Fig 4. Block diagram of 2004 amp, Sheet 1.
Fig 5. Phono amp, Sheet 2.
Sheets 1 to 5 show the 2004 preamp :-
Sheet 1 = general block diagram for the preamp.
Sheet 2 = Phono input, Q1 2SK369 j-fet driving into cathode of SE V1 6EJ7
frame grid pentode wired as a triode used as "grounded grid triode" in a
V1 6EJ7 drives a passive RIAA eq filter.
V2a+b = 12AT7 µ-follower second phono gain stage. The output from V2b
cathode is to an input at 2 pole 6 position rotary wafer switch.
Sheet 3, 4 and 5 appear below.
Sheet 2 phono amp.
In 1996 I was very favorably influenced by 'The Preamp Cookbook' written in 1988
by Allen Wright who had discovered that vacuum tubes were indeed wonderful
for analog audio signals. 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
Allen's 1988 idea for phono input was to use a high transconductance j-fet 2SK147
to drive a triode in cascode. You could get gain up up to 800x at quite low THD and
with about -20dB less noise than when using a well chosen 12AX7.
Allen's Four Valve Preamp of 1988 was a landmark design where he combined
the best aspects of a j-fet with tubes. In 1988 Vinyl was still extremely popular despite
digital technology invasions.
The Hitachi 2SK147 with Pdd rating 0.6W is no longer made but has identical properties
to Toshiba 2SK369 except for Pdd of 0.4W. These delicate TO92 package devices
are rated for max Vds = 40Vpk and max Id = 30mApk and for class A idle condition
Eds should not exceed 20Vdc, and Idc should not exceed 7mAdc which gives
Pdd = 0.14Watts. The input gate noise is less than 0.1uV, lower than a good signal
triode with 1uV - if you have a good sample.
2SK369 is available from
http://wescomponents.com.au at Ashfield Sydney for about $1.20 each.
Unlike bjts which have low noise at very low collector currents of say 0.2mAdc,
2SK369 has its low noise at 5mAdc. Gm = 40mA/V, and Rd = 80k, so its signal
parameters are like a pentode with Gm 40mA/V, Ra 80k and µ = 3,200.
To get near these figures using tubes, you would need about 12 x 6AU6 in
parallel, but because of the low Vd-s rating of 40V max Vo = about 10Vrms.
Because of high Gm, the 2SK369 is an ideal current source to drive the cathode
of a grounded grid triode in an input stage. The triode cathode input resistance
is usually its anode load / voltage gain and is usually less than 1k0.
with load of 1k0, 2SK369 has open loop gain of 1,000 x 0.04 = 40, and with phono
cartridge signal of say 0.5mV its Vdrain Vo is 20mV at low THD and very low noise.
Following the work by j-fet, the triode stages handle higher signal levels without high
THD, all without any GNFB. Therefore a low
noise MC phono amp can be made without using any loop NFB for RIAA eq or for any
If you have
ever examined the late Allen Wright's website, you may have come
the "white paper" which has Allen telling us the secrets about getting good sound
from vinyl. There is a good read at :-
This link still works in 2015.
not the same as cascade. Cascade is where there are two common
cathode gain stages with the second tube grid driven by the anode of first tube.
Cascode has two tubes or other devices in series, and some basic info about
this is here.....
Fig 6. Cascode basics with 2 triodes......
build this exact schematic which could be used for the input
for Moving Magnet phono cartridge. The B+ must be rather high at +338Vdc
because we have 2 triodes plus a dc carrying RLa load in series.
If anyone does build this, they must bias the heater Vdc supply at about +65Vdc
so that the max Vdc rating between heaters and cathodes, ( often 90Vdc )
is not exceeded.
Overall cascode gain
is not large at 102. This was calculated based on the given
Gm and Ra for 1/2 a 6DJ8 with the Iadc = 3.5mA and Ea at +120V.
So expect gain in your circuit between 90 and 110, and variation is because Gm
and Ra vary slightly for sample tube.
