[AMRadio] Class AB and B audio XFMRS

Jim candela jcandela at prodigy.net
Sat Mar 11 16:54:10 EST 2006

John E. Coleman (ARS WA5BXO) wrote:

>	I have been doing some experimenting here on the side and have found
>a interesting thing that I've seen before but forgotten about.  I thought I
>might share it here with you guys.

Geof W5OMR Wrote:

I can't believe that no one else has replied to this, on the reflector.

From: "Patrick Turner" <info at turneraudio.com.au>
Subject: Re: SE Output transformer design idea
Date: Monday, August 25, 2003 8:06 PM

Jim WD5JKO pasted in the following:

This from an old thread about single ended audio transformers, and how some
folks deal with the high DC flux from the class A audio output tube. In many
respects this is the same problem we AM'ers face with push pull modulators
where the mod transformer has to bear the RF final plate current without
core saturation. Pay attention to the stuff about using a DC constant
current source into one winding to buck the flux from the Class A output
tube plate current. It may not be practical, but it sure is interesting...

smoking-amp wrote:

> "Ian Iveson" <ianiveson.home at virgin.net> wrote in message
> > "smoking-amp" <donbowen at att.net> wrote
> > > ...[loads of seemingly sensible stuff about B and H]...
> >
> > Thanks, Don.  I almost posted my own essay on domains, but yours
> > probably makes more sense.  I got to the paragraph where I was
> > explaining what we both call a kind of "avalanche" effect.  I was
> > having difficulty there because it is a stable process, not like a
> > real avalanche that sustains itself with no further excitement.
> Yes, your right, I maybe overdid the analogy a bit to be more graphic,
> it should be stable, alignment of other domains increases field to
> make flipping easier.
> > Then I got to the bit where the curve levels off towards saturation
> > and couldn't remember why.  Still can't.  Similar to a population
> > curve...these sigmoids arise from some fundamental contradiction but
> > dunno which one in this case.  Glad you got me out of that one.
> I think Tim has got this, just running out of domains to flip as
> saturation sets in. Hardest ones to flip go last.
> >
> > I remember an experiment in which the secondary of a transformer was
> > monitored with sufficient resolution to see the individual flips.
> > Rather like a sloping line on an old low-resolution computer screen,
> > it is fairly clean at 45 degrees, but very lumpy at shallow angles.
> I recall doing a similar experiment in high school once, we just
> connected a high turns coil wound on an open U core to the input of an
> audio amp and then brought up a bar magnet to the core. Could hear the
> domains flipping, sort of like the sound of crunching plastic wrap.
> >
> > But I don't get your argument that the effect centres itself around
> > any bias point.  I can see that there will be hysterisis about any
> > established point, but the curve is steeper so it should be less,
> > surely?  Flipping a domain is easier or harder depending on what
> > proportion is already flipped.
> I may have jumped to conclusions here.  I need to look at one of the
> real magnetics books (the transformer books don't get down to this
> detail) here to see what really happens in minor hysteresis loops.
> Definitely there is hysteresis at any DC operating point, how the U
> varies with DC and AC magnitudes here needs to be checked. This
> morning when I got up I realized that I was relying on hand waving
> arguments here, so I went and measured a real SE transformer. Tested
> with a Hammond 1640SE. Used a Heath digital LC meter on the primary
> and connected an IRF230 Mosfet as a variable DC current source to the
> combined secondaries. This measurement turned out to be a lot more
> difficult than I expected. Had to get the current source very quiet or
> it generated effects on the meter reading. Had to use an isolation
> xfmer on the meter and another on the power supplies for the current
> source. Still found some interaction from connecting the current
> source with no current. I now plan on purchasing one of those pocket,
> battery powered, digital LC meters. Also found that hysteresis in the
> SE transformer caused inconsistencies in readings versus DC current
> between ascending current measurements and descending current
> measurements. Finally went and got a variac to demagnetize the SE xfmr
> with a slow dropoff of AC voltage. Then did two sets of ascending DC
> current measurements which agreed (demagnetized before each set). Here
> are my results: 0 Amp DC -> 27 Henry ,  .05A -> 27H , .1A -> 27H , .2A
> -> 26H , .4A -> 25H , .8A -> 24H , 1.2A -> 23H , 2.0A -> 22H , 3.0A ->
> 21H  Pretty apparent that U is dropping with DC current here, even
> when diluted out by an air gap.

