Deep Space
845se55 July 2008

Photo 1, The two 55W amps on my
bench.
After a few months of research
and development , I completed
construction of a pair of 55 watt class A monoblocs each using a pair
of
paralleled 845 triodes for single ended operation. These amps
were a very
difficult handcrafting challenge but I think my discerning customer who
waited
nine months will never find any better sounding amplifiers with such
natural
clarity and fidelity. During tests one evening a friend and I
were
reduced to tears with some good recordings. Only the best audio gear
achieves a
real emotional impact similar to well performed live un-amplified music
which
has always been my “gold standard” for audio quality.
The amps also have excellent
technical
performance.

Photo 2.
The first prototype monobloc
had all the power supply
components and audio amp components on the one chassis. It was found
that the
chassis weight for one channel went over 42Kg, and it became extremely
difficult to move easily, and all the parts had to be crammed together
which
brought a lot of problems in getting rid of the heat generated. So I
adopted
the same principles I used in my 300W monobloc amps and have two
chassis per
mono channel with power supply on one and audio circuit on the other,
and with
each chassis connected with very heavy duty umbilical cabling. The two
chassis
construction reduces the weight of any chassis to manageable levels
without too
much risk of spinal injury. It also gave me room to place all hot
running
resistors inside a heatsink on one end of the chassis top instead of
underneath, and this helps the amp stay cool even on hot summer days.
Each audio amp chassis weighs
approximately 30Kg and is
520mm deep x 230mm wide x 280mm high. Each power supply chassis weighs
16Kg and
are 280mm deep x 200mm wide x 250mm high. Total amplifier weight for
two
channels is about 92Kg.
Rectifiers are all silicon,
and the power supplies always
run cool and can be placed on the floor. The audio amp chassis can be
placed on
a bench or stand above the power supplies so the front on/off switch
can easily
be reached, and you can keep an eye on the tubes.
Photo 3.
This shows a close up
view with 2 x KR Audio 845
output tubes, 3 x Sovtek EL84 triode drivers, and 1 x NOS AWV 6CG7
input.
Initial tests were done using "expendable" Chinese made 845 which
measured well and sounded well during tests.
The amps are made to work with
any brand of 845. The KR have
cathodes needing 10V at 1A, while Chinese 845 need 10V at 3.3A, so the
KR tubes
run cooler than the Chinese types, if they both have the same anode
dissipations.
The KR845 may have a slightly more detailed sound than the cheapest
variety of
the Shuguang Chinese 845. I leave the final judgment on sound quality
of KR
tubes to other people. The KR tubes certainly look better made than
their
Chinese cousins, although the Shuguang 845 is a closer copy of the
original RCA
845 and other old ancient brands. Both measured very similarly low
distortion
levels, had the same biasing voltages and bias currents and power
outputs in
the same circuit. The power supply has 3 taps on the Vac cathode heater
ac
winding, and the cathode heating voltage can be finely adjusted for the
correct
level by choice of taps. Chinese 845 require a higher current,
therefore the
highest tap is used to obtain 10Vdc from the choke input supply, and KR
can use
the lowest voltage tap for the lower heater current.
RCA or other brands of NOS are
so rare now that it’s
pointless trying to find any. Because the KR tubes have 20 watts less
heat in
their cathodes, they could be idled at a full 100watts of anode power.
But the
KR 845 are 3 times the price of the Chinese types so I have set up the
KR to
run at about 75 watts of anode dissipation. I hope they last well. The
Chinese
845 might be purchased for usd $150 and this is good value. The KR845
are
listed at over 3 times this price, so I don't wish to take the risk of
idling
the KR tubes with too much anode dissipation. See the notes about the
operating
conditions below.

Photo 4.
The amp has its top cover
removed in my workshop. You can
see the hand made heatsink to the left which encloses all the main hot
running
resistors. Moving right you see a solenoid choke for the CLC filter for
the dc
power applied to all input and driver tubes. Under the solenoid are two
potted
chokes for the two choke input dc cathode heating in each 845. In the
center
back is a large potted output transformer with 72mm stack of 51mm GOSS
E&I
lams. Center front shows some of the 470uF filter caps in main anode
supply
rails and a 60H choke which is part of the 36mA dc anode supply to the
3 x EL84
working in triode mode. The 845 'Johnson' tube sockets are recessed and
there
are 4 McMurdo 9 pin sockets for smaller tubes. White labels with black
lettering is used to indicate what goes where to avoid people swearing
and
cursing because they cannot read tiny lettering in the gloom of a
listening
room.

Photo 5.
"Beneath the bonnet"of each
audio amp chassis, top
left, you see the entry and terminations of the incoming umbilical
cables.
Towards the right is a heatsink for three diode bridges for dc heater
supplies,
then rail discharge resistances, then underside of Johnson tube
sockets, and
far right you see the compact wiring of the 4 tube sockets for input
and driver
stages. bottom left is the active protection board to shut down the amp
if an
output tube mal-functions, then towards the right there is 60,000 uF in
four
caps for 845 heater filtering, then rail discharge resistances. All is
genuine
point to point wiring with mainly hardwood terminal strips well sealed
with
varnish. Hookup wire is 1mm thick pvc insulated multi strand copper
chosen for
very long term reliability and with very generous current ratings and
much with
an additional shrink wrap layer of insulation added where voltages are
over +
or – 500Vdc.