Output Rout from V1b anode is determined mostly by RLa load 17k5, because
the V1b Ra becomes effectively high, about 175k, because its Rk is the Ra
V1b acts like a current source of 175k working on all DC and C&R coupled loads
in parallel. Cascode gain is roughly proportional to the total RLa load values.
Effective V1b Ra' may be calculated :-
Ra' = Ra + ( [ µ + 1 ] x Rk ), where Ra is for V1b and Rk = Ra of bottom tube.
So in this case, Ra' = 5k0 + ( [ 33 + 1 ] x 5k0 ) = 175k. If V1a had its Rk of 1k0
unbypassed, then V1 Ra' effective = Ra + ( [ µ+1 ] x Rk ) = 5k0 + 33 x 5k = 39k
and the Ra' for V1b will increase to 1,331k, or 1.33Meg, and gain of V1b
becomes closely proportional when V1a total RLa is only 50k.
Rout from cascode = Ra' // RLdc, in this case 175k // 27k = 23k4.
If V1a Rk was unbypassed, Rout at V1b anode = 1,331k // 27k = 26.4k.
My triode cascode circuit shows cap coupled R4 50k. This R value is typical
of a the Rin to an RIAA network 1kHz. Passive RIAA networks have varying
input Z, with higher Z at 20Hz and slightly low Z at 10kHz.
If the cascode stage shown does drive a passive RIAA network then R&C values
chosen for the network must include for effective Rout of the stage, in this case
Most inexperienced ppl get this part of circuit design hopelessly wrong!
The SNR in triode phono circuits is mainly determined by the unavoidable grid
input noise and shot noise which is rarely less than 1uV in good samples of
most small signal triodes such as 1/2 12AX7 or 6DJ8. If an MM cart has
Vo = 4mV, then SNR = 1uV / 4mV = 0.001 / 4 = 0.00025 = -72dB. Much of
the triode noise spectrum will be below 1kHz. Noise in other parts of a
possible preamp may reduce this SNR by +6dB to - 66dB. The noise of the
triodes is just below noise of an unmodulated vinyl groove so that when amp
gain is turned up with cart in a quiet groove you hear the vacant groove, and
with very little amp noise able to be heard. With the MM cart, the gain will rarely
ever need to be turned up high enough to hear the vacant groove noise, hence
vinyl is an acceptable true hi-fi source providing the grooves are clean and free
of muck which produces crackles & pops. The vacant groove noise probably
has some noise from the tape source used when cutting the record. A quiet
recording venue or hall where there is total audience silence may still have noise.
Grid input noise is theoretically inversely proportional to square root of Gm.
The higher the Gm, the lower the noise. So if 2 x 1/2 6DJ8 are paralleled, noise
reduces x 0.707 ( in theory ). The 6DJ8 has much higher Gm than 12AX7, yet in
practice noise is often no lower than many 12AX7.
With 2SK369 j-fet, the gate input noise is commonly below 0.1uV, and the vacant
vinyl groove noise dominates, even when gain is turned up.
I've never known a phono amp with tube input to give acceptably low noise with any
MC cart. For silent periods during classical work you don't want to hear sausages
on a barbecue in the background. Phono input tubes age like all others and their
noise can increase 4 times in 4 years of constant use. Gas entry over time is the
main cause. .
Cascode circuits with 6DJ8, 12AT7, 6AQ8, 6BQ7 were very useful in RF circuits
for TV and FM tuners. The medium low input noise of these triodes exists in both
cascode and cascade triode stages. Usually the triode noise is noticeably
lower than in a pentode because of the "partition noise" relating to the g2 screen.
This disappears if the pentode is triode strapped. Almost anything is better than
an EF86 used in a phono stage as it was done in 101 preamps of the 1950s.
Following the first use of vinyl for 33.33RPM records in 1948, and 45RPM in 1949,
Denon invented the famous DL-103 Moving Coil magnetic cartridge in 1962.
