The above preamp picture was taken in 2000.
It  was initially built in 1994 as my first hi-fi DIY project after a 30 year period
when I was not involved in handcrafting any electronic equipment except
several pairs of speakers.
The original schematic had a 12AX7 SRPP phono input stage with passive
RIAA, then a second phono gain stage with 12AT7 SRPP and a following
line integrated stage with 12AU7 SRPP.

The chassis sides are formed from a 1.2 mm sheet brass channel section
made by a local metalworker, 50mm vertical, 25mm legs top and bottom.
I made up a rectangle with 4 pieces of channel with mitred corners,
locating them at right angles by clamping then to a  pre-cut timber former.
Joins were filed carefully until mitres joined with less than 0.5mm gaps.
Internal angles were fitted internally to each corner and 4 countersunk
4mm nuts and bolts placed to hold corners tight without the timber former.
Corners of brass were then soldered carefully, and filed down for round
corners to eliminate any sharp edges. Linishing with reducing grades of
wet and dry sandpaper removed excess solder and file
marks.

The chassis top for the tube part of the amp is 1.6mm aluminium.
Some changes meant some kitchen formica benchtop material was used
over the top of aluminium, quite OK for a prototype.
There is a 1mm sheet steel box to completely enclose the power supply,
and both the HT and heater transformers are potted to reduce stray magnetic
and electrostatic fields. These measures are important to suppress stray
magnetic fields on a chass where a sensitive phono amp is to be located.

After a couple of years it became necessary to re-build this preamp with
convenient features to facilitate A-B testing of other equipment also developed
before 2006, so two pairs of 12AU7 output buffers  and two dual gain controls
were added to give two pairs of L and R outputs so two pairs of power amps
could be connected for the A-B evaluations. This allows me to keep the levels
for two different power amps exactly the same while using a switch from the
speakers to either power amp outputs.

The original amp had genuine dual mono channels but with ganged volume
controls. There were two PTs and two CLC power supplies, but I found
there was no need for such complexity on a single chassis and just the one
supply with a separate filament transformer is now used. There is one smaller
choke included in a passive CLCRCRC filter for the B+, and a regulated DC
supply was added for all heaters. There is more than enough individual RCRC
filtering of the B+ shown in the power supply schematic to ensure each channel
is working just as separately as they would if they were each completely
monobloc construction.

Schematic 10 Tube preamp, as it was in April 2000.

A few people have emailed me to say they built this preamp, or to get additional
information about the power supply. I did not provide exact power supply details
because I felt anyone building such a preamp should at least be able to design
and build a power supply of their own, and I did have some generic power
supply designs in my power supply pages.
Power supplies do NOT affect the sound quality if built to a reasonable level of
quality.
There is now a suitable power supply schematic further down this page,
see the details for the 2004 10 tube preamp power supply.

In about 1996 I was very favorably influenced by some of the ideas
promoted by Allen Wright who wrote a provocative book,
'The Preamp Cookbook' in 1988 to make a little money and to spread
the gospel about preamp construction with tubes.

Allen departed from this world in February 2011, but his website is still worth
a read for information about preamps at http://www.vacuumstate.com

I began to think seriously about how to make a suitable low noise preamp for a
moving coil cartridge that I thought I should try because many people had told
me that generally MC were better than MM.

Allen began in 1988 to use a high transconductance j-fet such as the 2SK147
in cascode with a triode tube to give a gain block of up to about 800x at quite
low THD and with about 20dB less noise than when using a well chosen
12AX7 in a well made amp. Allen's Four Valve Preamp of 1988 was a
landmark design where he combined the best aspects of a j-fet with tubes.
The j-fets such as 2SK369, 2SK147 have low noise, high gm, and are very
suitable for cascode use with tubes. The j-fet only has to amplify the tiny
input signal about 20 times, and the distortion is low even without any NFB
since the signal is so low. Triodes then can take over amplifying the signal
to a higher level, and since their dynamic range is 20 times that of a j-fet
the distortion is still low at signal levels below 10Vrms, so loop NFB need
not be used for RIAA eq or for any other reason.

