I took your advice, Partick, and empirically selected the best RC network,
observing transient pulse response on the oscilloscope (with the NFB
temporarily disabled).
The best damping (of the OPT leakage + speaker inductance) took place with
R=15K and C=4700pF -- no spikes, no humps, just nice nearly exponential
drooping of the tops of the pulses. Of course, the Rola speaker will not
be
changed, so no more fiddling.
Your assessment of the transformer was on the spot. It is crap: 0.7ohm
secondary DC resistance, 330ohms primary resistance. Total losses referred
to primary 1,5Kohm (series). That was measured on the RLC meter @[EMAIL PROTECTED]
1KHz
with
the secondary shorted.
Overall NFB will be about 12dB. It is applied to the cathode of the first
12AX7 stage.
To provide stability at high frequencies I use the following "trick",
which
I can share.
I drop a small capacitor from the plate of the output tube to its grid.
About 10pF, but it depends. It works exactly like a classic compensation
capacitor in old op-amps, like LM301, creating one dominant pole from two
staggered ones: grid and plate.
Regards,
Alex
"Patrick Turner" <info@[EMAIL PROTECTED]
> wrote in message
news:481AF4A8.5832CC98@[EMAIL PROTECTED]
>
>
> Alex wrote:
> >
> > Thanks, Partick.
> >
> > Your reply is constructive and informative, unlike the other one.
>
> Sometimes my "other one" can be terse, and abrupt, tactless, and
> inclined
> to hurt everyon'e feelings, but hey, my tolerance and patience is
> strained at times....
> >
> > My case is not a Hi-Fi amplifier, but a radio with a single-ended
stage.
OPT
> > is quite lousy. It is designed for Rl=7.5K (for 6M5, EL80 or 6F6 tube
at
> > 250V). Its leakage inductance, referred to primary is as large as
165mH --
> > the windings are not interleaved.
>
> Yes, the OPT is dead awful.
>
> What I said about providing a resistance load in PP amps is also valid
> for SE types.
> The Zobel is to ensure the tube has a load at HF even though the tube
> de-couples from the secondary
> because of the increasing reactance of leakage inductance.
> Its very much like using a Zobel across a speaker to neutralize
> the series inductance of a voice coil.
> You *want* the impedance of the speaker to be resistive if possible,
> so that the crossover series L will indeed filter out HF from a bass
> speaker.
>
> >
> > By briefly looking at impedances, it looks like the amplifier will not
be
> > able to efficiently reproduce more than 8kHz anyway (at 1/2 maximum
power
> > level, without NFB).
>
> That may be plenty, if its an AM radio.
> >
> > So if I select Zobel C=3300pF and R=7.5K will sort of "tune" the
primary
to
> > 8KHz, maintaining nearly active impedance all the way to ultrasound.
Load
> > curve will be rather more linear, than elliptical, there will be no
> > pronounced parallel resonance and the distortion will not be soaring
at
high
> > frequencies.
> >
> > Is that correct?
>
> That is all about right.
>
> You may find it difficult to apply more than 15dB of NFB.
> 12dB might be enough though, but you wouldn't bother trying to extend
> HF response with pure resistance loads to much beyond 10kHz, -3dB point.
>
> In the case you have, the reactance of the C can be about equal to
> reactance
> of leakage inductance and series R can be about 1.41 x ZR or ZC,
> roughly...
>
> 12dB NFB around a pentode is usually enough to reduce the Rout
> of the amp including the OPT winding resistance so that the
> end result gives you an amp which has Rout lower than a triode with a
> low loss OPT.
>
> Its not uncommon to find radio OPTs which have 25% winding losses!
>
> If the speaker is in the radio cabinet with the radio chassis it'll be
> unlikely
> any other type of load will be used, so you won't have to fiddle
> endlessly around
> to "critically damp" the circuit after FB is applied to ensure it is
> stable with any value of pure C loading
> because such a load won't ever be connected.
>
> Sometimes the S winding is wound onto a core first with P winding on
> after.
> If there is a gap between core and P winding available, say 1.5mm, maybe
> you can sneak in
> another secondary with the same turns as the existing, doesn't matter if
> the wire is thinner,
> as long as the added winding has the same voltage as the existing S
> winding.
> This will reduce LL from 165mH to maybe 60mH, and reduce winding
> resistance losses, which are usually
> highest in the S windings on radio OPT.
>
> Its a pain to wind maybe 120 turns on by threading wire around and
> around, but I did it once
> with a pair of Rola OPT meant for 6BQ5, and then 20dB global NFB was
> fine.
>
> The sound was glorious with sensitive full range "ceiling" speakers.
>
> I had to be very dedicated.....
>
> Patrick Turner.
>
>
>
> >
> > Regards,
> > Alex
> >
> > "Patrick Turner" <info@[EMAIL PROTECTED]
> wrote in message
> > news:481AD9C5.95437630@[EMAIL PROTECTED]
> > >
> > >
> > > Phil Allison wrote:
> > > >
> > > > "Alex"
> > > > >
> > > > > I noticed that in most cases a capacitor or a RC circuit is
connected
> > > > > across the primary of a tube OPT. It is apparently to compensate
for
> > > > > the leakage inductance of the transformer.
> > > >
> > > > ** Partly for that PLUS the fact that most speakers have a
sharply
> > rising
> > > > impedance curve at high and supersonic frequencies. A nominal 8
ohm
> > > > speaker at 250 Hz can easily be 100 ohms at 50 kHz.
