LOAD MATCHING 4A, PP BEAM TETRODES or PENTODES.

Last edited 2017.  This page 4A is followed by 4B and 4C to keep page size low.

Fig 1. Ra curves for a typical 6550 with no loadlines.
Basic description.
Table 1. PP Class A to AB loads for 6550 or KT88 or KT90, various idle Ea & Ia.
Pure class A RLa-a, class AB1 Hi-Fi RLa-a, Class AB Lo-Fi RLa-a, all with max Po given.
Table 2. PP Class A to AB loads for EL34 or 6CA7, various idle Ea & Ia.
Pure class A RLa-a, class AB1 Hi-Fi RLa-a, Class AB Lo-Fi RLa-a, all with max Po given.
Table 3. PP Class A to AB loads for KT66 or 6L6GC, various idle Ea & Ia.
Pure class A RLa-a, class AB1 Hi-Fi RLa-a, Class AB Lo-Fi RLa-a, all with max Po given.
Table 4. PP Class A to AB loads for EL84 or 6V6, various idle Ea & Ia.
Pure class A RLa-a, class AB1 Hi-Fi RLa-a, Class AB Lo-Fi RLa-a, all with max Po given.
Fig 2. Class A RLa loadline 5.67k for one 6550 in pair with RLa-a = 11k3.
Class AB1? or Class AB2? Notes about the two classes.
Fig 3. 2 x 6550 PP loadlines, 13k0 class A, 3k8 class AB1.
Steps 1 to 15 to plot loadlines in Fig 3.
Fig 4. PP 6550 with RLa-a 8k3 for good hi-fi.
Steps 16 to 22 to plot loadlines in Fig 4.
Fig 5. PP 6550 with RLa-a = 26k8.
Steps 23 to 27 to plot loadlines in Fig 5.
Zobel networks and critical damping.
Fig 6. PP 6550 with Ea +600V, Eg2 +300V, Class AB1 RLa-a 5k3, Class A RLa-a = 29k3.
Steps 28 to 39 for loadlines in Fig 6.
--------------------------------------------------------------------------------------------------------
Fig 1. Ra curves for 6550 beam tetrode with fixed Eg2 = +300Vdc.

Fig 1 shows 6550 Ra curves with Ea taken up to 1,000V and well beyond the old data sheet limit of
about 600V because nobody in the tube making or printing industries enjoyed the expenses of ink, paper
and labor. All data sheets and books like RDH4 should have all been produced on A4 pages. The curves
are digitally cleaned up old GE curves which I extended from 600V to 1,000V by continuing the nearly
horizontal Ra curves for -30V, -25V, -40V to right side with the same slope to allow plotting full RLa load
lines where Va swings up to 1,000V. For 6550 with Ea at +600Vdc, Ea swing could be higher than +1,000V.
The data curves for Russian made 6550 and KT88 are not much seen anywhere, and do not have the drawing
quality which was normal in 1960 where factories could afford the staff and equipment to draw good curves
because many more tubes were made in 1960s than during the last 20 years.

I have found the diode line resistance is slightly higher than UK or USA NOS tubes made before 1980.

Table 1. 6550, KT88, KT90, for class AB1.
The best Ea & Ia idle conditions for Hi-Fi should be chosen for best music and long tube life.
 Tube type Number Pda rating  1 tube tetrode W Pda factor for idle Pda W Best idle Pda W Ea at idle Vdc Iadc at idle, 1 tube, mAdc Class A only. RLa-a 1.9 x Ea / Ia Class A Max Po anodes W Class AB1 Hi-fi RLa-a 0.8 x Ea/Iadc Class A & AB1 Hi-fi max Po anodes W Class AB1 Lo-fi Min RLa-a  Ia pk 450mA Class A & AB1 Max Po anodes W Eg2 Idle +Vdc 6550, KT88 42W 0.6 25.2 350 72 9k3 24 3k9 10        41 2k2 4.9    56 300 6550, KT88 42W 0.6 25.2 400 63 12k0 24 5k1 10        47 2k6 5       69 330 6550, KT88 42W 0.6 25.2 450 56 15k3 24 6k4 10        49 3k2 5       76 330 6550, KT88 42W 0.6 25.2 500 51 18k6 24 7k8 10        52 3k6 5       82 330 6550, KT88 42W 0.55 23.1 550 42 24k9 22 10k5 9.3       49 4k0 3.5   101 350 6550, KT88 42W 0.5 21.0 600 35 35k6 20 13k7 8.4       46 4k5 4.4   111 375 KT90 50W 0.6 30.0 350 85 7k8 28 3k3 11.9     43 2k2 7.9    56 300 KT90 50W 0.6 30.0 400 75 10k1 28 4k2 11.8     59 2k6 7.3    69 330 KT90 50W 0.55 27.5 450 61 14k0 26 5k9 10.9     51 3k1 5.7    79 330 KT90 50W 0.55 27.5 500 55 17k3 26 7k3 11.0     53 3k6 5.4    88 330 KT90 50W 0.5 25.0 550 45 23k2 23 9k7 9.7       53 4k0 4.0  101 350 KT90 50W 0.5 25.0 600 42 27k1 23 11k4 9.7       55 4k5 4.0  111 375
This table has the same loading for 6550 and KT88, and I have chosen idle Ea max at +600V.
Ea could be higher with wider swings of +/-Vapk but my experience told me that Ea over +600V could
lead to arcing at the tube socket or at OPT insulation. All octal based tubes have practical limits and to
operate at higher Vac safely, the should have an anode top cap connection like the 807. There was a TT21
transmitting tube which was exactly like a KT88, but with anode top cap. Eg2 was only +300V at pin 4.
In class AB2 with direct coupled cathode followers driving the grids, a pair of TT21 could safely produce
140W of class AB2 power. But if you do want up to 140W, just use 4 x KT88 / 6550 in class AB1.
And you will get more initial class A Po.