Although my original 10 tube preamp was an excellent performer with a
Shure V15 MM cartridge with high output, I eventually bought myself a
Denon 103D moving coil and replaced the Shure. The Denon has only
about 1/16 of the Shure output, ( 0.4mV ) and the 12AX7 is just too
noisy no matter how carefully one selects the brand of tube.
I found  that even a paralleled 6DJ8 was too noisy even though calculations
and theory suggested this should not be so.
So I decided to try Allen's basic idea of cascode phono input stage.

Since 2004, my prototype preamp has a much changed
schematic with many improvements :-

Each channel,
Phono input, 2SK369 j-fet driving into cathode of SE 6EJ7 frame grid
pentode wired as a triode used in a 'cascode' circuit.
RIAA passive filter,
12AT7 µ follower in second phono gain stage,
12AU7 µ follower line level gain stage which is fully bypassable,
12AU7 µ follower tone control stage with Baxandal feedback circuit,
giving gain = approx 0.94 with +/- 9 dB of max cut and boost at 100Hz
and 10 kHz.
The tone stage is fully bypassable, ( and is seldom used except when testing
speakers and sources to gain a very approximate idea of response problems ).
Balance is controlled by carbon track pots immediately prior to the output
buffer stages.
Gain is controlled by carbon track pots placed immediately before the output
cathode follower buffers which are 2AU7 with transistor constant current sinks.

Most of the time the line stage and tone controls are not used, with the input
signal applied directly to the gain controls, and the following buffers, so that
long cables to power amps across the room may be used, which saves having
a remote volume control.

For each channel, this 10 tube preamp has one MC input, which can be
used for MM input if the low gain circuit setting is chosen by removing a
bypass capacitor in the j-fet source circuit. 
There are 4 line level inputs, and one record output. 
There are two volume controls per channel with separately buffered outputs.

Slightly varied 10 tube pre-amp, 2004.

The 10 tube preamp of 2004 above has undegone some changes from the
picture right at the top of the page taken in 2001.
The switches and some RCA sockets originally mounted in the top plane
of the chassis in 2001 have all been moved to either the rear panel or front
panel, and their quality is improved since I used some high quality NOS
rotary wafer switches instead of the cheapest Taiwan made types I began
with.
The holes left behind have been covered over with well glued kitchen bench
top plastic laminate with a blue pattern, although it looks grey in the picture;
photos often don't tell the full story, and this was taken years before I got
a digital camera.

The schematics of the 10 Tube Preamp 2004 are in five separate sheets
which follow :-

Block diagram Sheet 1.

 
Phono amp, Sheet 2.

If you have ever examined the late Allen Wright's website, you may have come
across the "white paper" .pdf which has Allen telling us the secrets about
getting good sound from vinyl. There is a good read at :-
http://www.vacuumstate.com/fileupload/SP_15_Article.pdf

All good gurus or preachers intersted in extending the little pleasures in our lives
should be granted a fair hearing, but what they say may not all be considered the
absolute true gospel, because nobody is 100% correct about everything.

The underpinning foundation for Allen's phono stages is the use of the "cascode"
gain block with two active elements in series in what is sometimes called a
"totem-pole" series arrangement. Cascode is really the use of the anode
output of one triode to drive the cathode of a second triode with a "grounded grid".

Cascode is not the same as cascade.

To explain cascode just a little.....

Anyone could build this exact schematic and it might be useful for the input
stage of a Moving Magnet phono cartridge.

The gain achieved is not spectacular at 91x compared to having two tubes
in cascade where the anode output of one triode feeds the high Z grid input
of a second triode, giving gain of perhaps 780x. Bean counters always settle
for the cascade, because there's more gain per dollar.

But the cascade has some saving features. It has very low Miller capacitance
and gives ness noise than say a pentode such as an EF86 as they were often
used in what were rather poor MM input stages.
The gain of the cascode is much affected by the load because the Ra of V1b
or effective output resistance becomes high. It may be calculated :-
Ra' = Ra + ( [ µ+1 ] x Ra ) = 5k0 + ( [ 30 + 1 ] x 5k0 ) = 160k.
The RL delivering DC to V1b anode is 27k, so total Rout = Ra' // RLdc = 23.1k.
When calculating R&C values for a following RIAA network load this actual Rout
must be factored in correctly. Most ppl get this part of circuit design hopelessly
wrong.

