There are more ways than one to use tubes to make a fine amplifier.
I show a picture of one of the RCA monobloc amps brought to me to be radically
upgraded to make it reliable for another 60 years, and to give better music.
Fig 1.

This monobloc amp was known as RCA Orthophonic High Fidelity amplifier.
The general condition of the two amps was fair, with all transformers and chokes
in working order. The pair of amps were purchased years apart when stereo
listening became popular. There were some minor differences in transformer
case size and connection terminals but they both had the same schematic using
2 x EF86, 2 x KT66, and GZ32 rectifier. I have fitted an RCA input socket and
4mm semi recessed banana sockets for speaker cables which must have
banana plugs only. all other odd holes have been blanked off with sheet metal
only 1 of two octal sockets for preamp power retained, but nothing is connected
to it inside because the amps are to be used with a non RCA brand and stand
alone preamp. The amps were lethal with no bottom covers on chassis. I have
seen people get a severe shock when moving old tube amps while they are turned
on. Perforated metal bottom covers were fitted and with support feet. The paint
work on one amp was much affected with rust and both amps were entirely
repainted matt black.
Fig 2 below is a copy of the original 1960 schematic which I have cleaned up
slightly. Unfortunately, the original scan which I found elsewhere online was
typically poor quality, and don't have time or any motivation to re-draw the original
because I will never try to build the same schematic because my own schematics
work better. The original RCA schematic is in RDH4, and inspired Dynaco to use a
very similar schematic with a single 7199 / 6AN8 pentode-triode for all their amp
Fig 2.

This RCA amp has nearly all its voltage gain produced in V1, a single EF86. The R4
220k is bootstrapped via C3 from V2 cathode to much increase the effective V1
anode RLdc load. It is a local loop of positive feedback. Pentode gain varies
proportionately with anode RLa so V1 voltage gain is much larger than without the
V2 EF86 is triode connected to make a concertina phase inverter to produce
+/- 25.0Vrms at anode and cathode to power the grids of KT66. The equal Va and
Vk at V2 with opposite phases depend on equal RLa and RLk. The Rk = 40k, but is
in parallel with R4 40k to make 20k total Rk. The RLa dc = 20k, so although V2
looks like is has different RLa and RLk, they are virtually equal because of the

The dual EF86 schematic produces quite high THD because the V1 pentode must
produce slightly higher voltage than V2 Vk, about 28Vrms maximum. The circuit
gives less open loop bandwidth and more phase shift than my circuit with low
triodes such as 6CG7. But the EF86 was commonly used originally because
it must have been cheap to make, and it consumed a tiny amount of power and
produced huge voltage gain all of which very much suited company accountants
who needed to keep shareholders happy. 

But in 2016 someone with a pair of these amps still in original condition asked me
how to strap the OPT to suit 8r0 speakers. There isn't any info on strapping on the
original schematic. The original above was in early version of this RCA amp, probably
made around 1955. But the owner of these amps has bought another RCA amp in
about 1960 when stereo sound was possible from LP. The two amps looked identical,
but chassis height in the later amp was 5mm higher, and both had THREE sec
windings, instead of only 2 in the original schematic. It could probably be assumed
that both early and later amps had identical numbers of turns in OPTs, but at some
point, RCA decided to use 3 sec windings which then ensured better HF response
and lower winding losses when strapped for 8r0 speakers, and probably ensured

Fig 2A.
Fig 2A shows an abbreviated amp schematic to show the strappings for OPT sec
windings for early/later am versions.

It also shows how RCA gave owners the function of the "variable damping control".
VR1 can be twiddled for the best sound, and the output resistance of the amp could
be varied by applying Global Current Feedback between being "positive" for low Ro,
and "negative" for higher Ro. At some setting of VR1, the current FB was neither
positive or negative, and neutral. Another company, Bogen, used a similar damping
control, but hardly anyone else bothered because most rational engineers knew there
was only ONE correct design aim, to have Rout as low as possible while maintaining
IMHO, the RCA arrangement is utterly useless, and 99% of listeners would turn the
VR1 adjust pot with a screw driver and not hear any sound change.
So when I rebuilt these amps, I abolished the variable damping idea.

The above amps have been reformed with this much better schematic.... 