Right in the middle are two
rows of terminals which allow a
tech to reconfigure the output transformer secondary to suit either 3-6
ohms or
6-or-more-ohms.

Photo 6.
Here is the under chassis view
of each power supply with its
bottom cover removed. There is a small 5VA auxiliary mains transformer
to the
top LHS. This supplies 12Vdc power to the active protection and dc turn
on/off
circuit. The power transformer has a board on the RHS with 48 terminals
for
taps and ends of windings. The range of voltages available allow for
the use of
many different output tubes and configurations in future if 845 become
scarce
or unavailable. Suitable alternative tubes are 4 x KT90 in parallel
SEUL, using
the same output transformer but with a different winding arrangement on
the
primary. Or two x 13E1 can be used, or various push pull arrangements.
So if
ever there are no 845, the amps can be altered by a skilled tube
technician,
and sonic purity and great sound can be maintained without compromise.
Instead
of recycling the amps at the scrap metal dealer, if all the
transformers remain
intact the amps can re-built into something else, and even a six pack
of EL34,
6L6GC, 5881 could be used.
But in 20 years, I may not be
around....

Photo 7.
This workshop picture shows
the rear of an audio chassis
(left) beside its power supply (right). It would be difficult to make a
mistake
with the umbilical cable plugs because one is painted bright red, and
so is its
socket. If the absent minded audiophile makes a mistake by reversing
the plug
positions, the amp cannot be turned on and no damage is sustained at
all.
On the amp chassis there is
provision for bi-wiring or
having two speakers into the 4mm bind posts which are glued into a
plywood
block to protect them from breakage, and ensure that connections to
speakers
are only possible with leads that have 4mm banana plugs. Binding posts
that
rely on a wire poked through a hole in the post and then with a knob
turned
tight are an unreliable connection and possible sonic horror so I won’t
let
anyone use them with my amps. Quad thought the same way in about 1950
with
their Quad-II amps which sold in large numbers. Input socket is a
Cardas RCA
input for a single ended input only, shown recessed on the left.
On the amp rear side to the
right, the umbilical cables
emerge and are soldered into the amp to avoid connection confusion, and
the
problem of having too many plug and socket connections.
Cable length is 1.5metres, so
the power supplies may sit on
the floor, out of sight, out of mind, and away from any other gear,
while the
audio chassis may be on a bench 900mm high to allow easy access to the
on-off
switch, and to keep an eye on the tubes.
The rocker type on-off switch
is recessed to avoid damage,
and it switches low voltage 12Vdc. The actual mains switching is done
with
relays within the power supply. Thus 240V mains wiring is not brought
into the
audio amp chassis and so there no diode switching noise spikes or hums
from
where the on-off switch is so closely situated to the audio input
circuits.
Schematics.
I don't mind publishing
schematics for free on line. People
think I am crazy, but nobody is going to copy what I have done and make
a
profit because these amps will have a cost of production far in excess
of most
cheap nasty toy like amplifiers which one can inspect around the
Internet. Most
amp makers who make larger amp numbers in a factory all are frightened
to show
the secrets behind their creations. They don’t want you to know how
they
managed to get the cost of production to down to a very tiny fraction
of what
they want you to pay. Nobody will copy my designs of amps and be able
to make a
profit without employing a bean counter to remove the quality to
cheapen the
cost of production, thus stealing and ruining your music, and lessening
the amp
reliability. I have never obtained a review from magazines such as
Stereophile,
or paid the huge sums to advertise in that magazine, and make only low
volume
productions. So copy cat makers are never going to copy me, because
they cannot
offer a fake amp at a low price because the design has become world
famous and
priced way above most peoples' ability to pay.
Warning.
There
are high voltage
potential differences of up to 2,000V within each amplifier when
operational.
Only trained and experienced technicians should attempt to examine the
working
circuits or build the circuits shown in the schematics.
Sheet 1.
Input signals enter the input
V1 6CG7 with both halves
paralleled. There is a high pass CR input filter with C1-R1 to give a
pole at
5Hz. This keeps out dc in sources and extremely low frequency signals.
The
MJE350 transistor might seem to be quite out of place in a tube
circuit, but
acts as a passive component which supplies V1 anodes with a non
changing dc
current or what is called a constant current source, CCS. It is not a
perfect
current source, but close to one and with a real
source impedance of several
megohms at least. Therefore the
MJE350 cannot include any sonic signature in the signal path. But with
the
MJE350, V1 anode load is effectively only the following cap coupled
biasing
resistance R11, 180k. The Ra of V1 is about 5k, and the RL is 36 times
greater,
and when triodes are loaded with RL many times the Ra, they give the
best
sound, and the lowest possible distortion measurements. If you were to
replace
the CCS with a simple resistance of about 39k, THD/IMD would maybe
increase 3
fold. The THD of V1 is mainly all 2H, but it will add to that of the
output
stage because of the relative same phase of the 2H, so to minimize THD,
the CCS
transistor helps to lower distortion
and maximize the voltage
gain of V1.
Any brand of 6CG7 may used,
and my favorite is Siemans NOS
made in
For greater input sensitivity,
6922/6DJ8 could be used for
V1. One would have to use 2k7 R4 grid resistors at each grid because
the
6DJ8/6922 does tend to oscillate at around 200MHz if you parallel the
two
halves and did not use separate grid R. The cathode biasing resistor,
R5, would
need to be reduced in value until Ea measured about +120Vdc.
