LOAD MATCHING 5.
MORE ABOUT BIG BEAM TETRODES, 6550, KT88, KT90, KT120.
Initial page in 2012, edited Dec 2017.
Graph 1. KT120 AB Po vs RLa-a, beam tetrode, Ea +500V, Eg2 +500V.
Graph 2. KT120 AB loadlines, beam tetrode, Ea +500V, Eg2 +500V,
Graph 2A. Copied Ra curves for TungSol KT120.
Graph 3. KT120, KT88, AB loadlines, beam tetrode, Ea +535V, Eg2
+410V, both regulated.
Graph 4. KT120 + KT88 AB Po vs RLa-a, beam tetrode, Ea +535Vdc,
Eg2 = +410Vdc, both regulated.
Graph 5. KT120+KT88 AB loadlines, beam tetrode, Ea = +600Vdc, Eg2
= +410Vdc, both regulated.
Graph 6. KT120+KT88, AB Po vs RLa-a, beam tetrode, Ea = +600V, Eg2
= +410Vdc, both regulated.
Graph 7. KT120+KT88 AB Po vs RLa-a, +40% UL Operation.
Graph 8. KT88 AB Po vs RLa-a, beam tetrode, Ea = Eg2 = +400Vdc,
high initial class A Po.
Graph 9. KT120+KT88 AB Po vs RLa-a, TRIODE, Ea = Eg2 = +500Vdc.
High possible Class A.
Graph 10. KT88 Ra curves for Screen Eg2 vs Ia for Eg1 = 0V.
In 2012, I tested samples of KT120, and was keen to see how they
compare with more common EH KT88
and EH 6550, and less common EH KT90.
In 2017, I re-examined and re-tested the KT120 and KT88.
Many details and observations are in the text.........
Graph 1. KT120 Po vs RLa-a, beam tetrode operation, Ea =
Eg2 = +500Vdc.
Graph 1 shows test results for 2 x KT120 with a good quality PP
OPT with low winding losses, using a
1kHz sine wave. The B+ to CT OPT and for Eg2 was regulated at
I found it was impossible to obtain the high audio power if B+ and
/ or Eg2 rails sagged below +500Vdc.
Typical B+ rails are not actively regulated. B+ could
+550Vdc at idle, and remain at
+550Vdc during short term clipping when using a pink noise test
signal where there is occasional clipping
of some wave peaks. Pink noise should be bandwidth limited to -3dB
poles at 20Hz and 20kHz. If not,
and your meter can read from 2Hz to 500kHz, your meter will be
affected by the LF, and DMM don't like
LF measurements. With bandwidth limiting, the analog meter still
changes its meter reading, ie, the meter
needle wobbles above and below what you can say is the average
reading between wobble peaks.
Such is pink noise which contains all F which have randomly
changing amplitude and phase.
You may read the Vrms average, and watch it on a CRO. The average
wave peaks will be at about 1/3,
( -10dB ), of the height of the Vrms where clipping occurs
somewhere at more than 1second intervals.
This indicates maximum amp power is produced at the clipping
peaks, but average power levels are
about 1/10 of the maximum level So if the amp makes 100W max,
average pink noise level Po = 10W.
Pink noise is similar to music but amplitudes of music F are not
flat, because our ears don't perceive the
same levels of power for all F, and ears are less sensitive at
bass than at say 2kHz. Bass F between
20Hz and 120Hz energy can have 4 times the levels of all other F
Using a sine wave to test an amp examines its behaviour much more
critically than any music or pink
With a sine wave test right up to clipping, and with Ea and
Eg2 well regulated at +500Vdc, I found
it quite dangerous to test any of the family of KT120, KT88, 6550,
KT90, with Eg2 at +500V. At these B+
and Eg2 levels KT120 or KT88 can make maximum possible Po with
RLa-a 1k5 to 3k0. With RLa-a = 2k0,
and at idle, Ig2dc for a KT120 was 5mAdc. But at 130W the Ig2
increased to 30mAdc so Pdg2 = 15W,
nearly double the Pdg2 rating of 8W. At such high Po levels the
Pda has gone above 70W, and you may
expect the tubes to overheat and self destruct. So don't ever
expect 130W continuously from any of these
tetrodes where the Ea = +500Vdc.
KT120 survives the huge non linear increase of Ig2dc better than
KT88, but not for much longer.
The sag of B+ rail does reduce Po max slightly if it is
less than -10%, say from +550Vdc to +500Vdc.
The Idc from PSU to 2 anodes may increase from 90mAdc up to
450mAdc at max Po, a change of
+360mAdc, so an amp which has -50Vdc sag of B+ rail has PSU Rout =
139r. All the cheaper amps
including guitar amps may have PT with high Rw to allow the B+
sag, but more expensive hi-fi amps
have better PT so that Rout for a pair of KT120 may be only 50r.
If there are 8 x KT120, Rout should
be 12r0. It is not easy to achieve this, and the PT must be rated
for far more Idc power at idle.
Fortunately, music seldom is at the max continuous sine wave level
so usually the PT is good enough
if its VA rating is above twice the VA rating for idle. Remember
the idle VA includes heater power and
drive amp Po.
If Eg2 rail is allowed to sag say -50Vdc, it reduces Iadc
by gm g2 x V change and at high Po levels the
gm g2 is higher than the 0.8mA/V at idle levels, so the reduced
Eg2 at high Po levels makes tubes act like
they are biased for nearly class C with Ia cut off at 0V wave
crossings. So at high Po, there is very high 3H
from crossover distortion, 10%+. But this still occurs even where
Eg2 does not sag so it appears the
screens at a constant high +Vdc collect a large amount of
electrons while the anode swings to a low Va pk.
It is a strange effect to behold, and just shows the tube
performance is falling apart at such high Po levels.
And at such high Po, the Vg1 peak value is more than the Eg1 -Vdc,
showing that Eg1 is swinging further
than than it seems possible and there is grid current and the
coupling caps a slightly charged up to cause
some extra -Vdc bias at g1.
The huge increase in Ig2dc at high Po can be over twice
rated maximum Ig2 Idc so that Pd g2 is way
above the safe rating, and if the Po is continuous at high Po
level with high Ig2 dc, the tube will overheat
When overlaying the Vg1 Vac wave on Vo wave at OPT sec output, on
a dual trace CRO, the two waves
remain close until about 1/2 max Po, ie, where Vo = -3dV below
max. But the Vo wave then sags towards
max possible Po showing the Iac isn't being linearly delivered to
anodes and some of it disappears as
high peak screen currents every time the anode swings to low Va
and with high Ia.
Such naughty screens, robbing the anode of their current !
If GNFB is used, the g1 signal is boosted to
keep where Vo increase linear, which causes Ig2 at screens to
increase even more, so little change to max
Po level occurs and the screen overload increases while the NFB
tries to force the action to be linear during
the the last 30W up to Po max with a low RLa.
But at low levels the NFB makes a huge improvement to how the tube
No tests here use GNFB, or 20% local cathode feedback.
I am examining the power beam tetrode's wild behaviour before
being civilised by NFB.
The screen Eg2 rail should be regulated in hi-fi amps for
beam tetrode or pentode mode, with or
without CFB windings on OPT. Hi-fi amps can all have a lower Eg2
well below Ea This regulation should
make Eg2 constant with music until some high levels waves begin to
clip when average Po is only 1/10 of
the maximum possible. During 18 years as audio tech, I saw no
hi-fi listeners ever caused their amp to clip
their amps were 10W rated or less.
Hi-fi use means high Iac peaks at screens only occurs
intermittently and the energy stored in screen bypass
caps prevents Eg2 from sagging more than a volt or two. Such
intermittent high current peaks will not overheat
or damage the screen and the tube. All screens should have a
series R of say 270r 1W to a common bypass
C which may be fed Idc from R from anode B+. If Ig2 increases to
say double the idle value, there would be
some sag of Eg2, but this reduces current in shunt regulator
devices so the Eg2 stays remarkably constant.
In my 8585 amp, I show a shunt regulator for the 8 screens of 8 x
KT90. See 8585-amp-october-2006.html
See Fig 3, PSU, 1/4 down the page.
In 8585 the Eg2 = +330Vdc, with anode B+ at +480Vdc. A 3k6 x 10W
feeds Idc to all 8 screens of KT90.
I have KT90 idle Ia = 33mAdc so they run cool at Pda 16W. I found
having Ia any higher made zero difference
THD, bearing in mind I have 12.5% CB windings. Normal Ig2 for each
g2 is 3mAdc, just less than 10% of Iadc.
