SE32 with 13E1, CFB, 2012 version.

Image1, May 2012.
Mr Turner has more to say about SE32 amps....

patrick1-small-may2-2012.jpg

Schematic 1, SE32, April 2012 version......

schema-se32-amp-sheet1-13e1-33percent-cfb-25april2012.gif

The above Sheet 1 Mono Amp has similar operation to earlier SE32 from 2008.

The differences are :-
+750Vdc rail
and 25k RLa for dc to EL34 is replaced with 60H Choke + 5k
resistance from B+ rail = +509V. EL34 Iadc increased from 16mA to 24mA.
EL34 THD reduced due to high RLa ohms and to more linear working point.
EL34 Rk has been fully bypassed, to lower the effective Rout of the stage,
thus improving micro-dynamics at higher F. Gain is slightly increased.

The 68k RLa for dc to 6SL7 is replaced with CCS using 1 x MJE350, thus
much increasing the Ia and anode load ohms to increase gain from 48
to 64, while much reducing THD.

13E1 cathode R&C bypass network and CFB OPT winding is changed.
Eg2 increased from 170Vdc to 200Vdc.

Active automatic bias voltage adjustment is applied to 13E1 grid to allow for
variation in samples of tubes selected for use and for tube aging and variations
in static grid current causing grid bias change and elevated Ia.

Many other minor changes and tidying up of circuitry.   

The open loop voltage gain of the whole amp is increased by a total of about +4.5dB.
The amount of global NFB has been kept the same at about 9dB by means of changing
the GNFB resistance divider to reduce ß, ie, the fraction of output fed back to V1 cathode.
The 2008 amp needed 0.88Vrms input for clipping with 5 ohm load, but now only 0.57Vrms
input is needed so that preamps are not needed and signals from a CD source may fed
directly to the amp via a 20k log pot mounted in a metal box with source selector switch,
in what is also known as a "passive preamp", ( which is misleading, because nothing
is amplified or passed through any active device within such a box. )

A fraction of the EL34 cathode bias voltage Ek at point Z is taken to the protection circuit
in case of an EL34 having bias failure.

The 13E1 cathode biasing RC network has been arranged so R&C network is grounded
with local CFB winding on OPT now between cathode and R&C network.
A fraction of the Ek is taken from Point Y to the active auto biasing and protection
circuit input points shown on SHEET 2 PSU.

The variable bias voltage for 13E1 grid from BC327 on SHEET 2 is applied at point X.

C16 is added to bootstrap the bias Rg R22, 47k. This resistance then effectively becomes
a much higher value of ohms at signal F of 280k, so its loading effect on EL34 becomes
negligible. This reduced the 2H of EL34 enough to give a welcome reduction of overall
2H at high speaker load values where 2H of the EL34 adds to the 2H of the 13E1, so
that 2H cancelling does not occur. The effect on 2H of bootstrapping or not bootstrapping
R22 was measured.  The bootstrapping gave better overall results than not bootstrapping
so it has been retained. The increase in screen supply voltage gave slightly less THD with
low value loads, but the Rk at 13E1 needed to be increased from 185 ohms to 232 ohms
to get a slightly higher Ek so that the centre value grid biasing voltage remains at 0V with
Ia at 150mAdc. There is a slight reduction in maximum PO but the 13E1 should last
longer and be less likely to overheat.
Using a CCS with a transistor to replace R30,R31 Rk was considered but thought to be
unnecessary over-engineering. 

Schematic 2, SE32 PSU, April 2012 version......

schema-se32-psu-sheet2-13e1-33pcnt-cfb-25April2012.gif

The above Fig 4 PSU sheet 2 schematic needs little explanation for the basic rail
voltages supplied by the main large power transformer.
There is a second small auxiliary 5VA power trans shown which powers the Bias+Protect
schematic. There are two inputs to the circuit from resistance dividers in the cathode
R&C biasing networks of EL34 and 13E1.

The active protection is based on using a faction of Ek of EL34 or 13E1 to turn on a
sensitive gate SCR, C106D, if too much Idc flows between cathodes and 0V. The large
mains transformer has its mains primary winding switched open with a relay operated
by the SCR. The "on" blue LED goes out and red "fault" LED turns on. Owners can try
to reset the amp by just turning off, waiting 2 seconds, then turning back on.
But if an underlying problem remains present, the amp will just turn off again.
Over many years I have discovered how much damage has been stopped in many
amps where tubes have failed or shorted speaker leads or faulty speakers have been used.

The new addition in this bias+protect schematic is the use of Q1, a pnp BC327 transistor
which constantly controls the Idc in 13E1 in addition to the R&C cathode biasing network.

The Y input from 13E1 R&C bypass network is taken from join between R30 & R31
on amp sheet 1 and applied to a base of Q1, BC327, through a base current limiting R of
470 ohms.
This voltage normally about +8.2Vdc. The BC327 emitter has 680 ohms to a fixed +12Vdc.
The BC327 collector has 4k1 taken to a -17Vdc supply rail.
The bias current for the BC327 is 5mAdc.

The 13E1 grid bias voltage is derived from the collector voltage of Q1, and sent out
via path X to a 33k then to bias resistance of 47k
The Q1 collector load R1, 4k1 is bypassed with C1 4u7, thus the 33k and 4u7 form a
filter preventing high level audio signals at 13E1 cathode appearing at Q1 collector.
There is about 0.022Vac ripple voltage at -17Vdc rail and at Q1 collector, and this is
prevented from appearing at the 13E1 grid because join of 33k and 47k is bypassed
with the 2uF bootstrapping C16 on sheet 1. 
In effect, the whole arrangement shown prevents unwanted signal flows in 2 directions.

The gain of Q1 is highest at very low F, and only about 5.5.
If the Idc flow from 13E1 increases say 10mA, then Vdc at point Y will rise by +0.5Vdc.

This will appear at Q1 base and be amplified x 5.5 to cause a correction voltage of
-2.75Vdc to be applied from point X to the grid. If the tube transconductance = 20mA/V
then the reduction of tube current would be -55mAdc. In fact, the Q1 acts to regulate the
Ek Vdc appearing at 13E1 cathode, and thus keep Ek more constant than if a much larger
value of Rk was used, or if a an active cathode current sink were used.
For example, I tried a couple of CV6045 of unknown brand and with the ordinary cathode
R&C network used for the 13E1 made by ITT. The Ia went up to 180mA, and Pda went to
83Watts, which is TOO MUCH. With the active Q1 bias regulator, Ia remained less than
160mA, with a few -Vdc applied to grid, and Ek remained close to where it should be.
In other words, the Q1 circuit acts to provide DC feedback. The time constants chosen
for R and C parts ensure the bias trimming circuit is not unstable at some low F which
can so easily be a very real problem if the time constants are ill-chosen, or the the gain
of the transistor is too high to try to get absolutely perfect bias control. 

The arrangement for 13E1 cathode biasing and screen supply has all the current supplied
to the screen and through the shunt regulating zener diodes. So if any variations to Ig2
occur, then Ek will not change as a result. It the screen was ever to short circuit to
cathode, or more likely, the bypass caps between screens and cathode short
circuit, then maximum current through screen supply resistance R32, 23k5, sheet 1,
is 21mA, only 10mA more than normal. The tube would not work if screen supply voltage
reduces to the same as Ek at cathode.

Without the active grid biasing adjustment, the bias current of 13E1 is completely dependent on
the R&C biasing network. This R&C bias network does not have a very high amount of resistance,
so some help to avoid problems with biasing from one lone easy-to-get transistor is very welcome,
and in fact, forgivable, because try hard as you wish, but its presence is utterly inaudible, while
helping the music to be better sounding, just for you.

The other feature of my new biasing circuit has the biasing R&C network subject to all the screen
current plus the current through the zener diode shunt regulators. At idle, about 12mA flows from B+
to screen and zener diodes, During normal operation, the total of 12mA does not vary even if screen
current input were to vary between a normal 4mAdc at idle to 12mA at high levels.
The shunt regulators tend to keep Ek more constant than if they were not used, or if zener diode
current was not included in in the flow through the R&C biasing network as shown for the
2008 schematic. 

The other source of class A tube amp bias problems can be due to mains voltage variations.
Triode and UL amps are most prone to enduring higher than wanted Ia with abnormally
high mains voltages. If an amp has been designed badly with mains primary meant for only
110Vac to suit USA and 220Vac for everywhere else, then just 2 x 110Vac windings can be
in series or parallel to suit the national voltage. But here in Oz, I have often measured
255Vrms and when used with an amp with 220V transformer, the heater voltages rise from
the correct 6.3Vac to 7.3Vac, and Vdc can rise from the intended +400Vdc to +463Vdc,
and if the Idc flow was a normal 70mA in an output tube with a well regulated bias voltage
supply, then Ia might rise from 70mA to 85mA, and tube Pda rise from 28Watts to 39Watts,
and this can place tubes very close to their Pda limit. With many hi-end and low-end 
brand-name amps I have seen tubes fail due to overheating when the amps are used
here.
Therefore, it is WRONG to regulate grid bias voltage supplies because if the mains
voltage is too high, then you want the bias to increase with mains voltage increase
and this tends to compensate the effect of having a high mains voltage.
Alternatively, if the mains voltage is lower than normal, the grid bias voltage may be
less to allow more Ia to flow.
In my schematic above, the +/- 17Vdc rails are not regulated, and if the mains voltage
is high then -17Vdc rail becomes more negative, and tends to reduce excessive Ia.
At dc operation, the Ra of the 13E1 tube is very high, perhaps 10,000 ohms because
the screen is shunt regulated and kept at a constant voltage above the cathode
voltage. With 13E1 used in triode or UL, Ra is less than 1,000 ohms, and a rise
of 50Vdc in Ia would cause a rise in Ia of maybe 50mA which could make Pda
very excessive.
With my active biasing adjustment, and with shunt regulated Eg2, there will be few
problems with any variations in mains voltages.

The 2 amps in which this bias scheme has now been employed were brought to me
a couple of months ago for a service, the first since 2008. I advised my client that
more could be done to improve performance, so the 2012 schematics were
evolved.
The amps had been used very often, and the owner sometimes had them running
for days on end. A colleague told me the rated tube life for 13E1 is around 2,500 hours.
The first pair of 13E1 had lasted from 1997 to about 2005, and this last pair from 2005
to now, and from what I know, the hours of use have much exceeded 2,500.

All 4 tubes used since 1997 were ITT brand. In the two last tubes, the operation
of both were flawless and without reverse grid current or fading emission and both
were able to give a full 31Watts+ of power, and bias remained stable. 
However, the gettering in one tube shows considerable browning and aging, while in the other
the gettering has almost all become transparent but muddy brown, so both these tubes
will be replaced with NOS with nice clean bright silver gettering. The result of all the alterations
should raise the sound quality to be equal or superior to my SE35 amps with a quad of EL34
output tubes.

Gettering wear indicates tube condition....
Image 2, may2012.

Looks unused                              Looks a bit worn                    Looks worn out
13E1-worn-getter2-may2-2012.jpg

On the left side of the picture, there is a NOS 13E1, never been used.
The two other tubes were installed in around 2006, and they show classic signs of
gettering change from the clean bright silvering on unused NOS tube.
The metallic content of the worn tube gettering has partially combined with gas entering
the tube through join between base pins and glass and/or from slow release of gas
from within metal parts inside the tube. Both the worn tubes remain capable of
producing full power, but only one now has begun to have a slightly positive grid voltage
relative to applied grid bias voltage. The value of grid biasing resistance is a very
important design consideration, and I have seen some amps such as Quad-II with
bias R values much too high at 680k. Hence KT66 in Quad-II with slight tube aging
may have very unbalanced Ia because one tube has perhaps +5Vdc at its grid when
it should be less than 0.5V, which would be possible if the Rg was 68k instead of
680k. I chose Rg at 47k for 13E1, and much lower than many other makers might
use, but it HOLDS the bias down. Reasons for high value Rg are due to having
very week driver tubes such as EF86 in Quad-II with Ia less than 1mA, and
to get the wanted gain the load values much be kept high due to very low tube Gm.