The 6CG7 is an evolution of
the famous octal based 6SN7 but
with its two triode elements fitted into a nine pin tube, so the
technical
character is identical and it ensures the audio signal is initially
amplified
very linearly, while maintaining excellent musicality, micro detail and
warmth,
transparency etc that one enjoys with the best tubes when set up the
way I do.
8dB of global NFB is applied
from the output transformer
through R3&C2 to the top of R6.
C6, C7, R8, and R11 form a LF
gain stepping network to
optimize the LF stability and fidelity.
V2,3&4 are triode
connected EL84, each with individual
cathode biasing. Any brand of EL84 may be used, and you could have 3
different
brands together if need be because they each have their own cathode
biasing
network. I've fitted Sovtek which sound well, and mixing up brands or
using 3 x
NOS EL84 may or may not make a change. Three are used to produce what
becomes
ONE super triode with the ability to produce a maximum of 164Vrms of
signal
output with less than 2% THD, and with good gain, and wide bandwidth,
and with
combined Ra = 700 ohms only. The use of a 60H choke plus 7k to supply
Ia =
36mAdc total provides a high ac impedance anode supply load which
dissipates an
extremely small amount of ac power, so hence the excellent linearity,
because
like V1, RL is many times Ra, and RL approaches a CCS. The load seen by
the 3 x
EL84 is approximately 20k, and mainly due to the grid biasing
R28&R29 of
23.5k of the two following 845 grids. Again, RL = 28 x Ra.
There is zener diode shunt
regulated Vdc for V1 supply to
assist LF stability. Any noise in the zeners is filtered by R21 and C9,
and the
CCS MJE350 prevents any other noise entering V1 anode circuit.
Sheet 2.
Each 845 is set up in
conditions as follows, Ea = +1,060Vdc,
Ia = 70mA, cathode bias voltage = 150Vdc, RL per tube = 12k, and so for
both
the load is 6k.
The maximum drive voltage to
845 grids for clipping is up to
approximately 110Vrms containing 1.4% 2H from the driver stage. The
driver
stage anodes applies the drive voltage to the network of C16,17,18 and
R23, 24,
25, 28. This network transfers the signal safely from the EL84 anodes
at
+310Vdc to the 845 grids at -600Vdc. Coupling caps are 2.2uF each and
rated at
1,000V and the LF pole is at 9Hz.
The 845 anode current is
supplied from two rails, one at
+600Vdc, and the other at -624Vdc. This unusual arrangement reduces the
likelihood of arcing within the OPT between anode windings and earth
potential
secondary windings. The filtering of the two rails is by CLRCRC set up
to give
a damped LF pole, and many tests were done to ensure the continual
mains
voltage level changes and LF noise does not create LF resonance signal
which
then appears between grid and cathode of the output tubes, and
therefore does
not appear in the output in excess of 0.5mV if mains noise is
very
bad. The noise performance despite the twin rail use is extremely
good.
Despite so little global NFB these amps are the quietest tube amps I
have built
yet.
Two supplies of 10Vdc are
applied to 845 cathodes XX and YY
so that the small amount of 24mV of residual 100Hz noise is balanced by
R30&31, 32&33.
Sheet 3.
This shows the three simple
heater dc supplies used for ALL
tubes within the amp. The L3, L4 chokes used in the choke input dc
supplies for
the 845 cathodes are potted and do not cause any magnetic interference
in the
potted OPT on the same chassis. L2 is a solenoid type of choke in a CLC
filter,
so the Vac across the choke is tiny, and thus its change in magnetic
field is
negligible, so potting was not needed.
Sheet 4.
The main power supply chassis
have all the above within to
generate the positive and negative Vdc rails for the 845 and other
tubes, and
12Vdc for relay switch on and protection circuits. All the ac cathode
heating
voltages from the power transformer for the 845, EL84, and 6CG7 are
conveyed in
the umbilical cables to the amp chassis where they are rectified to dc.
Despite
all heater rectifiers being on the audio chassis, there is no resulting
diode
noise in the output. See notes below sheet 6 about power transformer
and iron
core component replacement.
A large number of rail voltage
arrangements are possible.
FUSES.
Caution! The amp must be
turned off and mains cables removed from wall
socket, and allowed 15 minutes rest after turn off before changing any
fuse
!!!!!!!!!!!!!!!!!!!!!!!!!!
Fuses or fuse wire links for
all windings are as follows:-
F1 Mains input, for 220V,
230V, 240V, 3A slow blow, type
3AG, and accessible at rear of PSU by owner. For 100V, 110V, 120V
operation,
mains fuse is 6A slow blow type 3AG.
All fuses below may only be
replaced by a technically
trained person.
F2, F3, F4. Three fuse wire
links rated for 10A soldered
into the underside of the psu chassis and covered with black polyester
sleeving
for 2 x 845 ac cathode heater windings and one other cathode heater
winding for
3 x EL84, and 1 x 6CG7.
F5, F6. Two x 3A slow blow,
3AG, soldered into place under
psu chassis for the two main HT rails of +660Vdc and -640Vdc, derived
from
voltage doubler rectifiers.
R58 Provides some protection
for the auxilliary small power
transformer under the psu chassis. This 1 watt resistor will burn out
if the
auxilliary transformer is shorted.
