This page has more information about details of a push pull 100W Ultralinear class AB1 monobloc amp.

I did build and sell a pair of 100W UL amps in 2004, and I composed the webpage about them at
monobloc 100W ULAb1 amp.
Emails from ppl around the world have propelled me to compose this new page with properly drawn
schematics using MSPaint instead of scans of drawings on paper. Between
2004 and 2014 I learnt
a more about tube amps so this page includes more than the 2004 page.

This page has notes about :-
Fig 1.
Schematic Sheet 1, 100W ULAB1 Amplifier, 1 x 6SN7 input, 2 x 6SN7 driver,
and 6 x EL34 in UL output stage.
This is a r
e-drawn schematic of 2004 with minor improvements.

Graph 1. Output power vs RL, with THD approx at clipping levels.

Table 1. Use of Hammond 1650T with real world speaker impedance values and how that
affects the sound quality.

Fig 2. Schematic Sheet 1, REVISED 100W ULAB1 amp, 1 x 6CG7 input, 2 x EL84 triode driver,
and 6 x EL34 / 6CA7 output stage.
This is a new schematic for 2014 with major tube changes and includes BALANCED FIXED BIAS.

Fig 3. Schematic Sheet 3, Protection, Delayed B+, Bias Balance Indicator, Clipping indicator.
This has a much improved amplifier management circuit to control behaviour of tubes during
times when they may become troublesome due to tube aging, short circuited speakers and cables,
owner stupidity, early random tube failure, use of dodgy NOS tubes maybe 60 years old,
etc, etc, etc. You don't want to see your house burn down do you?

Fig 4. Schematic Sheet 2, REVISED power supply schematic rated for 380Watts and shown here
for 6 x EL34 / 6CA7. This power supply can be applied for use with a wide range of output tubes
including 6 x EL34, 6CA7, 6L6GC, KT66, 807, 5881 in a 100W monobloc, or for TWO channels
each 50W with 4
x EL34, 6CA7, 6L6GC, KT66, 807, 5881.
The 380W PSU could suit 100W monoblocs with 4 or 6 x 6550, KT88, KT90, KT120,
or TWO 50W channels, each with 2
x 6550, KT88, KT90, KT120.

Fig 5. Schematic Sheet 1, 100W UL AB1 amp 1 x 12AU7, 2 x ECC99, 6 x EL34.

Fig 6. Schematic Sheet 1,
100W UL AB1 amp 1 x 12AU7, 2 x ECC99, 4 x KT120.

Fig 7. OPT bobbin winding details and core details.

Fig 8. OPT Primary and Secondary terminations and many available load matches.


Fig 1. Re-drawn 2004 schematic of 100W amp with 6 x EL34, with minor improvements.
The Schematic of Fig 1 is very similar to the 2004 100W UL Ab1 at
monobloc 100W ULAb1 amp
general description of how all the amps on this page work remain the same, although I recommend
the idea of using balanced biasing which is detailed below.

Anyone building a 100W PP amp as I have shown here MUST use use my schematic as a GUIDE ONLY.
Unless you can source exactly the same PT and OPT, you cannot make an exact copy of what I have
achieved, and the best most DIYers of manufacturers can do is choose PT and OPT which comply
with basic requirements. The ready made odd the shelf purchased PT and OPT will all have slightly
different properties to what I show in all my schematics. 

Very few ppl will try to wind their own 100W rated OPT or have one custom wound. I do have details
for a custom wound OPT at the bottom of this page.

But let us explore the power able to be produced by 6 x EL34 with different loads.
Graph 1.
Graph 1 is based on using 6 x EL34 with UL OPT with screen taps at about 40% in a UL amp.
The B+ is assumed to be regulated and fall no lower than +400Vdc at the maximum possible Po
generated by the tubes. In the real world, B+ rails are seldom regulated so that in real amps
the idle B+ might be 430Vdc, but it will drop to 400Vdc with with a continuous sine wave
signal at high levels, using silicon rectifiers and at least a 235uF C1 reservoir cap followed by
say 3H choke with less than 33r0 winding resistance and another C2 235uF. Having a charge limiting
resistor in series with HT winding and diodes or having tube rectifiers will give a much bigger B+
drop when input Idc may be 3 times the idle Idc when RLa-a is low for nearly all class B operation.
However, while average low levels of a few Watts will not cause B+ drop, brief transients in
music such as loud drum beats lasting a very short time may reach up to clipping levels. But the
duration of transients is low and the time constant between C2 cap and load is much longer so
B+ will stay up at the idle level, so the instantaneous power of transients will reach up to a power
level maybe 10% higher than the solid line curve on Graph 1.