The cascode input with fairly high Gm triodes like the 6DJ8 can give about 1/2
the noise of a 12AX7. Unavoidable grid input noise is proportional to 1/sq.rt Gm.
In other words, if you have 4 x 6DJ8 1/2 triodes in parallel, ie 2 full tubes, then
Gm would be 4 times higher, and noise would be 1/2 that of just 1/2 of a 6DJ8
as seen in the above basic phono stage. So for the cost of using 4 times the
number of triodes, noise only reduces -6dB, and all this is theory, because
in practice samples of tubes vary in their noise productions.

But for Moving Coil cartridges, the amp must be MUCH more silent than for
MM, and where 3uV might be OK at the grid input for an MM cart giving
3mV output for SNR = -60dB.
For an MC cart which makes just 0.3mV output, the SNR must still be -60dB
so noise at the amp should be under 0.3uV, and this is mostly impossible
to achieve using any arrangement of vacuum tubes. Not only that, tube noise
comprises steady hissing plus burst noise, plus lots of LF rumbles, and for
phono the LF gain with RIAA filter tends to allow the tube generated LF
trash pass through so hence all MM amps with tubes tend to have some amp
noise. As tubes age, their noise generation increases, and so one has to expect
to replace phono input tubes if used more often. However, with an MM cart
like Shure V15, rated for 5mV output, I found a 12AX7 used in a µ-follower
configuration to be excellent with regard to noise, and not any worse than
having a paralleled 6DJ8 input.
The MC cartridge was invented by Denon in about 1949, and their first
was about equal to a current production Denon 103R which I am using and
which puts out around 0.4mV. Back in 1949, the only way to get a good SNR
was to use a transformer to increase the 0.4mV to say 4mV, ie, use a 1:10 step
up transformer. The Rout of the typical MC is less than 20 ohms, so when
voltage is stepped up x 10, then impedance transformation gives signal source
R = 100 times higher at say 2k0. But such impedance is still quite low and noise
from a 20 ohm source may be 0.2uV, and the SNR at the cartridge is -66dB.
At the amp, the cartridge noise becomes 2uV, and the tube input may contribute
2uV, giving a total of 2.8uV, but the signal is 4mV, so SNR = -63dB.
By the time the signal goes right through the phono amp, expect SNR = -60dB.
The MC was the best thing to use in broadcast radio stations which could
afford to replace more expensive MC often, and where the high grade TTs
and amps could be frequently serviced, and used by adults only, and with
a steady hand. MC are fragile, but in general they sound better than MM.

The use of a high Gm j-fet such as 2SK369 and others solves the noise problem
and avoids the need for a step up transformer. Typical input gate noise of 2SK369
is 0.1uV.
The characteristics of a 2SK369 at an ideal Id = 5mA dc are like a pentode tube
having Gm = 40mA/V, Rd = 80,000 ohms, and µ = 3,200, but unlike the pentode,
the noise is 20 times lower !

When using a 2SK369 to work into V1b cathode input resistance of 751 ohms,
j-fet gain may be calculated as fet Gm x 751 = 0.04 x 751 = 30 times.
The V1b gain would remain unchanged at 23.3, so max gain overall becomes
699x. This then competes very well with the tubed cascade circuit, and offers
extremely low noise.
In fact, in practice, the gain of 699 would be too high. Total maximum gain
needed for an MC amp is 10,000x, or +80dB. This means that LF bass signals
at 0.03mV will become 300mV at the output, and 1kHz MC signals of 0.3mV
will also become 300mV allowing for the RIAA filter effect of reducing the
1kHz by 0.1x or -20dB. At the MC cart, signals at 20kHz might be 3mV,
and RIAA cuts these by -40dB so that they too become 300mV at the phono
amp output. So if the cascode stage gain = 250x, then following stage needs
to have gain of 40x to get a total of 10,000x.
The best place to put the RIAA filter is between the cascode stage and
subsequent gain stage.
In the 10 Tube Preamp phono section the gain of the fet + triode cascode
stage has been reduced by simple use of fet source resistors giving local current
negative feedback.

