Yes folks, solid state can be fun, too!
Most of us are probably aware of the drawbacks of the classical OTL
tube circuit.
High output impedance, hard to drive output tubes, DC imbalance at the
speaker connections, many tubes in parallel, high tube wear etc.
Last week, I was fiddling around with an idea I had befo swapping
the output tube stage for a pushpull MOSFET stage.
I started out with the usual floating paraphase splitter with 12AX7,
followed by a common cathode driver stage with 12AU7, driving a stack
of IRF530 MOSFETS, connected in series with the drain out on the upper
half and the source on the lower half.
This worked, but not very good. DC shifts, temperature problems etc.
Then I realized I had some N- and P channel MOSFETS lying
around...........

So I changed the configuration to SRPP input with 6DJ8 and
2SK135/2SJ50 complementary pushpull with the sources at the output.
At the anode of the lower half of the SRPP, I attached 2 caps to drive
the gates, and applied bias with a floating ground.
This floating ground goes to a DC balancing servo circuit, keeping the
output at +- 5 mV DC.
It turned out that an idle current of 400 mA per MOSFET was allowed
for, thereby putting the output stage in class A up until 10 watts.
There's zero global feedback, though I might be temped to add a little
to lower the output impedance some more.
OTOH, adding some more MOSFETS might be an interesting idea, too.
The supply voltage is +- 25 V and rigid smoothing is needed to
attenuate ripple voltage on the supply lines.
The transformer is capable of delivering 10 amps per channel.

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

On the test bench, it puts out nearly 20 Veff in 8 ohms, which equals
to 50 watts sine power, while the same voltage is maintained in 4
ohms.
An ideal companion to my Maggies !
Bandwidth is limited by a slewing network at the input to 200 kHz,
which is adequate.
THD is under 0,1 % near full power, and even lower at 1 watt.

In short, I'm very happy with this new contraption and though the idea
might not be new, it certainly is worth a try.

Next move will be a neat cabinet and some additional MOSFETs in
parallel, but I don't have the 2SK135/2SJ50 anymore, and apparently,
they're getting rare and expensive.
Does anyone know of a substitute for these transistors?
They don''t necessarily have to have TO3 housings, I should be working
on the heat problem anyway. HUGE heatsinks are needed!

--
Sander deWaal
Vacuum Audio Consultancy

Sander,

Magnatec (Semelab Group) has complimentary T03 lateral MOSFET transistors
which I understand use the same die technology as the 2SK135/2SJ50 series.
However, they are available as well in high power configurations, e.g. up to
16A continuous drain current w/ a PD of 250W. They also gave some 500W / 32A
SOT227 devices. Wow.

Semelab has a rep in the U.K: .

- Jon





From: Sander deWaal
Organization: Vacuum Audio Consultancy
Newsgroups: rec.audio.tubes
Date: Sat, 14 Feb 2004 21:58:37 +0100
Subject: A hybrid MOSFET OTL


Yes folks, solid state can be fun, too!
Most of us are probably aware of the drawbacks of the classical OTL
tube circuit.
High output impedance, hard to drive output tubes, DC imbalance at the
speaker connections, many tubes in parallel, high tube wear etc.
Last week, I was fiddling around with an idea I had befo swapping
the output tube stage for a pushpull MOSFET stage.
I started out with the usual floating paraphase splitter with 12AX7,
followed by a common cathode driver stage with 12AU7, driving a stack
of IRF530 MOSFETS, connected in series with the drain out on the upper
half and the source on the lower half.
This worked, but not very good. DC shifts, temperature problems etc.
Then I realized I had some N- and P channel MOSFETS lying
around...........

So I changed the configuration to SRPP input with 6DJ8 and
2SK135/2SJ50 complementary pushpull with the sources at the output.
At the anode of the lower half of the SRPP, I attached 2 caps to drive
the gates, and applied bias with a floating ground.
This floating ground goes to a DC balancing servo circuit, keeping the
output at +- 5 mV DC.
It turned out that an idle current of 400 mA per MOSFET was allowed
for, thereby putting the output stage in class A up until 10 watts.
There's zero global feedback, though I might be temped to add a little
to lower the output impedance some more.
OTOH, adding some more MOSFETS might be an interesting idea, too.
The supply voltage is +- 25 V and rigid smoothing is needed to
attenuate ripple voltage on the supply lines.
The transformer is capable of delivering 10 amps per channel.

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

On the test bench, it puts out nearly 20 Veff in 8 ohms, which equals
to 50 watts sine power, while the same voltage is maintained in 4
ohms.
An ideal companion to my Maggies !
Bandwidth is limited by a slewing network at the input to 200 kHz,
which is adequate.
THD is under 0,1 % near full power, and even lower at 1 watt.

In short, I'm very happy with this new contraption and though the idea
might not be new, it certainly is worth a try.

Next move will be a neat cabinet and some additional MOSFETs in
parallel, but I don't have the 2SK135/2SJ50 anymore, and apparently,
they're getting rare and expensive.
Does anyone know of a substitute for these transistors?
They don''t necessarily have to have TO3 housings, I should be working
on the heat problem anyway. HUGE heatsinks are needed!

--
Sander deWaal
Vacuum Audio Consultancy

PS: Be sure to protect those gates with devices such as zeners, ferrite
beads, stopper resistors, etc. or you'll have one hell of an RF oscillator
on your hands. Also note the high input capacitance of the devices.

IMHO, the Magnatec lateral MOSFETs should sound a lot better than the IRs.

I've got a box of T03 heatsinks with A/C fans (originally for the Hafler
DH-500) if you get in a jam.

- Jon

From: Jon Yaeger
Newsgroups: rec.audio.tubes
Date: Sat, 14 Feb 2004 17:09:05 -0500
Subject: A hybrid MOSFET OTL

Sander,

Magnatec (Semelab Group) has complimentary T03 lateral MOSFET transistors
which I understand use the same die technology as the 2SK135/2SJ50 series.
However, they are available as well in high power configurations, e.g. up to
16A continuous drain current w/ a PD of 250W. They also gave some 500W / 32A
SOT227 devices. Wow.

Semelab has a rep in the U.K: .

- Jon





From: Sander deWaal
Organization: Vacuum Audio Consultancy
Newsgroups: rec.audio.tubes
Date: Sat, 14 Feb 2004 21:58:37 +0100
Subject: A hybrid MOSFET OTL


Yes folks, solid state can be fun, too!
Most of us are probably aware of the drawbacks of the classical OTL
tube circuit.
High output impedance, hard to drive output tubes, DC imbalance at the
speaker connections, many tubes in parallel, high tube wear etc.
Last week, I was fiddling around with an idea I had befo swapping
the output tube stage for a pushpull MOSFET stage.
I started out with the usual floating paraphase splitter with 12AX7,
followed by a common cathode driver stage with 12AU7, driving a stack
of IRF530 MOSFETS, connected in series with the drain out on the upper
half and the source on the lower half.
This worked, but not very good. DC shifts, temperature problems etc.
Then I realized I had some N- and P channel MOSFETS lying
around...........

So I changed the configuration to SRPP input with 6DJ8 and
2SK135/2SJ50 complementary pushpull with the sources at the output.
At the anode of the lower half of the SRPP, I attached 2 caps to drive
the gates, and applied bias with a floating ground.
This floating ground goes to a DC balancing servo circuit, keeping the
output at +- 5 mV DC.
It turned out that an idle current of 400 mA per MOSFET was allowed
for, thereby putting the output stage in class A up until 10 watts.
There's zero global feedback, though I might be temped to add a little
to lower the output impedance some more.
OTOH, adding some more MOSFETS might be an interesting idea, too.
The supply voltage is +- 25 V and rigid smoothing is needed to
attenuate ripple voltage on the supply lines.
The transformer is capable of delivering 10 amps per channel.

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

On the test bench, it puts out nearly 20 Veff in 8 ohms, which equals
to 50 watts sine power, while the same voltage is maintained in 4
ohms.
An ideal companion to my Maggies !
Bandwidth is limited by a slewing network at the input to 200 kHz,
which is adequate.
THD is under 0,1 % near full power, and even lower at 1 watt.

In short, I'm very happy with this new contraption and though the idea
might not be new, it certainly is worth a try.

Next move will be a neat cabinet and some additional MOSFETs in
parallel, but I don't have the 2SK135/2SJ50 anymore, and apparently,
they're getting rare and expensive.
Does anyone know of a substitute for these transistors?
They don''t necessarily have to have TO3 housings, I should be working
on the heat problem anyway. HUGE heatsinks are needed!

--
Sander deWaal
Vacuum Audio Consultancy

"Sander deWaal"


Next move will be a neat cabinet and some additional MOSFETs in
parallel, but I don't have the 2SK135/2SJ50 anymore, and apparently,
they're getting rare and expensive.
Does anyone know of a substitute for these transistors?
They don''t necessarily have to have TO3 housings, I should be working
on the heat problem anyway. HUGE heatsinks are needed!



** Hitachi numbers 2SJ162 and 2SK1058 are similar TO3P devices.





............ Phil

is your schematic available somewhere?
I'm interested in DC servo part of it.
instead SK/SJ you can use 2SK1058 and 2SJ162 .
me thinks that some fellow RAT have them on his site for cheap .
cheers!

--
--
--
.................................................. ........................
Choky
Prodanovic Aleksandar
YU

"don't use force, "don't use force,
use a larger hammer" use a larger tube
- Choky and IST"
- ZM
.................................................. ...........................

"Sander deWaal" wrote in message
...

Yes folks, solid state can be fun, too!
Most of us are probably aware of the drawbacks of the classical OTL
tube circuit.
High output impedance, hard to drive output tubes, DC imbalance at the
speaker connections, many tubes in parallel, high tube wear etc.
Last week, I was fiddling around with an idea I had befo swapping
the output tube stage for a pushpull MOSFET stage.
I started out with the usual floating paraphase splitter with 12AX7,
followed by a common cathode driver stage with 12AU7, driving a stack
of IRF530 MOSFETS, connected in series with the drain out on the upper
half and the source on the lower half.
This worked, but not very good. DC shifts, temperature problems etc.
Then I realized I had some N- and P channel MOSFETS lying
around...........

So I changed the configuration to SRPP input with 6DJ8 and
2SK135/2SJ50 complementary pushpull with the sources at the output.
At the anode of the lower half of the SRPP, I attached 2 caps to drive
the gates, and applied bias with a floating ground.
This floating ground goes to a DC balancing servo circuit, keeping the
output at +- 5 mV DC.
It turned out that an idle current of 400 mA per MOSFET was allowed
for, thereby putting the output stage in class A up until 10 watts.
There's zero global feedback, though I might be temped to add a little
to lower the output impedance some more.
OTOH, adding some more MOSFETS might be an interesting idea, too.
The supply voltage is +- 25 V and rigid smoothing is needed to
attenuate ripple voltage on the supply lines.
The transformer is capable of delivering 10 amps per channel.

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

On the test bench, it puts out nearly 20 Veff in 8 ohms, which equals
to 50 watts sine power, while the same voltage is maintained in 4
ohms.
An ideal companion to my Maggies !
Bandwidth is limited by a slewing network at the input to 200 kHz,
which is adequate.
THD is under 0,1 % near full power, and even lower at 1 watt.

In short, I'm very happy with this new contraption and though the idea
might not be new, it certainly is worth a try.

Next move will be a neat cabinet and some additional MOSFETs in
parallel, but I don't have the 2SK135/2SJ50 anymore, and apparently,
they're getting rare and expensive.
Does anyone know of a substitute for these transistors?
They don''t necessarily have to have TO3 housings, I should be working
on the heat problem anyway. HUGE heatsinks are needed!

--
Sander deWaal
Vacuum Audio Consultancy

Sander deWaal"

The supply voltage is +- 25 V ...

On the test bench, it puts out nearly 20 Veff in 8 ohms,

Interesting that you can get 20VRMS with a +-25V supply...

My book says Vrms = .707 x Vpeak = 0.5(.707 x Vpp) = .35 x Vpp


to 50 watts sine power, while the same voltage is maintained in 4
ohms.

Do you thermally compensate the bias to the MosFets ?


Bandwidth is limited by a slewing network at the input to 200 kHz,
which is adequate.


More than adequate in my books.


Does anyone know of a substitute for these transistors?


Why not use MosFets from IRF ?

They have great MosFets for audio. Easy to work with,
predictable results. I've been using them for years. And they
are commonly available.

Syl

"Syl's Old Radioz"


Do you thermally compensate the bias to the MosFets ?


** There is absolutely no need to thermally comp the Fets concerned.

Hitachi lateral fets have a unique self stabilising bias
characteristic.


Does anyone know of a substitute for these transistors?


Why not use MosFets from IRF ?


** They do not make any lateral power fets.


They have great MosFets for audio.
Easy to work with, predictable results.
I've been using them for years. And they
are commonly available.


** Compared to the Hitachi ones Sander used IRF fets are not in the race
for ease of use in audio.

IRF fets have a large positive tempco so need carefully applied thermal
bias compensation.

Gate threshold voltages need to be matched and source ballast resistors
fitted when devices are operated in parallel.

They do not self protect when overheated and have no internal fuses to
limit damage to others if one device fails.

Also, the range of IRF P-ch fets which produce usable complementary pairs
with the N-ch ones is very limited.



............ Phil

Sander deWaal wrote:

Yes folks, solid state can be fun, too!
Most of us are probably aware of the drawbacks of the classical OTL
tube circuit.
High output impedance, hard to drive output tubes, DC imbalance at the
speaker connections, many tubes in parallel, high tube wear etc.
Last week, I was fiddling around with an idea I had befo swapping
the output tube stage for a pushpull MOSFET stage.
I started out with the usual floating paraphase splitter with 12AX7,
followed by a common cathode driver stage with 12AU7, driving a stack
of IRF530 MOSFETS, connected in series with the drain out on the upper
half and the source on the lower half.
This worked, but not very good. DC shifts, temperature problems etc.
Then I realized I had some N- and P channel MOSFETS lying
around...........

So I changed the configuration to SRPP input with 6DJ8 and
2SK135/2SJ50 complementary pushpull with the sources at the output.

The source follower config reduces the Cin of the mosfets
by a large amount, because most of the Cin of a mosfet is between the gate
and source.


At the anode of the lower half of the SRPP, I attached 2 caps to drive
the gates, and applied bias with a floating ground.
This floating ground goes to a DC balancing servo circuit, keeping the
output at +- 5 mV DC.
It turned out that an idle current of 400 mA per MOSFET was allowed
for, thereby putting the output stage in class A up until 10 watts.
There's zero global feedback, though I might be temped to add a little
to lower the output impedance some more.

In the case of the source follower in class A, each mosfet ses a load
of 16 ohms when an 8 ohm load is connected, so
if the Gm of ther mosfet is 0.9 A/V at 400ma of idle,
then the open loop gain is 14.4, and the gain reduction
is from 14.4 to 0.935, so that's about 14dB of sries voltage FB,
which will reduce THD enough in class A to allow reasonably good listening
at
a few watts.


OTOH, adding some more MOSFETS might be an interesting idea, too.
The supply voltage is +- 25 V and rigid smoothing is needed to
attenuate ripple voltage on the supply lines.

If you used 6 x mosfets, with say 0.8A each, idle diss
is about 20 watts per device, and thus you pull 120 watts from the tranny
at idle, and you need a very good heatsink,
something at least with 30 fins x 40 long, 2mm thick, from a plate
of 300mm x 300mm.
Then you could get up to about 55 watts class A.
The voltage swing max is about 23peak volts, 16.26 vrms.
At maximum class A power, the peak change in load current
is 4.8 amps, so 3.4 Arms, so the load value for 55 watts is
16.26 / 3.4 = 4.78 ohms.
With any load above 4.7 ohms, the power up to clipping will be all class
A.
With all loads below 4.7 ohms, power up to clipping will be class AB1.

The mosfet has a drain resistance of about 220 ohms, Gm = 0.9 A/v,
and hence a U = approx 200. Because mosfets have no current
input to their gates, they can be regarded in the same terms as tubes.
Hence gain, A, = U x RL / ( RL + Rd ).
With 6 mosfets in an aoutput stage in class A, each fet sees
48 ohms as a load when 8 ohms is connected at the output.
Therefore open gain is 200 x 48 / ( 48 + 220 ) = 36.
Notice how it isn't just Gm x RL = 0.9 x 48 = 43.2.
So with 6 mosfets in source follower, the gain reduction
is from 36 to about 0.973, so
the FB amount is over double what one gets with just a pair of fets.
The Cin still remains fairly low, but should be driven with a low
impedance anyway, unless lots of global FB is used,
when the driver stage simply needs to have a high current ability,
even though the driver amp has high output impedance.
This applies with transistor driver stages.

The transformer is capable of delivering 10 amps per channel.

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

Gee, I have never noticed that even good SS amps have better bass
than well concieved trioded KT88 amps.
And tubes still have more subjective transparency, imho.

Nevertherless, You are using a tube driver stage, which
may make a difference.



On the test bench, it puts out nearly 20 Veff in 8 ohms, which equals
to 50 watts sine power, while the same voltage is maintained in 4
ohms.
An ideal companion to my Maggies !
Bandwidth is limited by a slewing network at the input to 200 kHz,
which is adequate.
THD is under 0,1 % near full power, and even lower at 1 watt.

The claim for 0.1% at full power seems a little exaggerated for
the complementary six pack of output mosfets driven presumably by
a single ended triode having to make about 17 vrms of drive
at full power; it alone would have over 1%.

The fets themselves as a sixpack would perhaps have
0.1% thd since the source follower gain and thd reduction is
so great, about 37 times, which indicates the open loop
thd of the output stage might have been 3.7% before
the follower feedback is taken into account.
And about 4% at clipping for a sixpack of fets in common source
is about right.

Last time I tested ONE pair of Hitachi complementary pair of source
follower mosfets
with +/- 25v, I couldn;t get the thd lower than about 1.3%
with the load which gave the maximum class A power.

In most SS amps, the 0.1% you get could be reduced by say 50 dB
of global FB to about 0.0003%.
At 3 watts, it is difficult to measure the thd!

One could say then that the amp is a very honest and uncoloured amp
if operating in class A with lots of FB, like a wire with gain.

But your amp should be capable of 0.03% at 3 watts,
and this should be still quite inaudible, since
its only 1.5mV at 5 vrms output, and the spectra
similarly non spoiling, like one gets in a well concieved
all tube amp.


In short, I'm very happy with this new contraption and though the idea
might not be new, it certainly is worth a try.

Next move will be a neat cabinet and some additional MOSFETs in
parallel, but I don't have the 2SK135/2SJ50 anymore, and apparently,
they're getting rare and expensive.
Does anyone know of a substitute for these transistors?

Exicon mosfets should do very well, also
some made by IRF.
When paralleling the mosfets, use 0.47 ohm non inductive current
sharing source resistors.


They don''t necessarily have to have TO3 housings, I should be working
on the heat problem anyway. HUGE heatsinks are needed!

The chip inside the casings run hotter than the sink.
Always place the devices evenly along the sink, not huddled
together on an angle in the middle of the sink.
I have found that the area of the sink near the fets all together gets
hotter
than the ends of the sink say 120mm away.
But with say 6 fets spread along a 300mm sink, the lead lengths get
long enough to be inductive enough to cause instability at some
HF well into the RF territory, so
gate R of 330 ohms, and ferrite beads are needed,
and perhaps 33pF between each gate at each drain, and very, very careful
attention to earthing paths and rail bypassing.

I find a really well done mosfet amp is more difficult to make than a
basic tube amp.

If you don't need 50 watts, and 20 is plenty,
then a true PP approach can be taken with all NPN
or all PNP devices, as echoed in my circuit at
http://www.turneraudio.com.au/htmlwe...5050mosfet.htm
This uses a diff amp using transistors as the driver stage,
and it exploits the high collector impedance of the driver stage,
in two shunt FB loops which has an effect to reduce the Ro
and thd to about the same as a source follower stage.
But the drive voltage to the output fets is only 0.8vrms
and the PP action gives lower thd than an SE driver stage with
complementary pair.
I am about to change the transistor drivers to 2SK369 fets,
to reduce noise.

However, the connection of the output fets to a transformer
allows class A matching of 4 fets to 5 ohms, for 50 watts.
The way I chose was the hard way, and its far easier to
match the output fets using an auto transformer.
If the auto tranny has just one winding with a CT, and it has
taps in the centre of each half winding, then the turn ratio is 2:1,
giving an impedance ratio of 4:1, so then a 4 ohm load
is seen as 16 ohms by the output fets, and if there were only 2
fets, each sees 32 ohms, and if source follower is used,
then thd is very low since the gain reduction is high.
Then all one needs to do is make sure the drive amp's
thd is low, and you have a reasonable amp.
Only 40 watts needs to be dissipated in the heatsink.
The balanced drive is possible from an LTP, which
is inherently free of large amounts of 2H distortions.
The output stage using a pair of NPN devices is more linear than
a complementary PNP + NPN devices, although at a few watts of class A,
the advantages are negligible, since each device is operating as an SE
amp, and the thd is mostly 2H, and it cancels more at low level than it
does
when extreme current swings occur.

Not too many folks notice the difference between the 4 x mosfet
amp, and an amp with 6 x EL34, both producing
about the same class A power, and using a similar level of NFB.

I still think the tubes have the edge.

Patrick Turner.



--
Sander deWaal
Vacuum Audio Consultancy

If you have a heat distribution problem with your heat sink, use a copper
heat spreader
over the aluninum , this is common with high power RF
devices...................George




"Patrick Turner" wrote in message
...


Sander deWaal wrote:

Yes folks, solid state can be fun, too!
Most of us are probably aware of the drawbacks of the classical OTL
tube circuit.
High output impedance, hard to drive output tubes, DC imbalance at the
speaker connections, many tubes in parallel, high tube wear etc.
Last week, I was fiddling around with an idea I had befo swapping
the output tube stage for a pushpull MOSFET stage.
I started out with the usual floating paraphase splitter with 12AX7,
followed by a common cathode driver stage with 12AU7, driving a stack
of IRF530 MOSFETS, connected in series with the drain out on the upper
half and the source on the lower half.
This worked, but not very good. DC shifts, temperature problems etc.
Then I realized I had some N- and P channel MOSFETS lying
around...........

So I changed the configuration to SRPP input with 6DJ8 and
2SK135/2SJ50 complementary pushpull with the sources at the output.

The source follower config reduces the Cin of the mosfets
by a large amount, because most of the Cin of a mosfet is between the gate
and source.


At the anode of the lower half of the SRPP, I attached 2 caps to drive
the gates, and applied bias with a floating ground.
This floating ground goes to a DC balancing servo circuit, keeping the
output at +- 5 mV DC.
It turned out that an idle current of 400 mA per MOSFET was allowed
for, thereby putting the output stage in class A up until 10 watts.
There's zero global feedback, though I might be temped to add a little
to lower the output impedance some more.

In the case of the source follower in class A, each mosfet ses a load
of 16 ohms when an 8 ohm load is connected, so
if the Gm of ther mosfet is 0.9 A/V at 400ma of idle,
then the open loop gain is 14.4, and the gain reduction
is from 14.4 to 0.935, so that's about 14dB of sries voltage FB,
which will reduce THD enough in class A to allow reasonably good listening
at
a few watts.


OTOH, adding some more MOSFETS might be an interesting idea, too.
The supply voltage is +- 25 V and rigid smoothing is needed to
attenuate ripple voltage on the supply lines.

If you used 6 x mosfets, with say 0.8A each, idle diss
is about 20 watts per device, and thus you pull 120 watts from the tranny
at idle, and you need a very good heatsink,
something at least with 30 fins x 40 long, 2mm thick, from a plate
of 300mm x 300mm.
Then you could get up to about 55 watts class A.
The voltage swing max is about 23peak volts, 16.26 vrms.
At maximum class A power, the peak change in load current
is 4.8 amps, so 3.4 Arms, so the load value for 55 watts is
16.26 / 3.4 = 4.78 ohms.
With any load above 4.7 ohms, the power up to clipping will be all class
A.
With all loads below 4.7 ohms, power up to clipping will be class AB1.

The mosfet has a drain resistance of about 220 ohms, Gm = 0.9 A/v,
and hence a U = approx 200. Because mosfets have no current
input to their gates, they can be regarded in the same terms as tubes.
Hence gain, A, = U x RL / ( RL + Rd ).
With 6 mosfets in an aoutput stage in class A, each fet sees
48 ohms as a load when 8 ohms is connected at the output.
Therefore open gain is 200 x 48 / ( 48 + 220 ) = 36.
Notice how it isn't just Gm x RL = 0.9 x 48 = 43.2.
So with 6 mosfets in source follower, the gain reduction
is from 36 to about 0.973, so
the FB amount is over double what one gets with just a pair of fets.
The Cin still remains fairly low, but should be driven with a low
impedance anyway, unless lots of global FB is used,
when the driver stage simply needs to have a high current ability,
even though the driver amp has high output impedance.
This applies with transistor driver stages.

The transformer is capable of delivering 10 amps per channel.

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

Gee, I have never noticed that even good SS amps have better bass
than well concieved trioded KT88 amps.
And tubes still have more subjective transparency, imho.

Nevertherless, You are using a tube driver stage, which
may make a difference.



On the test bench, it puts out nearly 20 Veff in 8 ohms, which equals
to 50 watts sine power, while the same voltage is maintained in 4
ohms.
An ideal companion to my Maggies !
Bandwidth is limited by a slewing network at the input to 200 kHz,
which is adequate.
THD is under 0,1 % near full power, and even lower at 1 watt.

The claim for 0.1% at full power seems a little exaggerated for
the complementary six pack of output mosfets driven presumably by
a single ended triode having to make about 17 vrms of drive
at full power; it alone would have over 1%.

The fets themselves as a sixpack would perhaps have
0.1% thd since the source follower gain and thd reduction is
so great, about 37 times, which indicates the open loop
thd of the output stage might have been 3.7% before
the follower feedback is taken into account.
And about 4% at clipping for a sixpack of fets in common source
is about right.

Last time I tested ONE pair of Hitachi complementary pair of source
follower mosfets
with +/- 25v, I couldn;t get the thd lower than about 1.3%
with the load which gave the maximum class A power.

In most SS amps, the 0.1% you get could be reduced by say 50 dB
of global FB to about 0.0003%.
At 3 watts, it is difficult to measure the thd!

One could say then that the amp is a very honest and uncoloured amp
if operating in class A with lots of FB, like a wire with gain.

But your amp should be capable of 0.03% at 3 watts,
and this should be still quite inaudible, since
its only 1.5mV at 5 vrms output, and the spectra
similarly non spoiling, like one gets in a well concieved
all tube amp.


In short, I'm very happy with this new contraption and though the idea
might not be new, it certainly is worth a try.

Next move will be a neat cabinet and some additional MOSFETs in
parallel, but I don't have the 2SK135/2SJ50 anymore, and apparently,
they're getting rare and expensive.
Does anyone know of a substitute for these transistors?

Exicon mosfets should do very well, also
some made by IRF.
When paralleling the mosfets, use 0.47 ohm non inductive current
sharing source resistors.


They don''t necessarily have to have TO3 housings, I should be working
on the heat problem anyway. HUGE heatsinks are needed!

The chip inside the casings run hotter than the sink.
Always place the devices evenly along the sink, not huddled
together on an angle in the middle of the sink.
I have found that the area of the sink near the fets all together gets
hotter
than the ends of the sink say 120mm away.
But with say 6 fets spread along a 300mm sink, the lead lengths get
long enough to be inductive enough to cause instability at some
HF well into the RF territory, so
gate R of 330 ohms, and ferrite beads are needed,
and perhaps 33pF between each gate at each drain, and very, very careful
attention to earthing paths and rail bypassing.

I find a really well done mosfet amp is more difficult to make than a
basic tube amp.

If you don't need 50 watts, and 20 is plenty,
then a true PP approach can be taken with all NPN
or all PNP devices, as echoed in my circuit at
http://www.turneraudio.com.au/htmlwe...5050mosfet.htm
This uses a diff amp using transistors as the driver stage,
and it exploits the high collector impedance of the driver stage,
in two shunt FB loops which has an effect to reduce the Ro
and thd to about the same as a source follower stage.
But the drive voltage to the output fets is only 0.8vrms
and the PP action gives lower thd than an SE driver stage with
complementary pair.
I am about to change the transistor drivers to 2SK369 fets,
to reduce noise.

However, the connection of the output fets to a transformer
allows class A matching of 4 fets to 5 ohms, for 50 watts.
The way I chose was the hard way, and its far easier to
match the output fets using an auto transformer.
If the auto tranny has just one winding with a CT, and it has
taps in the centre of each half winding, then the turn ratio is 2:1,
giving an impedance ratio of 4:1, so then a 4 ohm load
is seen as 16 ohms by the output fets, and if there were only 2
fets, each sees 32 ohms, and if source follower is used,
then thd is very low since the gain reduction is high.
Then all one needs to do is make sure the drive amp's
thd is low, and you have a reasonable amp.
Only 40 watts needs to be dissipated in the heatsink.
The balanced drive is possible from an LTP, which
is inherently free of large amounts of 2H distortions.
The output stage using a pair of NPN devices is more linear than
a complementary PNP + NPN devices, although at a few watts of class A,
the advantages are negligible, since each device is operating as an SE
amp, and the thd is mostly 2H, and it cancels more at low level than it
does
when extreme current swings occur.

Not too many folks notice the difference between the 4 x mosfet
amp, and an amp with 6 x EL34, both producing
about the same class A power, and using a similar level of NFB.

I still think the tubes have the edge.

Patrick Turner.



--
Sander deWaal
Vacuum Audio Consultancy

"Syl's Old Radioz" said:

The supply voltage is +- 25 V ...

On the test bench, it puts out nearly 20 Veff in 8 ohms,

Interesting that you can get 20VRMS with a +-25V supply...

My book says Vrms = .707 x Vpeak = 0.5(.707 x Vpp) = .35 x Vpp

Plus AND minus 25 volts..........................and to be honest,
it's more like +-28V (absurd big tranny!).
According to your math, +- 25 gives 17,5 V.
My amp puts out 19,5 V, which I called nearly 20 V. Sorry!

--
Sander deWaal
Vacuum Audio Consultancy

"Choky" said:

is your schematic available somewhere?
I'm interested in DC servo part of it.
instead SK/SJ you can use 2SK1058 and 2SJ162 .
me thinks that some fellow RAT have them on his site for cheap .
cheers!

Thanks for all who responded substitute transistors.
I don't have the schematic online, but the principle is simple:
A lowpass filter with F= 1 Hz, comparator TL071 that feeds the ground
of the symmetrical bias circuit.
I'm pretty sure I've seen this servo schematic online somewhere, but I
forgot where.

--
Sander deWaal
Vacuum Audio Consultancy

Patrick Turner said:

with snips

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

Gee, I have never noticed that even good SS amps have better bass
than well concieved trioded KT88 amps.
And tubes still have more subjective transparency, imho.

Nevertherless, You are using a tube driver stage, which
may make a difference.

Maybe my tube amps weren't as good as I thought ;-)
Thanks for the math BTW, not my strongest point.

THD is under 0,1 % near full power, and even lower at 1 watt.

The claim for 0.1% at full power seems a little exaggerated for
the complementary six pack of output mosfets driven presumably by
a single ended triode having to make about 17 vrms of drive
at full power; it alone would have over 1%.

Blame my measurement setup.......I'll have a look at the calibration
again.
I was surprised too, but well........it sounds good!

HF well into the RF territory, so
gate R of 330 ohms, and ferrite beads are needed,
and perhaps 33pF between each gate at each drain, and very, very careful
attention to earthing paths and rail bypassing.

I did use ferrite beads and slightly different gate resistors as to
compensate for the gate-source capacity.
Without them, oscillations occured when driven with high frequencies.

--
Sander deWaal
Vacuum Audio Consultancy

Jon Yaeger said:

Sander,

Magnatec (Semelab Group) has complimentary T03 lateral MOSFET transistors
which I understand use the same die technology as the 2SK135/2SJ50 series.
However, they are available as well in high power configurations, e.g. up to
16A continuous drain current w/ a PD of 250W. They also gave some 500W / 32A
SOT227 devices. Wow.

Semelab has a rep in the U.K: .

- Jon

Thanks for the info, John.
Some others also pointed me towards the Semelab MOSFETs.
Others named the 2SK1058/2SJ162, which should be identical to the ones
I use now, be it in a different housing.

--
Sander deWaal
Vacuum Audio Consultancy

Jon Yaeger said:

I've got a box of T03 heatsinks with A/C fans (originally for the Hafler
DH-500) if you get in a jam.

Thanks for the offer, but I have plenty of that stuff

--
Sander deWaal
Vacuum Audio Consultancy

You're welcome, Sander.

Concerning the 2SK1058/2SJ162, Erno Borbley of Germany sells them in
carefully matched sets, obviating the need for current sharing resistors
(and the resulting degradation of sound that they entail). I've got a
complete set I bought from Erno in my parts drawer, unused since I got bit
bad by the tube bug, like a moth drawn to a brighter (filament) light . . .

Cheers.

Jon

From: Sander deWaal
Organization: Vacuum Audio Consultancy
Newsgroups: rec.audio.tubes
Date: Sun, 15 Feb 2004 18:03:00 +0100
Subject: A hybrid MOSFET OTL

Jon Yaeger said:

Sander,

Magnatec (Semelab Group) has complimentary T03 lateral MOSFET transistors
which I understand use the same die technology as the 2SK135/2SJ50 series.
However, they are available as well in high power configurations, e.g. up to
16A continuous drain current w/ a PD of 250W. They also gave some 500W / 32A
SOT227 devices. Wow.

Semelab has a rep in the U.K: .

- Jon

Thanks for the info, John.
Some others also pointed me towards the Semelab MOSFETs.
Others named the 2SK1058/2SJ162, which should be identical to the ones
I use now, be it in a different housing.

--
Sander deWaal
Vacuum Audio Consultancy

"Sander deWaal" wrote in message
...

Yes folks, solid state can be fun, too!
Most of us are probably aware of the drawbacks of the classical OTL
tube circuit.
High output impedance, hard to drive output tubes, DC imbalance at the
speaker connections, many tubes in parallel, high tube wear etc.
Last week, I was fiddling around with an idea I had befo swapping
the output tube stage for a pushpull MOSFET stage.
I started out with the usual floating paraphase splitter with 12AX7,
followed by a common cathode driver stage with 12AU7, driving a stack
of IRF530 MOSFETS, connected in series with the drain out on the upper
half and the source on the lower half.
This worked, but not very good. DC shifts, temperature problems etc.
Then I realized I had some N- and P channel MOSFETS lying
around...........

So I changed the configuration to SRPP input with 6DJ8 and
2SK135/2SJ50 complementary pushpull with the sources at the output.
At the anode of the lower half of the SRPP, I attached 2 caps to drive
the gates, and applied bias with a floating ground.
This floating ground goes to a DC balancing servo circuit, keeping the
output at +- 5 mV DC.
It turned out that an idle current of 400 mA per MOSFET was allowed
for, thereby putting the output stage in class A up until 10 watts.
There's zero global feedback, though I might be temped to add a little
to lower the output impedance some more.
OTOH, adding some more MOSFETS might be an interesting idea, too.
The supply voltage is +- 25 V and rigid smoothing is needed to
attenuate ripple voltage on the supply lines.
The transformer is capable of delivering 10 amps per channel.

How does this sound? Well, it sounds like a good triode amp, only with
more "authority".
Bass control is better than with my KT88 PPP in triode amplifiers,
while the transparency is about the same.

**Try using some modern BJTs. Lower distortion, at lower bias currents and
higher current capacity at lower cost. A bit trickier to drive, but more
musically satisfying.


--
Trevor Wilson
www.rageaudio.com.au

"Trevor Wilson" wrote in message
...

**Try using some modern BJTs. Lower distortion, at lower bias currents and
higher current capacity at lower cost.


** Given a pair of Hitachi mosfets or a pair of BJTs used as complementary
followers, the fets win hands down.

No need for any drive devices, any bias compensation, any thermal or any
short circuit protection beyond fuses.


A bit trickier to drive, but more musically satisfying.



** Spoken by a devout snake oil merchant and BJT amp bigot.





............. Phil

"Jon Yaeger"


You're welcome, Sander.

Concerning the 2SK1058/2SJ162, Erno Borbley of Germany sells them in
carefully matched sets, obviating the need for current sharing resistors
(and the resulting degradation of sound that they entail).


** Hitachi lateral fets do not need current sharing resistors.

In commercial amps using them they are rarely seen - designers only
add them to provide current sensing for a VI limiters.



............ Phil

"Phil Allison" wrote in message
...

"Trevor Wilson" wrote in message
...

**Try using some modern BJTs. Lower distortion, at lower bias currents
and
higher current capacity at lower cost.


** Given a pair of Hitachi mosfets or a pair of BJTs used as complementary
followers, the fets win hands down.

**Points:
* The MOSFETs require LOTS more bias current, when used in a zero Global NFB
system, to even APPROACH the linearity of the BJTs.
* The BJTs are more linear, anyway. From top to bottom.
* The BJTs provide more current, at lower cost, than MOSFETs (unless
quasi-comp is used).


No need for any drive devices, any bias compensation, any thermal or
any
short circuit protection beyond fuses.

**All true. BJTs are more challenging to implement, but ultimately a better
choice. Particularly in Zero global NFB designs.



A bit trickier to drive, but more musically satisfying.



** Spoken by a devout snake oil merchant and BJT amp bigot.

**Nope. Just the facts.


--
Trevor Wilson
www.rageaudio.com.au

"Trevor Wilson"

"Phil Allison"

"Trevor Wilson"

**Try using some modern BJTs. Lower distortion, at lower bias currents
and higher current capacity at lower cost.


** Given a pair of Hitachi mosfets or a pair of BJTs used as
complementary
followers, the fets win hands down.

**Points:

* The MOSFETs require LOTS more bias current, when used in a zero Global
NFB
system, to even APPROACH the linearity of the BJTs.


** Totally false assertion.



No need for any drive devices, any bias compensation, any thermal or
any short circuit protection beyond fuses.


**All true.


** So TW has no case left.

For high current boosters driven from a tube stage mosfets are clearly the
devices of choice.




.......... Phil

Sander deWaal wrote:

Jon Yaeger said:

Sander,

Magnatec (Semelab Group) has complimentary T03 lateral MOSFET transistors
which I understand use the same die technology as the 2SK135/2SJ50 series.
However, they are available as well in high power configurations, e.g. up to
16A continuous drain current w/ a PD of 250W. They also gave some 500W / 32A
SOT227 devices. Wow.

Semelab has a rep in the U.K: .

- Jon

Thanks for the info, John.
Some others also pointed me towards the Semelab MOSFETs.
Others named the 2SK1058/2SJ162, which should be identical to the ones
I use now, be it in a different housing.

The important thing for mosfets in a class A arrangement is that
as they get hotter, the idle current reduces, not increases.
The Hitachi 2sk134 I am using do this, and there is no need for
a heat sensing bias amp.
I run them at +34v rails, with each one of the quad seeing a 35 ohm load,
with 0.8 amps per mosfet.
The opt has a ratio of 35 ohms d-d to 5 ohms, ie, 2.65:1 turn ratio.

Patrick Turner.




--
Sander deWaal
Vacuum Audio Consultancy

"Phil Allison" wrote in message
u...

"Trevor Wilson"

"Phil Allison"

"Trevor Wilson"

**Try using some modern BJTs. Lower distortion, at lower bias
currents
and higher current capacity at lower cost.


** Given a pair of Hitachi mosfets or a pair of BJTs used as
complementary
followers, the fets win hands down.

**Points:

* The MOSFETs require LOTS more bias current, when used in a zero Global
NFB
system, to even APPROACH the linearity of the BJTs.


** Totally false assertion.

**Mope. It's a fact of life.




No need for any drive devices, any bias compensation, any thermal
or
any short circuit protection beyond fuses.


**All true.


** So TW has no case left.

For high current boosters driven from a tube stage mosfets are clearly
the
devices of choice.

**Nope. They're the EASIEST choice, not necessarily the best choice.
Naturally, it would be a smarter move to simply drive the Bases with
transistors, rather than tubes.


--
Trevor Wilson
www.rageaudio.com.au

"Trevor Wilson"

"Phil Allison"

* The MOSFETs require LOTS more bias current, when used in a zero
Global
NFB system, to even APPROACH the linearity of the BJTs.


** Totally false assertion.

**Mope. It's a fact of life.


** Spoken like someone parroting misunderstood information.




** So TW has no case left.

For high current boosters driven from a tube stage mosfets are clearly
the devices of choice.


**Nope. They're the EASIEST choice, not necessarily the best choice.


** Spoken like someone parroting misunderstood information.

TW has posted no case - so supplies nothing to debate.

His vague assertions are simply all false.

He has not one, single clue about amplifier design.

After all - he is a notorious snake oil merchant.





............. Phil

Trevor Wilson wrote:

"Phil Allison" wrote in message
...

"Trevor Wilson" wrote in message
...

**Try using some modern BJTs. Lower distortion, at lower bias currents
and
higher current capacity at lower cost.


** Given a pair of Hitachi mosfets or a pair of BJTs used as complementary
followers, the fets win hands down.

**Points:
* The MOSFETs require LOTS more bias current, when used in a zero Global NFB
system, to even APPROACH the linearity of the BJTs.

You are speaking out of context.
The discussion is about class A mosfet designs capable of being driven with
tubes.

A class A bjt output stage would require the same idle currents,
but would then have a large base current, so the output stage would have to be a
darlington
pair emitter follower set-up. And still the base current needs a stout driver
stage.


* The BJTs are more linear, anyway. From top to bottom.

Only because of the series negative voltage FB due to the follower
configuration.
The bjts have higher Gm than mosfets, and the open gain is higher,
therefore thd is less in a follower situation.
But without FB, bjt voltage linearity and output impedance is quite poor.


* The BJTs provide more current, at lower cost, than MOSFETs (unless
quasi-comp is used).

In class A situations, the amount of current being asked of the mosfets
under discussion is not excessive, and well within their capabilities.
It does not matter one iota if the bjt might be able to produce twice the
current of a mosfet.
And in any case, a sixpack of mosfets is capable of 24 peak amps
if each is rated fro only 8 amps.
In a class A situation with a max peak V swing of +/- 28 volts,
and a load which is wrong, say 2 ohms, then 14 amps is easy for the mosfets,
although the circuit would have reverted to class AB with 2 ohms,
to give 200 watts.




No need for any drive devices, any bias compensation, any thermal or
any
short circuit protection beyond fuses.

**All true. BJTs are more challenging to implement, but ultimately a better
choice. Particularly in Zero global NFB designs.

In this case the user of the mosfets wants to use no global FB,
and not even any FB around the driver stage.

The use of a sixpack of mosfets in source follower mode
in class A allows them to operate with high open loop gain, and the
local FB from follower operation reduces the thd about 36 times
from the approximate 4% max in class A in common source mode,
so in follower class A we should get less than 0.1%.
This thd declines with output voltage, and at 2 watts into 8 ohms,
or 4 vrms, which is 12 dB down from 32 watts into 8 ohms,
the thd level should be about 0.025%.
Such figures are entirely adequate.
It is possible to use an OPT with the mosfets, and a simple
toroidal auto type will do, and thus avoid using NPN and PNP,
thus reducing thd somewhat more, and get a better load match,
but the complementary action without an OPT is
linear *enough* when using class A.
In fact, using mosfets in SE class A with a DC load via a large inductor
or OPT is entirely listenable!
It may be possible to get less with bjts in class A, but who gives a toss?

The mosfets don't need thermal/bias controls, and have no base current,
and are thus easier to set up and drive than bjts which convey no useful
advantages in the class A OTL situation.


A bit trickier to drive, but more musically satisfying.

** Spoken by a devout snake oil merchant and BJT amp bigot.

**Nope. Just the facts.

Unfortunately, your facts, such as " but more musically satisfying.", is
anything but a fact,
and just a personal opinion.
We might all wonder what experience you have with designing,
building, and de-bugging any tube or SS amps recently.

Patrick Turner.

**Nope. They're the EASIEST choice, not necessarily the best choice.
Naturally, it would be a smarter move to simply drive the Bases with
transistors, rather than tubes.

Try rec.audio.transistors news group, you might get
a better reception for your ideas.

Patrick Turner.

Patrick Turner said:

The important thing for mosfets in a class A arrangement is that
as they get hotter, the idle current reduces, not increases.
The Hitachi 2sk134 I am using do this, and there is no need for
a heat sensing bias amp.
I run them at +34v rails, with each one of the quad seeing a 35 ohm load,
with 0.8 amps per mosfet.
The opt has a ratio of 35 ohms d-d to 5 ohms, ie, 2.65:1 turn ratio.

This is another thing to try, a OPT.
I ordered some 2SK1058/2SJ162 today, and will use 4 in parallel PP per
channel.
I might need to parallel the 6DJ8 as well, since it seems to me the
capacitive load as presented by the 8 MOSFETS might be a little heavy
to drive.
This also gives me the opportunity to split the heater's voltage and
lift the upper tube to upper cathode level of the SRPP.

--
Sander deWaal
Vacuum Audio Consultancy

Sander deWaal wrote:

Patrick Turner said:

The important thing for mosfets in a class A arrangement is that
as they get hotter, the idle current reduces, not increases.
The Hitachi 2sk134 I am using do this, and there is no need for
a heat sensing bias amp.
I run them at +34v rails, with each one of the quad seeing a 35 ohm load,
with 0.8 amps per mosfet.
The opt has a ratio of 35 ohms d-d to 5 ohms, ie, 2.65:1 turn ratio.

This is another thing to try, a OPT.
I ordered some 2SK1058/2SJ162 today, and will use 4 in parallel PP per
channel.
I might need to parallel the 6DJ8 as well, since it seems to me the
capacitive load as presented by the 8 MOSFETS might be a little heavy
to drive.
This also gives me the opportunity to split the heater's voltage and
lift the upper tube to upper cathode level of the SRPP.

Don't rule out the idea of using a pair of EL84 in triode as the driver
SRPP stage.
Use about 15mA of idle current, and Ea = 220 v.
But that means a 480volt supply.
But an EL84 in triode with transformer drive could be used.
Ra of the EL84 will be 2k ohms, so the 200 pF input C of the output stage
will have a pole at 400 kHz, quite high enough.
Gain is about 18, so for 1v in, the approximate correct drive voltage
is possible to the source follower comp pair output stage.

Patrick Turner.




--
Sander deWaal
Vacuum Audio Consultancy

Patrick Turner said:

Don't rule out the idea of using a pair of EL84 in triode as the driver
SRPP stage.
Use about 15mA of idle current, and Ea = 220 v.
But that means a 480volt supply.
But an EL84 in triode with transformer drive could be used.
Ra of the EL84 will be 2k ohms, so the 200 pF input C of the output stage
will have a pole at 400 kHz, quite high enough.
Gain is about 18, so for 1v in, the approximate correct drive voltage
is possible to the source follower comp pair output stage.

Thought about that, but the nearly 500 V ****es me off.
I've found a pair of ECC288s in my tube supply, they should be able to
do the trick just nicely with a lower voltage. The FETs are scheduled
to arrive on friday, so this weekend's gonna be FUN !

--
Sander deWaal
Vacuum Audio Consultancy

Sander deWaal wrote:

Patrick Turner said:

Don't rule out the idea of using a pair of EL84 in triode as the driver
SRPP stage.
Use about 15mA of idle current, and Ea = 220 v.
But that means a 480volt supply.
But an EL84 in triode with transformer drive could be used.
Ra of the EL84 will be 2k ohms, so the 200 pF input C of the output stage
will have a pole at 400 kHz, quite high enough.
Gain is about 18, so for 1v in, the approximate correct drive voltage
is possible to the source follower comp pair output stage.

Thought about that, but the nearly 500 V ****es me off.
I've found a pair of ECC288s in my tube supply, they should be able to
do the trick just nicely with a lower voltage. The FETs are scheduled
to arrive on friday, so this weekend's gonna be FUN !

The EL86 has U about 14, Ra = 1,400, and Ea could be 150V,
so a srpp needs only 350v. Easier than 500v?

The bootstrapped follower where the R between top tube and bottom
is say 5k instead of 1k makes the linearity better.
The top tube can be biased to about +200v, from a divider and 1M
bias R, and the signal from the bottom tube anode is then cap coupled with
1 uF to tyhe tom tube grid; the benefit is that the bootstrapped follower has
almost the same
low Ro as a CF, which is about 100 ohms with EL84/86 as a CF.
Since you won't be using global FB, its best to keep thd low of the driver
stage,
so that the tubes can do their best.
I am of the opinion that where tubes are used without corrective FB loops,
they ought to be set up and run as linearly as possible, and then they
sound very well.
But a well made IST is another option.
An LTP with CCS tail, with input to one side of the LTP,
other grid grounded, and 1:1 turn ratio IST with a primary CT winding to drive
the mosfets
would be the lowest thd option.
EL84 is best here because it has more gain,
and to get 18 vrms output from a-a, you need 1v g-g
so 1v is still needed if the input is standard unbalanced.
If you were to use EL86,
about 2v input is required.

The IST forms a high impedance load for the tubes;
if the inductance is 100H,
then ZLp = 63k ohms at 100 Hz,
and very much higher at 1 kHz,
so that for most of the band the Lp does not load the tubes to any
great amount, and the only load seen is the bias R to the mosfet
gates, perhaps only 100k ohms.
The mosfets can be biased with a divider between the +/- rails using
2 x 1M and a smaller R between the 1M to get the gates at the right
bias, about +1v and -1v, and then a pair of coupling caps,
say 2 x 1uF can be used to the drive point from the IST
or other bootstrapped follower etc.
Don't forget the series R gate stopper, and perhaps some zener limiting
diodes to prevent stray gate excursions more than the rail voltages.
Basically, you don't want stray tube type voltages applied to the
fet inputs, so these need to be limited in some way.

The output from the sources also ought to have diodes to the rails prevent
back emfs from sending current backwards into the fets.
SS amps don't like the speaker voltage rising more than the rail voltage.
Use some 5A 400v rated types for this.

Patrick Turner.






--
Sander deWaal
Vacuum Audio Consultancy

Patrick Turner said:

The EL86 has U about 14, Ra = 1,400, and Ea could be 150V,
so a srpp needs only 350v. Easier than 500v?

Certainly, and something to remember for the next project :-)

The bootstrapped follower where the R between top tube and bottom
is say 5k instead of 1k makes the linearity better.
The top tube can be biased to about +200v, from a divider and 1M
bias R, and the signal from the bottom tube anode is then cap coupled with
1 uF to tyhe tom tube grid; the benefit is that the bootstrapped follower has
almost the same
low Ro as a CF, which is about 100 ohms with EL84/86 as a CF.

The E288CC stage with 180 ohms Rk would yield 400 ohms, if my math
isn't off.
Still enough to drive 4/2 MOSFETS of the 2SK135/2SJ50 variety, I'd
think.

Since you won't be using global FB, its best to keep thd low of the driver
stage,
so that the tubes can do their best.
I am of the opinion that where tubes are used without corrective FB loops,
they ought to be set up and run as linearly as possible, and then they
sound very well.
But a well made IST is another option.
An LTP with CCS tail, with input to one side of the LTP,
other grid grounded, and 1:1 turn ratio IST with a primary CT winding to drive
the mosfets
would be the lowest thd option.
EL84 is best here because it has more gain,
and to get 18 vrms output from a-a, you need 1v g-g
so 1v is still needed if the input is standard unbalanced.
If you were to use EL86,
about 2v input is required.

Gain isn't a problem, my pre puts out 6 Vtt max on CD and 3...4 on LP.
My current tube amp needs about 2 V as well.
I'm very attracted to the interstage transformer option, especially
since I have some nice centertapped Unitrans on the shelf.
That's worth looking into, though I fear I'll have to take apart the
secondary CT connections wrt. drive phase.

The IST forms a high impedance load for the tubes;
if the inductance is 100H,
then ZLp = 63k ohms at 100 Hz,
and very much higher at 1 kHz,
so that for most of the band the Lp does not load the tubes to any
great amount, and the only load seen is the bias R to the mosfet
gates, perhaps only 100k ohms.
The mosfets can be biased with a divider between the +/- rails using
2 x 1M and a smaller R between the 1M to get the gates at the right
bias, about +1v and -1v, and then a pair of coupling caps,
say 2 x 1uF can be used to the drive point from the IST
or other bootstrapped follower etc.

In the current situation, they're biased from a floating +- 5 V supply
with LM317/337 and low-resistance voltage dividers to provide class
A/AB switching.
I want to keep this feature in the new amp wrt. lower heat production
when not listening (power is always ON).

Don't forget the series R gate stopper, and perhaps some zener limiting
diodes to prevent stray gate excursions more than the rail voltages.
Basically, you don't want stray tube type voltages applied to the
fet inputs, so these need to be limited in some way.

The zeners are there, 24 V across gates and gnd, before the gate
resistors.
Gate resistors are different for P- and N channel FETs, since the
gate/source capacity differs. Added ferrite beads, too.
I needed 470 ohms and 680 ohms for stable operation.
It turned out that class A bias allowed for lower values wrt.
oscillation, but class AB didn't ? ( I can't explain this?)

The output from the sources also ought to have diodes to the rails prevent
back emfs from sending current backwards into the fets.
SS amps don't like the speaker voltage rising more than the rail voltage.
Use some 5A 400v rated types for this.

Thanks for the tip, I'll be sure to put them in!

BTW. re. distortion: you said earlier the driver stage by itself
should have around 1 % THD, but I measured the driver stage again at 1
kHz and the reading was about 0,1 %.
Can you explain this difference?
Measuring setup consists of a Wavetek sine wave osc. with 0,005 %
distortion (specs), a homebrew nulling network with 70 dB
attenuation at 1 kHz and a millivoltmeter/oscilloscope.
And yep, I looked at the right scale :-)

I'll try to measure the attenuation at 2 kHz of the nulling network,
this might be a tad too high and would explain the lower figures.

--
Sander deWaal
Vacuum Audio Consultancy

Sander deWaal wrote:

Patrick Turner said:

The EL86 has U about 14, Ra = 1,400, and Ea could be 150V,
so a srpp needs only 350v. Easier than 500v?

Certainly, and something to remember for the next project :-)

The bootstrapped follower where the R between top tube and bottom
is say 5k instead of 1k makes the linearity better.
The top tube can be biased to about +200v, from a divider and 1M
bias R, and the signal from the bottom tube anode is then cap coupled with
1 uF to tyhe tom tube grid; the benefit is that the bootstrapped follower has
almost the same
low Ro as a CF, which is about 100 ohms with EL84/86 as a CF.

The E288CC stage with 180 ohms Rk would yield 400 ohms, if my math
isn't off.

Ro = 400 ohms is OK.


Still enough to drive 4/2 MOSFETS of the 2SK135/2SJ50 variety, I'd
think.

Since you won't be using global FB, its best to keep thd low of the driver
stage,
so that the tubes can do their best.
I am of the opinion that where tubes are used without corrective FB loops,
they ought to be set up and run as linearly as possible, and then they
sound very well.
But a well made IST is another option.
An LTP with CCS tail, with input to one side of the LTP,
other grid grounded, and 1:1 turn ratio IST with a primary CT winding to drive
the mosfets
would be the lowest thd option.
EL84 is best here because it has more gain,
and to get 18 vrms output from a-a, you need 1v g-g
so 1v is still needed if the input is standard unbalanced.
If you were to use EL86,
about 2v input is required.

Gain isn't a problem, my pre puts out 6 Vtt max on CD and 3...4 on LP.
My current tube amp needs about 2 V as well.
I'm very attracted to the interstage transformer option, especially
since I have some nice centertapped Unitrans on the shelf.
That's worth looking into, though I fear I'll have to take apart the
secondary CT connections wrt. drive phase.

Hmm, why take apart any part of the trannies?



The IST forms a high impedance load for the tubes;
if the inductance is 100H,
then ZLp = 63k ohms at 100 Hz,
and very much higher at 1 kHz,
so that for most of the band the Lp does not load the tubes to any
great amount, and the only load seen is the bias R to the mosfet
gates, perhaps only 100k ohms.
The mosfets can be biased with a divider between the +/- rails using
2 x 1M and a smaller R between the 1M to get the gates at the right
bias, about +1v and -1v, and then a pair of coupling caps,
say 2 x 1uF can be used to the drive point from the IST
or other bootstrapped follower etc.

In the current situation, they're biased from a floating +- 5 V supply
with LM317/337 and low-resistance voltage dividers to provide class
A/AB switching.
I want to keep this feature in the new amp wrt. lower heat production
when not listening (power is always ON).

????, not sure what your schematic is.



Don't forget the series R gate stopper, and perhaps some zener limiting
diodes to prevent stray gate excursions more than the rail voltages.
Basically, you don't want stray tube type voltages applied to the
fet inputs, so these need to be limited in some way.

The zeners are there, 24 V across gates and gnd, before the gate
resistors.

What happens if the negative drive voltage drops below 0V?
Normal practice is to have two x (zener + diode back to back),
say 5v zeners, between each gate lead and the output point.
This limits the voltage applied between gate and source.



Gate resistors are different for P- and N channel FETs, since the
gate/source capacity differs. Added ferrite beads, too.
I needed 470 ohms and 680 ohms for stable operation.
It turned out that class A bias allowed for lower values wrt.
oscillation, but class AB didn't ? ( I can't explain this?)

Class AB mosfets can be more unstable when bias current is low.



The output from the sources also ought to have diodes to the rails prevent
back emfs from sending current backwards into the fets.
SS amps don't like the speaker voltage rising more than the rail voltage.
Use some 5A 400v rated types for this.

Thanks for the tip, I'll be sure to put them in!

BTW. re. distortion: you said earlier the driver stage by itself
should have around 1 % THD, but I measured the driver stage again at 1
kHz and the reading was about 0,1 %.
Can you explain this difference?

0.1% is a good result for 18vrms output from a SRPP stage.
Normally one gets more than this.


Measuring setup consists of a Wavetek sine wave osc. with 0,005 %
distortion (specs), a homebrew nulling network with 70 dB
attenuation at 1 kHz and a millivoltmeter/oscilloscope.
And yep, I looked at the right scale :-)

I'll try to measure the attenuation at 2 kHz of the nulling network,
this might be a tad too high and would explain the lower figures.

The bridged T null circuit is the best simple circuit, but suited to
low impedance drives from outout stages.
I built a simple complementary emitter follower buffer with
less than 0.005% thd to measure thd of higher impedance circuits without
weighing them down with a load.

The bridged T network gives more than a -60 dB attenuation of the 1 kHz,
and leaves all the rest of the frequencies alone, when properly set up.
I have a switchable RC high pass filter also to remove hum, placed after the
bridged T and switchable, since hum makes it difficult to see the thd on the CRO
sometimes.

I like to *see* the thd, and measure it at the same time.

The bridged T consists of an air cored L of about 100 mh,
wound with 0.3mm wire on an empty solder spool.
Then the capacitance for resonance at 1 kHz will be 0.253 uF
This is doubled, and then the two caps of 0.506 uF are seriesed,
and connected across the L.
2 x 0.47 uF caps can be used with careful added trimmer caps
to make sure both sets of caps are equal value, and for the deepest null.
The resonance must be within a few cycles of exactly 1 kHz.
The active input is between one end of the LC circuit and ground.
Output is from the other end of the LC circuit and ground, and to a
high impedance meter/cro.
The join between the two seriesed caps is taken to ground via
two pots, one large value, 10k, and one small, say 1k,
and this gives you a course adjustment for a null, and fine adjustment,
because when trying to sift out 0.01% or less in any signal is difficult.
I have a low noise amp to amplify the output of the null filter by 20 dB,
so distortion as low as 0.005% can be seen on the cro
with a voltage sample as low as 0.5vrms.
We are talking about measuring down to 25 uV with some accuracy.
Trying to exceed this with simple SE test circuitry tends to
be too ambitious, and prone to noise and hum.
This distortion amp is a pair of low noise opamps with a 30 dB high pass filter
with a pole
at 1.5 kHz, so 1 kHz is further attenuated by 20 dB after the LC null filter.
There is a low pass filter active above 10 kHz, and the reduced bandwidth makes it
easier to see the the distortion on the cro, and be sure of what I am measuring.

My oscillator is a simple opamp based wien bridge type,
followed by a discrete component SS amp which only has gain
at around 1 kHz due to the feedback network employed,
so the max low distortion voltage is as high as 8 vrms.
There is a switched level attenuator to allow 4 ranges of output voltages
to be easily suited to the amp under test.
The output pot then gives easy fine adjustment of each range of output voltages.
The lowest range has the lowest thd at 0.002%.
For another 30 dB attenuation of thd in the oscillator signal I have a
simple 2 stage LC parallel tuned filter, but its insertion loss is 15 dB.
Nevertherless, if I want a volt with 0.0002% as a test signal,
then I have one.

It did take several attempts to get this thd detector designed and working
reliably.
Its contained in mild steel sheet box, with a remote power supply.
Don't be tempted to use any ferrite cored or other iron cored
inductors in test gear; it always adds 3H distortion.

Regards,

Patrick Turner.



--
Sander deWaal
Vacuum Audio Consultancy