The KT90 is similar to KT88 or 6550. KT90 class A Po is higher than KT88 because Idle Pda is higher,
and maximum class A Po is at anodes is about 46% of the idle Pda for all these types.

KT120 has rated Pda = 60W, so that Idle Pda could be 36W, so max class A Po from a pair could be 33W.
Thus we might say that KT120 would be the best tube for music but this is not automatically true; my customers
over 18 years heard quite enough fabulous music from a pair of rather unfashionable Sovtek KT88 or 6550
which were identical internally and which had a small "coke bottle" glass envelope and which were later promoted
as Tung-Sol 6550, in a marketing ploy. Nevertheless, people have told me that when they replaced their
6550 with KT90, and they may be replaced without re-biasing, they heard a sound change, but it was neither
good or bad, just another fine variety. KT120 can be plugged in to replace KT88 if you want to. KT120 cost
more than 6550 or KT88 but may be run in the same Ea & Ia conditions so they should last a long time because
their idle Pda is a smaller % of the maximum rated Pda so the operating temperature is lower which extends tube
life.

Table 2. EL34, 6CA7 for class AB1.
 Tube type Number Pda rating  1 tube tetrode pentode W Pda factor for idle Pda W Best idle Pda W Ea at idle Vdc Iadc at idle, 1 tube, mAdc Class A only. RLa-a 1.9 x Ea / Ia Class A Max Po anodes W Class AB1 Hi-fi RLa-a 0.8 x Ea/Iadc Class A & AB1 Hi-fi max Po anodes W Class AB1 Lo-fi Min RLa-a  Ia pk 350mA Class A & AB1 Max Po anodes W Eg2 Idle +Vdc EL34,6CA7 25W 0.7 17.5 300 58 9k8 16.5 4k1 6.9     31 2k7 4.5      40 300 EL34,6CA7 25W 0.7 17.5 350 50 13k3 16.6 5k6 7.0     34 3k3 4.1      50 350 EL34,6CA7 25W 0.7 17.5 400 44 17k3 16.7 7k3 7.0     37 3k8 3.6      59 375 EL34,6CA7 25W 0.6 15.0 450 33 25k9 14.1 10k9 5.9     33 4k4 2.4      67 400
The European designed EL34 pentode and American 6CA7 beam tetrode are considered identical
and deserves a separate table. They can nearly do the same job as 6550 or KT88. But they are more
fragile because rated Pda = 25W. They can swing quite high Ia pk current in class AB1. They do not like
to work in class AB2 where grids are driven positive which means grid input resistance becomes less than
a few thousand ohms rather than megohms where the grid is always kept negatively biased. I have allowed
the maximum class AB1 Po figures to be based on maximum Ia pk = 350mA. There have been some
foolhardy attempts my manufacturers to produce ever more Po from EL34.

The old Mullard data sheets suggest 100W from a pair is possible with Idle Ea = +800V, Eg2 = +400V,
Iadc = 25mA, and RLa-a = 11k0. I had one customer with a sample of this crazy idea where there were
8 x EL34 with Ea +900V, Eg2 +450V, and RLa-a was about 2k5, and total Po = 500W!
The amp owner was providing a PA system high service. He replaced EL34 quite often. The amp had only
one value of grid Eg1 bias applied to all 8 tubes and all tubes had different Ia. Mullard say idle Pda of 20W
is OK for such Ea & Ia conditions but the tubes can so easily overheat; I found the safe Pda at idle was 10W,
and I fitted 8 separate pots for bias adjustment and reduced Ea to +670V and Eg2 to a shunt regulated +400V,
and the 8 tubes easily gave 225W and was reliable, and music was just stunning.

Table 3. KT66, 6L6GC for class AB1.
 Tube type Number Pda rating  1 tube tetrode pentode W Pda factor for idle Pda W Best idle Pda W Ea at idle Vdc Iadc at idle, 1 tube, mAdc Class A only. RLa-a 1.9 x Ea / Ia Class A Max Po anodes W Class AB1 Hi-fi RLa-a 0.8 x Ea/Iadc Class A & AB1 Hi-fi max Po anodes W Class AB1 Lo-fi Min RLa-a  Ia pk 350mA* Class A & AB1 Max Po anodes W Eg2 Idle +Vdc KT66, 6L6GC 25 0.7 17.5 300 58 9k8 16.5 4k1 6.9     31 NA NA 300 KT66, 6L6GC 25 0.7 17.5 350 50 13k3 16.6 5k6 7.0     32 3k9 4.8      42 350 KT66, 6L6GC 25 0.7 17.5 400 44 17k3 16.7 7k3 7.0     31 4k7 4.5      45 375 KT66, 6L6GC 25 0.6 15.0 450 33 25k9 14.1 10k9 5.9     33 4k3 2.4      60 400
KT66 deserve a separate table because all the data suggests the peak Ia possible in class AB1 is
the lowest among large sized octal socket output tubes. KT66 has a fine reputation for sound, and is
* the asterisk indicates the maximum Ia for Class AB1 and with low Eg2 is not going to reach 350mA.
This is only possible if Eg2 > +375Vdc. I have estimated the peak Ia possible with the minimum RLa-a
listed while keeping the max Ia pk on the diode line. The available data for KT66 shows undefined diode
line value, and one may assume Rd = 220r. The KT66 comes more alive at higher Ea and higher RLa-a
with Eg2 > +375V.  The 6L6GC gives slightly more high Po with low RLa-a loads at lower Ea levels.

Table 4. EL84, 6V6 for class AB1.
 Tube type Number Pda rating  1 tube tetrode pentode W Pda factor for idle Pda W Best idle Pda W Ea at idle Vdc Iadc at idle, 1 tube, mAdc Class A only. RLa-a 1.9 x Ea / Ia Class A Max Po anodes W Class AB1 Hi-fi RLa-a 1.2 x Ea/Iadc Class A & AB1 Hi-fi max Po anodes W Class AB1 Lo-fi Min RLa-a  Ia pk 120mA Class A & AB1 Max Po anodes W Eg2 Idle +Vdc EL84,6V6 12 0.8 9.6 250 39 12k2 9.0 7k9 5.6     11 5k6 4.0      10 250 EL84,6V6 12 0.8 9.6 300 32 17k8 9.1 11k2 5.7     12 7k3 3.7      13 275
6V6 and EL84 give similar results with the same Ea & Ia idle conditions and loads. The best 6V6
made is now probably the JJ 6V6S, which can be set up with Ea & Ia up to about +400V and with Idle
Pda up to 12W, because their max rated Pda = 15W. The JJ is therefore a completely new tube if compared
to any NOS 6V6, or modern Russian copies. Unfortunately, the JJ 6V6S data sheets do not have the beam
tetrode curves!  They do have class AB2 TRIODE curves, which is entirely useless for those who want to
design an amp properly. Most ppl use JJ 6V6S in guitar amps where a pair can be expected to give 20W

EL84 or 6BQ5 made during last 20 years by Russians are about the same as original EL84 first made in 1957.
For hi-fi, UL operation is good with these small tubes.

Plotting load lines for PP amps may be done most simply using a single sheet of Ra data curves for one tube.
In RDH4, there are more complex explanations of using 2 data sheets for an output tube, with one sheet
inverted and the two sheets joined to show the combined PP action of both tubes. Nobody I know understands
the RDH4 method or tries to use is. But RDH4 also suggests the simplest method is by examining what is
in each of a pair of PP tubes using one data sheet with the Ra curves.

For class A PP amps, each output tube behaves almost identically to one SE tube in class A. However,
one tube of a PP pair is slightly loaded by the other because the transfer function of each is not linear.
Therefore the anode load for one tube in a PP pair is a slightly curved line. We may ignore this slight curvature
during design procedure to get a suitable load match. And we may draw the class A load line for a PP tube in
exactly the same as way as for a tube in an SE amp, as covered inloadmatch-2-se-beamtetrodes.html

Fig 2. Class A SE load RLa = 5.67k of one 6550 beam tetrode tube in a class A PP amp:-

The loadline RLa = 5.67k is for one tube of a pair in pure class A, and the RLa-a for the pair = 11k3 approx.

Notice that this set of 6550 Ra curves are for Eg2 = +250Vdc. The spacing of Ra curves look
fairly linear. But the Eg1 shown for each Ra curve increases unevenly :- 0, -3, -5, -7, -10, -12, -14,
-17, -20, -25, -30. The spacing between these numbers varies between 2V to 5V, and had the Eg1
been shown correctly with equal 3V steps between 0V and say -40V, then you would see the Ra
curves stacked closely at the bottom of curve sheet and much wider apart at top of curve sheet.
I believe that GE made the unforgivable mistake of trying to fool everyone into believing the 6550 was
much more linear than it ever could possible be. However, the knee of the Ra curve for Eg1 = 0V is
about right, and the Ia at knee = 350mA, and well above the maximum class A operation with class A
RLa load of 5,670r, shown by load line BAQC. The idle Q point =  +380V x 63mA, for a comfortable
idle Pda = 23.9W. I show the approximate calculated RLa =  0.9 x 380 / 0.063 = 5,428r, close to
the anode load BAQC = 5,670r. This single tube could be one of a pair of 6550 in a class A PP amp.
The nominal anode load for the pair is the primary load of the OPT which has a CT and has each end
taken to each anode. The nominal load between each anode = RLa-a = 2 x class A load for each
6550 = 2 x 5,670r = 11,340r. The OPT required for the 2 x 6550 would have to be 11k3 : 4r0, 8r0,
16r0.

The THD of the pair of 6550 without any NFB would be up to 4% at maximum class A Po = 22.5W
approx, and mainly 3H.

The performance will be vastly improved if the OPT is arranged to have 20% of primary winding for
local CFB. The load will not change at all, but the THD of output stage will reduce to 1% and the
effective Ra will reduce to less than triode connection.

IMHO, the pair of  6550 for pure class A PP with CFB would give sound equal to any other
exotic 22W amp using SE 845 or any other triodes, and it would be cheaper. But you won't easily
find a good OPT with 11k3 primary able to handle the 22.5W.
The Va-a = 504Vrms, and have primary resistance < 200r.

What is class AB1 and AB2?
In Class AB1 amps, the output tube grids are never forced to a positive Vg above the Ek, cathode Vdc.
The grid input resistance is megohms, but usually biased from fixed -Vdc via biasing resistance Rg, often
50k to 220k.
Therefore most output tube grids are driven by triode anodes of 12AU7 or 12AX7 using C&R coupling.

In Class AB2 amps, the the output tube grids are biased to a fixed -Vdc, and most of the operation has
Vg pk not exceeding the -Eg Vdc. But the output tube grids are driven by cathode follower triodes such as
12AU7 which have their cathodes directly coupled to output tube grids, with Rg resistance of 100k to say
-250Vdc rail. The cathode follower grids are biased to a slightly more negative -Eg, and they have high
Rin = megohms, and are driven by C&R coupling from the driver triode just the same as for AB1.
The cathode follower has low Rout of less than 1k0, so it will be able to drive output tube grids low input
resistance of between say 2k0 and 5k0 when the grid swings up about +20Vpk above the cathode Ek.
In other words, high level AB2 action requires power to be delivered to the output tube grids to get a
linear grid input Vac without being clipped in the case of normal class AB1 drive tube arrangement.

Class AB2 is never used in guitar amps or Hi-fi amps because it can easily lead to unreliability if output
tubes are over driven and the output tube grids overheat. But Ab2 was popular in 1930s with triode
output tubes, and also with early beam tetrodes with a fairly low Vdc rating for Eg2. When Eg2 ratings
rose from +300Vdc in a plain 6L6 to +450Vdc in 6L6GC, class AB1 worked just fine to swing Va pk
low with high Ia pk to get high class AB1 Po.

Class AB2 cathode followers to drive output tubes begin to suffer grid current overload when their grids
become more positive than their cathodes. This may be at Ia pk = 20mA, and the Vk swing is usually
enough to swing the output cathode some +20Vpk to give the range of AB2 operation. Below this limited
range, the cathode follower grids are driven by C&R coupling from driver triode anodes used the same way
as for class AB1.

In all class AB amps, the output Po has an initial amount of pure class A which is often less than 10% of the
maximum Po which produced by what is class B action where each tube works alone to produce the Po.
In the class A region, each tube shares the loading and each contribute nearly equally to power production
for all parts of a wave form. Above the region for class A, each tube takes turns to produce the power for the
positive or the negative wave peaks. If you consider a PP class AB amp with 6550 with nominal OPT RLa-a
load of 8k0, then you will get a few class A watts when the load is 4k0 for each tube. Between the limit for
class A action and the maximum possible class AB load, each tube is loaded by 2k0 which is the B RLa.

Fig 3. 2 x 6550 PP loadlines, 13k0 class A, 3k8 class AB1.

Fig 3 shows 2 x 6550 with idle conditions Ea = +420Vdc, Iadc = 60mA, for idle Pda = 25.2W
= 0.6 x maximum Pda rating of 42W, and this is suitable for class AB action.

For class A only, I recommend idle Pda should not exceed 0.7 x maximum Pda rating.

To analyze fully, we should draw loadlines for the the pure class A loading which we may accept as
the highest RLa-a that is likely to be used for only class A Po. And we need to draw loadlines for the
lowest class AB RLa-a for load for highest class AB Po. All class AB loads will be lower than the load
for class A only.

For Ea = +420Vdc, Idle Iadc = 60mA, Eg1 = -29Vdc, Eg2 = +300Vdc, and Idle Pda = 25.2W,
For only pure class A :-
(1) Plot idle point Q at 420V x 60mA. Plot point B on diode line at 120mA.
Read off the Ea at B = +18V. Max negative going swing for Ea = 420V - 18V = -402Vpk.
The RLa for each 6550 = Ea pk swing / Ia change = 402Vpk / 60mApk = 6,700r.

(2) Max positive going Ea swing = Idle Ea + Ea swing in (1) = +420V + 402Vpk = +822V.
This voltage is where Ia has reduced to zero mA.
Plot point D on Ia = 0.0mA axis at peak positive Ea swing at +822V x 0.0mA.

(3) Draw straight line from D to B and on further to Ea = 0V axis and plot point A.
This is the load line for pure class A in each 6550.
The RLa value = Va pk-pk / Ia pk-pk = 804Vpk-pk / 120mApk-pk = 6,700r. The line between
A and D should pass through Q exactly. If not, you have made a mistake, and you must start again.
The point C may be at maximum Ea positive swing without quite enough NFB, and may be ignored for now.
The class A load line ABQCD shows that Ia swings equally +/- 60mApk, and with a perfect 6550,
the +/-Va pk swing will also be +/-402Vpk.

However, note that Q is at Eg1 bias - 29Vdc. For maximum Ia for class A RLa 6k7, the +Vg swing is
from = -29V to -15V, ie, +14V, and -Ea swing = -402V. We can estimate the +Ea swing by seeing
where -Eg1 swing reaches -43Vpk. The point C is at -40V, and point D is at maybe -43V.
Therefore there seems to be little 2H generated.
But if you see where Eg1 varies +4Vpk to -25V, and -4Vpk to -33V, the Ea swings are approx
-170Vpk and +130Vpk, so it looks like there is considerable 2H at the lower Va level.

There will in fact be many harmonics in the Va from the 6550 with no NFB. Trying to use the curves to
estimate and predict all harmonic products without NFB and with NFB is quite a useless waste of your
time. You will only discover that a typical single class A 6550 will produce a range of toxic H products,
and when you measure a pair, any 2H will be found to have disappeared but there will be considerable 3H,
with increasing 5H, 7H, 9H at the level reaches clipping. Using 20dB GNFB will be found to reduce the
THD to less than 1% just under clipping.

(4) Draw Point F on the knee of the Ra curve for Eg1 = 0V.

(5) Drop a vertical line from Q to point H on Ia = 0.0mA axis, ie, at 420V x 0.0mA.

(6) Draw straight line from H to F and through to E on Ea = 0.0V axis, ie, at 0V x 450mA.

(7) Calculate the resistance value for line EH = 420V / 450mA = 933r. This is the class B RLa load
each tube has when the other tube has moved into cut off mode in class AB.

(8) Calculate the Nominal Class AB RLa-a = 4 x B RLa = 4 x 933r = 3,733r.
Calculate the class A load each tube sees for the initial class A operation prior to AB operation
= RLa-a / 2 = 1,866r.

(9) We wish to plot the load line for 1,866r for each 6550.
Calculate the theoretical Ia position of point J on Ea = 0V axis. Ia at J = ( Ea / class A RLa ) + Ia at Q
= ( 420V / 1,866r ) + 60mA = 285mA. Plot Point J at 285mA x 0V.

(10) Draw straight line from J through Q and on to point I on Ia = 0V axis.

(11) Draw point G where the line JQI intersects line EFH.

(12) The line GQI is the class A load for each 6550 in PP amp. The distances GQ and QI should be equal.
If not, you have made a mistake, so repeat your efforts without mistakes.

(13) Interpret results for class AB1 action.

The Va swing for maximum class AB Po is horizontal distance from Ea to point F = 420V - 50V = 370Vpk.
+/- 370Vpk occurs at each 6550 anode. The Va-a = 2 x 370Vpk x 0.7071 = 523.5Vrms, for a sine wave.
Max class AB Po = Va-a squared / RLa-a = 523.5V x 523.5V / 3,733r = 73.34W.

Initial pure class A Po = 0.5 x Iadc for one tube squared x RLa-a = 0.5 x 0.06 x 0.06 x 3,733r = 6.72W.

(14) What is the ideal load for the two 6550?

Calculations and loadlines have given 2 loads, 13k9 for 25W of pure class A, and 3k7 for 73W of class AB.

