Can I use triodes in place of Q13/14?
http://www.headwize.com/projects/sho...lmore4_prj.htm (hybrid
electrostatic headphones amp)
Since that's doing most of the voltage gain, seems to me a good place for
triodes.

Prune wrote:

Can I use triodes in place of Q13/14?

What would be the purpose of that ?

Graham

Eeyore wrote in
:

What would be the purpose of that ?

LOL, isn't this a tube group? Why not put more tubes in a hybrid circut?
Let me change the question then: is there any potential aural downside to
using triodes in that place?

[Prune]

Hmm. What about other tubes?

As for the bias, of course I'd have to adjust it from the previous stage.


flipper wrote in
:

On Sat, 01 Jul 2006 06:48:13 GMT, Prune
wrote:

Can I use triodes in place of Q13/14?
http://www.headwize.com/projects/sho...lmore4_prj.htm
(hybrid electrostatic headphones amp)
Since that's doing most of the voltage gain, seems to me a good place
for triodes.

Constant current source loaded 12AX7s would give a voltage gain of
about 95 where he's got 200 in that stage. Not sure what would happen
to his DC bias levels.

[Prune]

I was referring to US patents 3111630 and 3328711, and he has some updates
to that if you can read past the marketing nonsense at
http://www.fsaudioweb.com/listen/wolcottwp.html
I think I've mentioned it here in another context.

[Prune]

Fair enough; I intended to keep the topology mostly the same, just
adding another feedback line (crossed from the other side, since the
triodes would be inverting, and I want same-polarity feedback). The
main problem with positive feedback is stability. I'll throw this in
the simulator soon...


flipper wrote in
news
On Tue, 04 Jul 2006 10:41:34 GMT, Prune
wrote:

Well, I was thinking the gain limit could be overcome by band-limited
positive feedback (Wolcott Audio amps use this, there are some old
patents as well).

I was just addressing the question of substituting triodes for Q13/14
but you'll have to tell me what the result of redesigning the topology
is after you come up with it.


flipper wrote in
m:

On Mon, 03 Jul 2006 07:27:00 GMT, Prune
wrote:

Hmm. What about other tubes?

'Other' in what way? You said triode and 95 is about as good as it
gets with a CCS loaded triode

Pentodes will do more but they aren't triodes and you don't
generally
CCS load the plate. Or, put another way, they don't 'go there that
way'.

That's probably one reason why he's using solid state up to the high
voltage section.


As for the bias, of course I'd have to adjust it from the previous
stage.


flipper wrote in
m:

On Sat, 01 Jul 2006 06:48:13 GMT, Prune bobysgotguns_@_yahoo_.
_com
wrote:

Can I use triodes in place of Q13/14?
http://www.headwize.com/projects/sho...?file=gilmore4
_prj.htm
(hybrid electrostatic headphones amp)
Since that's doing most of the voltage gain, seems to me a good
place for triodes.

Constant current source loaded 12AX7s would give a voltage gain of
about 95 where he's got 200 in that stage. Not sure what would
happen to his DC bias levels.

[Prune]

Not necessarily; you can raise gain a lot more and then use a lot more
NFB. The inrease in distortion from one and the decrease from the other
is not a proportional tradeoff. See the patents.


flipper wrote in
:

On Tue, 04 Jul 2006 14:44:47 GMT, Prune
wrote:

Fair enough; I intended to keep the topology mostly the same, just
adding another feedback line (crossed from the other side, since the
triodes would be inverting, and I want same-polarity feedback). The
main problem with positive feedback is stability.

That and increased distortion.

I'll throw this in
the simulator soon...


flipper wrote in
news
On Tue, 04 Jul 2006 10:41:34 GMT, Prune
wrote:

Well, I was thinking the gain limit could be overcome by band-
limited
positive feedback (Wolcott Audio amps use this, there are some old
patents as well).

I was just addressing the question of substituting triodes for
Q13/14
but you'll have to tell me what the result of redesigning the
topology
is after you come up with it.


flipper wrote in
m:

On Mon, 03 Jul 2006 07:27:00 GMT, Prune bobysgotguns_@_yahoo_.
_com
wrote:

Hmm. What about other tubes?

