|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
173.174.66.72
In Reply to: RE: apparently amplifier tests conducted by the military posted by elblanco on March 17, 2015 at 20:46:45
Some interesting tidbits in that file...
"Practically all Williamson amplifiers tested showed a tendency to "ring" at about two or three cycles. This tendency was eliminated in all cases by changing the values of the coupling capacitance in the grid circuit of the output tubes from 0.25 mfd to about 0.02 mfd. The low-frequency response was not in any case seriously affected by this change"
Don't know if I've ever looked for ringing at such a low frequency. I wonder if it's an anomaly of the power supply...
--------------------------
Buy Chinese. Bury freedom.
Follow Ups:
Has everyone gotten a copy of thus report? It was done in the early-mid 50's in an effort to buy some small amount of hardware (for Navy sonar, I think) off-the-shelf.
I requested it in another life when I clearance high enough to request a lot of cool stuff, although I had no "need to know" for this report. I was working on Star Wars related sonar stuff. I think the only reason I got this report was because it was 30 years or so old at the time and no one had ever bothered to lower or de-classify it.
There was a lot of joking about me finding anything tube-related to request!
It really is the way to construct and author such a report. The proposed uses for the equipment were very clear if one were to read between the lines back then, thus the comments on stability, ringing and bandwith.
Peace,
DrRick
I just looked at the link. That's the report I remember.
Back then I was amazed to find it existed.
Peace,
DrRick
The ringing is caused by the large amount of global negative feedback, together with inadequate attention to the phase response and stability margin. To be fair, this was the early days when feedback was still new to most people. The extreme form of this instability is called "motorboating". :^)
The real problem is that the output transformer is inside the loop, and its inductance is a strong function of signal frequency and level. I have always guessed that this is the reason the Dynaco Stereo 70 has some feedback from the output tube plates, reducing from the transformer secondary while maintaining large amounts of total feedback.
That does sound like a plausible explanation. I checked the HK A500 that happens to be on my bench at the moment, but was unable to detect any sign of this phenomenon. Maybe the relatively low Z of the voltage doubler and the bandwidth of the OPTs are factors. I'd like to find an amplifier that does this so I can confirm for myself that the cause lies with NFB functionality.
--------------------------
Buy Chinese. Bury freedom.
It's been long time, I might be wrong, but IIRC Williamson himself mentioned this in the first article.
I just made another measurement in preparation for stowing this project away temporarily. The output section of the A500 has only 13dB NFB applied. That's considerably less than many amplifiers, and maybe the reason the effect being discussed doesn't appear in this model.
--------------------------
Buy Chinese. Bury freedom.
Some of that ringing can be corrected by changing the feedback bypass caps in value and I have even corrected it by using better filter caps in the power supply.When you lower the impedance of the power supply by using better caps,it has an effect on the higher frequency sq waves.
Nobody ever went broke underestimating the intelligence of the American public
H. L. Mencken
...I've noticed this phenomenon in a number of vintage amp designs, notably the various Williamson implementations. It manifests as an intermittent to continuous low frequency "bouncing" of the output at a couple of Hz with an amplitude of maybe 1/2V P-P. Easily noticible as woofer pumping. It can be induced by power line perturbations, large signal transients or even just physically pushing on the woofer which injects a low frequency signal into the output that the feedback circuit attempts to correct. While the PS may indeed be a factor, I was not able to significantly reduce the ringing by adding practically large amounts of decoupling capacitance betw stages. The cure for the designs I've worked with involved modifying the time constants of the coupling of two cascaded stages within the global feedback loop.
Typically, the ratio of time constants of successive stages in the unstable vintage designs was about 5:1. A rule of thumb even back then was that the ratio should be a minimum of 10:1. I've found that 20:1 to 25:1 to be most effective w/o significant affect on LF response or distortion. Some stable designs of the era achieved the staggering of time constants by using large coupling caps at g1 of the finals and small caps at g1 of the drivers. I like to reverse this with relatively small caps at the finals and large caps at the drivers. This seems to reduce the effects of output stage blocking under heavy drive conditions.
Steve, If following the 10:1 rule of thumb, this means that the time constant of the PS to the input and driver stages should be 10 times that of the PS to the output stage, or what? Probably this is obvious to the discussants, but not to me.
However, this is a great thread for learning from. Thanks.
The time constant rule of thumb applies to interstage coupling when those stages are contained within an overall feedback loop. The basic idea behind this is to stagger the RC time constants (or, effectively the Fc) of cascaded gain stages to avoid stacking of phase shift at or near one frequency. IMO, the power supply is not directly involved in this although a high impedance supply may incite the ringing of an inherently unstable design due to susceptibility to outside variation at low frequencies.
I'll look for a schmatic I can edit into here where the components of interest can be highlighted for clarity.
Thanks, Steve. Glad I asked the question, because it appears I was on the wrong track to understanding this. Someone else mentioned "motor-boating", which I think can be a consequence of PS filtering. That set me on the wrong path.
See attached image for time constants under discussion. In this example (Heath W-3AM) the time constant ratio is ~ 5:1.
An interesting aside is that when considered at the base units, the unit of the RC product is the second (s).