So total Ig2 for 8 KT90 is 24mAdc. The shunt regulator devices are
2 x BU208A high Vdc rated Bjts with MJE340
in Darlington pair mode to increase the hfe. Only 14mA flows in
bjts, with 3 mA in the resistance divider to control
base Vdc and the Idc in bjt. I found that high levels say over 50W
would cause slight Ig2 increase and Eg2 would
sag very slightly, which turns off Bjts so the screens can have
the 14mA so each could have 5mAdc. above 50W
it does not matter if Eg2 sags a bit at the 90W level the 8585
makes so effortlessly from each channel.
But if any tube suffers bias failure and has say 400mAdc Ia in its
last moments, its Ig2 dc may me 40mAdc and
that will make Eg2 reduce, and all tubes will have Iadc cut off
and amp goes silent. No damage occurs to any
parts. It is is one of two protection measures in that amp. There
is only 100uF to bypass all 8 screens, but I found
it all worked just fine, for hi-fi, where each channel never ever
needed to make more than 20W.
For all Eg2 regulators in hi-fi amps, series or shunt
types, the Eg2 needs to be lower than the B+ for
anodes so there is headroom for the regulation. If idle B+ =
480Vdc, Eg2 should not be above +350Vdc so that
if the B+ varies due to mains increase or decrease, Vdc can change
across resistance from B+ to shunt regulator
parallel to screens, or across regulator in series with screens.
The screens are important, but also fairly tough because they can
withstand Pdg2 well above the rating as long
as it is intermittent. KT90 has 6W rating, so if Eg2 = +350V, Ig2
can safely be 17mAdc, but screen wires would
be hot, so don't push your luck, and don't design the amp to allow
a constant 17mAdc.
KT120 have Pdg2 rating 8W, so with Eg2 +400Vdc, 20mA is
Screen current is about 10% of anode current except when Ia
increases hugely where Ea is low with low RLa-a.
Most data curves published now make no mention of how Ig2 can rise
non linearly high where Ea is low.
Graph 2A. Terrible data sheet quality for Ra curves for KT120.
I did find some very badly drawn data curves for KT10 and have
faithfully copied them from the .pdf data
sheets. Despite several hours at PC trying to use apps etc to
convert pdf image to .jpg, or .gif, I failed,
so, I spent an hour to redraw the curves in my great old copy of
The sheet size for V and mA is same as used for all tube Ea-vs-Ia
elsewhere on this page.
The vertical spacing of Ra curves shows gm g1 = 8mA/V at 500V x
At 100V x 300mA, gm g1 = 22mA/V and this is a typical variation of
gm g1 for many power beam tetrodes.
Notice the 3 lower Ra curves have sags in their shape, and for the
lowest, the Ea change from 0V to +50V
shows Ia reduces from 30mA to 20mA, and this is a region of
"negative resistance". Positive resistance is
where the V increase across R makes I increase. But for some
things, the opposite occurs.
The bottom 3 Ra curves look quite irregular below 100V and 100mA,
and this causes the very high THD
is loads which very high and where the Va swings into this region.
But for most loads, the operation is OK,
and similar to KT90, KT88, 6550, etc, but THD is higher than for
The tetrode gain is very high as the load becomes higher, so NFB
will easily tidy the operation very well.
From these curves, you could draw a B RLa load line for 1,400r
between 500V x 0.0mA to 357mA x 0V.
Theoretically, Va swing would be 454Vpk, and for RLa-a = 5k6, you
would get 73W.
The problem I do see is the straight diode line between 0V x 0mA
to 50V x 360mA, its R value = 139r,
and in all my tests I have never seen it so low, usually its above
However, Graph 2A looks remarkably similar to old 1960 curves for
KT88 with low Eg2.
All the lines on graph are made up by short straight lines but in
the real world the shapes of the knee shape
is never angular, they are always curved.
If you wish to see such curves, then one method uses a
ramped saw-tooth wave where Ea increases
linearly from 0V to say +500V over say 0.5 seconds before reducing
to 0V in 0.01 seconds. This is the
classic ramped saw-tooth wave, and when applied to a tube anode,
with its screen at a fixed Eg2 = say
+400Vdc, and Eg1 at a fixed -Eg1 bias Vdc, the oscilloscope will
show anode current increase to perhaps
400mA rapidly for between 0V and 100V, and then very little
between 150V and 500V.
By careful analysis of what is happening, aided by a PC, a very
accurate Ra curve can be generated, and
the Ra curve for where Eg1 = 0V should appear similar to what I
have below for load line analysis.
But be careful with such experiments. The Graph 2A shows the Ra
curve for Eg1 = -5V extending to
500V x 320mA, an operating point very well outside the Safe
Operation Area, SOA, for the tube.
The Pda at 500V x 320mA = 160W, but tube is rated for 60W. As the
Ia goes higher, the Ea used to test
them must be kept lower. Notice Graph 2A shows the top Ra curve
for EG1 = 0V only extending to
150V x 370mA for Pda = 56W, quite OK.
There is no need to damage a tube while you test it to find its
characteristic Ea vs Ia curves.
If you want to test tubes to find curves, try following up
the information at
The Audiomatica tube tester may not being made any more.
But there's also this resource,
The major benefit of regulated Eg2 is that variations of B+
anode supply do not change the Eg2 so Iadc
and idle biasing conditions remain stable. Mains can vary between
220Vrms and 250Vrms, causing unwanted
changes to anode B+ but if Eg2 is kept constant for that range of
mains change the idle Pda for each tube
will not vary enough to cause anyone to worry about tubes
overheating, or being not hot enough.
The normal level of mains in Australia here is 240Vrms. But
I have measured mains at 235Vrms to 255Vrms.
But to allow for all problems, and amp made for 240V mains should
work OK with 220V to 255V.
Mains may be 220V if shipped to China. The mains primary winding
should be in two 120V windings so that
it works OK in USA with 120V with both windings parallel.
The Vac changes of -20V or -8.3% or +15V or +6.2%.
Most amps have unregulated B+ rails. If B+ = +535Vdc with 240V
mains, B+ may move down down -8.3%
to +490Vdc, or up +6.2% to +568Vdc.
I measured the 2 KT120s with Eg2 regulated at +414Vdc, and
>B+ = +535Vdc, with each
Ikdc = 50.0mAdc,
and with Eg1 = -49.6Vdc, Idle Pda = 26.8W.
When I changed B+ up to 586Vdc, or down to 484Vdc, ie, +/-50Vdc,
the Iadc change was only +/- 1.6mAdc,
or 3.2mA total. This was done without change to grid bias -Vdc.
Therefore the Ra of each tube = 100V / 3.2mA = 31k.
But where anode B+ and fixed bias -Vdc are both not regulated and
only Eg2 is regulated, some surprising
If B+ reduces -8.3% to +490Vdc, bias Vdc reduces to -
45.5Vdc. The g1 is 4.1V more positive so Ia increase
= Vg change x gm g1 = +4.1V x 4.7mA/V = +19.3mAdc, so Ikdc =
59.3mA, and Pda = 490V x 59.3mA = 29W.
KT120 have Pda max of 60W, so the extra 2.2W will be OK.
If B+ increases +6.2% to 568Vdc, bias Vdc increases to
-52.7Vdc. The g1 is 3.1V more negative so Ia
decrease = Vg change x gm g1 = -3.1V x 4.7mA/V = -14.6mAdc, so
Ikdc = 35.4mAdc, so Pda = 568V x 35.4mA
= 20W. While this may seem OK, the amp is under-biased, and
initial class A reduces, and crossover distortion
increases. But at least nothing will overheat !
For biasing Output stages with regulated Eg2, the -Eg1 bias
circuit could have a zener diode included
somewhere to reduce Eg1 change to about 1/2 the percentage of
mains changes and thus keep idle Pda
close to constant for mains change between 220Vrms and 260Vrms.
Good design allows all amps work OK with 220Vac to 260Vac
mains. On page 3, RDH4, the author
says heater Vac should be 6.3V, +/- 10%. Thus if the heater Vac =
6.3vac with 240Vac mains, the change
to 220V or 260V will give heaters = 5.8Vac or 6.8Vac. The change
in heater Vac is less than +/- 10%.
But some amps made in USA or China have heaters which may be
6.5Vac with 220Vac mains.