> > > >
> > > > > However, I am not sure how to correctly calculate this RC
circuit.
> > > >
> > > > ** You are not alone ....
> > > >
> > > > > If I know my load impedance Rl,
> > > >
> > > > ** No you don't.
> > > >
> > > > You just know the " nominal " impedance = useless.
> > > >
> > > > What you need to make sure of is the amp does not break into HF
> > > > oscillations ( parasitic or continuous) when there is no load OR
when
> > > > under test with simulated ( or real ) speaker loads.
> > > >
> > > > The vast majority of tube amps have no need of such a compensation
> > > > etwork - as the OT has little leakage and the NFB reduces to low
value
> > at
> > > > supersonic frequencies.
> > > >
> > > > ...... Phil
> > >
> > > I don't know how many tube amps you have designed and or built or
> > > repaired
> > > but I've lost count. Many tube amps will oscillate at HF when
connected
> > > to
> > > a capacitor load.
> > >
> > > In order to gain unconditional stability in very many tube amps and
some
> > > tolerance
> > > of capacitor loads, or having no load at all, or having an inductive
> > > load wuth high Z
> > > at HF, is is VERY NECESSARY to place what is called a Zobel network
of
> > > R+C in series
> > > across one or more positions in the tube amp.
> > >
> > > The Zobel is more likely to be needed in an Ultralinear or beam
tetrode
> > > or pentode
> > > output stage where the gain rises as the load on the tube rises
because
> > > of the leakage inductance.
> > >
> > > The Zobel is less likely to be of any use in a triode output stage.
> > >
> > > Transformer series and parallel resonances between stray lumped
shunt
C
> > > and leakage inductance
> > > also confounds the designer, and the Zobel can act to damp the
resonance
> > > and the
> > > adverse loading they cause at the resonant F.
> > >
> > > So just what value for R&C are to be used in an output stage?
> > >
> > > Let us assume the amp is set up with an output tranny to power 5
ohms,
> > > but present
> > > a load to the tubes of 5k anode to anode.
> > >
> > > In most cases I have had to tame, generally I have tried TWO Zobel
> > > networks,
> > > one across each half primary of the OPT, ie, from CT to each anode
> > > connection.
> > > Where 5k is the a-a load, the minimum R value is 5k / 4 = 1.25k
because
> > > that's the class B load
> > > when the amp works in class AB.
> > > The next standard R value above is about right, say 1k5.
> > >
> > > The value of C should be high enough to not adversely load down the
amp
> > > too much at too low a frequency. Therefore I would use C which has
equal
> > > reactance to
> > > RLa-a / 4 at say 100kHz as a starting value. 0.001uF could be about
> > > correct.
> > >
> > > At 20kHz, the two x ( 0.001uF + 1k5 ) networks has Z = 16k a-a
approx,
> > > and this will not
> > > reduce the response into a resistance load very much.
> > >
> > > A typical amount of leakage inductance of a poor quality 5k OPT
coulod
> > > be 50mH in series with the
> > > 5k load. This is normal for very many tube amps.
> > > At 100kHz, 50mH has reactance = 32k, a huge amount of series
reactive
> > > impedance
> > > effectively in series with the load seen by the anodes.
> > >
> > > If a square wave with fast rise time is used to test the amp, and
the
> > > anode
> > > signals viewed, there can be a huge amount of overshoot and ringing
at
> > > ther anodes,
> > > even with a resistive load at the output.
> > > The ringing crap is transformed to the secondary, and when fed back
> > > can cause the amp to oscillate.
> > >
> > > The 1.5k plus 0.001uF will act to reduce the ringing at the anodes.
> > >
> > > Zobels used across the secondary usually have R = rated RL plus a C
> > > that makes the reactance of the C = R at 100 kHz.
> > >
> > > So 4.7 ohms plus 0.33uF are typical values.
> > >
> > > The other place for a Zobel are across the RL of the V1,
> > > so that where there is 47k load, you might use 4k7 plus 470pF.
> > >
> > > This will reduce the gain of V1 at HF, and reduce the phase shift
> > > where OLG > 1.
> > >
> > >
> > > But unless you have designed and built a lot of amps and know the
art
of
> > > trying things
> > > and observing, and ending up with R&C values that
> > > ONLY stop the amp oscillating, and load the amp minimally,
> > > so that the response with a pure R load is very litle affected, then
> > > chances are you'll get it all wrong.
> > >
> > > The obel networks make it possible to use ESL speakers without worry
of
> > > having a huge
> > > peak in the sine wave response between 8kHz and 32 kHz.
> > >
> > > With any value of pure C loading from 0.05uF 5uF, the peaking caused
> > > in the sine wave response due to having FB should not exceed 6dB.
> > >
> > > This will indicate a good stability margin.
> > >
> > > In fact pure C loads are rare. With most ESL, there is some C
alright,
> > > maybe 1uF to 5uF, but usually there is a series R equivalent to say
1.5
> > > ohms
> > > which provides enough damping to a tube amp which is in effect an
active
> > > bandpass filter
> > > with more than 6dB attenuation beyond the LF and HF open loop poles.
> > >
> > > See my website for numerous examples of Zobel networks
> > > in amplifier schematics.
> > >
> > > http://www.turneraudio.com.au
> > >
> > > Well damped amps do not sound bogged down by being overloaded at HF
by
> > > such networks.
> > > They usually have clearer sounding HF.
> > >
> > > Patrick Turner.
|