If the C coupling from the low impedance MC cart is not a high enough value,
then low frequency noise generated in the 47k biasing R1 will not be shunted by
the low MC impedance, but appear at the j-fet input gate and will be amplified
by the large amount of gain this amplifier has at low F since it is a phono amp
with RIAA de-emphasis curve filtering, which basically allows all the LF to
pass through but attenuates F above 50Hz.

Allen's original Audio Fidelity Integrated 4VP first made in 1988 used a 2SK147
in cascode with 1/2 12AT7, and this was followed by ordinary common
cathode amp using the other 1/2 12AT7, and global NFB was used for
RIAA equalisation. 2SK147 has been discontinued by Hitachi, but 2SK369
is a fine substitute, and costs less than $0.20c.
The line level stage in the 1988 preamp also used a cascode stage with 2SK147
with 1/2 12AU7. The whole preamp with its cascode line stage was considered
by some people to have too much gain overall, so that if you used a high output
cartridge, sensitive power amp, and sensitive speakers, the system was became
unusable because it would roar at you even when the volume control was set to
a very low level. I like to see variable phono gain and to use a fully bypassable
line level gain stage.

I also very much like the use of large electrolytic filter caps for rails of phono
stages to best stabilize the rail to prevent very low F variations in B+ being
amplified by the high gain at LF. Hence the C4 = 470uF.
There is  also C16, a 2uF polyester cap with very short leads between the B+
at between V1 and V2A anode supplies and 0V, to make sure the bypassing is
effective to high RF. I found that the circuit as shown tended to oscillate at
about 100MHz if the circuit board area for the j-fet plus V1 was larger than
50mm x 50mm in total area, and if leads to capacitors and track lengths were
too long, and even if there has been a strict adherence to star earthing.
Phono stages like this one need to be designed as if they were RF circuits;
small is beautiful in this case.
Bypass capacitors C1, C5,6,7,8,9 are very important. I also supplied the DC
to the heaters to V1 via RF chokes and bypassed the heater wiring well with
C to prevent parasitic oscillations. The RF chokes are 0.8mm enamel wire
wound as solenoids of one layer along a 30mm length if 10mm ferrite rod.

Tubes such as high transconductance frame grid pentodes like the 6EJ7 make
fabulous triodes for my purpose, but whatever you do, don't try to use them
as pentodes driven by a j-fet as shown. The extremely high resultant gain
WILL be impossibly unstable at some high RF and the tube will be weirdly
microphonic, and you'd think you had the bells of St Mary's Cathedral
connected to your amp.

V2a and V2b form a µ-follower, or bootstrapped follower as it used to be called
when invented in about 1943. This type of gain stage is my favorite because
of the clear sound, the nice measurements and simplicity. The efficiency is
good because there is no wasted current in separate anode resistances to supply
Ia dc or in cathode resistors to sink Ik dc in a separate cathode follower.
The set up as shown with a 12AT7 gives a very healthy gain of 36x or more even
with the V2a cathode R21 unbypassed. The load seen by the V2a anode is
approximately the open loop gain of the top tube x R20, plus R19 in parallel.
Effectively, the RL of V2a = 166kohms, which is over 10 x Ra for the 12AT7,
so THD is extremely low, and the much maligned 12AT7 can sing as sweetly as
any tube can.

I first tried this amp set for low gain to suit the MM ShureV15.
Then I changed to MC Denon 103D with the gain set high for MC.
After only 3 bars of Mozart I realized how much better MC could be.
I have retired the ShureV15.

In 2005, I developed a superior cascode MC amp circuit which is described in the
page http://www.turneraudio.com.au/preamp-rocket-phono-2005.htm

Integrated line level and tone control, Sheet 3.

This is the line gain stage and tone control stages. I like the 12AU7 as a preamp tube,
and it really sings in µ follower mode. There is some local shunt NFB to control
and reduce the gain to a sensible level since even a low µ triode like 12AU7 has
too much gain for most line level gain stages.
The balance control is incorporated in the shunt NFB path.

The line gain stage and tone control stage are both each bypassable to leave a
minimum length of signal path consisting of just volume control potentiometer
feeding output cathode followers.

Balance control is only possible when the line gain stage is used. I find that with
CD source I rarely ever use line stage gain, and I find balance control is never
needed.

I lent this amp to a customer for a fortnight and he used it for a week with tone
controls included, and he needed to be told that it could be deleted and he could
not tell when the tone control stage was included in the signal path or if switched
out of the circuit.

Gain pots, Ouput-buffers, Sheet 4.

The above pairs of output cathode follower buffers have transistor CCS 
to reduce any un-necessary loading effect by resistors carryin DC, and to
maximize open loop gain of the 12AU7, and hence minimize the thd, and
allow the load powered by the cathode follower to be a lower ohm value
than would otherwise be used.

Power Supply, Sheet 5.

Note that there is NO B+ regulation.
The anode current consumed by each channel = 20mA approximately, so with a
total of about 44mAdc and B+ across C4 = 290Vdc, then B+ power consumed
is about only 12 watts. This low amount of B+ power is easily filtered by
passive C, L and R components. Using  active tube regulation or active solid
state regulation tends to add a lot to complexity and I have found such things
create unwanted heat and tend to be unreliable. I sure DO NOT believe in silly
notions that tube rectifiers sound better!
This amount of B+ power is similar to that required by an old AM radio, and
in fact the type of power transformer found in many an old AM radio will
provide the B+ power needed by this preamplifier.

However, the 25VA radio transformer will probably not have the necessary
heater windings to enable 12.6Vdc x 1.8Adc to be generated, so anyone
building this amp should use a single transformer specially wound, or have
a radio transformer for the B+ and an auxiliary transformer for the heater
voltage with a 17Vac to 19Vac x 2A winding, with a 40VA rating.
Most of the power drawn by the amplifier is filament heater power.
Hammond Engineering of Canada also supply a range of suitable power
transformers.
My 10 tube preamp has the power supply within a steel box where there
are two power transformers as shown on the above schematic.
B+ is from a potted NOS Navy spec transformer and the low voltage
transformer is from Jaycar, an Australian parts supplier, and it is a general
purpose transformer that has several taps and allows up to 2A at 30Vrms
output. I potted this LV transformer in a steel can filled with dry sand to keep
it mechanically quiet and reduce stray magnetic fields.

The combined effect of the potting for both transformers and mild steel sheet
box for the power supply does reduce the stray magnetic fields just enough
to allow the very magnetically sensitive phono input circuitry to be placed at
only 450mm away from the power supply on the same chassis.
It is always better to use a remote power supply and umbilical cable, but in
my case I got away with the PS being close to the phono stage because of
the TWO layers of magnetic shielding. I used brass and aluminum for the main
parts of the chassis which looks nice, but plain thick mild steel is actually better,
especially if other devices such as CD players in plastic cases are located near the
phono amp, say on the shelf below the preamp.
I have seen even expensive CD players radiate stray magnetic fields that migrate
into phono amps like this and cause hum.

There is about 0.2Vrms of 100Hz hum at C4, and this is reduced by a factor of
0.001 by L1 and C5, and then again by the R2/C7 R3/C6 filters before each
channel by a factor of 0.015. Additional RC filters at the top right hand side
further reduce hum to utterly negligible levels.
The time constants of all the R & C filtering give good B+ rail stability that is
sufficiently immune to mains level fluctuations.
In fact, the amp can be turned off, then back on again after 2 seconds repeatedly
and there are no audible change to the sound, or slow wobbles in speaker
cones since so much stored energy is contained in the electrolytic capacitors.
Regulation of the B+ was not needed because even in phono mode the LF
noise was utterly mimimal, and since below 3Hz the total amp response reduces
steeply due to so many RC couplings between stages.

However, Q5, Q5 form a regulator circuit for the totally dc heater supply which
is biased at about +56Vdc so that the dc voltage difference between heaters
and cathodes does not exceed the ratings of 90Vdc and arcing or current leakage
from cathode to heater circuits is prevented. Such leakage is never fully preventable
regardless of the potential difference between cathodes and heaters. I have seen
old input tubes in amplifiers begin to have cathode-heater leakage which causes
hum to develop where the heaters have AC heating current. I have also seen
tube get a short circuit between heaters and cathode, but such faults are uncommon.
They are very easy to fix - just replace the tube.
To achieve good ripple rejection in the 1.8 amp dc heater supply, the easiest way
is to use a regulator rather than have an ungainly large choke and huge capacitors.
But please feel free though to use as much L and C and or R as you can afford.....