845 Anode fuses There are two
0.5A slow blow 3AG fuses
soldered between the bottom of R36 and each 845 anode in case the anode
current
exceeds 0.6Adc.
Sheet 5.
Most tube power amps don't
have any kind of active
protection against the eventual failure of one or more output
tubes. I have had to repair very many "good"
hi-end brand amplifiers that gave a lot of trouble due to poor design,
or
through mishap caused by owners, or malfunctioning speakers etc, etc,
etc. To
avoid smoke in the listening lounge room, and collateral damage to
other parts
within the amp, most amp makers do fit a couple of fuses. But fuses
provide
only partial protection, and they don't always blow when one wants them
to, and
owners are notorious for shunting them with something convenient like
aluminium
foil from the kitchen or using a 2A fuse instead of 0.2A which
encourages a
faulty amp to burn the house down.
Active protection is needed to
stop the smoke and damage and
to tell an owner when something is wrong, and if possible to shut down
the amp
and prevent fuses blowing.
Nevertheless, there are a fair
number of fuses fitted in
these amps and apart from the mains fuse they are all soldered into
place
because fuse holders are notorious for not holding a fuse firmly and
becoming
intermittent with dc flow. Fitting new fuses is a painful exercise
requiring a
tech with a soldering iron. The most likely problem in any tube power
amp is
the sudden or gradual unwanted increase in the idling dc current flow
in each
output tube. This current is sometimes called the anode bias current,
and it is
controlled by the voltage between the grid and cathode. But a tube can
change
its character as it ages or during some trauma such as caused by a
shorted
speaker cable, and despite the biasing voltage Vg-k, the bias current
may
increase to many times the idle value, with dire results if not dealt
with. The
above simple circuitry will shut down the amp in 90% of bias failure or
tube
failure cases, and if the failure was caused by some temporary fault,
it may be
easily reset to go again simply by switching off, then on again as
explained in
the text in the schematic above. Under normal operation, the 845 anode
current
is around 70mAdc at idle. This generates 150V across cathode
R34&R35.
Should the anode current ever rise to about 102mAdc, there will be a
cathode
bias voltage of 225Vdc, which means the Pda will have risen to about
100W, and
although this is the maximum rated Pda for 845, its plain wrong in
these amps
and due to a fault condition. So to give the fault condition it only
takes an
Ia increase of 32mAdc, or 45%, and you cannot rely on fuses to blow
with such a
small amount of current change, so active protection is the only
reliable way
to prevent Ia rising to perhaps 400mAdc in this amp. 400mAdc would
damage the
cathode bias resistors, and perhaps the OPT primary winding if the
condition
lingered for too long, and I have seen this happen in many amps brought
to me
for repair. The protection circuits have utterly no effect on the
sound.
Sheet 6.
The power transformer does not
run very hot because the
turns per volt ratio gives a B-max of less than 0.9 Tesla, and wire
sizes are
generous and rated for no more than 3 amps per square mm. The
transformer is
neatly layer wound with layered construction shown in the drawn section
through
the winding bobbin.
I give a two year warranty on
amp transformers. But if one
were to fail, and I was not around in future then there are no
standard
easily available replacements for any iron cored wound components in
these
amplifiers from any known commercial transformer winder. All are custom
wound
and may have to be ordered as a special order from perhaps Sowter
Transformers
located in the
winding a pair of trannies
with cores rated at 650W for cool
running.
An alternative is to rebuild
the power supply on a new
chassis to allow the use of multiple power transformers which may be
available
as trade stock from Hammond Engineering, through the Australian
dealers, EVATCO
in
unless they have been designed
to run with B <
0.9Tesla.
In the event that a
transformer is damaged and cannot be
used, it is possible for a suitably trained tech to disconnect it from
the
circuit, unscrew the external screws, and remove it off the psu
chassis.
The terminal board and top
sealing layer of resin and sand
concrete seal can be removed with chisel and hammer and remaining loose
dry
sand surrounding the transformer can be drained out. The transformer
should be able
to be removed from its pot and thus the pot can be re-used. To
dismantle the
wound transformer all bolted angles and bolts are all removed. The
grain
oriented silicon steel core can be salvaged by heating the transformer
in a
wood fire to a dull red to vaporize all plastics in the construction,
then
allowing it to cool down for a few hours. The burnt wire is cut away
for
re-cycling, and laminations should all fall easily apart and will be
ready for
re-use. A new transformer is then wound using a new plastic bobbin to
suit a
70mm stack of 51mm tongue laminations so that it will fit inside the
pot. The
newly wound transformer must be varnished while being wound or after
with a
soak and bake method. A new terminal board is made and fitted. After
testing the
new transformer it is re-assembled back into its pot and clean dry sand
used to
fill the pot except for the last 15mm. The sand must be thoroughly
vibrated and
settled. Spray-can varnish is applied to the sand surface and next day
the top
15mm of fill can be done using a 50-50 mix of epoxy (fibreglassing)
resin and
sand. This seals the pot and prevents dry sand running out. The spray
varnish
prevents liquid resin soaking into dry sand below. The completed tranny
is
re-installed into the chassis and connected up and tested. At present,
nobody I
know in
The power transformers and
other wound components such as
chokes and OPT have been designed to run cool, and all windings have
fuses, and
active protection is used against output tube failure.