If brought to Australia where I have measured 255Vac, the heaters
are 7.5Vac, and this too high so
0.56r x 10W would be used between each KT120 heater and the heater
winding. But I have modified several
ARC amps where B+ was supposed to be +420Vdc in USA, but with 250V
mains here, B+ was +470V, with
the two 110V primary windings in series. 450V is the Vdc rating
for electrolytic caps.
VT100 have UL taps and Eg2 is tied to Ea and with rise of Ea and
Eg2, I don't recall if -Eg1 bias was
regulated but I found 6550 in ARC amps overheat very easily and
often in Australia because idle Pda goes
To avoid risk of having electrolytics fail, it is better to
use a 1KVA transformer with input of 240V and
switched outputs for 254V, 247V, 240V, 233V, 226V, 219V, 212V. If
mains here is 250V, use mains switched
to 212V. Actual mains Vac at amp should become = 250V x 212V /
240V = 221V and B+ should be +422Vdc,
which is OK and all heaters should be 6.3Vac.
Calculation of Pda due to Vac + Iac operation :-
Total Pda all tubes = Anode DC power from PSU - Output power at
anodes to anode load.
An accurate formula for class AB Pda at any level is fiendishly
complex, and cannot be found anywhere
on the Internet.
But here I give the simple formula for class B amps where idle Idc
= 0.0mA, and perfectly linear tubes with
no anode resistance, Ra.
The formula works for Po above 60% of maximum Po.
Usually Pda is never excessive with Pda > 50% max.
Pda for both AB tubes = (1.8 x Ea x Va-a / RLa-a ) - ( Va-a
squared / RLa-a ).
In Graph 1, with 2 x KT120, B+ = Ea = Eg2 = +500V, and minimum
RLa-a for max Po = 1k5.
Va-a = 451Vrms, Po max = 136W.
Pda = [ (1.8 x 500Vdc x 451Vrms / 1,500r ) - ( 451Vrms squared /
1,500r ) ] = 270.6W - 135.7W = 135W.
The Pda for each tube = 135W / 2 = 67.5W. At this condition the
screen input current was 30mAdc,
so 15W, twice the Pda rating of 8W, so total tube heat = 82.5W ! ,
( but not including heater power = 6.3V x 2.1A ).
I found KT88 or 6550 could give nearly the same Po, and Pda also
exceeded 60W, and screen Pdg2
is also high.
Therefore KT120, KT88 or 6550 cannot
produce 130W continuously with Ea = Eg2 = 500Vdc.
KT88 made in 1960s and 70s had Eg2 rating +600Vdc, and Ea
rating of 800V. Ratings for 6550 were
never as high with Eg2 max = 400Vdc. I found Russian 6550 and KT88
are internally identical, so I
would never ever have Eg2 = +500Vdc for Russian KT88 or 6550.
The KT120 have Eg2 rating +650Vdc with Ea rating +850Vdc,
so they certainly can do more than
KT88 or 6550, and are more likely to survive. KT90 has
characteristics between KT88 and KT120.
I don't know of many ppl using KT120 in a guitar amp. But most
guitar amps meant for 2 x 6L6 or
EL34 have B+ at 450Vdc, Eg2 at +430Vdc, and KT120 might make
slightly more power with the
same speaker loads. During gross overloading of guitar amp output
tubes, the wave forms are a
series of square waves and during the grid current charges
coupling caps to increase Eg2 so tubes
work in class C mode. Pda ratings are exceeded, but not all the
time, so the tubes last fairly well, but
not as long as those in a hi-fi amp.
In nearly all guitar amps, Idle Ea is rarely above +450V,
and Eg2 is usually about 430Vdc, so there
is NO WAY anyone could get 135W of sine wave power with just 2
With a sine wave taken to onset of clipping, a pair of EL34 or
6L6GC will make 50W at OPT sec with
4k0 : 4r, 8r0, 16r0. At 50W, B+ may sag to +425Vdc, Va-a =
460Vrms, Input power from PSU = 88W,
Output power at anodes = 53W, total Pda = 35W, so each EL34 has
The Pda may be sightly higher before clipping, but it is always
below the 25W Pda rating.
Screens don't overheat. A pair of KT88 or KT120 will only make
slightly more Po to the same loading,
but will just last longer.
Graph 1 also shows Pda for ONE output tube at the left side of
graph for region of high Po.
To avoid the high Pda for 2 output tubes, the simplest solution is
to use 4 output tubes with the
same OPT, and same Ea but with lower Eg2 at say +400Vdc.
Graph 2. KT120 AB loadlines, beam tetrode, Ea +500V, Eg2
The Curve A is Ra for Eg1 = 0V, for KT120 with Eg2 =
500Vdc. The curve is also called the "diode line"
and negative going Va peak swings for class AB1 cannot be to the
left of this curve. The OPT I used for
the tests had ZR = 1,233 : 1, TR = 35.11 : 1. For example, if the
RLa-a was 3k0, then B RLa for each tube
in class AB mode = RLa-a / 4 = 3k0 / 4 = 750r. The minimum Ea and
maximum Ia for load 750r occurs where
the drawn load line of 750r intersects the Ra curve, at Ea 103V
and Ia 412mA.
Notice that the knee of the KT120 Ra curve is at Ea 150V x Ia
The curves were produced by using an OPT with low Rw and using Sec
loads between 0.567r and 12.9r
and careful measurement of Va at onset of clipping which occurs as
soon as the Va swing hits the diode line.
The Curve B is Ra curve for Eg1 = 0V, with Eg2 = 300Vdc,
copied from old 1960 data sheets for MOV KT88.
I have not one idea of how the the MOV company laboratory
technician achieved the data curves. It seems
to me Russians and all other modern producers don't know how to
test their tubes properly and many curves
you see online are suspiciously exactly like the 1960 curves.
Digital imaging makes it all look better, but with
less actual information.
Data sheet for MOV KT88 has knee of Ra curve for Eg1 = 0V
at about +150V x 450mA which is above and
outside the data sheet. Notice the slope of the Ra curve below Ea
50V. It is a straight line of 143r. Nowhere
could I find modern Russian made tubes with diode Ra curve slopes
less than shown above where below Ea
100V is is nearly straight line of 200r. This means Ea swing
cannot extend downwards as far as the 1960
KT88. The minimum RLa-a in 1960 would have been 4k0, with B RLa =
1k0, with minimum Ea 87V and max
Ia 412mA. Po = 85W which could easily have been achieved in 1960.
It is a pure co-incidence that same 1k0
load intersects the KT120 Ra curve at the same point.
If Eg2 is increased for KT88, the knee of Ra curve rises,
and the grid bias -Eg1 must become more negative.
Therefore Vg1 swing can be higher without g1 grid current so that
maximum Ia pk is higher so Va can swing lower
with a lower B RLa.
Conversely, if Eg2 for KT120 is lowered, knee of curve is lowered,
the bias -Eg1 will be less, so Vg1 cannot
swing as much to get such high Ia pk to a low B RLa to enable the
I show B RLa = 175r, what I call an illegal load; RLa-a = 700r.
This gives 85W but the quality of power is just
awful. But you get 85W with RLa-a = 4k0, and the THD is far lower,
and nothing overheats.
Tubes prefer high Vac change and low Iac change to get high Po.
Ra curves for different Eg1 -Vdc grid bias have not been included
because it requires much more sophisticated
test gear than I have time to make. All the Ra curves are not
needed to estimate the probable results.
Only Ea and Eg2 need to be known at the onset of clipping but but
at least Eg2 should be regulated when
testing with continuous sine waves.
The OPT I used for tests in 2012 weighed 10.5Kg, and load
ratio was 1,233r : 1r0 with all secs parallel,
or 1,233r : 4r0 with 1/2 the secs in series with the other half. I
used the OPT strapped to give ZR = 1,233 : 1,
TR = 35.11 : 1.
The OPT was meant for 8 x 6550 for about 240W to RLa-a = 1k5.
( each pair of 6550 would have RLa-a = 6k0. )
Winding resistance losses are low with high RLa-a, but
become high with low RLa-a.
RwP for each 1/2 Pri = 15r, for 30r for the whole Pri. RwS =
0.0355r, which is transformed to become 44r
across whole Pri so Total RwP+S at Pri = 30r + 44r = 74r.
But in class B, one of the tubes plus 1/2 of Pri is used for a
positive going 1/2 wave, and the the other tube
plus the other 1/2 Pri is used for the negative going 1/2 wave.
This may seem hard to understand, but each
Vac to each grid has opposite phase, and so each tube takes turns
to turn on during only positive going waves
at their grids. Each tube pulls the Ea towards 0V, to produce
power only during negative going Va swings.
For the instant while on tube conducts current while the other has
no current, only 1/2 the Pri is coupled to
load so TR is halved to become 17.555. Thus ZR = TR squared = 308.
Therefore Rws is transformed to
become 308 x 0.0355r = 10.9r at the sec. ( it is near enough to
The transformer may be considered without RwP or RwS and "perfect"
with all its winding Rw in series with
the primary winding input, and in class B, at each anode input
where Iac flows, the total RwP + RwS
= 15r0 + 11r0 = 26r0. If the load to input of this Pri = 500r,
then the load at perfect OPT = 474r, and this is
transformed at Sec to 474r / 308 = 1.264r.
In heavy class AB operation, the amp works similarly to a class B
amp, and the total winding loss
= RwP+S at Pri / total input RLa load, in this case, 100% x 26r /
( 474r + 26r ) = 5.2%.
Where the RLa = 375r, loss = 100% x 26r / 375r = 6.9%, for 250r
its 10.4%, for 175r its 14.8%.
But for initial class A mode, with both tubes acting like class A
single ended tubes, and with all the primary
being used for power production the RwP + RwS total for whole Pri
= 30r0 + 44r0 = 74r, so for where total
RLa-a = 2k0, the total RwP+S loss = 100% x 74r / 2,000r = 3.7%,
and this is approximately 0.7 times the
loss when working in high Po class AB. Most listening is done at
low levels, and winding losses do not
much increases above the level for class A.
Graph 3. KT120, KT88, AB loadlines, beam tetrode, Ea +535V,
Graph 3 shows test results gained in 2017 with KT120 and KT88
tested at the same time in same circuit.
The B+ = +535Vdc and Eg2 = +410Vdc and both were well regulated
with Vdc sag < -5Vdc at full Po.
The OPT is similar to what I used in 2012, with RwP+S in class B =
27r, and 69r for class A.
With Eg2 at +410Vdc, there was less alarming screen Idc at full
Notice the odd values used for RLa. But these were gained using a
dummy load with 4mm banana
sockets to allow Sec loads to be 6r0, 8r0, 12r0, 16r0, 26r0, 38r0,
62r0. The B RLa was calculated after
carefully measuring the RwS with a 300mAdc Idc in its low
resistance and RwP with DMM.
Graph 4. KT120 + KT88 AB1 Po vs RLa-a for Ea +535Vdc, Eg2 =
Graph 4 shows the Po vs RLa-a for KT120 and KT88 based on
measurements for Graph 3.
You can see that KT120 and KT88 give nearly equal performance in
any amplifier with KT120 only
giving a slight increase on Po at high levels with a low load. Pda
ratings are reached at 105W for 2k8
for KT88 and at 120W for 2k0 for KT120. But with a continuous sine
wave, both tube types would begin
to overheat because of much increased screen Pdg2. Both tube types
will survive with high output levels
in most music where the average level is 1/10 the clipping Po
level to ensure peaks in music do not clip.
But for guitar amp use, the tubes are often repeated driven into
gross overload and class C operation
and ONLY way to prevent tubes dying in the heat is to NEVER use
RLa-a less than 4k0. 90W is
available, enough for most rock musicians - in theory.
For hi-fi, loads higher than than 8k0 are best.
The main advantage of KT120 is that a pair can be used with idle
Ea of say +400Vdc, Ia 100mAdc,
Eg2 300V, in class A with 20% CFB, and RLa-a 7k2, for about 35W
pure class A. 2 x KT88 would have
70mAdc, RLa-a 10k2, pure class A = 24W. For most hi-fi enthusiasts
with sensitive speakers I can
suggest you may think angels brought you your music.
Graph 5. KT120+KT88 AB loadlines, beam tetrode, Ea =
+600Vdc, Eg2 = +410Vdc, both regulated.
Graph 5 shows same Ea vs Ia Ra curves for Eg2 = +410Vdc. But Ea is
raised to +600Vdc.
Idle Iadc can be 40mA for KT88, and 50mA for KT120.
Graph 6. KT120+KT88, AB Po vs RLa-a, beam tetrode, Ea =
+600V, Eg2 = +410Vdc, both regulated.
The 42W Pda rating is reached at 127W for KT88 with 3k2. 60W Pda
rating for KT120 is reached at
140W for 2k5.
I have not measured KT120 or KT88 with Ea 600Vdc. But where Va is
pulled low with high peak current,
the screen current will reach high levels above 27mAdc, but not as
much for where Eg2 = +500Vdc.
The alternative path to much more reliable high Po and more
initial class A is to use a quad of tubes
instead of a pair.
Graph 7. KT120+KT88 AB Po vs RLa-a, +40% UL Operation.
Graph 2 shows Ultralinear ( UL ) operation. For 2 x KT120, Ea =
Eg2 = 550Vdc and 40% screen taps
give a safe maximum AB1 Po of 115W for KT120 and 100W for KT88 or
6550. Max Pda per tube is
50W with RLa-a 2k5. Never ever use RLa-a < 4k0.
For almost all UL amps, the screens connect to taps on the
OPT primary so Eg2 = Ea. Many hi-fi
amps use the 40% taps for UL; RCA used 30%, Leak used 50%. The UL
taps lower the effective Ra
at idle condition to about 2k9 for Vac operation, THD becomes more
like triodes, but the Po max is
similar to beam tetrode with a fixed Eg2. With Eg2 = Ea, the
screen Ig2dc input is higher at idle than
is Eg2 is lower than Ea. With Vac operation, 40% of the anode Vac
is applied to g2 so where Va is
pulled low to say +100V with high Iapk swing 400mA, the screen is
pulled low to +370V so peak
screen input current is not as high as where Eg2 were to remain at
Thus the Eg2 and Ig2 and Pdg2 ratings for UL operation do not
apply for UL and screens can be at
a higher Eg2 than data suggests. In 1960, it was OK to use 6L6
which had Eg2 rating only +300Vdc
with UL connection with Ea and Eg2 at +450Vdc.
The same can be said about triode connected tetrodes and pentodes
where the screen is under
even less stress than UL.
99% of UL output stages have no active regulation of B+. Therefore
mains change levels which
cause B+ changes and fixed bias Eg1 changes also cause Eg2 to vary
equally as the B+ and
g1 -Vdc bias. So for DC operation, the Ra of each output tube is
equal to a triode. But the idle Pda
must include Pdg2. For KT88, typical Idle Pda+g2 = 25W, or max Pda
rating 42W x 0.6.
Thus Ia + Ig2 = 25W / 550V = 45.4mAdc, Ia will be about 40mA and
Ig2 about 5mAdc.
KT120 could have idle Pda = 36W with Ia 59mAdc, Ig2 6mAdc.
Calculate approximate Eg1 bias.
With UL, Eg2 is always = Eg2, and its the Eg2 which much
determines -Eg1 bias Vdc. The g1
biasing for UL is the same as for triode connection, and a rough
estimate of -Vdc g1 bias may be
Eg1 bias = ( Eg2 - [ triode Ra at 0V to 100V x Idle Ikdc ] ) /
Triode µ at Idle Q point.
For KT120, at low Ea and Eg1 0V, Ra = 930r. Ikdc = Ia + Ig2 =
65mAdc, Ea +550Vdc, µ = 7.2,
Ra = 1,200r.
Eg1 bias = ( 550V - [ 930r x 0.065A ] ) / 7.2 = -68Vdc.
With beam tetrode, Ea +550Vdc, and fixed Eg2 +400Vdc, neglect Ig2.
Ia = 60mAdc, and
Eg1 bias = ( Eg2 - [ triode Ra at 0V to 100V x Idle Ia dc ] ) /
Triode µ at Idle Q point.
Eg1 = ( 400V - [ 930r x 0.06A ] ) / 7.2 = -48Vdc
Calculated Eg1 is approximate, and if you build an amp, you will
have to adjust the -Eg1 to each
tube to obtain the wanted idle Ikdc and wanted idle Pda+g2.
For KT120 triode at idle Ea+Eg2 = +550V and Iadc = 65mAdc, Ra =
1k2 at idle. This assumes
mains input = 240Vrms.
The change of mains Vac can give interesting results if the B+,
and Eg1 rails are not regulated.
if mains increases from 240V to 260V, its +8.3%.
B+ will increase to +595Vdc, and if Eg1 was regulated, then for
each KT120 triode or UL, Ia+g2
increase = V change / Ra = 45V / 1,200r = +37mA. So Pda will be
595V x 102mA = 61W, quite
unacceptable. But with Eg1 rail not regulated, the Eg1 bias
increase = +8.3% = -68V to -74Vdc,
and the -6V increase makes Ia change = Eg1 change x gm g1 = -6V x
6mA/A = -36mA, and this
opposes the effect of raising Ea +45V. Therefore Pda will not
If Ea = 595Vdc, and wanted Pda = 36W, then Ia+Ig2 = 60mA.
Eg1 needed = ( 595V - [ 930r x 0.06A ] ) / 7.2 = -75V. This
calculation shows Pda increase is not
excessive where Eg1 varies by about the same % as Ea.
If mains reduces to 220Vrms, its -8.3%.
Ea = +504Vdc, a -47V reduction. Unregulated Eg1 change = +5.6V to
For the same Pda+g2 36W, Ia+Ig2 = 36W / 504V = 71mAdc.
Wanted Eg1 = ( 504 - [ 930r x 0.071A ] ) / 7.2 = -61Vdc.
Well, -61Vdc is close to -62Vdc, So Pda slightly reduces with the
mains reduction, and so far,
there is nothing to worry about.
Thus, DO NOT REGULATE Eg1 for UL or triode output stages
UL and triode output stages need more Vac to drive output grids
than pure tetrode use with
fixed Eg2. To get full Po the peak Vg must equal the Eg1 -Vdc
bias, so if Eg1 = -70Vdc, 50Vrms is
needed to each g1. Thus the driver amp must be able to make the
Vac without its THD exceeding
Instead of using UL at all, it is always better to use
between 12.5% and 20% of total primary
turns devoted to local cathode feedback windings. I consider 20%
the best % to use.
The Eg2 is fixed, and should be shunt regulated, and can be lower
than Ea. So for high Po,
Eg2 could be +400Vdc and Ea = +600Vdc and Po is similar to what I
show above in Graph 6.
Consider KT88 operation with a fixed Eg2 +400V and and 20% local
NFB with anode Po =
117W for 4k0. Va-a = The Vac for this = 684Vrms. The Va-k for each
tube = 342Vrms.
Consider one cathode has 20% of Va-k = 68.4Vrms+ = +96.7Vpk. Va =
273.6Vrms- = -386Vpk.
Minimum Vapk is at +600V - 386Vapk = +214V. Maximum Vk pk =
+97.7V. The Va-k minimum
= 214V - 97.7V = 116V.
Now between the fixed Eg2 +400V and cathode, the Eg2 to k minimum
= 400V - 97.7V =
302V. Each tube works similarly to a UL stage with 20% UL taps.
But unlike UL, the CFB has the
Eg2 well below the the level of +600V with UL, so the CFB causes
much less screen heating
and robbing of anode current than 20% UL or with pure beam tetrode
with fixed Eg2 at =400V.
I doubt it would be wise to use KT120 with 40% UL taps with B+ =
Ea = +600V.
But it was considered OK to have B+ = Ea = 550Vdc for KT88 with
40% UL taps to get 100W
safely, so +550Vc would be OK for KT120.
Ra, gm and µ for KT120.
For all tubes, µ = gm x Ra.
KT120 at Ia = 50mAdc, and Ea +500V, and Eg2 +400Vdc, Ra = 32k, gm
g1 = 4.7mA/V,
gm g2 = 0.9A/V, and max µ g1 = about 150. The µ of 150 is a number
for "amplification factor",
and with Ia dc feed from a constant current source or a large
choke with ac reactance > 5Meg,
the only loading of the tube is the is the Ra between anode and
cathode which is a constant
32k where the Eg2 to k is kept constant with no Vac. So for 1Vrms
at g1, expect Va = 150Vrms
with no other load. The ra is not an actual resistor, but any Va
change causes Ia change; thus
anode has transconductance gm a and for 150Vrms you find Iac 150V
/ 32k = 4,68mA, so
anode Gm = 4.68 / 150V = 0.031mA/V. Ra is a very ignored figure,
but it is always present.
For 20% UL operation, the screen has 20% of Va and it causes Ra to
become lower, and
it is calculated :-
UL Ra = Tetrode Ra parallel to 1 / ( UL fraction x gm g2 )
= 32k // 1 / ( 0.2 x 0.0009A/V ) =
32k // 5k55, = 4k7.
Without any load, UL µ = gm g1 x UL Ra = gm g1 x UL Ra = 0.0047A/V
x 4,700r = 22.1.
For 40% UL, UL Ra = 2.6k, for same idle Ea and Ia conditions.
Best use of 40% UL is for low Ea and class A where idle Ea = Eg2 =
420Vdc, and Pda+g2
= 36W so Ikdc = 86mAdc, and the tetrode Ra will be 28k, gm g1
maybe 6mA/V, gm g2 = 1.1mA/V
and the UL Ra = 2k1, with UL µ = 12.6.
The 40% UL does give an output stage with very good character with
Ra-a about same as RLa-a,
and triode like THD .
With 20% UL taps, each KT120 at idle has Ra = 4k7, gm = 4.7mA/V, µ
The 20% CFB effectively gives tube Ra' = UL Ra / ( 1 + [ UL µ
x ß ] )
= 4k7 / ( 1 + [ 22.1 x 0.2 ] ) = 867r, and this is less than
triode, which curves tell me is about 1k2.
For RLa-a = 8k0, and low level class A giving RLa = 4k0, the 20%
UL open loop gain =
= UL µ x RLa / ( RLa + UL Ra ) = 22.1 x 4k ( 4k + 4k7 ) =
Consider 20% CFB class A operation at about 5W level Va-k for each
KT120 = 100Vrms.
Va = 80rms-, Vk = 20Vrms+, Vg-k = 100Vrms / 10.16 = 9.84Vrms.
Vg-0V = Vk + Vg-k = 20Vrms + 9.84Vrms = 29.84Vrms+.
Therefore the closed loop gain, ie, overall gain with CFB
applied = Va-k / Vg-0V =
100V / 29.84V = 3.35.
The gain reduction from UL to CFB is from 10.16 to 3,36, so the
CFB connection is like a
UL amp with 10dB GNFB, and damping factor = 4.6, and I know ppl
who just don't bother with
any added GNFB with a 20% CFB stage with RLa-a 8k0.
Thus the KT120 works like it is a low µ power triode similar to
300B which has µ = 4.2,
Ra 800r, and gm g1 about 5.2mA/V. But 300B is a fragile old design
for triode that was first
made by WE in 1928 to provide enough power for movie sound in
movie theaters. Audiophiles
have been in love with the 300B for the past 90 years, but it has
much more limited abilities than
a KT120. And KT120 in triode mode is also quite something.
Driving CFB tubes.
With KT120 with 20% CFB, and to make 117W to RLa-a = 4k0, Va-a =
684Vrms, the closed loop
gain will probably be about 2.8, so Vg-g needed = 684 / 2.8 =
244Vrms, 122Vrms at each grid.
So the driver amp should be able to make about 130Vrms at
least to each KT120 grid.
Its not so easy to do this with low THD unless you have a choke
with CT + R between ends of
choke to to a pair of EL84 in triode mode, see my 300W amp driver
I also have CFB at 8585-amp-october-2006.html
The last version of 8585 had 4 x KT90, easily giving 100W per
channel with RLa-a = 4k4.
So each pair of KT90 are loaded by 8k2, and having easy work to
do, but the THD is very low,
and tubes needed to have idle Ik = 33mAdc, with Ea at +480Vdc
for idle Pda+g2 = 16W, and when
I serviced the amp after 5 years of daily use the tubes behaved
like new tubes, with a slight dark
blue glow, and no positive grid current at idle.
The owner had 3 stacked ESL57, fully restored, in parallel at
each channel. He said the sound
was just fine.
In 1957, The General Electric Co of the UK produced a book
"Audio Frequency Amplifier Design
with 17 schematics for amps from 5W to 1,100W. PP KT88 is shown
with 43% screen taps and
Ea = 550V giving 100W max with RLa-a = 4k5, but not much mention
is made of screen dissipation.
We may assume the GE amp recipe would be safe with a sustained
sine wave at clipping at 100W
with RLa-a load of 4k5. But with 2k5, I would suspect the KT88
would overheat and destroy
themselves when there is sustained excessive Pda. KT120 would
work for slightly longer with 2k5.
Some of my conclusions :-
1. For Ea up to +550V, KT120 give a small increase in maximum
AB1 power over KT90, KT88 or 6550.
2. At maximum possible class AB1 power, and regardless of what
beam tetrode is used, there is a
very small amount of pure class A1 power below the A to AB
threshold is reached.
3. The higher Pda rating of 60W for KT120 does allow for a
higher amount of sustained output power
into lower load values than tubes with lower Pda ratings.
However, there are limits to what is possible
no matter how many output tubes are used or what their combined
Pda max rating.
It must be remembered that a quad of 6550 or KT88 or a six pack
of EL34 or 6L6GC will more easily
do what a pair of KT120 will do.
4. High power is always possible with very low anode loads if
the screen voltage is raised to a highest
possible voltage. But the highest possible screen voltage could
result in screens overheating at high
power so tube is damaged. Thus you rarely ever see guitar
amps with B+ rails exceeding +470V.
Where Ea exceeds +470V as in the Ampeg 300W SVT with 6 x 6550
for 300W max, the screens are
at a much lower Eg2 produced by a separate lower Vac HT winding
to prevent much Eg2 sag at high Po.
But inevitably, Ea and Eg2 can sag a bit with a sustained sine
wave signal. Each 6550 has a high value
series g1 "stopper" resistor of 47k from the cathode follower
driver to try to limit overheating if ever grid
current flows, and also limit HF response to no higher than ever
wanted, and to stop RF oscillations if a
tube does overheat.
5. Consider a pair of KT120 with B+ anode supply = +700V, and
screens at +350V. Let RLa-a = 6,200r.
Loadline analysis shows Va pk = 600V, Va-a 848Vrms, Po max =
116W. Max Ia pk = 390mApk,
and Pda per tube = 28W. Efficiency = 66%, good for a class AB
The Pda is not always highest at clipping for a low load RLa-a.
If the 2 x KT120 produce 90W for 6k2, Pda per tube is 31W.
A perfect class AB amp with devices with no resistance or limit
to Va swing, so peak Va swing = Ea,
the class AB efficiency max = 78%. With KT120 with perfect
character, expect 157W to 6k2, and Pda
per tube = 22W.
But only 116W is possible with real world tubes, Pda per tube =
Pda does NOT increase linearly with Po level, and can be higher
for slightly lower Po for the same RLa-a.
But if the Eg2 is raised to +400Vdc, and RLa-a = 4k0, we could
get Po = 162W.
Input dc power = 253W so Pda per tube = 46W. Pda may go a little
higher as Po is reduced before rapidly
reducing to the idle Pda at low Po levels.
At all levels the Pda is less than the KT120 60W Pda rating. So
using higher Ea and less Ia swing is better
than the situation with Ea = +500Vdc, = Eg2 and RLa-a = 2k0 for
If you want lots and lots of power, and you want reliability,
use lots and lots of tubes, and never expect
more than 75W from one pair of beam tetrodes in the group of
6550, KT88, KT90, KT120. Never expect
more than 45W from 1 pair of EL34, KT66, 6L6GC, 807, or more
than 20W from a pair of EL84 or 6V6.
6. If you want a reliability with tubed high power class AB
amps, you need several protection measures.
There should be separate grid bias pots for each output tube,
10k, wire wound.
Use LED indication of bias current status at idle. In an ARC
VT-100 I totally re-engineered, there was a
red+green LED for each tube alongside the bias adjust screw for
that tube. If Idc was too high,
LED glowed red, if Idc was too low, it glowed green, and if it
did not glow the Idc was just right,
but if they all glowed slightly dull green, Idc was a little
low, but OK. Having slightly low Idc has never
killed a tube.
During class AB power production, red LEDs will light up
indicating increased average Iadc in each
OP tube. The amp was used for hi-fi, and the to get the LED to
glow red the sound had to be deafening.
In PA and guitar amps, you may assume max Idc from PSU to be up
to 4 times the idle PSU power.
2 x KT120 might each have Ia + Ig2 = 50mAdc at idle for 100mAdc.
But at full Po into lowest RL value,
total Idc may be 400mA. So unlike the situation with hi-fi amps,
you cannot use idle Idc current sensing to
protect a tube amp. However with red-green LEDs, if a tube turns
an LED to red at idle while others are
unlit, then it indicates a problem, and action can be taken,
which may mean a tube replacement during a
break in the gig. If you ignore red LEDs then shit happens.
There have never been any amplifiers made which have a circuit
built in to detect when a load value
is too low. I have never got around to building a differential
amp with 2 bjts which has a fraction of Va
applied to one input and a Vac signal from a current sensing R
in speaker return path. It need only be
0.1r0. Depending on ratio of the two Vac signals, a low output
load, such as a shorted speaker lead or
a jammed voice coil can be made to make a relay turn off the
amp, and have it stay turned off until the
problem is fixed. This protection would have prevented countless
repairs I did to solid state amps
damaged because of bad speakers or shorting speaker cables, and
bad dopey owners.
Sumwunn@sumwear.com.uv must have invented a small simple circuit
which samples Vac in a load
and Vac across a load and automatically can give a digital
read-out for the ohms resistance of the load.
It thus could be used to turn of an amp if load is too low, and
trigger a short audio message :-
"Do not play more music because fah-khan load is too fah-khan
But the reason such protection has never been used commercially
is that amp makers want you to
Blow Up Your Amp Soon, so you will upgrade to a new one. Its the
BUYAS mindset of greedy companies
at work. Some amps are made to work with lowest load of 4r0. But
they sure may blow up with 2r0 when
Mr Sum Idiote cranks up the volume.
The 2r0 load should immediately be detected and make the amp
turn off within a second if it detects a
low load, even at low levels.
Tube amps do tend to last longer than transistors when used
briefly with loads that are too low, but for
hi-fi amps, I always fitted active protection circuits which
turned off the amp if the average Idc at cathode
of one or more output tubes doubled from say 50mAdc to 100mAdc.
This often happened at idle when
tubes aged, and began conducting too much Idc. Many owners
thanked me for this feature because it
indicated when they needed to replace they output tubes.
7. Real benefits of KT120 would be most possible for where LOW
power is desired for hi-fi and where
clipping will never occur, and the power is nearly all pure
class A. The maximum pure class A possible
from any two beam tetrodes or pentodes is about 45% of the total
Pda at the idle condition, so that if two
KT120 are used with idle Pda at a safe 40W each, Pda total = 80W
and max class A PO = 36W, and
this is substantially above a pair of KT88 with safe Pda total =
60W, and class A PO max = 27W.
For class A, RLa-a is always higher than for class AB. Peak Ia
swing in the class A amp is never more
than twice the idle Ia.
But AB amps have maximum peak Ia pk much higher than idle Iadc,
with higher Pda than at idle, so max
AB idle Pda should be lower than 0.6 x Pda rating, lower than
for class A at 0.7 x Pda.
For example, consider KT120, Ea = Eg2 = +400V, Ia = 100mA, and
RLa-a = 5k6. Peak Va swing = 360V,
and Po = 46W, AB1, with the first 28W in pure class A1. The
loadline analysis shows the 60W Pda limit
line for KT120 well above the class B load line of 1,400r, so
there is no risk of tubes overheating.
Even if the load is reduced to 2,800r, The B RLa is only 700r
and max AB1 PO = 68W, and there is still
no risk of tubes overheating. From this it is possible to use
KT120 at an idle Pda = 40W reliably so that
Ia = 100mA, Under such conditions, distortion is minimized, and
music may sound much better than if the
amp were set up with Ea at +550V, and Ia at idle of 50mA, and
RLa-a = 5k6 and initial class A max = 7W.
Back in 1955, many people used a pair of 807 to make 80W in
class AB2 with Ea = +600Vdc,
Eg2 = +300Vdc, and cathode followers directly driving grids
which become low input resistance when
+Vg pk rises above cathode Ek. If you wanted 160W, then just use
4 x 807. The 807 has anode top caps
which allowed the high Ea swings safely without risk of arcing
from anode to adjacent heater pin 3 of an
octal socket. So here we are in 2017 and I am sure a
pair of KT120 will cost a lot more than a quad of 807. But the
807 THD at 80W was 13% at without any NFB.
8. During the history of tube amplifier development since about
1920, using low idle bias Ia and high Ea to
give ever more Po became popular in a world which screamed for
more and more power in all areas of
human existence. I have a 1983 copy of the UK Radio
Communication Handbook, 5th Ed. On page 9.32,
there is a 140W PP amp schematic for a modulator for an RF
transmitter. It uses one 12AX7 with 2 triodes
producing a balanced Va to drive a 12AU7 which has its anodes
driving the pair of grids of two TT21.