The overall performance...
The bandwidth is 3Hz to 100kHz and the preamp will drive any known solid
state or tube power amp inputs.
The output impedance of the cathode followers will allow the use of 10 metre
long interconnect cables and direct connection of the 'record out' select switch
pole to a recording device such as a sound card, so the phono preamp stage can
be directly connected to the sound card if required.

For those interested in having a preamp built, tube choices will be from my
very limited NOS stocks of tubes after full testing for noise and microphony.
Excellent sounding NOS Siemans tubes will be happily supplied providing you
can afford them, and find a supplier. NOS Siemans produced what I think
give the best subjective sound quality in terms of dynamics, detail,
bass/treble balance, vocals, musicality, bloom and warmth.
The subjective quality of tube amplifiers differ with tube choice.
All the preamps built have at least 5Hz to 100kHz bandwidth, and typically
measure 0.02% thd at 1Vrms output. Usually less than 0.1Vrms is needed
to drive power amps at normal listening levels, and thd/imd from the SET
pure class A triode circuitry is proportional to output voltage, so at low
output voltage levels the thd/imd is so low it is quite inaudible.
The measurements do not indicate a correlation to the marvelous sound
quality heard, so small triodes have a happily mysterious character within,
because triodes which measure similarly at thd/imd levels <0.02% can give
different sonic signatures.
For example, in a test with 3 other friends present, we all agreed that in a line
stage being tested for the afternoon, The order of preference for the 6CG7 tubes
tried in a line stage placed Siemans NOS first, Australian Miniwatt NOS second
by a nose, NOS Mullard third by a chest, and recently made russian EH6CG7
a long body length behind, and struggling.
But for difficult female vocalists, ( my respects to Celine Dion ), the Mullards
were the most beneficial to the sound, although nothing will save your ears from
Kylie Minogue, except the "off" switch.
We don't know why the EH6CG7 sounded so rough; perhaps the batch was
made on friday when the staff are impatient to be off to their weekend dacha
lodge in the forest where a bottle or two of Vodka is handy.....

With all my amps, I only use point to point wiring between tag strips I often
make myself.

Sometimes I use circuit boards which have 1.2mm solid copper wire links
hooked through pre-drilled holes on an 8mm grid pattern. Component leads
are surface soldered to wire links. Such boards offer reliable operation and
allow boards to flex or bend without getting a cracked dry joint seen so
often in printed circuit boards with thin tracks and small amounts of solder
at joints.
I use mainly Wima polypropylene capacitors for the signal path couplings
and 1% x 1 watt Welwyn metal film resistors or good quality 1W rated metal
film R from reputable makers in Taiwan.

I do NOT believe that Auricaps or many other brands of coupling capacitors
sound any better. I did trial a line stage in 2005 where Wimas were in one
channel and Auricaps were in the other. A customer friend and I used the
same mono sound source through each channel in turn with me trying to
trick my friend when I asked him to say which channel was better. After
about 6 changes A to B at identical levels and using 2 different recordings,
my friend could not pick any change or state any preference which was better
than chance, ie, he liked the Wimas just as much as the Auricaps.
I myself certainly could not hear any difference let alone a "better" sound
with Auricaps.

However, my friend proceeded to have me replace all the Wimas in his
preamps and power amps with Auricaps.

I will always consider that my customers are always right, and work as directed,
but I don't myself think I am missing out on better sound because I have not
used Auricaps. 

I don't believe in many myths about special parts. The circuit topology and design
and careful tube choice are far more important to the sonic signature.
The main reason to use better quality parts is reliablity and tolerance quality.
For example a cheap Taiwan made dual gain potentiometer may have 15% different
levels at the -20dB gain setting but otherwise work perfectly for 3 years.
Then it begins to make noises. So I won't use less than an Alps Black pot which
is 20dB more expensive but they will last 40 years with good matching on their tracks.
Best value switched volume attenuators and source switches are made by DACT.

Should you have any further special requirements, please feel free to ask me.

Back to Index Page