Graph 1 shows the power levels at clipping with secondary loads between 0.0 Ohms and 30 Ohms,
using an OPT with 333:1 ZR, and nominally 2k7 : 8r0, and using fixed bias, and using a continuous
sine wave signal between 400Hz and 1kHz . The dashed line curve shows the Po if the OPT had
zero winding resistance.
The solid line curve shows Po with a typical OPT that has 10% winding losses for where
RL = 1.5 x load for maximum possible Po. In this case, with a 5 Ohm sec load, the OPT
primary anode load = 1k7, and maximum possible AB1 power occurs and winding losses are
about 14%.
But with sec load = 8 ohms, winding losses are about 10%.
Winding resistance losses is highest when amp is in class AB with low RLa-a, and where
each 1/2 of the primary conducts current alone, and the tube load = 1/4 RLa-a, so if RwP + RwS
was say 80r and the tube load is 500r when RLa-a = 2k0, the winding loss =
100% x 80r / ( 500r + 80r ) = 13.8%. If the amp worked in pure class A with load say 8k0,
the winding losses become so low they may be neglected.
The approximate THD % levels mainly comprising 3H, 5H, 2H, 7H near clipping are shown.
There is a straight line A - A. Below this line the amp works in pure class A1.
Below the line the amp works in class AB1.

The EL34 are biased to give Pda = 15Watts which seems very low, but they have less tendency
to overheat, and they last a long time.
The discerning listener will hear the best hi-fi when the 6 EL34 have a speaker of about 11 ohms
and where RLa-a = 3k7, and maximum AB1 Po = 62Watts and the first 15 Watts are in pure class A1.
This assumes the listener owns two speakers each at least 88dB/W/M and likes his loudest busy music
to average 85dB SPL. Each of two amps needs to only make 0.5Watts to achieve 85dB average levels.

11 ohm speakers are virtually unobtainable. To allow the best results of 11r0, the secondaries
should be able to be re-arranged to give ZR = 2k0 : 2r0, 4r0, 8r0, and possibly 12r0.

This means when a 4r0 speaker is used with 2r0 outlet, the RLa-a becomes 4k0, and you can
get the magic 62 Watts. If the speaker Z has a minimum of say 2r5, then the amp will cope.
For highest Po possible, the 4r0 outlet is used, and 100W + is available.

Having said all that about load matches, I really don't expect many audiophiles or anyone else
will understand. But I have tried to talk about facts.

Now OPTs which have all these choices are not commonly available. Many are made with
"tapped secondary" which has 4r0, 8r0, 16r0 taps and where the whole sec winding equals
16r0, 8r0 = 0.7 of the turns, and 4r0 = 1/2 the turns. The tapped secondary is actually sometimes
4 sec windings in 4 separate sections among primary sections, and all 4 sec sections are simply
paralleled, so no adjusting of straps on OPT terminals is needed. But the winding losses are much
higher than where many windings are strapped to alter the ZR, and all windings are used
with each strapping.


DIYers might just buy a Hammond 1650TA.
It is 14lbs, or 6.4Kg, with load ratio, ZR = 1k9 : 4r, 8r,16r, rated for 120Watts.
This means that the Va-a = 474Vrms with RLa-a = 1k9. The peak LOAD voltage negative swing
= 335V at each 1/2 primary, but not including winding resistances.

In high AB1 power when RLa-a = 1k9, and sec RL = 4r0 at sec, there is a total of 53r in series
with each 1/2 primary load which is 1k9 / 4 = 475r. So load on tubes = 475r + 53r = 528r,
and winding losses are 100% x 53r / 528r = 10.0%. Now the peak current in 1/2 primaries
= 335V / 475r = 0.705 Amps. This peak current exists in the Rw = 53r, which includes the Rw
of a 1/2 primary + sec 4r0 Rw x effective ZR between 1/2 pri and 4r0 turns.
So the tubes must produce an additional amount of voltage swing = peak load current x effective Rw
= 0.705 x 53r = 37pk volts. The total peak Va swings at each anode = 335V + 37V = 372V.
The load line analysis might show Ea minimum at +80V, for the Ia max and on Ra line for Eg1 = 0V.
Therefore the Ea must be at least 80V + 372V = +452Vdc when this is the low Ea minimum due to
PSU resistances. So idle B+ might have to be +480Vdc, and this is getting excessive for EL34!
For Pda = 15W, the Idle Ia = 31mAdc, and Ig2 will be maybe 6mAdc.
The RLa-a including winding resistance can be worked out to be about 1,970r, and to get the
120Watts, the Rw must be considered. It means that each pair of EL45 is effectively working
with RLa-a = 5k9, and produces 40Watts. Make sense? I hope so.My calculations of anode
dissipation for each EL34 at Po max = 33Watts, and this exceeds rated max. If somebody loads the
am with 2r0 instead of 4r0, the tubes will overheat easily at less than full Po. Don't say I have not
warned you.
Of course its easier to just use Idle B+ = +425V at CT of OPT, and settle for maximum possible Po
to be maybe 90Watts including B+ sag and winding loses. That's because 90Watts is plenty for most
ppl, and the difference between 120W and 90W is only 1.2dB in SPL.
Using 6 x 6550 / KT88 seem a far better solution for 1650TA and to get 120W, and they won't mind
idle B+ = 480Vdc.
90W should be possible with 4 x 6550, like a VT100 ARC amp.

I have said 3k7 is the best RLa-a for 6 tubes and good AB1 hi-fi. And BTW, for those only needing
35Watts or less can have RLa-a = 8k0, and then all the Po is pure class A1, and nobody can argue
this isn't very good hi-fi as long as the available power ceiling is not exceeded. The Pure class A1
working region offers the least THD and IMD and the best damping factor, and widest bandwidth
and maximum effectiveness of the GNFB.

To get RLa-a = 3k7 with 1650TA the sec loads need to be 7r8, 15r6, and 31k2.
For pure class A1 max = 35Watts, RLa-a = 8k0 and the 3 sec loads needed are 16r8,
33r7, and 67r4. So, pure class A is impossible with normal available speakers.
Good hi-fi is not available with loads under say 6r0, and if a 4r0 speaker is used at 4r0 outlet,
and its minimum Z = 2r5 then the RLa-a = 1k2, and performance is disgusting, IMHO.


The Hammond 1650RA could be used, rated for 100W for 5k0 : 4, 8, 16, with tapped sec.
Idc rating is 318mAdc, more than needed. Weight is 12lbs, 5.45Kgs.  Its rated for 30Hz to 30kHz.
When using 5k0 : 16r0 loading, and class AB1, losses = 4%, after calculating from the Hammond
.pdf data sheet.
But with 5k0 : 4r0, losses are 7.8%. But use of 4r0 speakers means actual minimum Z
could be 2r5, and then RLa-a becomes 3k1, and this is close to the ideal 3k7.
Screen taps are at 42%.
So after all I have said, the Hammond 1650R seems like a good match for 6 x EL34, ( or for 4 x KT120.)
A single KT120 can do what 2 x EL34 can.
To get 100W AB1 with 5k0 : 4r0, the peak load swing = 707Vrms. rather than work through tedious
Rw calculations, just add 10% voltage for Rw and then you'd need Va-a = 777Vrms so there is a peak
swing at each anode of 549V. Peak load Ia = 0.39A, and you will need Ea at about +582Vdc, which is
an unsafe B+ for EL34! So don't expect 100Watts into 5k0 RLa-a with EL34. But when Po = 50Watts
with 5k the Va-a drops to 500Vrms, and a lower B+ is possible. 
But if the 4r0 speaker is moved to 8r0 outlet, then RLa-a becomes about 2k6 and with B+ of about
+425V at full Po the power will be near 100W.
But where RLa-a = 5k0, and with B+ = 430Vdc, each pair of tubes sees a 15k0 RLa-a and Po = class A.
Certainly the use of 4 x EL34 , 6550, KT88 or KT120 should be OK for "enough" good power, without
any worry you don't have enough Po.

In an amp as I suggest, at least 15dB of global NFB is required to improve the performance, I can assume
that 20dB GNFB might be used, if you are really clever, and while obtaining unconditional stability
and bandwidth from 20Hz to 65kHz with a resistance load, at full 400Hz Po and without core saturation
above 20Hz.

My conclusions:-
1. If you insist you MUST HAVE 100 Watts for a "4 ohm" speaker with Hammond 1650TA,
AND you don't care about the amount of class A for the initial few Watts, then use 4 x 6550 / KT88
at least.

2. Hammond 1650RA seems to be a more usable OPT for hi-fi and large amounts of Class A1.

3. Hammond should forget having a 16r0 outlets. It is mainly useless, because so few "16 ohm" speakers
are made and instead they should OPT secondaries suitable for 2 ohms, 4 ohms and 8 ohms which
means they have to revise all their OPT designs but of course, they won't.

3. Many Audiophiles would MUCH PREFER a high amount of class A power because they may only
ever use a few Watts maximum. They prefer 6 x EL34 because re-tubing cost is lower; EL34 make
good sound. Therefore a "16 ohm" speaker can be used on the 4r0 outlet of 1650T and if speaker Z
minimum = 10r0, maximum Po = 63 Watts of Ab1 and first 24 Watts being class A Po. Now the amp
operates optimally. The THD is low, bandwidth high, damping factor excellent, etc. Very few ppl have
16r0 speakers, unless they are specially made using identical 4r0, 6r0, 8r0 drivers in series. 

4. The same audiophiles would much like to use "4 ohm" speakers, or some other brand where
the speaker impedance is even lower than 4 ohms, AND they want CLASS A power. The 1650TA
will disappoint them. But 1650RA will be fine.

Before building and amp like this, think everything through, because I will NOT be there to correct
your many inevitable mistakes unless you have a large amount of previous experience!

Be prepared to make a few changes of the listed values I have for L1, R10, C5, R12, C15, R29, C19, R57.
All these R&C parts must be confirmed to be effective for stability which depends on the properties
of the OPT which have a huge effect on how it works with NFB applied.
In 2004, the OPT I wound for my customer's amp I sold had more primary inductance and less
leakage inductance than the Hammond 1650T may have, so the values for parts listed are a
GUIDE ONLY. If you just use what I show and hope for the best without testing it properly and
optimizing the values, the amp may oscillate, misbehave, sing badly and smoke!!!!

Sometimes a DIYer needs to have a tech inspect and correct his mistakes. Most DIYers and
many so called audio techs and engineers have very little knowledge of how to gain unconditional
stability in amps, especially tube amps. The 100W amp is a real challenge.

Some recent thoughts this March 2014 had me thinking a better and simpler biasing
adjustment could be used as well as different input and driver tubes for better dynamics.

Fig 2. 100W ULAB1 amp, 2014, 6CG7 input, 2 x EL84 triode drivers, and balanced fixed biasing.

Fig 2 is very similar to Fig1, but V1 is a paralleled 6CG7 which operates identically to 6SN7
but has a mini 9 pin socket which allows more V1 changes if desired, such as 6DJ8, 12AU7,
12AT7, ECC99, or trioded EF80, etc.
But any change from 6CG7 shown here will require change to ohm value of R8 so that the
Ea shown remains the same at +150Vdc approx. Exceptions would be for 12AY7, which
would need Ia to be lower at 5mA and so R6, R7 also must be altered.

Fig 2 shows V2 & V3 as a pair of
EL84 in triode in a differential pair driver and phase inverter.
in triode are cheap and plentiful and make superb preamp tubes and driver tubes.

Each EL84 in triode is the equivalent of 5 paralleled 1/2 sections of 6SN7 or 6CG7.
The Ra of the trioded EL84 with Iadc at 15mA is about 2k2, so the loading effect of  multiple
paralleled output tubes is negligible so the music is better.

Fig 1 shows biasing with six bias adjust pots, one for each output tube to set the idle Iadc.
But adjusting grid bias voltage Eg1 on one tube slightly affects the Ia bias condition of others which
makes adjusting bias tedious because it needs to be repeated several times, and many audiophiles
get the whole process hopelessly wrong, and end up with some tubes far too hot while others are

Fig 2 shows a considerable change to having BALANCED FIXED BIAS. Rarely has any manufacturer
ever used this wonderful idea because most of them are backward thinking old fuddy duddies.

Fig 2 output stage is based on the idea of having the six EL34 working as three parallel pairs of tubes
with each pair set up for bias similarly to the one pair of tubes in my 5050 integrated amp.

Fig 2 has 3 pairs of EL34 and in each pair, the bias Eg1 voltage applied to the two grids is adjusted by
a single pot so that the anode and screen Idc of each EL34 become equal. If the bias pot is set at center
position, then each Eg1 = about -38Vdc. But Ikdc would be different for each EL34. If the balance pot is
turned one way or the other, the Eg1 of one tube goes more positive, while the other goes negative.

In all PP amps, it is most important to have very nearly equal Idc in each 1/2 of the OPT on each side of
center tap which receives the B+ current. The 1/2 primaries have Idc flows in opposite direction and
these should be equal so that the iron core of OPT does not become magnetized by any difference in Idc.
A small net difference in Idc makes a flow in one direction across all the primary, and because the PP
OPT has no air gap and a high permeability, it will saturate easily with unbalanced Idc. If this happens,
the music turns to mud, with high distortion levels.

But how does anyone know when all pairs of tubes in a PP output stage have equal Ikdc?

Two methods are used. One involves holding two probes of Vdc meter from one cathode to the other
in each pair, and then turn the balance pot until a reading of 0.0Vdc is attained.
Fig 2 shows 3 pairs of test points, tp4&5, tp6&7, tp8&9. These would be recessed test points for
2mm probes. These test points allow anyone to measure the Vdc between EL34 cathodes and 0V
and across the six cathode current sensing resistances of 15r.

The other method is to have the three pairs of test points connected to three differential amps
arranged so each amp drives two green LED. When balanced Ikdc is present, the two LED will
glow equally brightly. The LED tell an owner if balance is OK and if it is, biasing is OK, nothing
to worry about. If one tube conducts too much Ikdc, or not enough Ikdc, then only one LED will
light up and if their brightness cannot be equalized with the bias balance pot then the Ikdc cannot
be balanced so something is wrong with at least one EL34 of the pair, and a volt meter will confirm
it. A new tube is plugged in and if balance is easily restored, the problem is fixed.


Fig 3 below  is a schematic which shows the differential amps for balance monitoring and
other features which make life with an amp like this much more civilized.

Fig 3. Amplifier Protection, Delayed turn on, Clipping indicator, and bias balance indication. 2014
1. Bias balance indication :-
See the three simple bjt differential amps formed by Q6 to Q11.
Please consider just V4 and V5 which form one of the three parallel pairs of output EL34.
Their cathode Ek voltage at k4 and k5 are fed through 2k2 and 470uF low pass filters to
ensure mainly only Vdc with little signal Vac appears at tops of C6 and C7. The two Vdc
voltages are applied to Q6 & Q7 bases. Only a small amount of Vdc difference at bases
is needed to make a large difference in collector currents. So if the Vdc difference between
Q6 and Q7 bases = 0.1Vdc, one led glows brightly while the other may be unlit.
As the bias balance pot is turned, the base Vdc can be equalized and close to equal
collector current flows in Q6 & Q7 so leds glow equally bright, and Ikdc in each tube is equal.
Common emitter R12 makes the differential action possible by ensuring the current increase
in Q6 equals current decrease in Q7.
the 2k2 should be within 1% tolerance or else the small base input current base current will
generate different Vdc across these R and good balance will be impossible.
The bjts I used for trials of the amp circuits was PN100, a very common small signal TO92 npn
transistor costing less than 10c each, if a packet of 50 is bought. I found that randomly picked
pairs of PN100 all gave splendid balance despite some possible hfe differences. Always use
leds from same batch so when equal current flows the brightness is equal.
It is remarkably easy to see which led is brighter than the other, and how close the pairs of Ikdc
will be.

2. Protection :-
All tube amps are prone to fail when output tubes conduct too much current for long enough
to cause overheating and self destruction and this may cause expensive damage to an OPT.
Usually overheating is due to a speaker load too low
or there is a shorted speaker lead,
a voice coil jammed in its magnetic gap, or bias is adjusted wrongly. A minor proportion of tube
failure is due to spontaneous breakdown well before coming close to average tube life which
can be many thousands of hours.

In Fig 2, all the cathode current sensing resistors will produce a Vdc which will rise during normal
use when volume level is turned up beyond the region of initial pure class A. Very few ppl will
use more power than these first few Watts. But when volume is raised, class AB action begins
and Ikdc will rise and then Ek will rise.

The Vdc at the tops of 470uF C6 to C11 are all connected via six diodes to a rail which powers
the gate of a sensitive SCR, C106D. R8 and C5 offer additional signal filtering at SCR gate which
we want to only react to Vdc. If the gate voltage exceeds about 0.65Vdc, the scr *latches* on,
causing the relay to turn on and open the contacts which interrupts the mains energy to PT1
primary winding. So the amp gets turned off automatically if excessive Ikdc occurs in one or more

If the Ek of any tube reaches 1.2Vdc, it is enough to cause the SCR to trip. The Ek of 1.2Vdc
means that Ikdc = 80mAdc, which is 2.4 times the normal 33mA Idc at idle. If an EL34 tube ages or
becomes gassy the Eg2 bias may not control Ia which then increases up to 250mAdc. This means
Pda + Pg2 = 102Watts and the anode will glow red hot and screen wires may melt down and glass
may soften and B+ may short circuit to 0V thus blowing a fuse. To avoid the pyrotechnical display,
the SCR trips when Ikdc has reached 80mA. Pda = 33.6 Watts, and just above the maximum
sustained Pda + Pg2 of 28 Watts.
Excessive signal Iac could cause overheating. Say the OPT has 2k5 : 8 ohm configuration.
It is difficult to over heat tubes if the speaker load is always 8 ohms or higher. But if a 3 ohm load
is used and volume turned to just over clipping with a sine wave for more than 3 seconds, the
amp will be turned off because Ek will have risen to 1.2Vdc.
But suppose there is a short circuit in a speaker cable or within a speaker. When volume is
turned up with music signal there is no output voltage but there is very high output current
which raises Ek very quickly and the the amp is immediately turned off.

There is no direct protection against an intermittent short circuit where levels are low.
For example, I once repaired a Quad-II amp which had been powering an old Quad ESL57 with a
midrange panel which began arcing when output voltage went above 1.5Vrms. The speaker
impedance reduces to about 1.5 ohms during arcing, and this heated up KT66 which went red hot
and overheated the PT. Smoke billowed forth, and a bad smell, and luckily the amp was turned off
before major damage. I later put in protection circuits and the owner bought new speakers.
Dynamic speakers can become damaged with overheating by excessive levels when teenagers
are allowed near a volume control. Typically, a voice coil of a bass speaker will heat so much the
glue holding wires on aluminium former melt and the coil will then jam tight in the magnetic gap
between iron poles of magnet. This reduces the speaker impedance to the amp overheats and can
cook to death.

When Q5 scr is triggered, it becomes a very low resistance and turns on Relay 1 and led 1, which
indicates amp has been turned off and a fault exists. Relay 1 is a Double Pole Double Throw
and has two sets of contacts. One set is for the opening of the Neutral line and the other set
is used to open the negative -16V rail to the emitter resistors to Q6 to Q11. This prevents current
flow in bjts Q6 to Q11, and leds 2 to 7 turn off. So when the amp is turned off, the six green led are
extinguished and there is just one red led glowing to tell an owner something is wrong.

The power source for the protection circuit and the bjt differential amps comes from PT2 which is
at least 7VA. This transformer continues to work when a fault causes PT1 to be turned off. 

The amp may be "reset" by simply turning the mains switch off, then on again a couple of seconds
later. If the fault persists, the amp will turn off again.

I have repaired many solid state amps which have had their output transistors plus driver stages
severely damaged by heat from one reason or another. Tubes are slightly more capable of
overload for longer, but the only solution when overload occurs is to turn off the amp.
Unlike SS amps, there is no need for a relay to disconnect speakers because the OPT cannot
pass DC current to a speaker load.

Delayed B+ application :-
Immediately after turn on the mains input current is very high for 2 main reasons.

All tube heater elements have low resistance when cold. Higher than normal working current
flows until the heat raises the heater element resistance.  However, the heater initial current
is not so high that any current limiting is required.

There is much higher peak input current flow in mains winding during initial magnetization of
the PT core and the initial charging up of B+ rail capacitors.
These high "inrush" currents may require a high amp value for the mains fuse so it does not blow
at turn on. The high fuse value would allow mains input current to continue even if bias failure
and excessive Idc flowed in output tubes, so the high amp value fuse is not effective against
bias failure if the protection circuit does not work. To enable a lower fuse value that will blow
when other protection measures fail the inrush current to capacitors needs to be limited by a
series resistance in the HT winding or in the mains winding. Ive arranged the circuit in Fig 3
so the B+ will rise to 2/3 its full value within 5 seconds with 100r at 50W rating in series with
HT winding. After 5 seconds, the 100r is shunted by Relay 2. The B+ then rises to its maximum
possible voltage of about +460Vdc. The peak current input at turn on and after 5 seconds
when 100r is shunted is about 1/2 what would occur without the delayed relay. Hence a fuse
of useful value can be used. During the delay period, the fixed bias voltage establishes and
after about 10 seconds the tubes all begin to conduct but Ia is gently turned on, and B+ is
pulled down to +430Vdc for normal operation.

If the amp is turned off, then back on again after say 3 seconds, then heaters are still hot
and Ia pulls the B+ level down. But when turned on again the Delay Relay is activated again
and any surge in tube current or mains input current is avoided and B+ is raised slowly again
to maximum.

Delay Relay 2 is controlled by Q1 & Q2 Darlington pair which is turned on after about 5
seconds by base voltage delayed by R3 and C3 which have a time constant of 7 seconds.
When C3 voltage has risen
to about +10Vdc, the 8V zener diode conducts to turn on Q1 and Q2 and 100r x 50W is
shunted between points U and V. The diode across R3 15k allows fast discharge of C3
when amp is turned off, so that when turned on again quickly there is a delay before the
100r is shunted.

All these measures ensure maximum tube life and save repair expenses.

Clipping Indicator :-
Its not essential to know if a hi-fi amp clips, but it is handy, especially if one speaker has a
shorted cable. To do this, a sample of signal from V1 anode is fed through high resistance
network of C4, R4, R5. The bjts Q3 & Q4 form a high input resistance Darlington pair.
The V1 anode signal voltage will suddenly rise to a high level when the amp clips.
Then Q1 base voltage becomes high enough to turn on collector current which flows in D7
and R7 and red led1 is turned on. It will flash on the peaks of signal current to give a warning
that clipping has begun. Clipping with a high value speaker ohm load will not cause high
cathode current so the protection Relay 1 is not activated. Clipping the amp with high value
speaker loads does little damage, ( except to listeners ears ).
But if a low load of say 2 ohms is used, excessive input signals will easily cause clipping
and high Ikdc and amp will be turned off by Relay 1.

Reliability of Protection :-
If the PT2 becomes faulty and +/- 16Vdc PSU rails remain at 0V, the 6 green leds will not
light up after turn on, and the Delay Relay 2 will be turned on to shunt the 100r. B+ cannot
rise to full value. The normal EG1 bias voltage will be established, but Ia and Ig2 will not
rise to wanted values, and tubes will have low Pda and be over-biased for the low B+.
The amp distortion will be high, but it cannot be damaged if used.
There are more complex schemes to ensure the amp cannot turn on unless the protection
circuit rail voltages have been established. I do not think the extra complexity is necessary.
Protection circuits are needed, but they do not have to work very often, and the parts I use
are all easily replaced if needed.


The 100W amp needs a good PSU. I have come up with a 308VA rated design which allows
6 x EL34, 6CA7, KT66, 6L6GC, 6550, KT88, KT90, KT120. 
The larger octal 6550 to KT120 allow a higher idle bias Ia and Ig2, and thus there will be
more initial class A power than for EL34.

It will be found that a quad of
6550, KT88, KT90, KT120 will do at least the same work as
6 x EL34, and the same PSU can be used but with one less differential amp for balanced bias.

Fig 4.
The PSU schematic has B+ produced with voltage doubler rectifier using 6 amp x 100PIV rated Si diodes.
Many will want to use a bridge rectifier with HT winding with taps between 300Vac and 400Vac, but then
the Delay Relay 2 is exposed to over 250Vac, and most easy to get relays don't have voltage ratings
above 250V. 
I don't have much more to say about the PSU I have drawn here because anyone can see
the text within the schematic.


Fig 5. 100W UL AB1 amp with 1 x 12AU7, 2 x ECC99 and 6 x EL34 and balanced biasing
Fig 5 is very similar to Fig 2 but input is 12AU7 and LTP is formed with 2 x ECC99 which
I have never used but which should work as well as EL84 in triode.