Your measurments are typical for a well gapped core.
The inductance remains fairly constant, with DC change, AND AC change.
With no gap, the variation of inductance would be far greater.

> Would need to do a further test with
> variable AC level in measurement of inductance to see if U picks up
> and peaks at some AC level before one could conclude my earlier claim
> of just moving the whole inductance curve with DC operating point.
> Maybe will get to testing that yet.

The more tests you do, the more you will understand the capabilities and
properties of the SE transformer, which is so baffling to most ppl.

> 3.0 Amps DC on the full secondary
> here would correspond to 409 mA on the primary when the turns ratio is
> taken into consideration. The 1640SE is rated for 200mA max. DC bias.
> M6 steel.

When I build an SE amp, I have to set the air gap in the core optimally.

I need to run the amp at the maximum mid frequency voltage, a dB below
clip, and roll the frequency down to say 20 Hz, where there shouldn't be any
but to where probably the effect of the primary inductance in parallel with
the load
has reduced the load value.
The distortion seen is due to tube overloading, not core saturation.
Reducing frequency further should cause more tube distortion, a reduction of
output voltage, but no short circuit knocking noises in the OPT due to
This is the case, even with 16 dB of FB trying to correct against the errors
created at these LF.

Measuring the inductance with a tube in place is easy.
It should be done away from regions of distortion.
Without any RL connected, and with a 10 ohm cathode
current sensing R in the cathode circuit, the AC and DC current can be
for various signal levels of input to the SE tranny.
So you can measure the voltage and current, and find the L impedance, ZL =
Va / Ik,
and from this you calculate Lp = ZL / ( F x 6.28 ) ,
where 6.28 is 2 x Pi, for all equations to work.

This method is more informative than fiddling around with CCS on the
output winding.

But the use of a CCS through the secondary to oppose the
primary DC flow is one way to null some of the effect of the DC
in the core of an SE tranny.

I'd say parafeed and a cap coupled non gapped OPT
would be a better solution, but if the SE tranny is simply
made well enough, there is no need for parafeed, or trying
to force a gapped SE tranny to behave better at LF.

As you shoulod know, the gap reduces inductance dramatically compared to
the same turns on a non gapped core.
The absense of DC magnetisation in a gapped core makes little difference to
the inductance
in shunt with the load.

> >
> > Van der Veen says don't worry about it...nearly all the effects
> > cancel out with a decent transformer.
> >
> > cheers, Ian
> Yes, a low output impedence from triode output, distributed output
> topology, or global feedback should fix up xfmr effects nicely unless
> the xfmr is really designed badly.

And cathode bias means the self regulation of bias currents in the PP
circuit is good.

On my PP amps I have sub circuits driving leds to indicate
if the DC imbalance drifts more than 5 mA. In normal use, there is
negligible drift, and leds stay extinguished.
But at high levels, with clipping, they start to flash.

>  Also, a note to Ronald or David, I didn't mean to sound like I was
> trashing SE designs, obviously I am playing around with them too in
> distributed load (Quad) topology, have Hammond 1640SE and 1628SE xfmrs
> here (using secondary as cathode winding).

If you only have the secondary for the FB coil, best
results are with 6BQ5/EL84, ( or a few in parallel ) since this tube has
higher gain and the cathode FB actually does something useful.
The CFB should ideally be between 10% of total primary voltage,
like Quad II, or up to 20%.
If the OPT has a ratio of 5k to 8 ohms, the voltage ratio is
25 :1, and CFB is only 4%.

> Just takes a bigger xfmr
> for SE. I want to compare a real SE design sound to the psuedo SE
> design (P-P using a triode in the feedback loop) I mentioned in an
> earlier post. If the pseudo SE sounds like real SE then could make
> high powered ampls.

I'd go for normal FB, well done SE amp.
Psuedo SE? very imprecise, unless you really went hard with R&D....

> for more usual low efficiency speakers. Another
> more extreme variation would be to try triode feedback in a SS ampl.,
> but don't want to cause a firestorm over this. Undoubtedly, hard
> clipping issues will limit the quality of this.
> Don

Patrick Turner.

No virus found in this outgoing message.
Checked by AVG Free Edition.
Version: 7.1.375 / Virus Database: 268.2.1/279 - Release Date: 3/10/2006

More information about the AMRadio mailing list

This page last updated 18 Oct 2017.