The input tube is a 6CG7 twin triode with both halves in
parallel.
The parallel connection gives a single triode with Ra = 5k, and µ
= 20.
Its DC supply is via the MJE350 wired as a constant current source with
a very high finite resistance of many megohms. Transistors wired like
this
have no negative effect on the sound, and
increases the fidelity because
the constant current source acts as a passive high impedance current
supply.
The triodes do all the
actual work on the signal amplitude.
The input tube distortion of this stage is reduced about 10dB compared
to
using a resistance to deliver dc to the tube because any such
resistance
becomes part of the ac signal load, and the
higher the value of anode load ohms,
the more linear a triode becomes.
The 6CG7 is electronically the same
as a 6SN7, and the data curves for either
tube are the same.
The input triode has its own shunt regulated anode supply voltage.
Its DC heater supply is derived from rectifying a Vac source in the
amp.
This input tube can provide wide enough bandwidth to power the long
tail pair
using a pair of EL84 in triode mode.
Alternative input tubes could be used for an input tube, but
the heater supply
( sheet 8 ) may have to be changed to suit because twin triodes such as
12AT7, 12AU7, have different pin layouts to 6CG7 and 6DJ8. 6BX6 or 6AU6
sharp cut off pentodes may be strapped as triodes and used for V1. 6BX6
has gain
up to about 45. But heating currents will change, so there may need to
be some
circuit changes.
Just about all tubes other than 6CG7 could be used with the CCS anode
supply as
shown but the cathode R4 would have to be changed so that the correct
anode
+Vdc voltage appears.
Where anyone might use an input tube with different gain, the global
NFB
network R21, R22, C8 may have to be changed to maintain the same
amount of applied global NFB. Where there is a zobel network of R&C
from
Point C to 0V, using a different input tube may require zobel values to
change.
I doubt there is a better input tube than the 6CG7 or 6SN7
if people prefer octal
tubes.
The Sieman's NOS 6CG7 are *very* good sounding, if you can ever buy any
without having to sell your house to raise the funds. But I find my NOS
stock
of Australian made 6CG7 to be quite excellent.
The long tail pair, ( LTP ) or otherwise known as a
differential amplifier has two
EL84 wired as triodes because these small power pentode output tubes
can equal
the
performance of five paralleled halves of a 6CG7 and give slightly
better linearity.
This LTP stage acts to convert the
single ended unbalanced output of V1 into
two exactly equal output voltages of opposite phase to drive the output
circuit.
Such an LTP stage can produce well over 100Vrms between each EL84 anode
and 0V, or over 200Vrms anode to anode
in two phases.
In my amp each EL84 needs to produce 75Vrms at its anode to drive each
output
tube grid at clipping, ie, 150Vrms anode to anode.
A maximum of 8Vrms is needed between each EL84 grid to produce maximum
required output from EL84. In this case, 8Vrms is generated by V1 6CG7
and is
applied to one EL84
grid while the other EL84 grid is grounded.
Both EL84 input grids are biased from the voltage established at C5.
Note that C7 and R11 form a low frequency step response in the open
loop
frequency response character. This helps stabilize the amp at bass
frequencies and
improves recovery
on overload, reduces saturation effects, and improves the
margin of stability at LF. Bass sounds natural, tight
and well controlled.
There is no HF step response network shown because it was found that
this
amp did not need such a network. There is flawless unconditional
stability at HF,
but some zobel
networks in the output stage did reduce the inevitable
slight square wave overshoot into purely capacitive loads.
The V2/V3 LTP uses a choke of 500 Henrys with a CT in the anode
circuit as
well as the resistor loads of 6k9 to each anode circuit, R17&R18,
and R19&R20.
This technique
with a choke is a unique feature I try to use in all amps where a
choke is used to increase the AC impedance of the element through which
Idc
is supplied to the triode anodes. The combination of choke plus
resistance has a
minimum combined low impedance
equal to the 6k9 plus low choke wire
resistance at very low frequencies. So a useful amount of dc current
may
be
provided to each EL84 anode without needing a very high B+ supply
voltage.
The effect of the 6k9 also acts to allow useful gain at very
low F below 10Hz
where there is a very low audio signal content in
music, while also avoiding the
phase shift caused by L shunting Ra, so the L&R act as a gain
shelving network
and stability with NFB is OK. At F above 10Hz the 500H choke reactance
increases at
6dB/octave and becomes a very high reactance load.
500H has Z = 31.5k at 10Hz, 315k at 100Hz, and much more for most of
the
audio band until its self capacitance begins to reduce its reactance.
The resistance in series with the choke prevent the
chokes's capacitive
reactance causing phase shift at the anode because the 6k9 is several
times
the Ra of each EL84 in triode.
Therefore the choke plus the series R provide a very high ohm loading
at
each EL84 anode and therefore the voltage swing can be maximised
without
the distortion
which would occur if the choke was not used.
Therefore the resistance load value for each EL84 is mainly the
capacitance
coupled bias resistors used for each output tube.
In my amp I have 6 x 120k ohm grid bias resistors on each side of the
the PP
output stage.
All the bias supply ends of the 120k resistors are bootstrapped to the
cathode
feedback winding through Cc, Sheet 2 below.
Thus the anode load experienced see by each EL84 of the the LTP is
approximately
40kohms,
low enough to ensure good ac balance but high enough to ensure THD of
the
driver LTP stage is about -10 dB lower than if the dc was brought to
the
EL84 anodes via
resistors which would have to be approximately 15k, which are
is really too low to
get low THD.
The Ra of each EL84 as used = 2.2k approx, so the load of 40k = 18 x Ra
which
ensures very low THD.
Because the Ra of each EL84 with 15mAdc is only 2.2 kOhms, this LTP
drive
amp gives a wide bandwidth when driving 6 output tubes. The output tube
Miller capacitance is very low because a fixed bias
voltage is applied to all
the output tube screens.
A transistor constant current source is used to feed the
commoned cathodes
of the EL84. This type MJE340 transistor acts as a purely passive
manner,
and has no discernible effect on the sound,
and acts to maintain the balance
of LTP output voltage phases within 1%.
Balance with a constant current sink such as the MJE340 used here
remains
excelent even if the EL84 were very poorly matched.
The CCS ensures the signal tube current through anode loads and tubes
is
the same for each of the tubes and loads on each side of the LTP and
the balance accuracy is
determined by the closeness in load ohm values.
A pentode tube could have been used for CCS instead, but
there would
have been no benefits.
In other versions of this LTP with EL84 I have tried using R&C
cathode
biasing networks for each EL84 and found this
gives closer balance of dc current to each EL84 without any detriment
to
signal operation.
Unbypassed Rk may cause the Ra to become much higher, lower the
stage
gain, demand twice the drive voltage from V1, and reduce the
bandwidth thus
reducing the sonic dynamics. In my circuit here with
cathode feedback in the output stage, the
output tube grid signals are
twice the levels of using the plain ultralinear configured amps.
I have found the idle current balance in EL84 remains constant
over many
years since the EL84 are set up with a very low amount of Ia compared
to
when used in an output stage where they usually run at Ia =
40mA, but here
they have Ia = 15mA, so the EL84 should have a long life without
problems.
The THD from such an LTP is mainly 3H and below 1% even at 100 Vrms
from each anode.
In this circuit a maximum of 75Vrms is required and 3H is typically
less than
0.4%. Some 2H is generated if the two EL84 are not matched, but it is
usually
less than the 3H generated. To minimize 2H each
EL84 may be reverse
positioned to get the least 2H by means of phase cancellation between
V1
and themselves.
At normal levels and because of the global NFB the overall THD
is
quite negligible because THD is about proportional to output voltage.
A previous version of this driver had zener voltage regulation
for the bias voltages
for the CCS base and the grid voltages, but that was found to be
unnecessary
because these voltages and that of the V1 triode is supply from a zener
regulated
supply of +320V, see Sheet 4.
There some obviously acceptable other tubes that could be used
in this circuit.
EL86 are pentodes which will work with their B+ supply to the choke
at 10%
lower than shown because
their Ra is only 1.4k in triode and you get a slightly wider Vswing at
the anode.
But EL86 triode gain is only 1/2 that of EL84 since µ = 11, so
when producing
75Vrms from each anode, about 17Vrms is needed to drive the live input
grid
of the LTP, and
EL86 are not manufactured any longer. The 6V6 could also
be used and the gain is very similar to EL86, but
the Ra of the 6V6 triode is
twice that of EL84, but nevertheless sufficiently low to provide
excellent
driving power to a multi-tube output stage. EL34 would be far better
than 6V6.
The use of octal socket tubes would probably look better than
what I have used;
6V6, EL34, 6SN7 are plentiful and many NOS varieties exist, and I would
suggest all
give excellent music.
Since 2006, instead of V1 operating as a single
ended stage as shown,
I have tried using a pair of triodes in an LTP with
cathode CCS.
This allows for balanced drive to the same following LTP stage using
EL84.
The CCS used in the EL84 cathode circuit may be replaced with a fixed
resistance,
say 5k6 to a negative supply rail, say -150Vdc.
The signal input is fed into one grid of the input LTP and
global NFB
fed into
the other grid of the V1 LTP.
I have tried this on several re-engineered amplifiers, ARC VT100, and
Dynaco
MkVI, and was rewarded with astounding music and excellent technical
measurements.
The other possibility for my drive stage involves the use of
boostrapping the anode
load resistors of the EL84 anodes to suitable taps on the main OPT.
McIntosh use this technique in MC275 and
other amps.
Theorectically, the bootstrapping is a mild application of positive
feedback and
thus the effectiveness of applied global NFB used is lessened.
My idea of the choke avoids the loss of effectivness of NFB, so with my
circuit
less NFB is needed to achieve the
same measured outcome with bootstrapping. Bootstrapping is a much
favoured
design feature among company accountants, but I don't employ any
accountants,
and I employ a choke instead !
However, I am not all against bootstrapping where it results
with a very small
amount of PFB, as is the case with bootstrapped biasing resistances
seen in the
Sheet 2 output stage below....
SHEET 2, OUTPUT
STAGE FOR 300W AMPS.
The output stage looks complex, but it is mostly just
repetition of a basic idea.
The R&C part identification seems strange but all resistors and
capacitors
in similar functions for each tube are just labled with the same number
for R,
and same letter for C; I am
sure any tech will get used to the idea.
The two balanced outputs from the LTP driver amp are fed to two
rails
from which there are 6 coupling caps of 0.47 uF, ( Ca ), each to couple
each output tube. R1 2k2 grid stoppers are used on each tube to prevent
RF oscillations. Each output grid is biased with R2 120k, and all 6 on
each side of the
PP circuit are taken to a -14V fixed bias supply via R6 4k7.
The -14Vdc supply is shown on
Sheet 8.
Each output tube has 15Watt cathode bias resistors comprising three
parallel 1k5 x 5W to make R3 500 ohms. The cathode bias voltage will
be about +18V, so total fixed plus
cathode bias equals a grid bias = -32Vdc.
This method gives good self regulation of the bias and saves having so
much heat dissipation in cathode resistances.
The anode supply is about +500V which may vary +/-5% if mains
input
voltages vary.
A fixed voltage is supplied to the screens via separate R4 330 ohm
screen
stoppers.
At no time did I find that the tubes wanted to oscillate at some
frequency well
above the audio band. HF Oscillation is impossible while output tubes
remain
at their normal
operating temperature. In the case of bias failure one or more
tubes may rise in temperature
and break into HF oscillations. But ANY bias
failure will always trigger my active protection circuits
which turn off the
power supply to prevent tube oscillations or damage to tubes or other
parts.
The 6 cathodes on each side of the PP circuit with their RC
bias networks
500r + 1,000uF are taken to the ends of the cathode feedback
windings
but via 1.67 ohm current sensors,
R5. Voltages across T-U and V-W are
used to work dynamic bias stabilizing
circuits shown on Sheet 5.
Capacitors of 470uF ( Cc ) are used to bootstrap the applied fixed bias
of
-14V so that the 6 x 120k bias resistances have an effective loading
value
of between 40k and 60k on the LTP
driver anode circuits. This mild form
of bootstrapping helps the driver amp generate minimum
distortion.
Resistors of R6 4k7 form negligible loading on the CFB winding, but
allows
the bootstrapped -14V bias voltage to be supplied to the 6 x
120k bias R.
The Output Transformer has 5 sections of primary windings
interleaved
between 6 sections of secondary windings and provides bandwidth at 250
Watts at 15Hz to
270kHz with resistance loading.
The OPT has a 110mm stack of GOSS E&I laminations. Not all
the possible
HF bandwidth is used when global NFB is added, and bandwidth is reduced
with GNFB to 65kHz for stability reasons. The 12 secondary windings can
be arranged to give waste free and
uniform current densities in all secondary
windings to match from 1,200 ohms anode to anode to
either 2.5 ohms or
5.6 ohms.
It is a somewhat complicated task for a non technical person to change
output
transformer matching, so the default setting is the 5.6 ohm load match.
The tube load value was chosen so that it is equivalent to having 7,200
ohms
per pair of output tubes. This load could be as low as 3,600 ohms
without
causing undue increases
in distortions or loss of power so when the amp is
set for 2.5 ohms load match, 1.4 ohms could be
driven. When set for 5.6
ohms the load could be as low as 2.8 ohms without a
problem.
Some speakers have dissappointing impedance curves with dips in Z well
below their nominal claimed Z but this amp design will handle them all
with ease.
Maximum power output of approximately 350 watts would be when the load
is about 3 ohms using the 5.6 ohm load match. When 8 ohms is used the
amp gives 200 watts of output and with a very
high % of class A power.
The dominant idea with this output stage is to produce a large amount
of
class A power but with a good reserve of class AB power for loud
transients,
and for where speaker Z is low.
The function of the 300 watt output stage is no less than having six of
my
50 watt amps all wired in parallel, or like having 15 x Quad II amps or
19 Williamson amps in parallel.
But should the voltage at one or more cathodes ever rise to
levels resulting
in a possible early tube failure, ie, from 35mA to say 64mA, then the
cathode voltage will rise from
about +18V to +32V, and the active
protection circuit will turn off the anode supply.
The
protection circuits are shown on Sheets 5, 6 and 7.
The only stabilizing zobel network needed is the 4.7 ohms +
0.22uF across
the output terminals. Thus at 154kHz, the reactance Ce 0.22uF = R7 4r7
so as frequency rises
above 160kHz there is an increasingly resistive load
across the output terminals.
Any value of capacitance across the output terminals and
without any
parallel or series resistance load does not provoke any HF
oscillations.