After winding so many power
and output transformers and
chokes during the last 12 years of commercial operation, not one has
failed, so
I have never needed to repair any of my work.
Much time has been spent on
making the iron cored items in
these amps.

Sheet 7
The output transformer is
layer wound and varnished with
Wattyl 7008 polyurethane two part varnish generously applied to
windings as the
transformer was wound. It is potted in a galvanized iron pot and filled
around
with dry sand or roof pitch as chosen. See the notes above about
transformer
replacement. The output transformer is more difficult to wind than the
power
transformer because of the number of fine wire turns around the large
size core
size. There is no known ready made replacement type available from any
commercial winding specialist although one may possibly be ordered as a
special
from Sowter Transformers located in the
The anode resistance of the
two 845 in parallel is 1,100
ohms and when in parallel with an anode load of 6,000 ohms, the source
resistance = 930 ohms. Primary inductance is over 40H at 150mA dc
current. The
-3dB response drop at LF and due to primary shunting inductance occurs
at 4Hz
at loud levels normally used. At the 50 Watt output level the onset of
core
saturation occurs just under 20Hz. At 50W the HF -3dB point is above
30kHz, all
with zero global NFB. Some of the slight sag in HF is due to the
stability network
of R37, and C23 beginning to load the amp above 50kHz. Shunt
Capacitance from
the primary anode terminal No1 to the secondary is less than 3,000pF.
Leakage
Inductance has slightly less attenuation effect than the shunt
capacitance, and
the resonance between leakage L and Shunt C is above 30kHz. Primary
winding
resistance = 95 ohms for 2,760 P turns, 0.45mm wire. Secondary winding
resistance = 0.124 ohms for 5 x 72 turn secs in parallel, 0.9mm wire.
Sec
winding resistance referred to primary = 182 ohms, so total Rw = 95 +
182 = 277
ohms at the primary input. Winding loss percentage = 100 x 277 / 6,277
= 4.44%
with 6,000 ohms anode load with the OPT secondaries set up for 4.1 ohms
or 6.4
ohms.
LOAD MATCHES AVAILABLE.
5 x 72 turn parallel secs give 72 turns to match for 4.1 ohms, allowing
speakers nominally above 3 ohms.
4 x 90 turn parallel secs give 90 turns to match for 6.4 ohms, allowing
speakers nominally above 6 ohms.
4 x 72 turn parallel secs in series with 2 x 36 turn parallel secs give
108
turns to match 9.2 ohms, allowing speakers nominally above 8 ohms.
The signal current density in
every secondary winding
remains equal for the first two arrangements of secondaries but
slightly higher
for the 9.2 ohm set up. With such a high maximum power of 55 watts into
4 ohms,
there is usually enough power for any speaker regardless of its nominal
impedance. The 6.4 or 9.2 ohm setting should only ever be used if
speakers have
under 86dB sensitivity, 1W, at 1M and have impedance above 6 ohms and 8
ohms
respectively.
All speakers including ESL
types by Quad such as the ESL 63,
989, 2805 should be tried with the 4 ohm setting before changing the
setting to
a higher one. Many older high impedance speakers of say 16 ohms are
much more
sensitive than more modern types, and therefore require a tiny amount
of signal
voltage and power to play very loud, so there is no need to change the
impedance matching to the 9.2 ohm setting, and the 4 ohm setting will
be very
adequate. Most people will never use more peak power over 10 watts with
average
power well below this figure. Because the amp measures so well at 55
watts, at
an average power of 2 watts the distortion is well below audibility,
see the
notes below.

Sheet 8.
Six chokes per channel are
used for filtering and to prevent
the heat losses through alternative methods of filtering or active
regulation.
No solid state chip regulators are used because they become unreliable
when
used in circuits with such high voltages lurking about. There are a few
simple
zener diodes for basic shunt regulation. The main positive and negative
voltage
rails of over +600V and below -600V have a CLRCRC type of filter. Each
C is
formed with 2 x 470uF in series to make 235uF, and total C per rail =
705uF.
There is a resonance between the 4H choke and following 235uF at 5.2Hz,
but the
added 100 ohms in series plus the following additional 100 ohms plus
235uF act
to damp the peak in the resonance, thus making each voltage rail less
liable to
vary at LF below 10Hz due to mains voltage changes that occur
continuously, and
typically of +/- 20mV.
Sheet 9.
This shows the arrangement of
resistors used within the
aluminium heatsink at the rear end of the amp chassis. Cheap ceramic
bodied
wire wound types are used with their dissipated power being well below
the rating
for the resistance chosen. These are readily and cheaply available from
many
suppliers. A total of about 50 watts is dissipated in the resistors
shown, and
to keep them all cool and thus more reliable, the heatsink was built up
to
enclose the resistors behind a removable cover fitted with many fins
and which
springs tightly against the enclosed resistors. They are glued to the
3mm thick
aluminium heatsink fixed plate with Selleys 401 engineering grade
silicone with
a temperature rating of 200C. The screwed cover can be removed by
removing the
9 x 4mm metric machine screws. White heatsink paste is used between
resistors
and the cover. While it is possible than resistors might fail, it is
not likely
that they will fail. Resistors usually fail by fusing open, and these
can be
prized off the heatsink with a chisel, and a new one fitted. It is a
messy job
to replace any resistors, and a tech needs to know what he is doing,
but at
least the resistors are easily accessible once exposed to view.