Thanks for going to all that trouble to illustrate the principle. I get it now.
Steve
The W4s are probably one of the worst offenders in that regard because of the iron.
Nobody ever went broke underestimating the intelligence of the American public
H. L. Mencken
Sadly, this is true for the W4.
Peace,
DrRick
He said in Valve Amplifiers that it was a waste of power to have the amp slowly oscillating at 0.5Hz or so and firmed up the power supply to remove it. He said that most amps that he measured did oscillate at very low freq. It was a surprise to me.
While regulation is desirable, it's not practical in vintage designs. In new/DIY designs, regulation may be possible but it adds expense and size, esp at higher power levels. Real world implementation of regulation typically involves solid state devices that in my experience are not as reliable as the tube circuits they supply....kinda brute force too. I prefer circuit/component solutions if possible. As always YMMV.
I threw it out there because I do not regulate B+ at all times and was wondering what happened when you do. I have only regulated B+ on drivers not power tubes.
It's not all that practical in new designs either, unless you stay all class A, which sort of defeats the need in the first place.
The demands placed on such a regulator in a class ab amp are pretty nuts, but at the same time, a choke input filter will get you close.
Hello,
Can you elaborate on this?
My undertsanding might be wrong, but I always thought a class A amp was where there was the greatest benefits to have a regulated B+.
Since in class A the PS is in the signal loop, the lower Z it has, the better. Of course, as Steve posted below, the regulator's error amp is an other amp that follows the supplied stage's signal demand.
Thakns,
Joris
In a class A amp, provided the output transformer has sufficient inductance, you aren't going to see huge deviations in power supply demand.
If you have an AB push-pull amp, when you leave class A and move towards class B, the current draw from the power supply changes drastically. (Regulation helps here)
This isn't to say that a fully regulated class A amp is a bad thing, it's actually easier to do and will sound different, but it's not solving the obvious issue that class AB amps have.
I think a good case could be made that an active regulator is a feedback amp that has to operate down to DC with great stability compared to an audio amp that only has to run down to 10Hz or so. Regulating an amp's supply is sorta like cascading two amp systems.
that and I have yet to need more than 3 stages, including the linestage. usual practice is two stages, plus a linestage. just no need for even a 5-20 style 3-stage.
cheers,
Douglas
Friend, I would not hurt thee for the world...but thou art standing where I am about to shoot.
Interesting Steve.
i might try this my self as I have one EL34 PP that do the same thing's in the LF . it sound great but do have some woofers pumping not much but you can see it.
thanks for the post
LT
It may not be possible to use small caps. at the I/Ps of the "finals", due to grid to ground resistance limits. This is especially true, when "fixed" bias is employed.
As for blocking, I recall reading where 1 fellow was going to implement "PowerDrive" using ZVN0545As, even though positive grid current was not permitted. A rather high valued grid stopper prevented significant g1 current from flowing.
Eli D.
...Ex: Heath W-3AM Original circuit uses 0.01uF/470K at input to driver (6SN7) for fc~34Hz and 0.25uF/100K at input to finals (5881 w/1K "stoppers") for fc~6.4Hz. Ratio of tc~5. It bounced at LF. Revised circuit uses 0.47uF/470K at input to driver for fc~0.72Hz and 0.1uf/100K/1K at finals for fc~16Hz. Ratio of tc~22. No bounce and nearly identical LF distortion @-3dB (~10W) of max rating with 20Hz THD easily meeting original spec of ~0.3%.
Interestingly, the revised values are exactly those of the Craftsmen C500A, an inherently stable triode Williamson design and they're similar to Dave Gilespie's stability enhancing revisions to the Heath W4, an ultralinear Williamson using 6SN7s and 5881s but with a lower quality OPT than the W3.
"0.1uf/100K/1K at finals for fc~16Hz"
This seems preferable to me also, but it does raise the issue of low frequency phase shift. I'm skeptical as to whether such shifts are audible, but many DIY audiophiles do whatever they can to minimize the effect. There's definitely a conflict of interest in resolving the problem.
--------------------------
Buy Chinese. Bury freedom.
...can be significantly reduced thru judicious use of negative feedback. See attached plot from D.T.N Williamson's article in Wireless World. I'd also predict LF "ringing" @ 1-2Hz based on this plot. The circuit this plot is derived from has a tc ratio of about 1:1, much less than the 10:1 rule of thumb.
I love the way it goes through a 180 deg shift at about 2 Hz. I can practically hear the OPT screaming for mercy... :)
--------------------------
Buy Chinese. Bury freedom.
-but in general if the power supply has timing constants that are higher than that of the amplifier circuit, there will be some instability in the amp (LF resonance) as a result.
These days we are more likely to fix that with bigger filter caps, but actually cutting off a little higher with a smaller coupling cap is the better move IMO as the cap will have less effect otherwise on the sound- its easier to build a good small cap than a good big cap.
"its easier to build a good small cap than a good big cap"
And in addition, smaller coupling caps can produce a much shorter overload recovery time. I'm always careful not to use caps that are larger than necessary for this purpose.
--------------------------
Buy Chinese. Bury freedom.
+1
Peace,
DrRick
~!
The Mind has No Firewall~ U.S. Army War College.
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: