On Fri, 06 Jun 2008 22:41:08 -0500, flipper <flipper@[EMAIL PROTECTED]
> wrote:
.....snip.....
>
>> That's probably because I have quite a bit of overall
>>negative feedback,
>
>How much now, and before the PFB increase?
>
The global negative feedback is set by "beta = B", at mid-band its
set by the resistor (10K) that goes from the first cathode to the
output terminal, and another resistor (470 ohms) that goes from the
1st cathode to ground. That sets B=0.0448 . This doesn't change with +
feedback.
The open loop gain before +feedback was about 152. The open loop
gain with +feedback is approximately 267.
Using the closed loop gain formula (G=A/(1+BA)). I calculate a
closed loop gain of 20. My measured closed loop gain is 22.
BA, the "loopgain" (B=beta-the feedback network ratio, and A=open
loop gain) is 6.8 with no +feedback, and about 12 with +feedback. This
is the number that reduces all the extraneous junk that is produced
inside the feedback loop.
>Also, I'm curious. What triodes are you using and how much gain are
>you getting now?
>
Before +feedback, the 1st stage (5751) is about 39. After +feedback
it's about 65. 2nd stage (6SN7) shows a gain of 14, with or without +
feedback.
>> and the phase ****fts from the zero's (coupling
>>capacitors) were made worse by lowering the one stage's coupling
>>capacitor. If I'm not mistaken, the output transformers primary
>>inductance is effectively a "zero" (an "s" on the bottom of the
>>transfer function ?, or inductors in parallel act like caps in
>>series). So you'd have 3 zero's in the feedback loop, even if you
>>discount the power supply. That works out to almost 270 degrees of
>>phase ****ft... if you have any degree of gain at those low
>>frequencies, the thing is going to oscillate.
>> By decreasing a coupling capacitor, the phase ****ft will get worse
>>at a given low frequency, and make stability worse (for the same
>>amount of feedback).
>
>Well, except it isn't passing the signal through, unless you have so
>much PFB it's still got gain down there.
>
>> So by decreasing that capacitor in the positive feedback loop, the
>>local loop became more stable, but the global loop became more
>>UNstable!
>> Since the transformer is pretty well a constant, overall low
>>frequency stability would improve by increasing the values of coupling
>>capacitors, until the transformer "dominates" the "zero's".
>
>Hmm. Well, I'm no math whiz when it comes to poles but having the amp
>trying to drive the OPT lower than it can pass seems backwards to me.
>
>That will cause large voltage swings in the amp, as it tries to
>'force' things through, and it often causes HF bursts. Speaking of
>which, the HF bursts can cause motor boating too although I presume
>you're not having those or else you'd see them.
>
>Where is the LF it's 'amplifying' coming from? Strikes me that it
>might still be B+ filtering and it's just more sensitive now that gain
>is higher.
>
At the upper end of the amplifier's response you have to compensate
the feedback loop by adding a capacitor in parallel across the
resistor (between cathode and output). At high frequencies you have
added an extra "zero" in the B(Beta) network to compensate for the
poles in the amplifier (A). In an AC amp we suffer from "zeros" at
low frequencies. So..... we could add a pole in the compensation (B)
network, similiar to what we did at the high frequencies.
I think you are correct about the B+ being the problem.... I think
the supply is adding "zero" behaviour at the low frequencies, and the
larger A (open loop gain) is making things worse. BA must never have a
value more than 1 when its phase is 180 degrees, so I have a number of
options...
1- regulate the B+ (eliminates power supply "zero" behaviour,
reduces phase ****ft) .... a zener maybe?
2- lower the amp gain (A) - no local +FB - keeps BA from getting
close to 1, but increases distortion
3- put a pole in the feedback network, thus ****fting the offending
phases the other way.
If you were brought up using RDH4, you would be using attenuation
skirts (multiples of 6db/octave) instead of the more abstract poles
and zeros. They work out much the same, except I find poles and zeros
much easier to deal with - both mathematically and from an overall
"systems" approach. I notice a LOT of arguments about NFB, and it
seems that most of the fuss is over terminology, and the language of
the concepts. My feedback theory isn't all that hot.... but I've
managed to keep it simple. I generally always do a few quick
calculations, to avoid all that shunt-series confusion. I'd sooner
trust the math than my memory!
Time to play some more with the amp!
-Paul
|