These drive the primary of an OPT with 16k8 primary load and
secondary load is 7k0.
But OPT could be 16k8 : 4r0, 8r0, 16r0.
The operation description has idle Ea = +1,000Vdc, Eg2 =
+300Vdc, and idle Iadc = 35mAdc and Pda does
not increase at any Po up to 140W where Va-a must be =
1,533Vrms, so at each anode Va = 767Vrms.
But if Va = 767Vrms, then Vapk = 1,084Vpk, completely impossible
unless Ea was say +1,140V.
GEC data says Ea can be 1,250V. I calculated Pda at 140W was 24W
per tube at 140W and efficiency
= 74%, excellent for any class AB amp.
Eg1 fixed bias applied to both tubes = -40Vdc. Text under the
schematic says Iadc should not exceed 35mA
at idle, which puts idle Pda = 35W. While Ea at idle may
+1,140Vdc, instant power would be 140W, but at a
constant 140W, non regulated B+ might sag to +1,000Vdc, so Va-a
may be 920Vpk, Va-a = 1,300Vrms,
Po = 100W. This sag of B+ is mentioned. I think using a quad of
KT88 with Ea at +500V to make 140W
would be less dangerous, and easier to do.
You must not try to use more than +700Vdc at pin 3 on an octal
socket. There is high risk of arc to adjacent
earthy pin 2 for heaters. But pin 4 is for g2, and that may be
at +450Vdc. The next pin 5 is for grid g1,
and has a -Vdc bias of about -40Vdc to -50Vdc, but the socket
must be kept clean and free of pollution to
prevent tiny leakage Idc to grid bias resistors which may be say
I have serviced amps where arcing occurred from pin 3 to 2 and
with quite low Ea = 350Vdc = because
of socket pollution. This was more a problem when so many more
ppl smoked tobacco indoors.
The 1960s, Mullard advised us that 96W was possible with 2 x
EL34, and these used B+ = 900Vdc
generated by a voltage doubler using the new silicon diodes with
negligible 'on' resistance. So +450Vdc
was easily produced for Eg2.
I repaired one with 8 x EL34 and which gave 500W, ie, each pair
made 125W. There was almost no initial
class A Po. When I got it from owner it had 4 different brands
of EL34, all with very different Pda with the
same single -Eg1 bias. It regularly smoked. 2 sockets has burn
damage between pin 3 and 2 because
of arcing. But the owner admitted he'd replaced 2 x 12" x 8r0
speakers of the total of 12, in two bins.
Dynamic tweeters had fused open years before. He'd rewired all
speakers in parallel, not knowing what
an Ohm was. Both bins in parallel gave load of 0.65r, when it
should have been 6r0 for both bins.
He'd used the amp for PA indoors for years, and maybe Po max was
50W, and tubes regularly failed.
But I found a tap on HT winding which gave B+ = +670V, and amp
gave 225W with a correct load.
I put in LC filter for Eg2, did numerous circuit mods, put in 8
bias pots, one for each tube, and 8 test points.
The owner's son of 13 learnt how to adjust the bias. I fixed the
speakers, installed horn tweeters, equalised
bass and treble levels and installed box braces and polyester
fill. Later I went to a gig for
Australian - Philippine Association and the sound was just
magnificent, detailed, no distortion, with warmth,
and everyone was having a ball. I also made him a solid state
amp with an op amp which limited gain
so that when the kids got onto stage to sing, their distance
from the microphone was far less critical so the
quiet singers were heard OK and the loud ones didn't overload
In 1990s, Musical Reference sold stereo amps with 2 x EL84 which
made 35W per channel with B+
= +700V and Eg2 = +350V. I repaired one where Anne Idiote had
replaced 200mA fuses between common
cathodes and 0V
with 2A fuses. Where a human mistake is at all possible, then a
small % of ppl will make that mistake.
So the fuse never blew, but the tubes sure did. Analysis shows
the 35W was possible, and that's what
All these contraptions went up in smoke all too easily. The EL34
amps were a favourite in big churches
where a priest gave thunderous sermons about sin and damnation,
while us boys gorked at the rich girls
in their Sunday best which our parents could not afford for our
sisters, who gorked about for good looking
boys. One of the speakers would fail, the amp soon followed, and
smoke poured from the amp enclosure.
One thing was certain, Satan was definitely listening to the
priestly bullshit !
To improve insulation at tube sockets, a small rubber O-ring
around pin 3 of octal output tubes could be
used to seal the tube base to the socket to prevent a corona
formation and effects of accumulated dust,
moisture, pollution allowing leakage currents from pin 3 and
eventually an arc. But tying down EL34 with
a plastic base is difficult. The classic guitar amp "tube base
grippers" with springy thin steel can be rather
useless because ppl bend them flat to get a tube out, and tubes
then are not held in well, and can fall out
especially where most combo guitar amps have their chassis in
the top with tubes hanging upside down,
likely to fall out over time with vibration. Best tube holders
for output tubes have a metal bracket fitting
around top of glass with two springs pulling tubes towards the
chassis. O-rings would help. Also, some
silicone needs to be squirted around pin 3 lug at tube sockets
under chassis to prevent arcs to pin 2 or 4,
or to anything else.
KT120, KT88, 6550 have metal base sleeves to which you can
solder some thick copper wire loops.
Insulated wire can be used to tie down the tube on two sides
through 2 chassis holes. The socket base
metal ring is connected to pin 1 of tube base plug pins. Many
ppl connect pins 1 and 8 for cathode and
g3 suppressor grid of EL34 and other tubes. You must never
assume you need to connect pin 1 to 8 for
the cathode. ARC amps had the weakest spring tension of all the
tube sockets I saw during 18 years of
repairs. 6550 could and did regularly fall out of their sockets,
being mounted horizontally. It was in 6550
that I learnt how to tie down the tubes.
Old 6L6 had the same electronic specs as 807 for max Eg2 =
+300Vdc. The later 6L6GC has higher
Eg2 rating of +450V, which allowed use in guitar amps where Ea =
Eg2 = +450Vdc. The high Eg2
allows Va to have high Ia pk = 380mA, and up to 68W from anodes
to 4k0. But never with a
continuous sine wave, because of Ea and Eg2 sag. But 50W is
For Hi-Fi, a quad of 6L6GC, 807, KT66, 6CA7, EL34 can easily
give 60W AB1, with 30W
of possible pure class A1 with Pda at idle of 18W per tube. This
will be just as reliable as using a pair
of KT120, KT90, KT88 or 6550 to obtain the same power. But KT120
would be the best at safely
sustaining idle Pda of 36W each for total Pda = 72W, equal to
Pda of 4 x 6L6GC.
Graph 8. KT120, AB Po vs RLa-a, beam tetrode, Ea = Eg2 =
+400Vdc, high initial class A Po.
Graph 8 shows AB Po vs RLa-a for a pair of KT120 with Ea = Eg2 =
400Vdc, Iadc = 100mAdc each,
idle Pda = 40W. Use of 43% UL or 20% CFB will give similar Po
results, but with less THD & IMD than
pure beam tetrode. Total Idle Pda = 80W for 2 x KT120, and with
RLa-a = 7k0, max initial pure class A
Po = 35W, and anode efficiency = 44%. The Po has much less THD
and IMD compared to the the
first 35W from a pair of KT120 set up to make 135W in virtual
Because its now 2018, and everyone feels terrible guilt about
climate change even when huddling
over a 1kW heater somewhere in a Canadian winter, I suggest the
use of a switch to increase -Eg1
grid bias to reduce max Pda at idle to 20W, and you'll still get
9W class A, and high class AB.
Only a few guitarists like low Po pure class A amps. I worked to
optimise many of the amps they
bought. High power was not always wanted, and if they wanted
high THD they could produce all
the THD in input gain stages and amplify it without driving the
output tubes hard, so the 30W they
have was plenty for a small venue. They often have several
effects boxes and do not want to lose
the effects by over-driving an output stage. There are smart
musos and dummies who just make noise,
and some muso amps have only one gain control at input and with
no "master" gain control before the
But all Hi-Fi listeners are allergic to any distortions. They
find the high class A from tubes to be better
than anything else they have tried. Just exactly why is also
mysterious, but I know what I was told
by so many customers over my 18 years repairing and
manufacturing amps and speakers.
Most Hi-Fi listeners use no more than 0.25W average Po per
channel. They want fine music without
ear damage or a divorce. If their speakers have 90dB/W/M
sensitivity, then 0.25W makes 84dB SPL.
Two channels gives 87dB SPL. If average Po = 0.25W, peak Po
could be 25W, with SPL 104dB.
I do know 5W is entirely inadequate for speakers rated for
87dB/W/M. I have a 5W SE amp with
1 x EL34 in triode with speaker rated for 93dB/W/M in kitchen.
In the small room it is fine for mono
sound. a single KT120 would definitely be better for more than
But 2 x KT120 can use any popular off the shelf OPT such as a
Hammond 1650P with 6k6 : 4,8,16r,
and rated for 60W. But you would not aim for high AB1 Po. 1650P
has 43% UL taps and is happy for
low Po most of the time with class A from KT120 with idle Pda
40W max, KT90 with 33W, KT88 and
6550 with 30W. but most ppl could never tell if Pda was 25%
lower. The lower idle Pda gives much
longer tube life.
For those wanting to KT120 to replace EL34, 6L6, 6L6GC, 5881,
KT66 to obtain more pure class
A1 power, idle Idc must be higher and heaters need 2.1A instead
of 1.6A for EL34. Many amps will
not have sufficiently rated power transformers. Tube rectifiers
such as GZ32 will fail. So the same
B+ Iadc should be used as in the original amps. If GZ32 are
replaced with Si diodes the B+ can be
higher and KT120 will produce higher AB Po, with same class A
Po. I have used KT88 and KT90 in
Quad-II amps meant for KT66 and found the sound improved, even
with the original toy like OPTs.
Heater power for 2 x KT66 = 16.4W, input EF86 add 3.9W to make
total 20.2W. We may assume
all transformer windings dissipate 5% of the wanted power as 2W
of heat in windings. Quad-II were
designed to have AM-FM tubed radio tuner and type 22 preamp
connected, so using KT88 or KT90
is OK without the tuner or preamp. But KT120 does NOT belong in
old Quad-II amps.
Anode current for 2 x KT66 / EL34 etc is limited to about
150mAdc max. The HT winding resistance
of many PT in old amps can be quite high; in Quad-II it is 125r,
to limit peak charge currents in tube
rectifiers. I found using LESS Iadc in old Quad-II worked just
fine where all the other PSU mods were
done including use of KT88 etc. The idle current and class A Po
max IS NOT the only thing to be
concerned about to get the best music with an old amp originally
designed for 22W.
KT120 could be well used in Quad-II-Forty amps which were first
produced by Quad in the mid 1990s.
KT88 were standard output tubes. But the HT winding for the 5U4
rectifier is 390-0-390Vrms, giving
B+ = +428V at 161mA. If Si diodes are used, B+ = +525V then
Eg2-0V = about +510V and some
adjustment of the two cathode biasing resistors is needed so
that Ek will be about +53V, and Ea
= +467V. Pda could be 35W with Iadc = 76mA, so Rk = 700 x 10W.
The Quad-II-Forty was designed
by Andy Grove but built in China at a time when the ONLY thing
China knew about tube amps was
how to do a nice looking paint job which in fact was a very
fragile quality of paint. Quad-II-Forty has
a HT winding that has no taps, and its Vac is too high, and 5U4
rectifier was designed for low peak
charge current using choke input ( L+C ) type of PSU. The
5AR4 / GZ34 was always a better tube
rectifier for Quad-II and allowed 33uF + 33uF instead of
original 16uF+16uF caps.
But see my Quad-II power amp mods page at quad2powerampmods.html
See other pages for other amps at re-engineeredamps.html
shows a long list of amps I modified
for better performance.
Graph 9. KT120+KT88 AB Po vs RLa-a, TRIODE, Ea = Eg2 = +500Vdc.
High possible Class A.
Graph 5 shows KT120 compared to KT88 in class AB1 PP triode. At
idle, Ea = +500V, Ia = 50mAdc,
Pda = 25W. All these tubes make about 22W of pure class A to
KT120 can have idle Pda 70mAdc each and they can make 27W pure
class A for 11k5, and higher
Maximum triode AB Po is produced with RLa-a between 2k0 and 2k5,
but nobody sane would
ever waste tubes trying to get high AB Po with triodes with very
low RLa-a. Nominal ideal RLa-a
would be 8k0, where initial 10W to 20W is pure class A with 32W
to 35W of class AB. it should
sound very well.
For your continued education..........
Graph 10. Ra curves for Screen Eg2 vs Ia for Eg1 = 0V.
Graph 10 Ra curves have been reproduced from 1960 MOV data
sheets for KT88 which appear
to have been very poorly drawn and published. I have reproduced
the curves here which show Ea
vs Ia where Eg1 is kept at 0V. These Ra curves are approximate.
You can drive a tetrode or pentode with Eg1 kept fixed, and
apply input Vac to screen, so for
example, if the idle point was at Ea 300V and Ia 125mAdc, with
Eg1 = 0V, then Eg2 would be +100V.
If Eg2 increases +75V the Ia = 250mA, and if Eg2 decreases -75V,
Expect Ia = 20mA. The transfer
curve overall is not very linear, but for this range of Ia
change the current THD = 5.5%, mainly 2H
and THD at Ia change < 20mA would be quite low. The average
screen g2 gm = Ia change / Eg2
change = 230mA / 175V = 1.3mA/V. If you look Ea 200V vertical
line, then for Eg2 change from 250V
to 300V gives I change 285mA to 375mA for gm = 90mA change / 50V
Eg2 change = 1.8mA/V.
Now if you have not fallen asleep, take a look at curves at basic-tube-4.html See Fig
5, and left side,
about 1/2 way down the page for 6550 in single ended class A
beam tetrode mode. Idle conditions :-
Ea = +380V, Ia = 67mAdc, Eg2 = +237V, Ig2 = 5mAdc, Ek =
+22.8Vdc, Rk = 317r. Eg1 bias = -22.8Vdc.
The graphs for Va vs THD are drawn using logarithmic scales for
both axis, so a straight line indicates
linear transfer function.
Solid curve C-C is for g1 drive with fixed Eg2, and is one of
the most linear curves I observed in this
group of tests conditions.
Dashed curve D-D is for g2 drive with bias at +237Vdc, with
fixed Eg1 connected direct to 0Vwith
Ek giving Eg1 bias = -22.8Vdc.
Curve D-D is even more linear than for g1 drive. For 9W, Va =
201Vrms, and I calculated g2 Gain
= 4.28, so for 9W, you need Vg2 = 201Vrms / 4.28 = 47.0Vrms.
This is not a high drive Vac but it
needs to be from a cathode follower of say 1/2 6CG7 with Rout
< 750r, because the screen needs
Idc feed and has finite input Z of maybe 12k. The tube is
working in triode mode where cathode and
g1 give a current source which can be linearly changed by Eg2
But this form of triode has virtually the same Ra as for beam
tetrode with fixed Eg2, and and local CFB
could not be effective and so GNFB is needed if NFB is to be
used. But basically, g2 has properties
you should know very well.
Most "Ea vs Ia characteristics" for beam tetrodes and pentodes
give Ra curves for varying Eg1 bias
with a fixed Eg2 between +200Vdc to+350Vdc. If you look at above
Ra curve for Eg2 = +350V,
it should be the same for Ra curve you would get if Eg1 was
fixed at +350Vdc, and for Eg1 = 0V.
You may wonder how anyone would test the tubes to get Ra curves,
or examine them to see how
long they will last. I searched Google for how they found the
data curves in 1960, and found nothing,
which means many ppl are making tubes and selling them, but not
one ever gets the curves for
what he makes. You can see how I got my curves above, but the
other method is to apply short
time Vac from 0V to 1,000V at any anode, and measure anode
current with Eg2 and Eg1 at the
wanted levels. But there's a lot of work to calibrate such Vac
and Iadc measuring. In 1960, there
were tube powered curve plotting machines for very many
different things, but all that expertise
has now been lost and forgotten. A few guys have made curve
plotting attempts for a few tubes,
but tetrodes and pentodes present some special challenges.