IMHO, these are two extremes we do not need to use, and an RLa-a "midway" between these two loads
would be better because we do not need the whole 73W of class AB Po, 40W would be plenty.
And we would not mind if the initial class A Po was say 15W instead of the miserly 6.7W with RLa-a 3k7.

(15) Use simple arithmetic to determine the ideal load to match a pair of 6550.

We can be sure that any RLa-a between the 2 values found so far may be calculated from the data we have
for one of loads, 3k7, or 13k9.

We have found that we can get 6.72W class A with RLa-a 3,733r, and 73W of maximum class AB.

The 6.72W = 0.5 x Iadc squared x 3,733r = 0.0018 x 3,733r. For this example pair of 6550, Idle Iadc is the
same for all load values, and the RLa-a load for pure class A may be easily calculated for any amount of Po
up to to the load for only pure class A = 13k9.

So what RLa-a will give 15W? RLa-a wanted = wanted Po / ( 0.5 x Iadc squared = 15W / 0.0018 =  8,333r.
Thus the Class A RLa for each tube = 4,166r, and the class B RLa = 2.083r.

Let us see what that looks like as load lines......

Fig 4. PP 6550 with RLa-a 8k3 for good hi-fi.

(16) We begin with RLa-a = 8,333r. Calculate RLa for each 6550 = RLa-a / 2 = 4,166r.
Calculate Ia at K = ( 420V / 4,166r ) + 60mA = 160.8mA.
Plot point K at 160.8mA x 0V.

(17) Calculate Ea at point L = 4,166r x 160.8mA = 670V. Plot L at 670V x 0.0mA.

(18) Draw straight line K to L, and this should pass through Q exactly.

(19) Calculate the class B RLa = RLa-a / 4 = 8,333r / 4 = 2,083r.
Calculate to plot point M. Ia at M = Ea / B RLa = 420V / 2,083r = 201.6mA. Plot M at 201.6mA x 0V.

(20) Draw straight M to H, and plot point N where MH intersects diode line. NH is the B RLa loadline.

(21) Plot point O where NH intersects line KQL. OQL is the class A RLa for each 6550, and OQ should
be equal to QL.

(22) Class A Po total = 15W, and max class AB Po = 38W.

THD will be not much above the THD for 13k9 for only pure class A. For most ppl, 38W Po max is "enough".

Fig 4 shows a good ratio balance of class A Po : class AB Po, 15W : 38W. The class AB load for hi-fi is not
critical and can vary +/- 25% without worry. Ideally, the class AB OPT should have nominal load ratio
= 8k3 : 4r0, 8r0, 16r0.  Speaker loads always vary at least +/- 25% above / below the nominal value,
and a "8r0" speaker can be expected to vary between 6r0 and 10r0 for the range of frequencies where most
power is needed, usually 80Hz to 800Hz.

Many ppl are tempted to use OPT for 6k6 : 4r0, 8r0, 16r0, and you would get 12W of pure class A for 6k6,
but if 8r0 speaker has dip to 6r0, then RLa-a becomes 5k0, and class A Po max = 9W.

But if the 8r0 speaker is moved to 4r0 outlet, its 6r0 low value makes RLa-a = 10k0, and then you get 18W
of class A, and speaker load of 8r0 makes RLa-a = 13k2, and gives 22W class A. If speaker load is 10r at
4r0 outlet, then RLa-a = 16k5, and it is impossible to get the full class A Ia swing of +/- 60mApk for the class

The Po for RLa-a values higher than for maximum possible RLa-a = 13k4 may be found graphically by
plotting the loadline. Let us consider RLa-a = twice RLa-a for maximum possible class A1 Po.
The higher RLa-a = 26k8, and therefore RLa for each 6550 = 13k4, and if you plotted this loadline we
would get.....

Fig 5. PP 6550 with RLa-a = 26k8.

(23) Start with RLa-a = 26,800r. RLa for each 6550 = 13,400r.
Q is at +420Vdc x 60mAdc.

(24) Calculate position for point S = ( Idle Ea / RLa ) + Idle Idc = ( 420V / 13,400r ) + 60mA
= 31.3mA + 60mA = 91.3mA.
Plot point S at  0.0V x 91.3mA.

(25) For ALL RLa-a higher than the RLa-a 13k4 for maximum possible class A, the +/-Ia change
will never reach 60mA.
Draw straight line between S and Q.
Plot point R where SQ intersects diode line. Point R is at + 13V x 90.3mA. R to Q represents
the negative going Vapk swing = -407Vpk.

(26) The positive going Vapk swing = +407Vpk. Calculate max Ea = Idle Ea + positive going Vapk
= 420V + 407V = +827V.
Calculate the +/- Ia swing for the Vapk swing = 407V / 13,400r = +/-30.3mA.
Calculate the Ia at Ea = 827V = Idle Ia - Ia swing = 60mA - 30.3mA = 29.7mA.
Plot point T at 827V x 29.7mA.

(27) Draw straight line line from S to T and continue to above Ea = +950V.
This line should overlay the line S to Q, and pass thru Q. If not, you have made a mistake,

The straight line SRQTU is the load line for 13k4. The Ea can swing negatively to point R at +13 x 91mA.
The difference between Ia point S and T = 0.97mA, and difficult to plot accurately. Va swings positively
to point T at Ea +427V. The Ia at point T is 29.7mA. The +/- Ia pk swings = +/- 30.3mApk - providing
the NFB keeps THD very low.
Each 6550 generates +/-407Vpk for its RLa load of 13k4, so it makes 6.18W.
The theoretical Po max with RLa-a = 26k8 is 12.36W, about half the Po for maximum possible class
A for RLa-a = 13k4.

Where RLa-a is above the RLa-a for maximum class A Po, there cannot be any class AB. However,
in the real world, speaker load values change widely, and may be be much less than the nominal ohm value
at one or more frequencies so that it may be impossible to make the amp always work in pure class A at
all frequencies. Most speakers of hundreds I measured had a minimum load value at a narrow range of F
and not less than 1/2 their nominal RL value. The class AB capability usually copes well with the lower
speaker RL without causing perceptible distortion.

Let us assume the OPT for this pair of 6550 is a nominal 8k3 : 4r0, 8r0, 16r0.

If you had a 16r0 speaker and it measured 16r at all frequencies, and you connected it to the 4r0 outlet,
then the RLa-a = 33k2.
Maximum possible class A is with RLa-a 13k4, 24.2W, so with 33k2,
expect Po max = 24.2W x ( 13.4 / 33.2 ) = 9.8W.

This Po is very much less than the maximum possible class AB Po of 38W with RLa-a = 8k3.
I found many people were quite happy with being able to get only 10W from their amps, especially if the
speakers had sensitivity > 95dB/W/M. Using a 16r0 0r 8r0 speaker connected to 4r0 amp terminal is OK
providing the amp is unconditionally stable, ie, it will NOT oscillate at any frequency between 0Hz and
1MHz without any load connected, or with any high value load, or with any kind of reactive ( L or C ) load.

Pure beam tetrode or pentode THD varies with load, and a total of at least 20dB of NFB MUST be used
to tame the generally high THD and very high amp output resistance without NFB.
I found cathode feedback windings on OPT plus some GNFB gave the best results which pleased all my
customers. See the 8585 amp

Notice the loadline SRQTU extends to over +950V. Without NFB, and a high RLa-a, it will be found that
the positive Va swings will go much more positive than the negative Va swings and THD becomes very high.
With sufficient NFB, as the load increases in R value, the gain of 6550 approaches its µ, so gain A in the
equation for amount of NFB applied increases, and thus THD is kept low even though gain without NFB
becomes more non linear with rising RLa-a value.

Without any load at all, gain A can be close to µ, perhaps 200, and very often this means the amount of
NFB may exceed 40dB which will cause oscillations at LF and HF unless gain shelving networks are used
to reduce gain and phase shift below 20Hz and above 30kHz, and usually also with a Zobel network across
each half OPT primary, and perhaps across the secondary to prevent HF oscillations. The C+R Zobel
networks across each 1/2 primary make the load of each anode = RLa above say 50kHz where many
speaker loads have become a high reactance because they are inductance loads, and often do not have a
C+R series network to make their load like a resistance at above 30kHz.

In this example where nominal hi-fi RLa-a = 8k3, the Zobel across each 1/2 primary could be 1nF 630V
rated + 4k7 10W. At 34kHz, the impedance of the series R+C has reduced to 6k6, and at 100kHz, is 5k0,
thus limiting output tube gain and phase shift at F where oscillation is likely. If phase shift exceeds 180 degrees,
and gain of amp exceeds 1.0, the amp will oscillate. This can often be seen to be the case if you connect 0.22uF
across outputs on many old amps - they then begin oscillating at F between 40kHz and 140Hz.It is because the
NFB network feeds back a fraction of output and the phase shift changes the FB to be positive voltage FB.

I have always used "critical damping R+C networks" in all the power amps I repaired, reformed, or
built from scratch. All my schematics show what was needed to get my amps to always be unconditionally
stable.

A pair of 6550 can easily generate 100W+ AB1 with Ea = +600V, Eg2 = +300V, and RLa-a = 5k3.
But with very high RLa-a, say 40k, each tube has 20k and the possible class A swing will be up to 1,200V.
So even my extended graph may not accommodate all the possible Va load swings where Ea is 600V.

Fig 6. PP 6550 with Ea +600V, Eg2 +300V, Class AB1 RLa-a 5k3, Class A RLa-a = 29k3.