'Other' in what way? You said triode and 95 is about as good as it
gets with a CCS loaded triode

Pentodes will do more but they aren't triodes and you don't
generally
CCS load the plate. Or, put another way, they don't 'go there that
way'.

That's probably one reason why he's using solid state up to the
high
voltage section.


As for the bias, of course I'd have to adjust it from the previous
stage.


flipper wrote in
news:krpca2hautfvadspeckp14km2ojffef7la@4ax. com:

On Sat, 01 Jul 2006 06:48:13 GMT, Prune bobysgotguns_@_yahoo_.
_com
wrote:

Can I use triodes in place of Q13/14?
http://www.headwize.com/projects/sho...?file=gilmore4
_prj.htm
(hybrid electrostatic headphones amp)
Since that's doing most of the voltage gain, seems to me a good
place for triodes.

Constant current source loaded 12AX7s would give a voltage gain
of
about 95 where he's got 200 in that stage. Not sure what would
happen to his DC bias levels.

[Prune]

Any reason the EL34s are not run in pentode mode?


flipper wrote in
:

On Wed, 05 Jul 2006 13:23:45 GMT, Prune
wrote:

Not necessarily; you can raise gain a lot more and then use a lot more
NFB. The inrease in distortion from one and the decrease from the
other is not a proportional tradeoff. See the patents.

I've read the patents but claims are not always reality and I'm always
a bit suspicious of free lunches. Would be happy to see some
supporting math, though, as there isn't any in the patent.

However, even if the claims perform as implied you aren't raising gain
a lot more to add a lot more NFB. You're in the situation of adding
gain just to get back to the NFB that's already in the original
circuit. And the split load bootstrap doesn't do you any good either,
as far as gain goes, as that's just a substitute for a CCS, which
is/was already there.

Lacking a detailed analysis my gut tells me that the secret, if there
is one, lies in the cross coupling and subsequent nulling in the PP
transformer.

flipper wrote in
m:

On Tue, 04 Jul 2006 14:44:47 GMT, Prune
wrote:

Fair enough; I intended to keep the topology mostly the same, just
adding another feedback line (crossed from the other side, since the
triodes would be inverting, and I want same-polarity feedback). The
main problem with positive feedback is stability.

That and increased distortion.

I'll throw this in
the simulator soon...


flipper wrote in
news
On Tue, 04 Jul 2006 10:41:34 GMT, Prune
wrote:

Well, I was thinking the gain limit could be overcome by band-
limited
positive feedback (Wolcott Audio amps use this, there are some old
patents as well).

I was just addressing the question of substituting triodes for
Q13/14
but you'll have to tell me what the result of redesigning the
topology
is after you come up with it.


flipper wrote in
news:ddnha2t2i53nu2e9vjj3de8mfbo6renbnv@4ax. com:

On Mon, 03 Jul 2006 07:27:00 GMT, Prune bobysgotguns_@_yahoo_.
_com
wrote:

Hmm. What about other tubes?

'Other' in what way? You said triode and 95 is about as good as
it gets with a CCS loaded triode

Pentodes will do more but they aren't triodes and you don't
generally
CCS load the plate. Or, put another way, they don't 'go there
that way'.

That's probably one reason why he's using solid state up to the
high
voltage section.


As for the bias, of course I'd have to adjust it from the
previous stage.


flipper wrote in
news:krpca2hautfvadspeckp14km2ojffef7la@4a x.com:

On Sat, 01 Jul 2006 06:48:13 GMT, Prune
bobysgotguns_@_yahoo_.
_com
wrote:

Can I use triodes in place of Q13/14?
http://www.headwize.com/projects/sho...?file=gilmore4
_prj.htm
(hybrid electrostatic headphones amp)
Since that's doing most of the voltage gain, seems to me a
good place for triodes.

Constant current source loaded 12AX7s would give a voltage
gain
of
about 95 where he's got 200 in that stage. Not sure what would
happen to his DC bias levels.

[Prune]

Thanks for the answer. The link works fine, I just checked...

One thing I'm wondering is if this circuit may not be violating one of
Nelson Pass' patents, http://free.patentfetcher.com/GetPatentPDF.php?
f=Pats/US/53/76/US5376899.pdf
I know the patent is valid since Texas Instruments recently licensed it.
Since Gilmore is selling amps with this topology, if this falls under
the patent it would be a violation.
He has fully differential amplifier with NFB from both sides, connected
in such a way that some of the error signal would go in same polarity
through to the other side (in this case through the connection at the
tail of the input differential pair) and cancel out at the output since
it's common mode. Or am I missing something?


flipper wrote in
:

On Thu, 06 Jul 2006 10:40:54 GMT, Prune
wrote:

Any reason the EL34s are not run in pentode mode?

For the linearity, so he said.

An equally good answer would be, because that's what works.

The link doesn't work anymore but, if memory serves, he had a P FET as
follower driver to the cathode of a grounded grid EL34 which, in turn,
ran into a CCS.

Now, a pentode wants to provide a current that is set by the
grid-cathode voltage independent of plate.

So make it a pentode and you end up with two current mode devices
arguing with each other. The pentode insisting, with all it's might,
that the current *is* going to change, regardless of what happens to
plate voltage, and the CCS insisting, with all it's might, that by
golly it ain't and that it'll swing plate voltage to whatever it takes
to make it so.

Put another way, you end up with infinite gain into an infinite load
resulting in infinite instability.

If you're thinking you can pick up the lost gain here, don't, as
you'll likely end up with an oscillator rather than an amplifier.

Again, if memory serves, he had quite modest gains in the front end
stages so you might have better luck getting it there.

Why do you want to change the FETs to triodes?

flipper wrote in
m:

On Wed, 05 Jul 2006 13:23:45 GMT, Prune
wrote:

Not necessarily; you can raise gain a lot more and then use a lot
more
NFB. The inrease in distortion from one and the decrease from the
other is not a proportional tradeoff. See the patents.

I've read the patents but claims are not always reality and I'm
always
a bit suspicious of free lunches. Would be happy to see some
supporting math, though, as there isn't any in the patent.

However, even if the claims perform as implied you aren't raising
gain
a lot more to add a lot more NFB. You're in the situation of adding
gain just to get back to the NFB that's already in the original
circuit. And the split load bootstrap doesn't do you any good
either,
as far as gain goes, as that's just a substitute for a CCS, which
is/was already there.

Lacking a detailed analysis my gut tells me that the secret, if
there
is one, lies in the cross coupling and subsequent nulling in the PP
transformer.

flipper wrote in
m:

On Tue, 04 Jul 2006 14:44:47 GMT, Prune bobysgotguns_@_yahoo_.
_com
wrote:

Fair enough; I intended to keep the topology mostly the same, just
adding another feedback line (crossed from the other side, since
the
triodes would be inverting, and I want same-polarity feedback).
The
main problem with positive feedback is stability.

That and increased distortion.

I'll throw this in
the simulator soon...


flipper wrote in
newsuqka2lafpl3c2pbuvdiqsge5ja18n7la5@4ax. com:

On Tue, 04 Jul 2006 10:41:34 GMT, Prune
wrote:

Well, I was thinking the gain limit could be overcome by band-
limited
positive feedback (Wolcott Audio amps use this, there are some
old
patents as well).

I was just addressing the question of substituting triodes for
Q13/14
but you'll have to tell me what the result of redesigning the
topology
is after you come up with it.


flipper wrote in
news:ddnha2t2i53nu2e9vjj3de8mfbo6renbnv@4a x.com:

On Mon, 03 Jul 2006 07:27:00 GMT, Prune
bobysgotguns_@_yahoo_.
_com
wrote:

Hmm. What about other tubes?

'Other' in what way? You said triode and 95 is about as good
as
it gets with a CCS loaded triode

Pentodes will do more but they aren't triodes and you don't
generally
CCS load the plate. Or, put another way, they don't 'go there
that way'.

That's probably one reason why he's using solid state up to
the
high
voltage section.


As for the bias, of course I'd have to adjust it from the
previous stage.


flipper wrote in
news:krpca2hautfvadspeckp14km2ojffef7la@ 4ax.com:

On Sat, 01 Jul 2006 06:48:13 GMT, Prune
bobysgotguns_@_yahoo_.
_com
wrote:

Can I use triodes in place of Q13/14?
http://www.headwize.com/projects/sho...?file=gilmore4
_prj.htm
(hybrid electrostatic headphones amp)
Since that's doing most of the voltage gain, seems to me a
good place for triodes.

Constant current source loaded 12AX7s would give a voltage
gain
of
about 95 where he's got 200 in that stage. Not sure what
would
happen to his DC bias levels.

[Prune]

While the first figure in the patent has independend amplifiers, it is
just an example and the patent is not limited to that. Indeed, an LTP-
based version is also covered by the patent; see the Aleph-X schematic
http://www.facstaff.bucknell.edu/esa...es/axschem.jpg (also
covered by another Pass patent, on modulated current sources).

Any 'positive feedback' one might wish to argue takes place in the
long tail pair is the same 'coupling' that has been in every long tail
pair since time immemorial so if it's 'the same' it would invalidate
the patent from prior art, not the other way around.

Well yes, except that what the patent covers is the combination of this
process with negative feedback from both sides of a fully differential
in/out amplifier. Up to the patent, that this effect is in an LTP
doesn't seem to have been a major consideration in reducing distortion.

If prior art was really invalidating the patent, TI's lawyers wouldn't
have bothered to license it.

BTW, when simulating such a circuit in SPICE, injecting a pulse at the
output to represent distortion, the simulation clearly shows the effect,
and a pulse appears on the other side of the circuit with the same
polarity.

I'd like to see if this happens with the electrostatic amp, but the EL34
model I found isn't working right when cathode driven...


flipper wrote in
:

On Wed, 12 Jul 2006 19:53:59 GMT, Prune
wrote:

Thanks for the answer. The link works fine, I just checked...

Yes. I was using the link in your replies and hadn't noticed word wrap
had screwed it up

One thing I'm wondering is if this circuit may not be violating one of
Nelson Pass' patents, http://free.patentfetcher.com/GetPatentPDF.php?
f=Pats/US/53/76/US5376899.pdf
I know the patent is valid since Texas Instruments recently licensed
it.
Since Gilmore is selling amps with this topology, if this falls under
the patent it would be a violation.

It isn't the same topology.

He has fully differential amplifier with NFB from both sides,
connected
in such a way that some of the error signal would go in same polarity
through to the other side (in this case through the connection at the
tail of the input differential pair) and cancel out at the output
since
it's common mode. Or am I missing something?

Nelson's patent has two 'independent' (identical) inverting
amplifiers on the input, not a long tail differential. It has two
current sources under them (one for each as they're 'independent' and
identical), not one. And it has a single resistor 'coupling' the two
separate inverting amplifiers, not separate resistors going to a
common long tail CCS.

Any 'positive feedback' one might wish to argue takes place in the
long tail pair is the same 'coupling' that has been in every long tail
pair since time immemorial so if it's 'the same' it would invalidate
the patent from prior art, not the other way around. And the series
resistors do not participate in any positive feedback 'coupling'
(other than to, perhaps, make it worse) so their existence, while
visually similar, is functionally different. They are there for the
negative feedback node, not 'positive coupling'.

Lastly, the patent is essentially a single gain stage and I'm not sure
it would translate to multiple stage global feedback even if the rest
were similar, nor do see a claim for it.

The manner in which you describe the Gilmore circuit raises another
point, though, in that one cannot patent a 'goal' nor a 'result'. One
patents the means to accomplish it so, while it might be fortuitous if
the Gilmore circuit achieved the same results by other means, that,
alone, would not violate the patent.

----------------

To back track a bit to the earlier discussion on positive feedback and
distortion, I didn't mean to leave the impression that you can't
reduce distortion by combining positive and negative feedback. You
can, but It also increases distortion. Contradictory? No.

The traditional use of positive feedback, to reduce distortion, is
based on the premise that one, or more, stages are low distortion with
a 'problem' high distortion stage(s) (usually the output). Now, you
can reduce distortion with negative feedback but you need excess gain
to do so and, since extra gain stages are not 'free', positive
feedback to a low distortion stage increases it's gain, and
distortion, so you can then apply negative feedback. The problematic
stage's distortion is reduced at the expense of increased distortion
in the low distortion stage (from the positive feedback) but since the
low distortion stage is 'so much better' than the problematic stage
the amplifier's overall distortion is improved. You always get the
distortion increase on the positive feedback stage but the
'improvement' only happens if you've got one significantly worse, or
better, so you can trade one for the other. It isn't a 'free lunch',
you're just shuffling the meat around on the table

What I meant regarding the Wolcott patent was in relation to the claim
that using positive and negative feedback is such an 'improvement'
over prior art and I'm saying that, if his topology is, it's likely
due to the cross coupling and subsequent cancellation in the PP OPT
rather than the combination of positive and negative feedback, per
see, since that technique was well known and routinely employed. In my
opinion, the cross coupling would be the 'new thing'.

And back to your idea of using triodes, the problem is you lose gain
and while popping in some positive feedback might get it back it would
do so at the expense of distortion *somewhere*, compared to the
existing circuit. How much I don't know because it depends on where
you'd stick it and how good each of the existing stages already are
but it can't possibly be 'equivalent' because you have less inherent
gain before applying the positive/negative feedback (unless we get
into an argument about whether triodes are 'inherently more linear'
than what they're replacing).

It also happens to be, assuming his description is correct, that where
you lose the gain is where he's generating the lion's share of it,
around 200, while the input stages are a relatively modest 5 and 4. So
it seems to me that 'positive feedback' would be acting on the most
vulnerable stage, hence my suggestion to increase the input stage gain
to compensate.. Their distortion would go up but at gains of only 4
and 5 one would think it must be incredible small to begin with,
compared to where the 'problem' is.

My only concern would be potential biasing problems because he seemed
to be particularly worried about offset drift.

flipper wrote in
m:

On Thu, 06 Jul 2006 10:40:54 GMT, Prune
wrote:

Any reason the EL34s are not run in pentode mode?

For the linearity, so he said.

An equally good answer would be, because that's what works.

The link doesn't work anymore but, if memory serves, he had a P FET
as
follower driver to the cathode of a grounded grid EL34 which, in
turn,
ran into a CCS.

Now, a pentode wants to provide a current that is set by the
grid-cathode voltage independent of plate.

So make it a pentode and you end up with two current mode devices
arguing with each other. The pentode insisting, with all it's might,
that the current *is* going to change, regardless of what happens to
plate voltage, and the CCS insisting, with all it's might, that by
golly it ain't and that it'll swing plate voltage to whatever it
takes
to make it so.

Put another way, you end up with infinite gain into an infinite load
resulting in infinite instability.

If you're thinking you can pick up the lost gain here, don't, as
you'll likely end up with an oscillator rather than an amplifier.

Again, if memory serves, he had quite modest gains in the front end
stages so you might have better luck getting it there.

Why do you want to change the FETs to triodes?

flipper wrote in
m:

On Wed, 05 Jul 2006 13:23:45 GMT, Prune bobysgotguns_@_yahoo_.
_com
wrote:

Not necessarily; you can raise gain a lot more and then use a lot
more
NFB. The inrease in distortion from one and the decrease from the
other is not a proportional tradeoff. See the patents.

I've read the patents but claims are not always reality and I'm
always
a bit suspicious of free lunches. Would be happy to see some
supporting math, though, as there isn't any in the patent.

However, even if the claims perform as implied you aren't raising
gain
a lot more to add a lot more NFB. You're in the situation of
adding
gain just to get back to the NFB that's already in the original
circuit. And the split load bootstrap doesn't do you any good
either,
as far as gain goes, as that's just a substitute for a CCS, which
is/was already there.

Lacking a detailed analysis my gut tells me that the secret, if
there
is one, lies in the cross coupling and subsequent nulling in the
PP
transformer.

flipper wrote in
news:j9gma21apud2sh7gd41nss5fiq81bfmed1@4ax. com:

On Tue, 04 Jul 2006 14:44:47 GMT, Prune bobysgotguns_@_yahoo_.
_com
wrote:

Fair enough; I intended to keep the topology mostly the same,
just
adding another feedback line (crossed from the other side, since
the
triodes would be inverting, and I want same-polarity feedback).
The
main problem with positive feedback is stability.

That and increased distortion.

I'll throw this in
the simulator soon...


flipper wrote in
newsuqka2lafpl3c2pbuvdiqsge5ja18n7la5@4a x.com:

On Tue, 04 Jul 2006 10:41:34 GMT, Prune
wrote:

Well, I was thinking the gain limit could be overcome by band-
limited
positive feedback (Wolcott Audio amps use this, there are some
old
patents as well).

I was just addressing the question of substituting triodes for
Q13/14
but you'll have to tell me what the result of redesigning the
topology
is after you come up with it.


flipper wrote in
news:ddnha2t2i53nu2e9vjj3de8mfbo6renbnv@ 4ax.com:

On Mon, 03 Jul 2006 07:27:00 GMT, Prune
bobysgotguns_@_yahoo_.
_com
wrote:

Hmm. What about other tubes?

'Other' in what way? You said triode and 95 is about as good
as
it gets with a CCS loaded triode

Pentodes will do more but they aren't triodes and you don't
generally
CCS load the plate. Or, put another way, they don't 'go
there
that way'.

That's probably one reason why he's using solid state up to
the
high
voltage section.


As for the bias, of course I'd have to adjust it from the
previous stage.


flipper wrote in
news:krpca2hautfvadspeckp14km2ojffef7l :

On Sat, 01 Jul 2006 06:48:13 GMT, Prune
bobysgotguns_@_yahoo_.
_com
wrote:

Can I use triodes in place of Q13/14?
http://www.headwize.com/projects/showfile.php?
file=gilmore4
_prj.htm
(hybrid electrostatic headphones amp)
Since that's doing most of the voltage gain, seems to me a
good place for triodes.

Constant current source loaded 12AX7s would give a voltage
gain
of
about 95 where he's got 200 in that stage. Not sure what
would
happen to his DC bias levels.

[Prune]

In the simulator I tried replacing the N-channel MOSFETs of the third stage
with 6DJ8s. Gain is lower but my source provides sufficient output to
still drive this amp to full swing. My question is how to optimally set up
the operating point of the triodes, as I don't know much about tubes. I
was getting large distortion until I pulled the 6DJ8 cathodes and the
bottom of the 2nd stage that feeds them 30 V below the -400 rail, as well
as decreasing the resistors on the current source at their anodes. Now
it's better, but not perfect. Suggestions on how to do it best? It would
be very nice if I could get tubes to work in this stage as well.

[Prune]

Hi flipper, thanks for the help.

flipper wrote in
news
Best gain you're going to get with CCS loaded 6DJ8s is around 30 to
33.
His open loop gain was 4,000 (4x5x200) and 1,000 closed. You're at
650, or so, open loop... unless you've increased the gain on the
front end.

My source can do 2.7 V on each side of the differential inputs. That
means I'd need gain under 300 to get full swing.

Q13/Q14 are enhancement mode FETs, meaning they operate with the gate
above source. Tubes operate with the grid below cathode potential.

Right, but I can always add a more negative supply by referencing the DC
I get from another winding to the negative rail.

Making it virtually impossible you could just pop them in and have it
magically work.

Well I didn't expect it to be that easy. That's why I'm asking here how
to best bias this.

I'm curious how you decided what value those unmarked resistors were
to begin with.

From Gilmore's original sketch; I simply forgot to add the numbers when
I redrew the schematic for Headwize. They are 100 ohms.

Now
it's better, but not perfect. Suggestions on how to do it best? It
would be very nice if I could get tubes to work in this stage as well.

Well, you need to pick an appropriate operating point for the tube and
adjust everything so it's biased there.

http://www.turneraudio.com.au/tube-operation1.html
http://www.turneraudio.com.au/tube-operation2.html

[Prune]

Here's my attempt this far in the simulation. I used 6DJ8, and to
partially make up for lost gain I decreased feedback (200K to 250K), as
well as pulling the triode cathodes and 2nd stage bottom another 70 V,
adjusting the second stage Vbe multiplier gate resistor. The 100 V zener
for the second stage supply was decreased to 80 V; the CCS resistors went
from 100 to 67 ohm. That runs the triodes at about 15 mA each with swing
between 55 and 130 V above the new more negative supply. I left the P-
channel followers' drains at -400. So this seems to work. Not sure how I
can generate the needed 60 mA -70 V to reference from the -400 V yet, maybe
use a doubler fed by the low voltage of the front end.

[Prune]

http://img145.imageshack.us/img145/9211/exrk5.png