Sheet 10.
Here we have the layout for
rugged cables used to get power
from the power supplies to the amp chassis. When the amp chassis are
examined
with a copy of the above, just exactly how everything is set up becomes
less
confusing. The octal plugs at the ends of cables are permanently
connected. All
wires are soldered into the hollow pins of the plugs. There is no
access to the
wire ends leading into the hollow pins of the plugs. If a pin is broken
off a
plug, the whole plug is made useless. The only solution is to cut the
plug off,
and rewire a new octal plug onto the lead as shown above. A cheap type
of octal
plug from RS components could be used to make a new plug. The original
plugs
were made using only the bottom base from an 8 pin tube plug. This had
the
sides ground off so the base fits neatly inside a 30mm long piece of
PVC
electrical wiring conduit tubing with about 25mm internal dia. The
central
keying spigot of the plug has a 4mm threaded rod inserted to reinforce
the
spigot which otherwise will all too easily be broken off accidentally
by a
careless owner, leaving no way to correctly locate the plug into the
socket at
the power supply, and therefore promoting many bad tempered experiences
while
trying to make the amplifiers work. Both plugs MUST be plugged in
correctly for
the amp to be able to be turned on.
There are some inbuilt safety
features of the umbilical
cables. While plugged in there is little danger. The danger from a
shock is no
worse than any normal 240V wall socket plug used for many other
household
appliances. It is impossible to turn on the amps with only one octal
plug
plugged in, or with plugs reversed, i.e, and red into black socket,
black into
red socket. However, if somebody were to wrench one of the power supply
cables
from the power supply while the amps are turned on, the amps will turn
off
immediately. If any person were to immediately grab the pins of the
plugs after
removing them, then they would be connected to the live stored voltages
within
the amplifier chassis. Diodes have been placed to prevent the flow of
current
from the major amp voltage rails and thus prevent a shock. The 845
cathode
heater supplies are biased at the cathode voltage of -450V but at turn
off
there is a relay to reduce this voltage to less than 40V within less
than 0.5
seconds, so it would be very unlikely to experience any kind of shock
unless
one tried desperately to do so.
It would be unwise to allow a
pet dog to chew on cables.
Most animals will get message from a wire they chew, and learn to leave
them
alone. The cabling used is particularly rugged industrial grade cabling
with
thicker PVC insulation than used for high power 240Vac rated cables.
The highest
voltages are carried in the two thick black cables while very low
voltages are
carried by the orange cable.
For all things we cherish,
practiced care is the best
insurance.
Place the power supplies on
the floor behind the amps and
speakers and well away from any likelihood of being tripped over by
passing
traffic. Although the amps have been made fairly ruggedly, don't drop
one, or
allow it to be pulled off a bench. When moving the amps, turn them off
and wait
until they have cooled down for 10 minutes. The umbilical cables can be
unplugged, and coiled up and tied up to the rear carry handle on the
amps.
Don't let trailing cables get under your feet, or catch on anything.
Sheet 11.
So how clean is the signal
from these amplifiers?
I'd like to say it is much
cleaner than most other tube
amplifiers. Do you ever wonder why the nitty gritty technical aspects
of tube amplifiers
are hardly ever mentioned as you surf around the Internet? Does anyone
else
bother to publish curves like these? Ever wonder why?
There is usually too much to
be ashamed of if makers told
you the whole story, and because the technical character of the amps
concerned
is usually very poor. Many makers realize that nobody cares about the
technical
aspects as long as the sound is good. But I know good sound is only
possible if
the amps are technically very good.
A large part of my living is
earned by re-engineering very
poorly designed hi-end brand amplifiers made by brainless designers who
are
eternally optimistic and in constant denial about the woeful aspects of
their
products.
NOISE.
As supplied, the 845 amps have extremely low noise level that can only
just be
heard if an ear is held tight against a midrange speaker. Noise of any
kind
measures less than 0.35mV.
DISTORTION.
The graphs above in sheet 11 show THD levels between half a watt and
clipping
and for various load values. The vertical and horizontal axies are both
logarithmic so it is easier to see low distortion levels at low power
levels.
THD is mainly 2H, with some 3H, 4H and 5H well down.
The amps will comfortably give
huge sound levels into any
type of speaker over 3 ohms including ESL. THD is about 2% at an
"illegal" power output of up to 60 watts at just past clipping. THD
is 0.5% just under clipping, but the first 10 watts of power at any
load over 3
ohms produces less than 0.15%. The least THD at all levels occurs with
a 7 ohm
load, due to the natural second harmonic distortion cancellation that
occurs
between the driver and output triode amplifier stages. THD artifacts
generated
during normal loud listening never rise above 0.05%. IMD artifacts
consist of
those harmonics related to the second harmonics of fundamental tones.
Such
"2H" related IMD harmonic products are the subjectively "least
worst" type of distortions.
If you have 2 volts at the
output with 4 ohms, the power
level is one watt, thus producing 89dB SPL into most modern European
made
speakers with average sensitivity. if this was the average level, your
wife
will tell you to turn it
down! It is loud.
If the
THD was 0.05%, then the distortion voltage is 0.001 volts or 1
millivolt. If we
could listen to that 1mV of distortion played through your 89dB/W/M
speakers
and without the wanted undistorted music, we would find the sound of
the
distortion would not be audible, or about as loud as nervous rat
sneaking
across the floor to get past a sleeping cat.
Way back in about 1953,
exhaustive tests by the BBC and
others revealed that listeners were unable to discern any distortion
present if
it measured below 0.5% and if the system response was hi-fi bandwidth.
It must
be remembered that in 1953, transducers such as microphones, speaker
drivers,
record cutter head amps and record playing and tape recording and radio
transmission and reception introduced 10 times the 0.5% on a routine
basis, and
with noise levels that might wake the dead. Nowadays we have the
artifacts created
by solid state and digital processing, and the music still battles to
get
through all that recording gear. From the above graphs for various
loads,
notice that the distortion is low where we listen to music, but it
increases
with power. If one were able to accurately record a gunshot, and replay
this
recording through these amps to try to reproduce the sound as it was
heard
without ear muffs, the amps would be forced into clipping, ( maximum
possible
power levels ) and into high distortion, but it is of no importance
because
such noise is so short lived, and only noise, and not musical, and the
distorted noise won't alter the idea that someone may have been shot.
So the
high ceiling level of power is good for "transients" and not much
else. Nobody is going to listen to deafening levels of anything musical
for
very long. We like to sit at least 20 metres away from the local
orchestra in a
hall because the sound among the players can be deafeningly loud, and
never how
Beethoven or Mozart intended. And by the way, to get the sound of a
pistol shot
in a room properly reproduced without an amp or speaker clipping, a
wall full
of speakers and 1,000 watts would be needed. A single grand piano
played as
loudly as possible sounds overwhelming up close, but majestic when some
distance away. With 50 watt amps, if we want that majesty, it is up to
us to
have sufficient speaker sensitivity or get more powerful amps. Majestic
bundles
of cash are required, and perhaps a change of wife as well though :-).
The use
of TWO 845 per channel raises the power ceiling and lowers the
distortion
compared to when only one 845 is used per channel. This should allow
complex
orchestral music to be enjoyed, not just a lone cello or piano.
BANDWIDTH.
At ordinary loud levels bandwidth is 5Hz to over 50kHz, which is
slightly
better than the amp with no global NFB, as quoted above.
DAMPING FACTOR.
Non technical people might think about mops and buckets of water when
the words
"damping factor" are used. Not one amplifier ever made is perfect.
Amplifiers are like a car which gets you from A to B, but the more
passengers
you have the slower the car goes. The perfect car would behave the same
with 4
people aboard or with just the driver. Now amplifiers don't actually
"slow
down" like a car might when the load increases, but the output voltage
from any amp will drop in level as soon as a speaker load is connected,
and
continue to drop as more speakers are connected. We want a minimum
amount of
voltage drop because all speakers are made to operate ideally with a
non
changing amplifier voltage and regardless of the signal frequencies.
The lower the number of ohms
in the speaker, the greater the
voltage drop will be. This is confusing. The more people in the car,
the harder
the car works to drive uphill for a given speed. But in electronics,
the less
ohms there are, the harder the amp works. The God Of Triodes invented
audio
electronics and made the rules so difficult and absurd so that fools
wouldn't
try to muck around with electrical things. The basics about ohms,
current and
voltage can be learnt by Googling, and experimenting with basic
resistors and
gear at home on rainy sundays, using a low voltage ac supply of 50Hz
and a
voltmeter, and armed with knowledge about Ohm's Law.
Suppose an amp makes 6 volts
without any speaker connected
so that's the measured output voltage without a "load". So there is
no current flow between the speaker terminals with no load. Let us
connect a
speaker, and maybe we would measure the amp voltage drops down to 5
volts. Let
us suppose the speaker load resistance or impedance measured was 5
ohms.
Consider Ohm's Law. I = E / R, where I = current in amps, E = voltage
in volts,
and R = resistance in ohms. For our speaker, with 5 volts present the
current =
volts divided by resistance = 5 / 5 = 1 amp in this case. The voltage
drop at
the amp terminals was from 6 volts to 5 volts, with current changing
from zero
amps to 1amp. The amplifier "output" resistance, Rout behaves as if
we had a perfect unchanging voltage source but with a concealed mystery
series
resistor between this perfect source and the output terminal. We need
to know
what the mystery resistance value is.
To easily calculate this R, we
must apply Ohm's Law. R in
ohms = E in volts divided by I in amps, so R = E / I . So to Rout =
change in
load voltage divided by change in load current. In this case, Rout = 1V
change
/ 1amp current change = 1 ohm. There is in fact no real 1 ohm R or
perfect
voltage source within the amplifier, but the amp acts exactly as if
there is
such an Rout present and in series with a perfect voltage source with
extremely
low output resistance. We carry the concept into our minds by
considering model
of what is present, even though we really have a bunch of tubes and an
output
transformer.
There is a ratio between the
amplifier Rout, "output
resistance", also known as the Zout, "output impedance" and the
speaker ohm value, and it is called the Damping Factor, or DF. The
Damping
Factor = speaker load in ohms divided by amplifier output resistance.
The
higher this number, the better, but it should be above 4.0.