Fig 6 shows loadlines for one 6550 in a PP pair arranged for high Po with Idle Ea = +600V,
Eg2 = 300V. Keeping Eg2 low as +300V means that if the RLa-a was lower than shown by line WVYU,
the Ea could not swing above the Ra curve for Eg1 = 0V, because grid current will limit the grid drive Vac
unless the grids are driven directly by a cathode follower to achieve class AB2 operation. The Idle Pda is 24W,
or 0.57 x max Pda rating of 42W. Pda could 0.43 x max Pda, ie 18W, if the Iadc was lowered to 30mA without
much increasing the the THD.

To plot the loadlines for high Po :-
(28) Plot the load RLa for pure class A.
Plot point B on diode line at Ia = 2 x Idle Idc = 80mA. B is at +14V x 80mA. The straight line between
B and Q is the class A Va pk negative swing = +600V - 14V = 586Vpk.
Calculate RLa = Ea pk swing / Ia pk swing = 586V / 40mA = 14,650r.

Note. The positive Va pk swing will also be 586Vpk ( theoretical ). But this means max Ea would be
at Idle Ea + Va pk =  1,186V, and this will not fit on the graph sheet.

(29) Calculate Ia swing for the available space on curve sheet. Max possible Ea = +1,000V.
Calculate Ia pk swing for 400V Va pk swing = 400V / 14,650r = -27.3mA. The Ia at Ea 1,000V
= 40mA - 27.3mA = 12.7mA.
Plot point C at 1,000V x 12.7mA.

(30) Draw straight line from C through Q and B and to A. This line should pass through Q and if not,
repeat calculations to eliminate your mistake. The line ABQC is the class A load line for maximum class
A Po for one 6550.

(31) Calculate RLa-a for pure class A only = 2 x RLa = 2 x 14,650r = 29,300r. Calculate maximum
class A Po = 0.5 x Iadc squared x RLa-a = 0.0008 x 29,300 = 23.4W.
( 11.7W of pure class A is produced by each 6550. )

(32) Plot loadlines for maximum class AB1 Po.
Plot point V on the knee of the Ra curve for Eg1 = 0V, at +52V x 408mA.

(33) Plot point U vertically below Q at +600V x 0.0mA.
Draw straight line from U to through V and and to W, at 0.0V x 450mA.
The line WVU is the lowest class B RLa you should try to use. If the B RLa is lower, the operation at
high Po continuously may easily cause tube overheating.

(34) Calculate minimum class B RLa = Idle Ea / Ia max on axis 0.0V, = 600V / 450mA = 1,333r.
Calculate the resulting class A RLa = 1,333r x 2 = 2,666r.
Calculate the resulting class AB RLa-a = 1,333r x 4 = 5,333r.

(35) Calculate Ia at point X on axis and for Ea = 0.0V = ( Idle Ea / class A RLa ) + Idle Iadc
= ( 600V / 2,666r ) + 40mA = 225mA + 40mA = 265mA. Plot X at 0.0V x 225mA.

(36) Calculate Ea position of point Z where Ia = 0.0mA. Idle Ea to Z = class A RLa x Idle Iadc
= 2,666r x 40mA = 106.6V.
The point Z will be at Idle Ea + 106.6V = +706.7V. Plot point Z.

(37) Draw straight line from X through Q and to Z. The line must pass through Q. Revise all calculations
if this does not occur.

(38) Plot point Y where line XQZ intersects line WVU. The distances YQ and QZ represent the -/+ Ea
swings for pure class A for each 6550.
The line X to Y is included to accurately draw the class A RLa but it does not represent any Ea change for
the RLa 2,666r.

(39) Calculate initial class A Po for both 6550 = 0.5 x Idle Iadc squared x RLa-a = 0.0008 x 5,333r
= 4.27W.

From the above, we see that RLa-a loads between 29k4 and 5k3 are all tolerated. The RLa-a to give
best hi-fi which gives good balance of maximum class A and AB Po may be calculated as average of the
5k33 and 29k4 = 34k7 / 2 = 17k4.
The class A Po will be 13.9W, and if you plotted the class AB load lines for RLa-a = 17k4, you would
find AB Po = 36W.

Some musician amps have been made to make 100W from a pair of 6550. Ampeg made an amp with
6 x 6550 which could produce 300W.

For all class AB amps, you will find it almost impossible to find any off-the-shelf OPT with enough
variable secondary windings or taps to get low loss wide bandwidth for low Po pure class A and high
Po class AB using any speaker load between 4r0 and 16r.

For example, for only pure class A Po from PP 6550 with Ea at +600V, the RLa-a = 29k3, and if the
speaker = 4r0, then load ratio ZR = 7,325 : 1, and the turn ratio TR = 85.6 : 1.
For max class AB Po, ZR = 5,333 : 4r0 = 1,333, and TR = 36.5 : 1, which is not so unusual.