In not a huge amount of time, the worn tubes with gettering nearly all oxidized,
tubes will develop increasingly positive grids, and at some point the gas inside
the tube cannot be further kept low, and the vacuum "hardness" is lost. 
Gas will increase, and tubes will begin to conduct much more Idc than usual and then
overheat, as determined by Pda = ( Ea x Ia ) Watts and they may melt down
internally, or cause glass to melt, or glass to crack, and perhaps cause damage to
PSU and/or OPT. Sometimes internal grid wires will warp and anode will be shorted to
cathode and then high PSU current is drawn and a mains or HT fuse will blow.
But all too often, amp damage occurs well before a fuse blows.

My active protection circuit will prevent such pyrotechnic melt down by turning
the amp off safely and automatically.

The active bias adjustment circuit will try to bias the grid more negatively as the Ik
increases due to a positive going Vdc, thus the tube life is extended maximally even when
tube failure is not far away. The biasing circuit cannot keep old tubes working with
enough Eg1 compensation, and eventually the protect circuit will work unless an owner
sensibly replaces the tubes before they degrade so badly.

What you cannot see in the picture above is that the very good looking unused 13E1
is completely useless. When I tried it the screen current was about 12mA instead
of the correct 4mAdc for the value of Eg2 with respect to Ek. Screen current is always
a worry with any multigrid tube, and of course in every batch of NOS tubes bought so
carefully from someone@somewhere there will be an occasional dud, and here is one.
But the tube caused no damage when tested in the SE32 because the high Ig2 causes
the shunt regulated Eg2 to just sag down to a lower level so the tube conducts less than
normal Ia, and nothing is damaged, although with only 60% of normal Ia, the power output
is much reduced and THD becomes high at high music levels.

So nobody should ever assume a newly made or NOS tube is working just fine after
plugging one in. Regular servicing is the answer, and maintaining a well paying kind
of respect for your local tube amp fixer-upperer. .

Image 3, close up of nearly worn out 13E1 tube...
13E1-worn-getter1-may2-2012.jpg

You can see how bad the worst of two worn tubes have become after years of wonderful music.
There is a reflection of a window on the top left of the picture, but very nearly all gettering
metal has been used up on the two sides of the tube. You can see the round gettering
rings which were fitted when the tube was made, and it was from these rings that the metallic
gettering was sprayed onto the glass internally under a vacuum during manufacture.

NOTE, Gettering changes shown above can occur in many other tube types, and one may
see where gettering is "nearly used up" and if the tubes have been used for 5 years it
is very wise to replace them. And NO, I do not have shares in tube making companies.

I suggest everyone Google "vacuum tube manufacture" to learn more about how
tubes are made.

Some images for SE32, 2012.........

Image 4. Two SE32 monoblocs on bench, without covers.
SE32-2onbench1-no-covers-may2-2012.jpg
Note capacitor enclosure bottom left. Easy access for setting for OPT load match on
board on OPT for sec winding terminations.

Image 5. Two monoblocs on bench without covers.
SE32-2onbench2-no-covers-may2-2012.jpg
This shows the layout of parts on top of chassis. Fully shielded tube is 6SL7, with 2 reddish tube
dampers fitted by owner ( which don't do anything IMHO ). EL34 and 13E1 have dampers also
fitted. 13E1 has copper wire used to stop tube falling out of socket which is possible due to
small socket size and weighty tube.
Iron wound components are OPT, left rear, Big mains PT far right, anode choke for EL34
at left front, nearest, and 4H B+ filter choke between PT and anode choke. Small 5VA
tranny for bias+protect circuit is under chassis. The six B+ 470uF x 400V rated electrolytic caps
are mounted in a case behind the mains PT. 

Image 6. Under chassis view.
SE32-underchassis-may3-2012.JPG

Image 7. Close up bias+protect board.
SE32-protect-board-may2-2012.jpg


Image 8. Close up 6SL7 and EL34 board. Rather crammed.
SE32-input-board-may2-2012.jpg

Image 9. SE32, covers on.
SE32-2onbench3-with-covers-may2-2012.jpg

Happy soldering ! happy swearing ! happy stopping of the amplifier smoking ! ,
and then happy listening!
:-)

To SE32, 2008.

To power amps.

To Index page.