevant then, or should a supply provide enough energy to handle these over the time taken for the averaging to occur? Or both?
"I personally consider Lcrit regulation to be one of the least necessary requirements to designing a high quality SET. In fact, if not done exactly right, it can seriously degrade sound quality."
Regarding the PS, care to provide some hints? I am listening!
Thanks TK.
Regards,
91.
Edits: 05/28/15
"Glad you had a chuckle. I'm not an EE - am I expected to know this, or is asking questions okay?"
I wasn't laughing at you, it just reminded me that we have a few members here who invented hi-fi. They're constantly "discovering" micro-minutiae that the rest of us should care about, often just by listening. :)
"So, the fluctuations do occur in SET amps playing music, but should average over time. Are the fluctuations relevant then, or should a supply provide enough energy to handle these over the time taken for the averaging to occur? Or both?"
The fluctuations must be averaged over time for the purpose of determining whether average current is really changing. However, they represent *instantaneous* changes, because they mirror what the source material is doing. In that respect, the power supply must be designed so that those instantaneous current changes don't create instantaneous voltage changes on the B+ rail. IOW, the PS must be able to hold it's voltage constant despite changes in average current, as well as the instantaneous current changes that result from amplifying music. The former is where regulation comes in. However, in a SET, average current changes are really miniscule, so I don't worry much about power supply regulation. The instantaneous current changes are a big deal though. Luckily, the supply only needs to exhibit a low Z output over frequency to deal with that. This is a reasonably simple goal to accomplish, and we have tools now that weren't readily available 20 years ago. The supply can be modeled in SPICE, and the simulation will let the designer "look" back into the supply to see how it will react to the amplifier's instantaneous current demands. LTspice (also known as Switcher CAD) is free to download online.
--------------------------
Buy Chinese. Bury freedom.
Thanks for the explanation TK - it is really helpful to me. I think I need to get to know SPICE, at least so I can do what you suggest.
Yeah, I suspected you were not laughing at me. I hear what you are saying regarding the invention of Hi-Fi.
Thanks again.
91.
It's not a misunderstanding so much a it's a gray area.
SETS have very little variation in average current as a function of power output, but only if the tube is fairly linear (by which I mean constant-mu over the operating region) AND the load impedance is high enough for linear behavior. If either or both of those criteria are dispensed with, then the average current will fluctuate significantly with the signal envelope. Envelope modulation is highly audible even though there are no widely accepted measures of it, so power supply regulation becomes important at least over envelope-power time scales.
I have never seen any good data on what those time scales may be. In the early solid-state days there was a big fashion for huge capacitor banks in the power supply to extend that time scale. Enthusiasm for that approach faded after a decade or so, so I'm not sure that tells us anything useful. That leaves good regulation or a regulator as reliable options - or design an amp with linear behavior in the first place!
Thank you very much for the explanation Paul - even I can glean some understanding. I am sure others also benefit from your generosity and helpfulness.
Regards,
91.
nada aqui
--------------------------
Buy Chinese. Bury freedom.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: