[GRe]

Has anybody tried or seen a SE topology as shown below?
(Put reader to Courier-New font to view the schematic)
I'm asking because for a fun project I'm trying to get
an idea whether the topology is viable or not.

T1 is a transformer without gap. To cancel the field in
the core, caused by the primary current Ip, a current
Is of opposite direction, with a current value of Ip*N,
is imposed upon the secondary by a constant current sink.
N is the winding ratio, V+ is a regulated voltage.


Ip -- T1 Is --
+----------) | (----------- V+
| ) | (
V1 | ) | (
--- ) | (
--- ) | (
--- ----o---) | (-----o-----||----+
Vi ----- - - | | C1 |
O B+ | |
| | |
| | |
| | |
+-----o | | /|
| | +---+ +-+ | Spkr
- +-+ |CCS| Is +-+ |
Cc - | | Rc +---+ | \|
| +-+ | |
| | | |
| | | |
+-----o--------------------o-----------o
|
--- GND


I have a pair of line matching transformers handy with
a 22:1 winding ratio on the end taps, Rsec. = 0.7ohm,
Rprim. = 188ohm and a max. power handling of 10W.
Onset of core saturation is at around 115Vrms at 50Hz.
Primary inductance measures 73H at 100Vrms and 50Hz.

With an 8ohm load total Ra, including loss resistance,
will be around 4K4 which is a good value for an ECL82
operating at a B+ of 200V and a plate current of 35mA.
For this plate current Is needs to be 770mA, then the
power dissipation in the transformer is about 0.65W.
I don't think this amount of power causes a problem.

What could be the pitfalls of this topology?

Rgds,
Gio

[Patrick Turner]

On Oct 7, 10:21*pm, "GRe" wrote:
Has anybody tried or seen a SE topology as shown below?
(Put reader to Courier-New font to view the schematic)
I'm asking because for a fun project I'm trying to get
an idea whether the topology is viable or not.

T1 is a transformer without gap. To cancel the field in
the core, caused by the primary current Ip, a current
Is of opposite direction, with a current value of Ip*N,
is imposed upon the secondary by a constant current sink.
N is the winding ratio, V+ is a regulated voltage.

* * * * * * * Ip -- * *T1 * *Is --
* * * * * *+----------) | (----------- V+
* * * * * *| * * * * *) | (
* * * * V1 | * * * * *) | (
* * * * * --- * * * * ) | (
* * * * * --- * * * * ) | (
* * * * * --- ----o---) | (-----o-----||----+
*Vi ----- - - * * | * * * * * * | * * C1 * *|
* * * * * *O * * *B+ * * * * * *| * * * * * |
* * * * * *| * * * * * * * * * *| * * * * * |
* * * * * *| * * * * * * * * * *| * * * * * |
* * * * * *| * * * * * * * * * *| * * * * * |
* * *+-----o * * * * * * * * * *| * * * * * | /|
* * *| * * | * * * * * * * * *+---+ * * * *+-+ | Spkr
* * *- * *+-+ * * * * * * * * |CCS| Is * * +-+ |
*Cc *- * *| | Rc * * * * * * *+---+ * * * * | \|
* * *| * *+-+ * * * * * * * * * | * * * * * |
* * *| * * | * * * * * * * * * *| * * * * * |
* * *| * * | * * * * * * * * * *| * * * * * |
* * *+-----o--------------------o-----------o
* * * * * * * * * * * * * * * * * * * * * * |
* * * * * * * * * * * * * * * * * * * * * *--- GND

I have a pair of line matching transformers handy with
a 22:1 winding ratio on the end taps, Rsec. = 0.7ohm,
Rprim. = 188ohm and a max. power handling of 10W.
Onset of core saturation is at around 115Vrms at 50Hz.
Primary inductance measures 73H at 100Vrms and 50Hz.

With an 8ohm load total Ra, including loss resistance,
will be around 4K4 which is a good value for an ECL82
operating at a B+ of 200V and a plate current of 35mA.
For this plate current Is needs to be 770mA, then the
power dissipation in the transformer is about 0.65W.
I don't think this amount of power causes a problem.

What could be the pitfalls of this topology?

Rgds,
Gio

Anyone is free to provide cancelled dc magnetization of the core by
means of applying a fixed positive supply voltage to one end of the
OPT sec, then have CCS from live end of sec to 0V and cap couple the
speaker.

But you are putting yourself to a lot of trouble just for 4watts from
ECL82 / 6BM8, although it is a good learning experience.

If I were you and you wished to get a blameless 4 to 6 watts I would
use an EL34 strapped as a triode then driven with paralleled 12AU7 /
12AY7 / 12AT7 or trioded EF80 / 6BX6.

Instead of a high current CCS at the sec I would use a choke feed to
the anode of the power tube, say 40H at 60mAdc, and then cap couple
the primary winding of your OPT using a 60uF motor start cap rated for
450Vdc. The sec is then conventionally set up with one end grounded,
and the live end taken to a speaker. Such a choke should be nearly the
size of the OPT and can have similar dc resistance for the winding as
the OPT primary. The number of choke turns should be about twice the
turns for the OPT primary.

The primary inductance of your OPT may well be also about 40H if you
are lucky. It will be in parallel with the choke inductance.
But more analysis is required on your part about the properties of
your OPT. A typical anode load might be 5k and the response is down
-3dB at bass when total XLp = ( RL in parallel with Ra ) = 5k // 1.3k
= 1k, so if Lp was 20H, the LF pole = 8 Hz.

You could also use a CCS feeding the anode with say 60mA from a supply
rail of twice the idle Ea of the power tube.
So if Ea was 300Vdc, you'd need a rail of +600Vdc so that the Va may
swing UP and DOWN about 200 peak volts.
To make a suitable CCS using solid state does take some doing and
knowledge about how easily solid state devices fail in these
circumstances, and some knowledge about having say two high voltage
rated power mosfets in series for the CCS pass elements and having
adequate diode clamps to prevent any stray back emfs from inductances
from blowing crap out of delicate SS input circuits.

A carefully air gapped choke is far easier to use once you have such a
choke.

The other alternative to a CCS on the secondary side of the OPT is to
have the arrangement you show in your schematic but replace the solid
state CCS with a choke plus RC parallel network. The low voltage V+
would be say +20Vdc. The choke will need to be about 100mH, and air
gapped for up to 1Adc, and R = 19 ohms and adjusted for the wanted Idc
flow and the C =
10,000uF. The +20V rail needs at least 10,000 uF bypassing to 0V and
speaker drive cap could be 4,700uF.
The choke will not be easy to obtain, and may need to wound on a core
about the same size as your OPT and properly gapped and using say
1.5mm wire dia but at least there are not many turns needed.

Of course the end result of using transformers you have and which you
do not say are air gapped makes your exercize rather complicated and
fiddly to adjust just right compared to just buying say a pair of a
ready made SE OPTs from Hammond.

Option 3 may be to use a pair of OPT tubes such as EL86 and in series
and with B+ supply of 400V and then cap couple the OPT to the cathode-
anode junction. Philips produced a 10W amp with RL = about 1k and
pentode connection operation which requires a high L screen feed
choke. The idea was to allow use of a speaker with voice coil load of
about 800 ohms. But of course with time the extremely fragile voice
coil became easily fried, fused, or corroded, and nobody ever mourned
the passing of the Phillips silly idea to try to save the cost of a
proper PP OPT.

Patrick Turner.

[Patrick Turner]

On Oct 8, 9:14*pm, flipper wrote:
On Thu, 7 Oct 2010 13:21:41 +0200, "GRe"





wrote:

Has anybody tried or seen a SE topology as shown below?
(Put reader to Courier-New font to view the schematic)
I'm asking because for a fun project I'm trying to get
an idea whether the topology is viable or not.

T1 is a transformer without gap. To cancel the field in
the core, caused by the primary current Ip, a current
Is of opposite direction, with a current value of Ip*N,
is imposed upon the secondary by a constant current sink.
N is the winding ratio, V+ is a regulated voltage.

* * * * * * *Ip -- * *T1 * *Is --
* * * * * +----------) | (----------- V+
* * * * * | * * * * *) | (
* * * *V1 | * * * * *) | (
* * * * *--- * * * * ) | (
* * * * *--- * * * * ) | (
* * * * *--- ----o---) | (-----o-----||----+
Vi ----- - - * * | * * * * * * | * * C1 * *|
* * * * * O * * *B+ * * * * * *| * * * * * |
* * * * * | * * * * * * * * * *| * * * * * |
* * * * * | * * * * * * * * * *| * * * * * |
* * * * * | * * * * * * * * * *| * * * * * |
* * +-----o * * * * * * * * * *| * * * * * | /|
* * | * * | * * * * * * * * *+---+ * * * *+-+ | Spkr
* * - * *+-+ * * * * * * * * |CCS| Is * * +-+ |
Cc *- * *| | Rc * * * * * * *+---+ * * * * | \|
* * | * *+-+ * * * * * * * * * | * * * * * |
* * | * * | * * * * * * * * * *| * * * * * |
* * | * * | * * * * * * * * * *| * * * * * |
* * +-----o--------------------o-----------o
* * * * * * * * * * * * * * * * * * * * * *|
* * * * * * * * * * * * * * * * * * * * * --- GND

I have a pair of line matching transformers handy with
a 22:1 winding ratio on the end taps, Rsec. = 0.7ohm,
Rprim. = 188ohm and a max. power handling of 10W.
Onset of core saturation is at around 115Vrms at 50Hz.
Primary inductance measures 73H at 100Vrms and 50Hz.

With an 8ohm load total Ra, including loss resistance,
will be around 4K4 which is a good value for an ECL82
operating at a B+ of 200V and a plate current of 35mA.
For this plate current Is needs to be 770mA, then the
power dissipation in the transformer is about 0.65W.
I don't think this amount of power causes a problem.

What could be the pitfalls of this topology?

Rgds,
Gio

The biggest pitfall is you have, essentially, a push pull amp (albeit
odd looking) with one side of the push-pull (secondary) dissipating as
much energy as the other but doing no work. I.E. The CCS dissipates as
much as the tube. Think big heatsink.

I recall he wanted to use one ECL82 to get about 4W, say into 8 ohms
which is 5.6Vrms and = 8peak volts.

So the range of voltage movement means he'd need about 10V across the
CCS to allow the voltage to reduce down to 2V and up to 18V which is
OK. 10V x 1A = 10W and is slightly more Pd than the ECL86 pentode
section.



Purists might argue it's really a hybrid amp and not 'all tube'.

Ah, but the CCS has infinite impedance and cannot affect the ac
current change anywhere except by indirect means of preventing the
core from saturating, so the amount it could be considered to be
hybrid = 1 / near infinity = approximately zero.


You'd need the CCS to track tube current, which might pose some
motorboating problems due to feedback through the transformer and the
cathode bias capacitor time constant.

Well, with the added C coupling to load any GNFB could cause LF
instability because of the added time constant and hence phase shift.
But phase step networks should cure all that. The CCS shoud have no
effect because it has constant current.

A simple approach might be to (multiplying) current mirror the cathode
into the secondary.

AAAhhhhhh........

Patrick Turner.

- Hide quoted text -

- Show quoted text -

[Alex]

"GRe" wrote in message
...

Has anybody tried or seen a SE topology as shown below?
(Put reader to Courier-New font to view the schematic)
I'm asking because for a fun project I'm trying to get
an idea whether the topology is viable or not.

T1 is a transformer without gap. To cancel the field in
the core, caused by the primary current Ip, a current
Is of opposite direction, with a current value of Ip*N,
is imposed upon the secondary by a constant current sink.
N is the winding ratio, V+ is a regulated voltage.


Ip -- T1 Is --
+----------) | (----------- V+
| ) | (
V1 | ) | (
--- ) | (
--- ) | (
--- ----o---) | (-----o-----||----+
Vi ----- - - | | C1 |
O B+ | |
| | |
| | |
| | |
+-----o | | /|
| | +---+ +-+ | Spkr
- +-+ |CCS| Is +-+ |
Cc - | | Rc +---+ | \|
| +-+ | |
| | | |
| | | |
+-----o--------------------o-----------o
|
--- GND


I have a pair of line matching transformers handy with
a 22:1 winding ratio on the end taps, Rsec. = 0.7ohm,
Rprim. = 188ohm and a max. power handling of 10W.
Onset of core saturation is at around 115Vrms at 50Hz.
Primary inductance measures 73H at 100Vrms and 50Hz.

With an 8ohm load total Ra, including loss resistance,
will be around 4K4 which is a good value for an ECL82
operating at a B+ of 200V and a plate current of 35mA.
For this plate current Is needs to be 770mA, then the
power dissipation in the transformer is about 0.65W.
I don't think this amount of power causes a problem.

What could be the pitfalls of this topology?

Rgds,
Gio

It will work. V+ does not have to be well regulated (it can be an old
fashioned wall power pack), provided that CCS has high output impedance,
e.g., based on op-amp with power MOSFET.

And using an op-amp is very suitable because you can make secondary biasing
current exactly proportional to the cathode current (ideally would be better
to get it proportional to the plate current, but it is quite impractical to
place a current sensing circuity into the +200V rail). This will ensure good
ampere-turns cancellation if you change tubes or +B changes...

Now the pitfall.

The CCS will be dissipating about the same power as the main tube (ECL82 in
your case). Consider this: CCs has to regulate well under maximum possible
speaker voltage, which might be reaching say about 5V. Thus V+ shall be 5V
plus some headroom for the +V ripple, plus some headroom for the CCS
regulating MOSFET (say 3V) plus voltage drop on the current sensing resistor
of the CCS (say 1V) plus voltage drop of the secondary winding (0.55V). Thus
V+ will realistically be about 10...12V. With 770mA you will end up with
8...10W dissipation. Prepare a heatsing for your CCS MOSFET and the power
pack will be warm.

Other than that and if you can make a tracking CCS of an op-amp and a
MOSFET -- it is all yours.

And now the most fantastic extention of your concept. If instead of CCS
which cancels out only DC current, you add some AC component to the CCS, 180
degrees out of phase, you will end up with a class A push-pull. One leg of
it will be a tube, and the opposite leg -- the MOSFET scaled up 22 times in
current and scaled down 22 times in voltage. But the output power will be
doubled! Thus 8...10W of dissipation in the CCS MOSFET will not be totally
wasted.

Regards,
Alex

[Patrick Turner]

On Oct 9, 2:03*pm, flipper wrote:
On Fri, 8 Oct 2010 06:34:00 -0700 (PDT), Patrick Turner

snip,

So the range of voltage movement means he'd need about 10V across the
CCS to allow the voltage to reduce down to 2V and up to 18V which is
OK. 10V x 1A = 10W and is slightly more Pd than the ECL86 pentode
section.

That *is* "as much" isn't it

Yes, all things considered, it would be a good idea to have some
headroom.

Purists might argue it's really a hybrid amp and not 'all tube'.

Ah, but the CCS has infinite impedance and cannot affect the ac
current change anywhere except by indirect means of preventing the
core from saturating, so the amount it could be considered to be
hybrid = 1 / near infinity = approximately zero.

Actually, no, it isn't 'infinite'. It's a lot, all right, but not
infinite. However, that wasn't the point.

It's in the audio path.

We can argue it's transparent, and I might be inclined to agree, but
it's an active solid state thingamabob in the audio path so a 'purist'
might take umbrage with it.

Didn't say *I* would. Just threw it out for consideration.

I agree with you. I might say it IS HYBRID in the minds of those who
argue it to be so.
But I often use CCS anode loads for triode signal stages to avoiid the
triode having to produce power in a DC carrying resistor between anode
and a B+ rail. Such a "devious" practice makes the anode load maybe
20Ra rather than 3Ra and the TDH is reduced 10dB for free, and without
any external NFBb, although because the triode has more internal open
loop gain there is more internal truode NFB applyting itself hence the
reduction of THD with near maximal gain and near CCS loading with very
little current change.

Which is worse? - a triode sleeping with a whory bit of solid state
arse, or a simple country girl resistor?
Or do we insist the triode sleep with a big fat dame with an iron
character with plenty of turns?

A man can't win the arguments, but afaik, a CCS using a transistor or
mosfet *improves the sound*.


You'd need the CCS to track tube current, which might pose some
motorboating problems due to feedback through the transformer and the
cathode bias capacitor time constant.

Well, with the added C coupling to load any GNFB could cause LF
instability because of the added time constant and hence phase shift.
But phase step networks should cure all that. The CCS shoud have no
effect because it has constant current.

GNFB is a different matter. Tracking tube current means you have to
measure it and then that is fed to the secondary which is then coupled
back to the primary. That's potentially a feedback loop. Hopefully not
much of one since the pentode is a current source but, well, something
to keep in mind is all I meant.

OK, GNFB does generate some stability challenges.....

A simple approach might be to (multiplying) current mirror the cathode
into the secondary.

AAAhhhhhh........

Dern handy things, current mirrors Although, that large a
multiplier might pose a challenge or two. No matter what the mechanism
is it would need to be adjustable and then some process for
determining when you got it right. Maybe run a full power 20Hz pilot
tone and adjust for symmetry.

Btw, I almost suggested a choke loaded parafeed like you did in the
other post but that means expensive chokes and, like I mentioned
above, I presume the idea was to save the cost of such things.

The double voltage series push pull parafeed burns off about as much
(don't forget heater) as the secondary CCS so that would be to take
care of 'purist' objections to solid state thingamabobs.

If one does the amp with tubes in series it probably is best to use
only triodes. A suitable type which is cheap and doable by a beginner
is the 6CM5/EL36 which are dirt cheap because nobody wants 'em even
though they'll give a 2A3 a good contest.

6CM5 in triode has µ = 5 and Ra = 600 ohms. Pda for class A can be
18W.

The triode use means you don't have to muck around with screen
supplies.
And of course one can make a simple little SRPP with a pair of 6CM5 to
get about 7W and you only have to drive the bottom tubes. The load is
an OPT and cap coupled from the top tube cathode. Biasing is easy
although a heater winding for the top tube must be biased at the top
tube cathode potential.

Patrick Turner.

[Big Bad Bob]

On 10/07/10 04:21, GRe so witilly quipped:
Has anybody tried or seen a SE topology as shown below?

hmm...

has anyone ever considered a rotating 'phase selector' switch that
clicks over each time you power up the unit? The idea would be to swap
the phases on both primary and secondary each time you power up the amp,
thereby (to some extent) 'equalizing' the magnetization of the core in
both directions to prevent it from building up in one particular
direction (and killing your output). OK, so you'd have to make sure it
was a "one shot on power up" switch, but something like that isn't
impossible to make, like a solenoid that rotates the switch whenever the
unit energizes (you'd do this during warm-up so there's no arcing in the
switch).

I've seen the effects of magnetization before, having once 'flipped'
both output transformer windings (both pri and sec) on an ancient stereo
system that had a dual SE pentode output stage. Prior to the flip, max
volume was barely audible. After the flip, worked perfectly. The
stereo had a nice cabinet, hinge top with turntable (built some time in
the early 1960's I think). Lost track of it years ago, so I don't know
if it still works.

[Patrick Turner]

On Oct 10, 5:40*pm, Big Bad Bob BigBadBob-at-mrp3-
wrote:
On 10/07/10 04:21, GRe so witilly quipped:

Has anybody tried or seen a SE topology as shown below?

hmm...

has anyone ever considered a rotating 'phase selector' switch that
clicks over each time you power up the unit? *The idea would be to swap
the phases on both primary and secondary each time you power up the amp,
thereby (to some extent) 'equalizing' the magnetization of the core in
both directions to prevent it from building up in one particular
direction (and killing your output). *OK, so you'd have to make sure it
was a "one shot on power up" switch, but something like that isn't
impossible to make, like a solenoid that rotates the switch whenever the
unit energizes (you'd do this during warm-up so there's no arcing in the
switch).

I've seen the effects of magnetization before, having once 'flipped'
both output transformer windings (both pri and sec) on an ancient stereo
system that had a dual SE pentode output stage. *Prior to the flip, max
volume was barely audible. *After the flip, worked perfectly. *The
stereo had a nice cabinet, hinge top with turntable (built some time in
the early 1960's I think). *Lost track of it years ago, so I don't know
if it still works.

You have raised an interesting issue of core behaviour. What you are
saying is that SE OPT cores gradually become magnetized in one
direction, and the core mysteriously inhibits signal passage by means
of presumably core saturation or collapse of primary inductance in a
way similar to that caused by a group of shorted turns somewhere.

After working in hi-fi repair industry for 16 years I have never
encountered the phenomena you speak of. I have dealt with very many SE
amplifiers and old radios with SE OP stages and found none to have
core materials requiring phase reversing to "re-juvenate" audio
performance. While my electronics was a hobby when I was a teenager
in the 1960s I never found any of the many SE OPTs I fiddled with
needed reverse phasing to improve audio. A friend and I often used a
6V6 and SE OPT from some old radio to modulate the RF cathode current
of another 6V6 used for an SE RF output stage. With cystal mic and
headphones we made our own radio telephone, albeit on an illegal
frequency band - until the fad wore off and we discovered girls, jobs,
and Life.

One would think I might have noticed misbehaving core materials but
afaik all core materials never become permanently magntized in one
direction.

Patrick Turner.

[Big Bad Bob]

On 10/10/10 01:04, Patrick Turner so witilly quipped:
One would think I might have noticed misbehaving core materials but
afaik all core materials never become permanently magntized in one
direction.

I'd actually read about this happening before, if I remember correctly
(it was a long time ago) that a big problem with class A circuits is
slow magnetization of the core. My solution was to reverse the current
flow rather than replace them (which happened to work), and being in
High School at the time I didn't have much cash to refurbish it
properly. The transformers themselves were relatively small, so I'm
guessing that they were made of cheap or easily magnetized materials, in
lieu of ferrite or high quality iron.

[Patrick Turner]

On Oct 10, 8:12*pm, Big Bad Bob BigBadBob-at-mrp3-
wrote:
On 10/10/10 01:04, Patrick Turner so witilly quipped:

One would think I might have noticed misbehaving core materials but
afaik all core materials never become permanently magntized in one
direction.

I'd actually read about this happening before, if I remember correctly
(it was a long time ago) that a big problem with class A circuits is
slow magnetization of the core. *My solution was to reverse the current
flow rather than replace them (which happened to work), and being in
High School at the time I didn't have much cash to refurbish it
properly. *The transformers themselves were relatively small, so I'm
guessing that they were made of cheap or easily magnetized materials, in
lieu of ferrite or high quality iron.

Reversing the DC flow direction should not hurt anything with regard
to core performance. Wikipedia has some info about "anisotropic" core
behaviour which deals with cores that have a preferred magnetizing
direction of magnetization but apart from mention of this I cold not
find any other info after Googling "permanent magnetization
transformer cores". I didn't see any specific mention of slow/fast
core permability reduction over time in class A SE OPTs. I would have
thought the iron molecules relax fully each time the set is turned off
when there is no DC present.

But there is one slight issue of reversing of anode-B+ connections
because the active end of the primary may go to the primary end at the
largest turn length rather than the smallest turn length so the shunt
capacitance may change to perhaps 30% more which may affect the
stability if there is NFB.

In cheap old radio OPTs there is often only 1 section of P and one
section of S, often with P wound on first and the anode connection
made at the bottom of the bobbin and closest to the core so that the
shunt C is quite low because the bobbin base is 2+ mm thick. The other
end of P is at a steady B+ potential and may be close to S winding and
C does not matter.
In such a case when reversing the anode and B+ connection there is a
big increase in shunt C from anode to 0V, much more than 30%.

Patrick Turner.

[Big Bad Bob]

Patrick Turner wrote:
In such a case when reversing the anode and B+ connection there is a
big increase in shunt C from anode to 0V, much more than 30%.

Apparently the laws of unintended consequences apply here, too.