This page contains.....
Informal chat about the 2323 amp development, use of the 6CM5 or
EL36 and its Ra, gm,
best triode operating points.
6FW5 can be used as direct replacement for 6CM5 but without top cap
Schematic for 23 watt channel, explanation notes,
for power supply with notes and 2006 picture of the amp.

What the heck is JBS-1?
It stands for Junk Box Special One. An old bloke who was to move from
house to apartment sold me a power tranny and two output trannies
and a choke which they didn't find the time to use in a hi-fi project in 1960.
The iron was meant for a a stereo amp using a pair of 6GW8 in UL PP,
and giving about 12 watts with wind behind it.

The transformers were all Ferguson branded which were made in a Sydney
suburb of Chatswood back before the 1960s when Australia still manufactured
electronic products. Most Ferguson OPTs were definitely of their "lo-fi" range.
They produced three OPT model ranges, lo-fi, medium-fi and hi-fi with hi-fi
OPTs rarely ever selling because they'd cost twice the lo-fi price and hardly
anyone took hi-fi seriously, or were too poor, or both. The iron cores are low
permeability crap with high losses which gave high LF distortion in OPTs
and lots of core heating in their power trannys. 

I though a little trio-delic amping would be appropriate and proceeded to test
the iron. Choke was good, but not many Henrys and I eventually didn't use it.
Power tranny ran scorching hot after 2 hours unloaded, Hmm, crummy old iron
in this one, and not enough turns per volt. So I stripped it apart, added lots of other
old iron I had laying about from the junk and re-wound the tranny with more
turns per volt, to give lower Bac. Ah, it run not so hot now, and it would cope
with 4 x 6CM5 at the same idle current as in 6GW8 in UL, but with higher
Ea for triode.

The OPT were low fidelity types. Its always been my bad luck to buy old gear
from guys who were hi-class young misers back in 1960, and who probably had
almost zero idea about what quality was, or that wide band OPTs were necessary
for hi-fi. Nevertheless, although the bandwidth of the OPT wasn't good, and with
saturation occurring at about 35Hz at 12 watts into 8ka-a, and HF cut off at about
20kHz with the 6CM5 in triode.
Obviously to me the OPT also didn't have enough turns per volt, and didn't have
enough interleaving.
The dc resistances were a bit high, and the amount of iron a bit small, but heck the
iron was cheap, and I made room on the chassis to increase the iron if I ever had
the time and acquired better "free" or "spare" OPT.
I was able to stabilize the circuit with critical RC damping networks when 12dB
of global NFB was applied.
Usually with any given OPT, if you have the same dc idle current for larger more
powerful tubes than the tubes it was designed for the OPT will cope OK with the
increase in ac signal voltage and the increased signal current, and will not be
stressed even with twice the applied ac voltage, giving 4 times the original maximum
power intended.
But with twice the signal voltage applied into the rated load the frequency of
saturation will double, so that instead of saturation occurring at say 40Hz it will
be 80 Hz at the higher voltage.
Saturation in OPT is a voltage related phenomena.
So while the amp may have saturation at 40Hz occur at 300 Vrms anode to anode
which is near 12 Watts into 8ka-a, at 2 Watts Va-a is only 126Vrms, so saturation
is at about 16Hz, and below any normal music frequencies, so the music will not
suffer at all because 2 Watts of power is all most people actually use on most days
unless teenagers are present at the volume control.
I was asking to get twice the rated power from the old Ferguson transformers,
so output voltage would rise
to about 1.4 times the design figures, and although Fsat would become higher,
the amp wasn't ever going to be used at high levels.
The 6CM5 triodes would provide greater control of bass and better midrange
dynamics as well as dynamic range.

6CM5 in triode gives = 4, Ra = 600 ohms, gm  =  6.6mA/V.

These figures mean that the anode voltage can swing quite low and unrestricted
by the Ra for Eg = 0V.
I easily got 23 watts AB1 per channel even with such lossy OPT; maybe copper
losses are 12% or more; I never bothered to check that, but the power I got with
Ea at only +375V
was more than a pair of KT66 in triode. 
I run Ia = 50mA, so anode + screen dissipation = 18.7 watts which is ok because
these tube don't go cherry red until about 28 watts is reached. 
Their low data rating is because of the TV use conditions.

There are guys who have extracted 200+ watts with such tubes in RF amplifiers
in class C! I didn't want power lost in cathode resistors with the required 60V
bias across the approx 1.1k Rk.
for each tube so for the class AB fixed bias was used.
I had some old 6CG7 laying around to make a driver input and long tail pair phase
inverter/driver. From these old pulls I used the best measuring ones.
At first I made the JBs as just a power amp but later it became integrated,
and useful to lend out to people curious about tube sound while I fixed their
stuffed solid state amps.

People didn't want to give it back when I rang to say their solid state amps
were repaired.

Anyway, I thought if such a "garbalogical" amp could change people's minds
about their horrible budget Cambridge and Creek amps, then what would
they think with something made with parts which were not junk?

( Then I found out as time went buy,
but ppl's wallets rarely agreed with their brains ).

So the JBS has earned its stripes, and I learnt one of the best kept secrets
about the 6CM5 :-
Although it is a beam tetrode it will sound very good as a triode.
When I first built it and compared it to a pair of Quad-II amps I had
repaired, the 6CM5 seemed to sound better.
The 6CM5 was meant for line output in tv sets, and capable of as much
cathode current as a KT88. But anode dissipation is limited to around
18 Watts in a class A or AB situation, so with two running Pda = 36 Watts,
about 15 Watts of pure class A was available if a high RL was used.
6CM5 could be used as a single ended triode with about 4k anode load for
about 5 watts with Ea = 375V and Ia = 50mA, and methinks it will out
perform 2A3!

Don't try to use the tube in beam tetrode mode; it is too non linear like
most high gm tubes. But In the 1950-60 era Phillips did produce a couple
of PA amps for supermarkets in the 1950s and 60s which used either 4 or
6 in tetrode mode to make 80 or 120 watts respectively. These horrors were
set up as low bias amps tetrode amps with Ea = 330V and low idle bias
current for nearly class B, and they had a lot of NFB. I had a couple of
these amps given to me and they are only suited for "100V lines" and
the OPT has no ability to have the secondaries rearranged to match
4 to 8 ohms.
The Phillips amps worked OK because women disappeared into early
shopping mall stores to and didn't re-appear till hours later looking strange
and dazed and having spent all their husbands earnings on appalling junk
they didn't really need - such was the seductive effect of muzak via tubes!

Ultralinear can be used where the screen taps are at 40% or more.
Ea should never be more than +375V
when in UL or triode, or else the
grid bias needed becomes too negative. I found the tube can runaway
out of control if fixed grid bias exceeds -70Vdc.
So don't be tempted to run them with Ea = 425V or 450V.

Anyway, over the years I have had a lot of enquiries about the 6CM5
schematic which didn't appear on the last edition of the website; only a
picture of the amp as a straight power amp appeared.
So due to demand, here are the schematics of the amp and power supply
and a fresh picture of the amp with the cover grille removed.




Schematic of
            2323 triode amp.

This schematic is a straight forward amplifier which takes advantage of the
nice triode characteristics of the 6CM5 or EL36 to make a PP triode class
AB amp with around 25 Watts output.
Another tube that could be used is the very little known 6FW5 which is an
octal tube and exactly like 6CM5 electronically.
and this is a blessing because top cap
connections are dangerous because people leave them exposed  and some
victim will reach over and touch the +375V and maybe its your grand child.
Maybe there are some stocks of 6FW5 lurking somewhere. There would not
be many on Oz but would be a lot in the US. Meanwhile, the 6CM5/EL36
would have to do. There used to be a 6CM5 in every second TV set until HV
transistors were finally able to be made with a reliable outcome. As solid
state took over from tubes the 6CM5 was one of the last tubes to be retired
in favour of solid state.

Because there is not a really high B+ for the driver stage, V3/V4 to work
from and because the driver has to produce up to about 47Vrms at lowest
possible distortion I took the "dead" grid of the V3/V4 LTP to -25V because
it was available from the zener string which regulates the bias voltage and
which appears in the power supply schematic.

Note the zobel networks used to stabilize the amp even with only 12dB
of global NFB. Such zobels tame the gain and reduce phase shift at HF
where otherwise the gain and phase shift would allow oscillations at HF.
Depending on the OPT chosen and its shunt capacitance and leakage
inductance, the values of R and C shown will not necessarily be used.
Unless you know how to build an amp with NFB so it is critically damped,
you will find your amplifier may oscillate, especially with a 0.22uF cap
on the output without any R load.

So values of  C6 & R12,  C11 & R28, C12 & R29,
C15 & R31, and  C14 & R30 all have to be trimmed
values which ensure stability at HF!!!

There is also a LF gain / phase shift correction network with C5 & R11.
This should always be used regardless of how much inductance the
OPT primary has.

Schematic 2323
            triode amp power supply

This is very close to the original power supply.  If your chosen power
transformer does not have a bias winding then you may use a separate bias
transformer using one taken from a canobolised solid state amp. I have a
few old small transformers with a 240 primary and a few windings which
can be used to make a suitable bias voltage by a voltage doubler or
quadrupler, and then apply RC filtering. The filtering of the bias circuit
above is a bit excessive, but I like quiet bias supplies which retain their
charge when the amp is turned off then on again while cathodes are still hot.
The bias voltage is also shunt regulated but really need not be regulated because
when the mains voltage rises say 10%, The B+ will tend to rise since the B+ is
not regulated. However the combined parallel Ra of all four output tubes is 150
ohms only, so the B+ rise with a change in mains voltage will not be much,
since the source resistance of the B+ is probably much more than the Ra
or the 4 tubes in parallel.
To counter the rise in B+, if the grid bias is also allowed to increase with
a mains rise then the tubes tend to be biased to conduct less current by
increased grid bias. I used to be fanatical about regulating and smoothing
bias circuits but it isn't always necessary in a triode amp. But I also wanted
shunt regulation of the voltage applied to control the constant current source
Q1 MJE340 for the cathode current to V3, V4.  Again, this isn't strictly
necessary if a voltage divider is used to replace the zener string and large
value caps of 1,000 uF are used as bypass caps. Such effective bypassing
with humungous cap sizes is legitimate, does not spoil the music and the
caps for the voltage wanted are small and cheap.

The B+ rail is not regulated. The transistors MJE340 and BU208A make it
look like it is regulated but what is there is an "electronic ripple reducer",
otherwise called a capacitance amplifier but really what the bjts actually do
is act as a giant emitter follower with the two bjts connected as a darlington
pair, with very low output resistance and very high input resistance at signal
frequencies. When I first built the amp it was regulated with a string of zeners
to hold the MJE340 base at about +377 V.
Some years later a friend asked me to lend the amp to him to try with a pair
of horn speakers he'd built which were about 105dB/W/M efficient.
Unfortunately there was hum to be heard and I altered the regulator to what
you see above and the hum disappeared. In fact there is only 10mV of hum
at the bjt output emitter, although ripple voltage at the input of the collectors is
about 11Vrms. So the above active hum reducer acts with an attenuation factor
of about 0.001. To get the same amount of attenuation with a CLC filter with
C1 = 50uF as shown, ( C2 & C3 in series, ) and C10 = 100uF, the needed
choke has to have XL = 1,000 x XC, and if C10 = 100uF, XC at 100Hz
= 16 ohms, so XL must be 16,000 ohms so L is 25H which would have been
10 times more costly than a couple of R&C and two cheap and common bjts
mounted on some scrap aluminium for a heat sink.
The mechanism of the hum persisting in the above emitter follower is
because there is some ac collector hum current current since the ripple of 11V
works into the collector resistance which is perhaps about 10k. HFE of the
pair of transistors is about 200, so about 0.005 mA flows at the base and
since the base resistance is 1k, about 5mV of hum must appear at the base.
The passive filtering of R2 & C6 and R5 & C8 attenuate the 11Vrms of ripple
from the top of C2 by 0.00011 and in any case the resulting 1.25mV of hum
at top of C8 is phase shifted by nearly 180 degrees and probably
counters the hum from the collector to base resistance path.
So about 1.0 mA of ac hum current flows in the collector and so some base
current must flow and since there is a 1k series base resistor, R7, there is a
tiny hum voltage at the base. The 1k is needed to allow the base voltage to be
quickly pulled down by the current in the four 1N4007 diodes in series between
base and output if the output is ever shunted to ground in a fault.
The action of R8, 2.7 ohms acts to allow a voltage of 1.4V to be generated
across itself when I out = 0.5A approx. If this occurs the threshold voltage of
2.5V across the 4 diodes is exceeded and the base is pulled towards ground
and the excessive current should cause the mains fuse to blow.
The resistors R 2 and R5 which total 300k have a dc base input current flow
so that when 230mA flows in the collector-emitter path, base current is about
0.16mA, thus base voltage is about 50V below the voltage at the top of C2.
The transistors don't seem to be able to be killed in the circuit even if a short
circuit occurs.
Complicated? sure is, and much has to be considered and
included in a high voltage regulator or ripple reducer or else the use of
solid state as slaves to the tubes lets the smoke out of the devices, and once
it has come out, it won't go back in!
If the input to the collectors is shunted to ground then current stored in C10
could flow backwards through the bjts. This means instant death to a bjt if the
reverse voltage from emitter to collector exceeds about 6V!! So hence the
diode between base and collector allows any back flow to harmlessly bypass
the bjts. There is a 100 ohm series R4 rated at 10 watts on the heat sink.
there is normally about 23v across this R and about 5 Watts is dissipated.
About 25V is across the bjt, so 5Watts is dissipated in the bjt.
With an increase in current, the voltage across the 100 ohms increases,
and voltage across the bjt reduces, so short circuits kill the resistor, not the bjts.
The maximum current flow possible from the 425V at top of C2 is limited
to 4.25 amps. This is less than the BU208A maximum collector emitter
current rating. Series pass element regulators should always have this feature.

Maybe you are not impressed with the idea of enslaving power transistors
to the whimsical current desires from the tubes. OK, use a choke filtered
B+ instead. See my pages on power supplies and follow your nose to where
I discuss the benefits of using large capacitances and smallish chokes
to get a CLC input circuit with very low ripple with 470uF, 2.5H choke
and 470uF with diode rectifiers. It means though that the B+ winding with
a full wave bridge needs to be about 284Vrms. About 142Vrms is needed
for a doubler circuit with the two input caps at 470uF each which need to
be rated for 250V each, but the same 2.5H and 470uF still follow; the 470uF
can be 450V rated where there is approx 375V across the caps.
Selecting a power tranny with slightly too high HT winding is better than
selecting one which gives a disappointing low B+ voltage.
Having say +400V B+ at 230mA for the 2323 can easily be reduced by placing
series R between the diodes and caps to be charged to reduce the peak
charging currents in high value electros with silicon diodes.
Using a B+ = +350V will give much less power. When I wind transformers
for such amplifiers I usually have about 4 taps at about 15Vrms steps in from
the ends of the HT winding to accommodate different output tubes and be
able to adjust the B+ voltage without using too many series trimming

There is shunt regulation of the anode supply to the preamp input stage using
a string of zener diodes at the first stage of the power amp.
This prevents any possibility of LF instability.

There is no active protection on the amp because it is a Junk Box Special
and I have a lot of spare 6CM5, and if I fuse an OPT then I won't cry too
I'll just have to wreck another few Sundays to make replacement OPTs
which would withstand a short circuited or saturated tube for longer than
the existing trannies would. The wire used in 1950/60s OPT was usually
very thin and prone to easy fusing. The use of thin transformer wire
was an unscrupulous capitalist plot to maximize meager company profits
and have you come back to buy again later.
( But when transistors displaced tubes in 1960, no wonder people happily
threw out their parked and defunct tube amps. And the strange thing was
that although the cost of production of early SS amps was much cheaper
they initially sold for very little less than the tube amps they replaced,
and of course ppl needed two amps, because stereo had arrived ).
If I was keen I would place 200mA fuses into each cathode circuit on
each output tube. These should blow if a tube saturates and thus save the
OPT but maybe not. I am wary of fuses because they don't always blow
 when you want them to. See my other pages on active protection.

The amp in 2006. It gave no troubles after 11 years, and in 2009 it still
goes fine. Even the black felt pen markings on transformer tape is easily readable.

JBS-1 amp, 2006.

2323 integrated stereo triode amp

The power transformer at the rear has a 75mm stack of 38 tongue E&I lams
using questionable core material. Even with the B reduced from about 1.1
Tesla used in 1960 to 0.9 Tesla the core still gets quite warm in hot weather,
but not excessively hot.
I have piles of this iron from old trannies I have stripped down but I will
now only use it in chokes.

The damned ugly aluminium "thinge" in the middle of the chassis at the
front is an early attempt of mine to farnarkle solid state devices to work
as willing slaves to their masters which are the tubes surrounding the
( Farnarkling is a Sunday R&D process which means you keep trying to
get something to work and not let smoke out before dinner time. Usually
such things take to dinner time on Monday to get right. )

The "Thing" is a bit of channel section with a heat sink insulated and
mounted inside the channel. The high voltage transistors with B+ on
the collectors are also well insulated in their mounting so they are double
insulated from the channel which is at 0V. Gradual breakdown of insulation
between bjts at 400V and metal heat sinking at 0V is almost inevitable
at some time and I have fixed a few commercially made amps with
such problems such as a high end Centrepoint amp.Using TWO
insulation washers and plenty of silicon sealant around the TO3
bjt and where the base and emitter pins project through the 5mm
holes in the sink metal is essential to prevent little tiny leaks of current
which can destroy a solid state device so easily. I found the size of the
ripple reducer to be a bit excessive, but at least it runs cooler than if I'd
been a miser and made it too small. 

Keep the home triodes burning!

Back to Index Page.