Replacement of all output tubes and bias adjustments :-

Amp is cold, and has been turned off for at least 1/2 an hour.
1. Remove all 6 old output tubes.
2. Adjust all 3 balance pots to center position.
Make sure preamp is turned off, or gain is turned down to zero.
Make sure a speaker is connected.
3. Plug in 6 new EL34.
4. Turn on power amp.
5. Watch each and make sure all 6 have heaters glowing.
Notice that all 6 green leds should glow equally bright at turn on.
6. Leds will begin to change brightness after 20 seconds and then some will turn off,
then on again. This is expected behaviour before any balance pot is adjusted.
After 30 seconds you may see only 3 leds of the 6 alight.
7. Adjust one pot until you see the two nearest leds glow equally bright.
Then adjust the next pot and the next until you see all leds glowing equally.
The amp is ready to go. For those worried about the real tube condition, they may use a digital
voltmeter set to Vdc and measure test points k4 to k5, k6 to k7, k8 to k9.
All 3 measurements should read less than 0.05Vdc, and if not. adjust balance pots more finely.
Then measure from all 6 test points to 0V and you should read 0.495Vdc +/- 0.5Vdc.

8. After an hour you may find one or two led of a pair has faded to less bright than its other.
re-adjust balance pot to equalize brightness. Further variation should not be needed for days
or weeks later. You do NOT have to re-adjust pots each time you use the amp, there will always
be some variation in idle currents until some time has passed but after an hour all led should
glow equally brightly.

9. Music may be played after the first 30 minutes.
But next day you can play music a minute after turn on.
10. It will be noticed that very high sound levels with some clipping may cause leds to change
brightness. This indicates pairs of tubes have temporarily unbalanced Ia because at clipping
the Ia becomes At sensible lower levels the should be little change in led brightness and even
though the average value of Iadc flow through tubes has risen in class AB1 working, it should
be the same in each balanced pair of EL34.

Severe overloading may cause red led1 to turn on and all 6 green led to turn off.
The protection circuit is telling you "enough is enough you idiot!"
Suppose V4 of a pair V4&V5 begins to conduct too much Iadc and gets hotter than other tubes.
The led which represents V4 will glow bright, the other for V5 will turn off. Turn amp off,
let it cool down, remove hot tube, replace with a new spare, then turn on again and try to
re-balance the pair.
No adjustments should be needed with other 4 x EL34.
If balance is still not possible for V4 & V5, move the replacement V4 to V5, and put removed
V4 back in.
THEN balance should be possible. And the tube you have removed probably has gone open,
and is not conducting at all, except for its heater.

Many audiophiles are very uncertain about biasing their amps and they sometimes get all
mixed up about which tube they biasing, which pot, and cannot use a voltmeter.
They don't know what to do when a tube finally becomes faulty towards the end of its life.
My past customers here just ring me up, and I advise them over the bother. It seldom
happens with my protection circuits because they get used to coping, and with the bias
balance indication, all an audiophile really ever needs to do is "turn all pots until all green
leds glow equally".

If I was to make a 100W UL monobloc now, I'd consider using 4 x 6550, KT88, KT90,
KT120 instead of 6 x EL34, and in fact 4 x KT120 offer the best solution. If the amp is
designed for KT120, and has B+ at +425V, then 6550, KT88, KT90 may also be used
with the same Eg1 biasing and same Ia and Ig2 and Pda+Pg2 of 23.3Watts at idle,
and Po will be very nearly the same. The KT120 need 2.1Aac for heaters, 0.3Aac
more than the others.

Fig 6. 100W UL AB1 amp with 1 x 12AU7, 2 x ECC99, 4 x KT120.
The sound quality from this amp should be little different from using 6 x EL34.
The pair of KT120 on each side of PP output circuit can produce a maximum total peak
current in class AB1 = 1.3Amps. 3 x EL34 would manage 0.75Amps. This means that the
KT120 will drive a lower speaker impedance. However, the total idle Pda for 6 x EL34
= 6 x 425V x 0.029A = 74Watts. Total Pda for 4 x KT120 = 4 x 425 x 0.05 = 85Watts,
so there is very little difference in the maximum possible pure class A which would
be 33Watts for EL34, and 38Watts for KT120.

I've shown only one PSU schematic and one protection schematic but there are 4 amps
on this page so If you make one of the amps, you need to make minor adjustments to the
PSU and protection circuits. Remember that the balanced biasing networks for EL34
and 6CA7 will produce a lower negative bias and cannot be used for KT66, 6L6GC,
6550, KT88, KT90.


The basic OPT specification is for 100W+ rated with UL taps between 25% and 50% and for
ZR = 2k5 : 2r5, 5r0, 10r0.
The secondary ohms are lower than nominal 4r0, 8r0, 16r0 because nearly all loudspeakers have a
minimum ohm load value which is lower than the nominal one mentioned by the maker.
Fig 7.

Fig 7 has details for a suitable OPT. It is a drawing for winding the bobbin and gives details on the core,
the winding wire and insulation and winding patterns.

The total height of the wire and insulation is :-
Primary, 16 layers at 0.52mm oa dia wire = 8.32mm
Secondary, 4 layers at 0.99mm oa dia wire = 3.96mm
0.05mm insulation P-P, 11 layers at 0.05mm = 0.55mm
0.6mm insulation P-S, 8 layers at 0.6mm = 4.8mm.
One 0.6mm cover layer over completed winding = 0.6mm
Total winding height = 18.23mm.
The CORE window = 25mm x 75mm and the available bobbin window = 22mm x 71mm.
The turns must be neatly wound on in layers without ANY wires crossing over others and subject
to forming a shorted turn in future.

There are multiple secondary windings which can be arranged in various way to change the load match,
while maintaining the same wide bandwidth, and winding loss % for all values of nominal loads, and
maintaining unconditional stability with a global NFB loop which does not need to be adjusted when
changing OPT load matching strapping pattern.

There is no acceptable way to use the 8 secondary windings on this OPT arranged as a "tapped secondary".

An OPT with a tapped secondary needs to have the same number of secondary sections, each with
enough turns to provide taps to 4 amp output terminals, for 2k5 primary RLa-a : Common, 2r0, 4r0, and 8r0.
If the primary remains unchanged as I have it above, and low winding losses are to be maintained,
each tapped secondary would need to have 2 layers of 72t x 0.9mm Cu dia wire, so that both layers of
144t = 8r0, a tap at 100turns = 4r0, and a tap at 72 turns = 4r0. The trouble is that if double the number of
secondary layers were used, the total winding height would increase to 19.89mm made possible by reducing
the 0.6mm insulation to 0.3mm, and risking arcing from P to S, and also doubling the shunt capacitances

and slightly increasing leakage inductance.

But then most winder tradesmen will find difficulty fitting the total winding height to less than the available
bobbin height of 22mm because as the turns are put on, they do not lay flat but tend to bulge up a couple of
mm when wire is bent around the bobbin. 

The OPT would need total re-design to include tapped secondaries.

The core material should be E&I Grain Oriented Silicon Steel laminations 0.35mm thick.
Most OPTs are made with this material but the makers usually place the Es and Is into the wound bobbin
with each in alternate directions. This gives the finished core a permeability, µ, of between about 9,000 and 17,000.
While this ensures the primary inductance is a huge Henry value, if ever the Idc on each side of primary
becomes slightly imbalanced, then the core may easily saturate with the net Idc flow in one direction.
The core does not need µ to be above about 3,000, and to achieve the lower µ the Es and Is must be gathered in
bundles of say 10 to 25 laminations and then all bundles are inserted to core in one direction while the next bundle
is inserted in the opposite direction. This is called "Partial air gapping" where there is no actual complete air gap.
Laminations are thus not maximally interleaved.
A minimum complete gap with all Es facing the same way and butted to all Is would reduce the µ to about 1,000,
which is too low to have sufficient primary inductance for a PP amp. The wanted µ is 3,000, and the only way to
achieve this is by trial and measurement by inserting the lams in say bundles of say 25 lams and then measuring
the inductance with 240Vac at 50Hz across the whole primary. A 10r0 series R is used to measure current, and
the reactance of the coil = Vac / Iac and the Inductance = Reactance ohms / ( 6.8 x 50Hz ).
From this measurement the core µ can be calculated with the formula for inductance and you'll have to visit
my Push Pull OPT design pages to find all that out.
I believe the lower µ achieved this way conveys the benefits of low distortion generated by the GOSS but
avoids the problems of Idc offsets, so sound is better.

Fig 8.
Fig 8 shows 5 different strapping patterns for the total of 8 secondary windings.
I suggest the pattern of connections from A to P be established on a board mounted on
the open frame E&I transformer bobbin, so short length secondary winding leads can be led to the
terminals.  There must be easy access to the secondary terminal board with easily removable
screwed transformer box covers, or some other hinged panel on transformer case, facing the rear of the amp.

The primary terminations can also be on a similar board but on opposite side of bobbin to secondary winding
leads. There will be little need to access these connections.

Some amp makers use 3mm thick fibre glass board mounted between the OPT and the chassis top.
This has an added benefit of providing a non magnetic gap between OPT and the steel top of chassis. 
This single board has the same size as the plan area of OPT and wire leads from bobbin can be brought down
to two rows of terminals in the board. Best terminals are not turrets, but 25mm long 2mm threaded brass rod with nuts
each side of board to allow wrap around soldering of wires to each 10mm of rod on each side of board. Unlike turrets
with rivet fixing, the brass rods will never come loose when threads are soldered.
The chassis top will have TWO slots maybe 25mm x 80mm, to allow protruding OPT connections to enter the
under chassis area but without taking up valuable space for other components. 16 secondary winding terminals are
required and the slow allows 2 rows of 8 terminals. Once the chassis bottom cover is removed, ALL OPT connections
are accessible and the wiring up of the circuit and servicing are both easy.
Quad-II had such a convenient arrangement.

I believe my method of using multiple secondary windings with wire links allows the vast majority of speaker
loads between 1 ohm and 45 ohms to all be driven by the tubes in high power Class AB1, or in low power Class A1,
to suit owner preferences.

soldering, and try not to get confused with too much information.

Back to Power Amps.

Back to Education and DIY Directory.

Back to Index Page.