Capless Parafeed (with apologies to SY)

[dsavitsk]

For some reason, this post from Gavin on building a Mini Monkey, or a Parafeed DRD amp with a 12B4 as the output tube has always intrigued me. So, I have slowly amassed all of the parts to build a version of it for headphones.

About to get started, I stumbled on SY's Heretical line stage, and it occurred to me that perhaps this might make a better output stage, and could eliminate the parafeed/ultrapath cap entirely.

So, I am sure there is something I am not getting right here, and the swing of the servo might be that something. But for a headphone amp where the voltage swing is minimal and the output tube's bias can be kept reasonable, perhaps this can work. Thoughts?

[dsavitsk]

Here are some actual parts for driving some 32 ohm headphones. These are largely selected as I have them around, so there is nothing magical about them.

The +/-25.2V is to run all 4 heaters in series and provide +/- voltage for the CCS and opamp (the opamp is good up to +/-30V). Biasing of the 12B4 is ~12V running at 25mA. If there isn't enough room to swing around the opamp's supply, it can be increased a bit with a resistor or diode in the heater supply to drop some volts.

[revintage]

Think it would be cleaner way do it like Dave Slagle. Instead of the choke you use a 1:1 bifilar followed by the transformer. Only one tube and no sand.

Maybe SY kills me now when I say that you should try to avoid CFs as well as cascodes !

[dsavitsk]

Originally posted by revintage
Think it would be cleaner way do it like Dave Slagle. Instead of the choke you use a 1:1 bifilar followed by the transformer. Only one tube and no sand.

Yes, I like Dave's design but it violates two constraints on this project -- the first is that I don't want to buy any more iron, or parts for that matter, and the second is that I have done plenty of single tube headphone amps, and I am looking for something different. At any rate, Dave's design still suffers from the final PS cap being in the signal path which I am trying to avoid here.

[Gary P]

Not to through a wet blanket on your fire, but in your current design the final cap is still in the signal path. The output stage current loop includes the 12B4, the transformer primary, and the final power supply capacitor. The servo and CCS biasing do nothing to remove the power supply from the output stage current loop.

[dsavitsk]

Yeah, I think you are right. Maybe switching to a sand rectifier will give me enough extra voltage to add a shunt reg to avoid this.

[kenpeter]

Sy's B+ series regulator servo kills two
birds. If you are into killing birds that is...

I'd be tempted to spice up an MJK style
antitriode (or anti-cathode follower as
this case may be) as an active current
source...

Just orient the current sense resistor
(not shown in your drawing) sticking up
out the top end of something P-Channel.
And tap the output halfway up that sense
resistor. Halfway between the Cathode
and P-Source.

The apparent halving of drive impedance
(sharing half with CCS as-if a Mu-Follower
turned upside down) should more than
compensate the small sensing resistance
added between cathode and OPT.

I don't know a suitable P-Channel device
offhand, so it may be a purely hypothetical
topology tweak to anti-follower the CCS...

[Michael Koster]

OK I did make a doodle yesterday while reading this and I have
played with the so-called anti-follower or maybe series-regulated
balanced drive cathode follower.

Since the CF circuit is stable as it is, it doesn't *need* a CCS for
regulation; the advantage is in the high signal impedance.
I would be tempted to have the servo control the gate voltage on
the MOSFET + cap couple ala gyrator. 2P25s are decent but maybe
there are some smaller ones. Coes is the only load on the signal as
in the case of the basic CCS load.

This is just a sketch of the concept. Output Z is approximately
Rp/2u + Rsense/2

Hope it's not too far off topic ;-)

Michael

PS This could solve a common problem in CF circuits, which is not
enougn pull-down drive strength, which causes asymmetric slewing
distortion. This is at least symmetric but still need good standing
current.

PPS Ken, 25 pcs of 6N170A arriving UPS on Tuesday... muahahahaha

PPPS maybe it would make a decent OTL

[kenpeter]

If you can make the quiescent drop across RSense be the
same as the tube's required bias, the input then shifts to
ground level... I don't know, that might be an advantage?

Don't matter if the servo is acomplished by B+ regulation,
triode grid bias, or the CCS. But it was pointed out earlier
that the triode's power supply cap is directly in the pull up
loop. I think some form of B+ stiffening is still a good idea.

Sy's B+ servo is perhaps more simple in that it doesn't
have to drive a voltage swing as great as the entire
output. And how to float power for that anti-cathode
servo op-amp so it can do so?

The main "advantage" I see in MJK's proposed drawing
of the anti-cathode servo is that it permits abuse of an
enhancment mode P device. As no suitable depletion
mode device seems to exist for P-CH at the moment.

I think you could bias the gate of an enhancement device
here with a keyfob battery. Such that anti-cathode is still
a viable CCS topology. And direct the actual servo back to
B+, where it can double duty for the required regulation.
Weather series or parallel regulated, either would work...
A matter of personal choice.

[kenpeter]

Originally posted by Michael Koster
PS This could solve a common problem in CF circuits, which is not
enougn pull-down drive strength, which causes asymmetric slewing
distortion. This is at least symmetric but still need good standing
current.

Yes, lot of stages with low impedance can't actually drive
a matching load anywhere but up... At least not linearly...
There is something to be said for having a predictable
minimum series resistance in the output. RSense would
likely be in the right ballpark for that.