Fig 3.
Fig 2 schematic uses a cascaded pair of differential triode amps, aka Long Tail Pairs, LTP.
These produce less THD when compared with the RCA original schematic or many others
from 1950-60.

Many PP class AB1 tube amps generate between 3% and 6% THD at -1dB below full power
any local CFB in OP stage or any GNFB network connected. The THD is mainly 3H,
5H and other higher H in declining %. In most most of these amps the input and driver
stages produces about 1/3 of the THD, and it is mainly 2H, 3H, 4H etc. The percentage level
of all these harmonic products declines in proportion to voltage output. A typical UL PP tube
amp without any NFB and with KT66 may make 3% THD at 30Watts, about 1% at 3 Watts,
and 0.5% at 1Watt, when output voltage is 0.18 x 30Watt level. 20dB GNFB will reduce the
THD to 0.3%, 0.1%, 0.05% respectively - providing all the tubes have not become defective
with aging/use hours.

It is important to make the input and driver stage to be as linear as possible to avoid even
low levels of
THD or IMD at all levels of operation. Adopting this approach has led to
widespread appreciation of the sound quality of a number of amps I have re-engineered
such as Quad, Leak, Dynaco, VAC, Manley Labs Snapper, ARC and others.

Are there any advantages with the Fig 2 schematic?

Low twin triodes are used with Ia = 5mAdc in each triode instead of 1mAdc in each EF86.
The higher Ia and internal local NFB in triodes plus their
low Ra make the THD/IMD lower.
Open loop bandwidth is so wide it must be tailored by use of LF stability networks C10&R15,
C11&R16. HF networks needed are R14&C9, C16&R33. The two existing 9 pin sockets for
EF86 are used for two 6CG7 to make a pair of cascaded LTP (differential amps) produce
lower noise floor and much less THD / IMD compared to the EF86.

The other major benefit with 6CG7 is that they will easily drive the 120k grid bias R22+R23.
The original use of 560k meant as tubes age they get a substantial +Vdc above 0Vdc
at grids because of "reverse grid current at idle" which turns on tubes, making them run hotter
and that makes the effect worse. So the Rg needs to be low. The use of TWO separate cathode
bias R&C networks means the two KT88 will have equal Ek and Ik for much longer than in original
amps with single R&C where Ia can vary by 50%, and THD increases by 200% before anyone
notices, and the hotter tube dies young.

The signal current Iac of each triode in each LTP has opposite phase so the net Iac from all
4 triodes measured at
the their B+ rail connection at top of 50uF C7&C8 is very low.
Therefore little bypassing is needed.

The B+ rail for the two LTP at C7&C8 is fed by 1k0 from a very much better filtered power
supply CLC filter which uses the original RCA choke plus two 235uF B+caps ( made up
with series 470uF caps ). The large C replace the original B+ rail caps of 10uF + 10uF
indicated in Fig 2. The 2 amps which came to me had additional 16uF+32uF B+ caps which
looked like replacements, or an attempt to reduce hum, which probably didn't work,
because hum was caused by something else more difficult to identify. I have 235uF at the
B+ rail at OPT CT, so the impedance between OPT CT and 0V is very low at LF so
this gives blameless performance in class AB1 at high levels with a modern speaker
of low sensitivity and low speaker Z.
Noise at the amp output was less than 0.3mV.
The RCA OPTs have 20% taps for the screens. This means the 3H and 5H produced by
KT88 are
not much reduced below plain tetrode connection with screens taken to a fixed
B+ rail. But tetrode Ra is reduced from 35k to 6k0, still much better than pure tetrode, but
Rout of the amp remains far too high to allow use without the considerable amount of
GNFB shown. There are TWO global NFB loops, and one is conventional series voltage FB
from output active terminal to V1b grid, which is the second input port of the V1a and V1b LTP.

The original RCA OPT does not have extremely low leakage inductance so HF oscillations
with pure C loading will occur when any global voltage NFB is used. To avoid this typical
behaviour with lower than optimum OPT design, I have placed a second global current NFB
loop between the small choke L1 and bottom of R3 100r at the resistance divider used for
the global voltage NFB loop.The choke is approximately between 0.4 and 1.0uH, and is 5
turns of 1.2 dia wire with coil about 10mm dia inside dimension, and it can be seen in Fig 7
below at bottom LHS just inside the 0V speaker terminal.
The reactance of L1 is negligible at most AF so it has virtually no effect below 10kHz.
But at 20kHz, and if L = 0.5uH, XL = 0.05r, and 0.5r at 200kHz. the NFB voltage applied to
V1b grid via R3 100r has 90 degrees of phase advance ahead of the output load voltage.
This compensates for the phase lag due to leakage inductance of the OPT. The L1 effect
is most effective when load current is high. Use of 0.22uF across output terminals and with
use of 5kHz square wave will show that ringing is minimized and and oscillations are
impossible. The L1 choke acts in conjunction with C3 470pF across R7 680r to give the
extra phase compensation in the NFB network to maintain stability at HF.

With the NFB network as shown, THD at 2Watts = 0.03%, Rout = 0.3 ohms.
factor with 4 ohms is good at over 13.
If the OP tubes were set up in triode mode, the amp would give rather low Po because
triode Ea swing
is limited by grid current and low B+. Fixed bias might then be an
advantage, and the 20Watts with 4 ohms may sound better. But I have never found
anyone to tell the difference between two identical amps, except for the use of UL taps in
one and triode in the other.

Fig 3 does not include the power supply.
I have not yet prepared a schematic for this. However, because these two amps are the
only RCA amps which have been completely reformed by myself, then any schematic is
only of benefit to someone who services this pair of amps, and because I might be dead
in a few years time, I may not be the person doing service. So, anyone who does the
service will have to draw the circuit as he finds it to be, a laborious task, but in any case
he will have to familiarize himself with parts layout and this is what takes time.

If any service person cannot figure out the PSU, then I would consider him or her to be
incompetent. However, they may well need the information on the audio amp because
it is much more difficult to understand, and very different to the original schematic of 1960.
The tube rectifier has been deleted. 1N5408 Si diodes are used instead. 2 pairs of 470uF
caps 350V rated and in series are used with the existing choke to make a very effective
CLC filtered
B+ rail which is marginally higher than in the original amps which used KT66
or 6L6GC. This better suits the KT88 which I have used.

From the available heater windings I have a voltage quadrupler rectifier using 470uF/35V
 and Si diodes to produce a well filtered Vdc rail for V1 filaments and -45 rail for V2 LTP

There is a small 5VA transformer (bottom right Fig 6) to supply a +16Vdc rail for the active
protection circuit and Bias Balance Indication LEDS, (2 green LEDs on front Fig 4).
There is a relay (Fig 6 right side), driven by the protection circuit so that if too much Idc
flows in either KT88 then amp is automatically turned off at mains, the 2 green LED are
turned off, and a single red LED turns on to indicate a fault and owners can see the
problem from across the room. Where each green LED lights up with equal brightness
 the bias current in each KT88 will be correct, and nearly equal to each other.
If the bias currents differ by more than 10%, the LED brightness will become dissimilar,
warning owners of a tube problem. The cathode bias Rk regulate Ia fairly well, but when
tubes age the Ia can begin to vary, and sometimes an output tube can draw much too much
Ia and thus cause an imbalance in Ia for the tubes, and in the OPT which causes much
higher than normal distortion to occur.

There are no bias settings or other circuit adjustments that an owner needs to make.
A 2A slow blow mains fuse is used beside the mains cable IEC chassis socket.

There is one gold plated 
RCA input socket, and semi recessed 4mm banana plug sockets
for output. There are no screw down binding posts which always allow speaker cables to
work loose giving intermittent operation and high distortions which often goes un-noticed
for years.

So speaker cabling MUST have 4mm banana plugs. 

The owner of the two amps came around for an audition when I had them ready for collection.
He was well delighted by the sound, and complete lack of any noise.
Some pictures of my work...
Fig 4.

Fig 5.

Fig 6.

Fig 7.

As a retired amp worker I don't do this kind of restoration work for the ultra low prices
I used to accept.

A huge number of hours is needed to get this standard of operation to meet modern
expectations of sound quality and reliability.

If YOU spend 12 months training yourself to do this work and while learning all you can
from this website, you too might be able to get old junk to perform better than many
brand-name amps now made.

Happy soldering.

To Re-engineered amps

To Index Page