It is still being made, and I have a DL-103 rated for 0.37mV output which I have
used since 2004. I found it took only 5 minutes use to discover it made better
music than a Shure V15. In the 1960s a good SNR with MC needed a step up
transformer to increase the 0.4mV to say 4mV, ie, 1:10 step up.
Data for Denon MC, DL-103R :-
Stylus: 16.5 µm diamond spherical tip.
Cantilever: Aluminum.Frequency Response: 20 ~ 45 kHz.
Output: 0.3 mV at 50 mm/s.
Output Impedance: 40 Ω.
Load Impedance: 100 Ω.
Channel Separation: Over 25 dB at 1 kHz.
Compliance: 5 x 10–6 cm/dyne (100 Hz).
Tracking Force: 2.3 ~ 2.7g (± 0.3g).
Weight: 8.5 grams.
If the Denon has 100r load and no trans, the total R = 28r, and noise will be much
less than any triode input noise. The j-fet with high Gm and very low noise avoids
ever needing a step up transformer.
Typical gate input noise of 2SK369 is mainly HF hiss with much less rumble and
LF sputter you get with tubes. So with RIAA filter reducing all input signals
by roughly 6dB/octave above 50Hz, the majority of j-fet noise is reduced, and
output amp noise is MUCH lower than any tube could provide.
High Gm j-fets such as 2SK369 are nearly always used with drain load R much
less than the internal dynamic Rd of say 80k. J-fet loading is similar to loading
pentodes where a load for EF86 might be 150k, while its Ra = 1M0 or more.
Because j-fets are "transconductance devices" with gates drawing no current and
drain current altered by a change of gate voltage, the parameters for a j-fet could
be specified similarly to tubes in terms of µ, Gm and Rd. While this is convenient
for triodes, j-fet Rd is affected by capacitances and Rd is really only the value we
find at low F where XC is many more ohms than Rd. In tubes, the amplification
factor µ = Gm x Ra, and in 2SK369, if Gm = 40mA/V, and Rd = 80k, then
µ = 0.04 x 80,000 = 3,200. Gain for tubes = µ x RL / ( RL+Ra ), or, if we neglect µ
gain = Gm x ( RL // Ra ). If Ra is many times RL then it may be neglected.
The effective Rd' of the j-fet is increased if there is an unbypassed Rs.
If Rs = 50r, then Rd' = 80k + 3,200 x 0.05k = 240k, very hi Z.
In a cascode circuit driving a triode, the high Rk is many times higher than the
triode RLa, so any THD produced by triode is much reduced by the local current FB.
The THD of the cascode stage is mainly produced by the input device. This THD
is mainly 2H which is highest when drain load < Ed / Id. If 2SK369 has Ed = +12Vdc
and Idc = 5mA, Ed / Id = 2k4. The open loop gain is 96.
If the drain load is increased I found 2H declines to a null with RLd = 6k3, and then
increases with some 3H as RLd increases. Typical THD with RLd = 2k4 increases
at a rate of 1% per 0.4mA Ia so at 7Vrms with 2k4 you may get THD = 7%.
This is far higher than what you get with a pentode at Vo 7Vrms with typical load
of 100k. But if gate Vin = 0.2mV, load current change = 0.0002V x 40mA/V = 8mA,
so expect 2H = 1% x 0.008 / 0.4 = 0.02%, and if there is 10dB of current FB with an
unbypassed Rs then THD < 0.008%. Let us not worry about such tiny things.
Calculations for 2SK369 driving the triode strapped 6EJ7 in Sheet 2 :-
The 6EJ7 triode data for Ea = +130V, Ia = 5.4mAdc give Ra = 10k6,
Gm = 5.2mA/V, µ = 55.
The RLa total at 1kHz = 15.66k.
6EJ7 triode gain A = µ + 1 / ( [ Ra / RL ] + 1 ) = 56 / ( [ 10.6 / 15.66 ] + 1 ) = 33.4.
Suppose Va = 57mVac. Then Vk input = 57mV/33.4 = 1.7mVac. The signal load
current Id = Ik = Va / RLa = 0.057V / 15.66k = 0.00364mA.
Therefore cathode input R = Vk / Ik = 0.0017V / 0.00364mA = 467r.
The cathode Rin = drain load for 2SK369. Assume link 1 Sheet 2 is used as shown.
This gives Rs = 50r. The total loading for 2SK369 = 469r + 50r = 519r.
The j-fet gain = Gm x RL = 0.04 x 519r = 20.76. Iac in triode and j-fet = 0.00364mA.
Vac across Rs 50r = 50r x 0.00364mA = 0.182 mV. Total Vds = Vd + Vs =
1.7mV+ 0.182mV = 1.882mV. Vgs = Vds / Gain = 1.882mV / 20.76 = 0.091mV.
Therefore, to make 57mV at V1 anode, we need Vgs = 0.091mV so the j-fet
open loop gain = 57mV / 0.091mV = 626.
We have 0.182Vac across 50r, so Vg-0V gate input = 0.182 + 0.091 = 0.273mV.
The 50r gives 10dB local current FB and final overall gain = 57mV / 0.273mV
= 209. Gain reduction factor = closed loop A / open loop A = 209 / 626 = 0.33
which is about -10dB NFB.
To verify these calculations by measurement of Sheet 2 depends on accurate
measurements of tiny Vac prone to noise and high inaccuracy. But you can what
you can expect. If you build two channels, and the same input signal is moved
from one channel to the other, and you have equal Vout from each triode, then you
have two good channels, and I have found that a couple of randomly chosen
2SK369 j-fets from a batch of 10 purchased will often give channels matched
within +/- 0.5dB, quite closely.
The 6EJ7 could easily produce Va = 5.7Vac at low THD. With all Vac 100
times higher than might appear in normal working, there is a better chance of
measuring Vac input and gain accurately.
Now there's a quicker way through this mass of figures. In cascode circuits where
bottom device is a current generator controlled by input voltage and its Ra or Rd
is well above 100 times its RL, the overall gain may be more easily calculated :-
Gain A = RLa of top triode x Gm of bottom device. How simple!
We have RLa = 15.66k, and j-fet Gm = 40mA/V, so gain = 15,660 x 0.04 = 626.
This is the same as all those figures above calculated. If you use other tubes for V1.
6BX6 or 6AU6 in triode, or paralleled twin triodes like 6DJ8, 12AU7, 12AT7, the
gain remains almost identical. 12AX7 or 12AY7 would be unsuitable because
when paralleled they still could not have idle Idc = 5mA without operating in
grid current regions to give high THD.
There are LF poles in Sheet 2 with C2 9uF+47k 0.37Hz, C10 0.47uF+470k 0.72Hz,
C13+1M 0.34Hz, C14 0.47uF+1M 0.34Hz. All these poles give theoretical LF
cut off of -3dB at about 2Hz. Above 2Hz, the sine wave response at Vo should
rise to a flat line until about 10kHz when a slow roll off rate begins. The amp gave
me the flattest sine wave response using R&C values as shown. The trimming R
in RIAA filter have been optimized to include possible capacitor values being 5%
incorrect. Most better C brands made after 1990 have C within +/- 2%, and R
values of 1% are often better than 1%.
Fig 7. Basics about SRPP compared to µ-follower.
Fig 7 shows SRPP and µ-follower gain stages using 12AT7.
The SRPP is similar to V2, in Fig 2, ( 2000 preamp ) above.
The µ-follower is similar to V2, in Fig 5, Sheet 2, 2004, ( 2004 preamp ) above.
So how do the two stages compare? which one is better?
The 2000 SRPP has highest Rout and THD and slightly lower gain.
We may conclude that where the RL > 2 Ra then the µ-follower will always has
lowest Rout and THD, and highest gain.
For the SRPP, if R3 is unbypassed, the 1k5 gives considerable local current FB
and I predict gain = 17. But the LCFB will reduced THD by -6dB.
The µ-follower can have its R8 bypassed for much lower Rout than SRPP.
Because the anode load of V3a is also higher than for V2a, V3a has low THD.
Below, in Fig 8 there are two more µ-follower stages used for optional gain
and tone control stages. These show some variation but you may conclude
the µ-follower is better than SRPP.
Both SRPP AND µ-follower are better than having normal common cathode
gain stage followed by a direct coupled cathode follower, for application
where the stages power the known internal amp loadings. With the series
connected triodes there is a large saving in anode power to the 2 triodes
because there is no DC power lost in an anode to B+ resistance or cathode
to 0V resistance for cathode follower. With so many triodes, it all adds up.
With series triodes, each triode gives an active load to the other.
I circuit as shown tended
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 with good "star" earthing.
Phono stages with high Gm devices must be treated like RF circuits, and "small
is beautiful" in this case, and expect to have to at least have all electros bypassed
with plastic caps and perhaps use ceramic 0.1uF plus have leads all shorter than
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.
In 2005, I
developed a superior cascode MC amp circuit which is described
Fig 8. Integrated
line level and tone control,
Fig 8, Sheet 3.
V3a&b is 12AU7 µ follower line level gain stage which can be switched in or out of circuit.
This allows input signal from line inputs or phono amp to connect directly to the gain control
pots or have signal increased +15dB or +10dB approx. I found it useful with low sensitivity
power amps and speakers, or where line levels was high ( from CD player ) or low, ( from
old AM/FM tuner etc ).
V4a&b is 12AU7 µ follower tone control stage with Baxandal feedback
circuit and able to
be switched in or out of circuit. Gain = approx 0.94 with +/- 9 dB of max cut and boost
at 100Hz and 10 kHz. This stage is seldom used except when testing speakers and
sources to gain a very approximate idea of response problems. Some vinyl or AM/FM
radio sources can benefit with tone controls.
Balance is controlled by pots prior to the line level gain stage. I saw little reason to ever
use balance controls and when the line gain stage is switched out there is no balance
control. All CD I have heard have acceptable balance.
The volume is controlled by pots placed immediately before 12AU7 cathode follower output
buffers which use cathode current sinks with bjts, to enable optimal Idc in triodes for high Vo
and low THD.
Most of the time the line stage and tone controls are not used. So line input from RCA
input terminals may be applied directly to the gain control pots ahead of 12AU7 output buffers
to reduce the number of am stages to an absolute minimum. The cathode followers convert
the highish Rout from volume pots to low Rout < 600r which allows long interconnect cables
of say 5 meters with C of 600pF without audio F losses.
Fig 9. Gain
pots, Output, buffers, Sheet 4.
Fig 9 Sheet 4 shows 1 channel and how I have dual volume control pots VR4 and VR5
so I can use 2 pairs of L+R cathode followers for 2 pairs of L+R outputs to facilitate
A-B testing of other equipment. Unless careful AB testing is done, how does anyone know
if any difference exists between power amps or speakers? I can have the same input
source but preamp feeds 2 pairs of L+R power amps, each level adjusted so that when
speakers are switched from one power amp to other there is no perceivable level change.
The only thing which may change, and possibly be noticed by a listener is the sound quality.
It is remarkable how often nobody can hear any change to sound of power amps.
The cathode followers have transistor
CCS between cathode and 0V for a high Z Idc sink,
aka CCS. This reduces loading by by resistors carrying 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 tolerated.
The only "sonic signature" of such a CCS is to improve the sound quality.
Fig 10. Power
Supply, Sheet 5.
The 2004 PSU is more sensible than the original 1994 amp which had two PSUs
including two PTs for B+ for two completely mono channels.
I found there was no need for such complexity on a single chassis and one B+
rail and one Vdc heater supply works perfectly. There is one B+ choke
included in a passive CLCRCRC filter for the B+. There is no need for B+
regulation except for the phono amp B+ rail. Heaters are all DC and regulated.
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, the B+ power consumed
is about 12Watts.
I have used the available 240-0-240 CT winding for full wave rectifiers with 2 Si
diodes. The actual VA rating for the HT winding should be 30VA.
It allows for overload or something shorting and a lowish value input mains fuse
To get about +285Vdc (working) and say +320Vdc unloaded, HT Vac could be
120Vac (unloaded) used for a doubler rectifier OR a 240Vac for a Si bridge.
It is easy to have B+ which is too low, and it cannot be raised, but if B+ is a bit too
high it can be reduced by some extra series R between Si diodes and caps or
further along the RCRC filtering.
I did not see a need for B+ regulation, although fanatics would have B+ regulators
for every stage of each channel. Feel free to design your own method of B+
regulation. The regulation is always something else which can go wrong, so
I try to avoid it. But when it is done right the regulators prevent LF noise in B+ rail
generated by the constant mains voltage level change because of people sharing
the Mains supply line and switching heaters, stoves, electric jugs, air cons et all.
Such noise can get to the signal path if the B+ is not very well filtered. I found the
use of CLCRCRC was sufficient to keep very low F out of the amp output when
using phono source where there is highest likelihood of seeing very LF noise at
amp Vo because the LF gain is very high with RIAA filter. This amp has low
enough LF noise, helped by RC coupling between stages.
In a later phono stage, the "Rocket", I did use regulation for the B+ applied to
phono stage. The Rocket is a stand alone unit needing a remote PSU and the
regulation guards against use of an inferior quality PSU made by someone else.
This amount of B+ power is similar to that required by an old AM
A PT found in many old AM radios will provide the B+ power needed by this
preamplifier. But it is always better to begin with NEW PTs. I am sure Hammond
make something suitable for B+. If a chosen PT does not have heater windings
capable of 23Watts to heat 10 preamp tubes a second PT must be used.
I found a suitable heater PT, 60VA from Jaycar, an Australian parts supplier.
It has several taps and allows up to 2A at 30Vrms output.
Most of the power drawn by the preamp is filament heater power.
The preamp has the power supply within a steel box with steel bottom
and removable top. The HT PT is NOS potted ex-Navy, and the heater PT
is potted in a steel can filled with dry sand to keep it mechanically quiet
and reduce stray magnetic fields.
effect potting and mild steel sheet
reduces stray magnetic
fields just enough to allow the very magnetically sensitive phono input circuitry
to be placed 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 the TWO
layers of magnetic shielding were enough to exclude effects on the phono stage
which has a very magnetically sensitive j-fet in input stage.
I used brass and aluminum for the main parts of the chassis which looks nice, but
plain 1.2mm mild steel is better especially if other devices such as CD players
in plastic cases are located near the phono amp. Keep power amps away
from phono inputs.
The heater Vdc is regulated by circuit around Zd1, Zd2, Q4, Q5.
The whole heater supply is biased at about +56Vdc so that the dc voltage
difference between heaters and cathodes does not exceed the ratings of 90Vdc
so arcing or current leakage from cathode to heater circuits is prevented.
Such leakage is rare in new tubes, but never fully preventable regardless
of Vdc between cathodes and heaters. Old tubes in amplifiers can have slight
cathode-heater leakage which causes slowly increasing hum to develop
where the heaters have AC heating current. I have seen old tubes get rare
shorting between heaters and cathode; it is easily fixed, - 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.....
Bandwidth is 3Hz to 100kHz, -3dB.
Rout < 600r, able to drive any known tube, solid state power amp, sound card.
Connections. I have plenty of RCA inputs, record output, and dual L+R outputs.
THD < 0.02% at 1Vrms output.
Wiring style is mix of point-point with tag strips plus a fibreglass board with
copper wire tracks and some surface mounting of R&C parts.
I used 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.