But suppose the amp Rout was 1 ohm, then DF = 4 / 1 = 4, and this is
barely
acceptable. With 6 volts without any load, and variations of 3 to 30
ohms, and
the speaker voltage will vary from between 4.5 volts to 5.8 volts, and
the
change in acoustic levels is only 1.5dB. Usually a good speaker
designer would
have assumed all amplifiers have Rout = 1 ohm or less. A DF = 10 would
be even
better though, so with speaker = 4 ohms, Rout would need to be 0.4
ohms. Any
increase of DF above 10 will cause no perceptible change to the speaker
response.
But suppose the amplifier had
a DF = 0.5. That would mean
that output resistance ( Rout) = twice speaker ohm value. And speakers
vary in
ohm value at different frequencies, and typically between 3 ohms and 30
ohms
for what might be a nominal "4 ohm speaker", which averages 4 ohms
between the most energetic band of musical frequencies between 100Hz
and 1 kHz.
So if the amp has Rout of say
8 ohms, and the speaker R
varied from say 3 to 30 ohms, the voltage at the speaker would vary
very much,
and so would the acoustic response of the speaker, and an amp with a
low DF
number is one that is like having a sound system with a very badly
adjusted
graphic equalizer or tone controls. Some speakers might sound passable,
while
others would sound dreadful! Most sound would be very colored.
With Rout
= 8 ohms, and speaker nominal R = 4 ohms, DF = 4 / 8 = 0.5 which is
very poor.
The output resistance of the
845 amps = 0.5 ohms, so DF = 10 with a 5 ohm
load. The small
amount of 8dB of global NFB used in the amp reduces the Rout from 1.1
ohms
without any NFB connected.
There is much said about
damping factors needing to be high,
but anything over 4 is acceptable to 99% of people, especially with
triode
amplifiers and as long as the speaker has been designed competently so
that
large impedance variations do not occur in a way that may cause
response
problems. A badly designed nominally 4 ohm speaker might have a drop in
impedance to 2 ohms, and although that may not cause a large perceived
response
problem it may cause much more distortion to occur at frequencies where
the
impedance is low, and this will damage the music, and maybe damage the
amp.
Many tube amps may have Rout =
1 ohm, and might be used with
ESL speakers with high impedance at LF and quite low impedance at HF;
Quad
ESL57 ranged from 33 ohms at 50Hz to 8 ohms at 1 kHz to 1.8 ohms at
18kHz, and
one might expect boomy bass and missing highs with amp Rout = 1 ohm.
But no,
the DF at bass is 50 / 1 = 50, and excellent, and its 8 at 1 kHz, and
in the
critical region of music there is very little difference in response
levels.
The ESL 57 was designed to operate from amps with Rout = 1 ohm. So if
you had
an amp with very high DF at all F and with Rout = say 0.1 ohm, the HF
might
become too prominent with ESL57. One might always add a 1 ohm series
resistor
to make the sound more bearable.
So one has to be flexible
about damping factors, and not be
too obsessive and that isn't easy for many audiophiles, especially the
hard to
please majority who cannot understand anything I have said above.
And for explanations for more
things that are inexplicable
such as negative feedback, please travel to elsewhere in my website to
become
truly confused by science.
My 845 amps depend on a large
amount of applied science and
calculations to give the best sound possible from tubes. I doubt very
much that
there would be any change to the sound if I were to use exotic
materials such
as 50% nickel in the output transformer cores, or pure silver wire.
Audio Note
in
and 50% grain oriented silicon
steel E&I core
laminations, and they used silver enameled wire. All sorts of claims
were
made about the superior sound
because of the exotic
materials but there has never been an ABX test between the Ongaku and
the same
circuit with mere copper wire and all GOSS core, and made by the same
man, and
honestly tested and reviewed.
Audiophiles make outrageous
claims about sound quality. Some
maintain it is the part quality or brand that matters most, with tubes,
capacitors, resistors, cables, solder type, all of which is not the
full story
which should include the choices of tube type, circuit used, and
measured
distortions and the amount of negative feedback and how it is applied.
Rarely
does one ever see an audiophile change output transformers. There are
miles of
plain old copper wire in there. They'll wax lyrical about pure choke
loading to
gain stages, but this often achieves higher distortion than when using
a simple
resistor only. At low typical listening levels there is considerable
iron
caused distortion which resembles the crossover distortion in SS amps
at low
levels. Chokes are only fabulous as I use them with a series
resistance, and
where the signal is high as in the above EL84 driver circuit and where
the Ra
of the tube is very low. Solid state CCS load on the input tube works
better
than any choke plus resistance, but is limited to where signals are low
because
of the fragility of solid state where high voltages lurk. Chokes
without the
series R as well reduce the bandwidth and increase phase shift at
extremes of F
and no good where even a small amount of loop NFB is used.
Without the numbers being
naturally very good, no quantity
of exotic minor parts or materials will make a lemon taste sweet. But
if the
numbers are really good, then some sonic gain might be made with choice
of tube
and capacitor brands.
I find that caps used in amps
make little difference, but do
make a difference in speaker crossovers, and my tip today is that you
should
only use polypropylene capacitors in speakers with high current
capacity such
as "motor start" polypropylene caps, and its best to never use any
bi-polar electrolytics. Most makers don't do this because of cost and
size of
poly caps. I only use polypropylene coupling caps in amps, and I cannot
tell
any improvement has occurred if an alternative brand of polypropylene
cap is
used. But feel free to experiment, it won't do any harm, if you know
what you
are doing.
More pictures.....

Photo 8.
Photo 9.