DIY Type 76 preamp first impressions
Janos
The day before yesterday I enjoyed the first fruits of my last 20
months of designing and building the 76 preamp.
The preamp is based on Joseph Esmilla's type 76 line stage, that is in
turn based on Berman's type 76 line stage in the Sound Practices vol
13, pg 27-.
The power supply is different from the one they use; I used parts that
I had in my "junk"-box.
Kyle was right, we could have fired it up at the audio club, to see
the smoke coming. The troubleshooting included the rectifiers
transformer that blew; it also killed the high voltage's diodes, and
the small piece of ga 26 silver wire, connecting the ground bus to the
RCA jack, smoked away, before the fuse could blow. And there was also
a blown power supply electrolitics cap, that was just too old to
reform - it was reforming fine till 110V, then it blew at 115V... lots
of smoke, stench, and mess to clean up...
Other than that, it fired up quite fine. The voltages were exactly as
expected, the cathode followers lined up correctly after the
amplification stage of the type 76 tubes.
The schematics:
The line stage is the same as Esmilla's and Berman's, except that I
use a 30K load on the 76, and 25K cathode bias on the cathode
follower. The 76 is yet unbypassed. The other change, is that I use
the 7193 as the cathode follower, that I got from Kyle, instead of the
6SN7. It is in fact one section of the 6SN7, with both the grid and
the plate on the top, with caps. Should be an excellent 6SN7 variety.
Esmilla's has a B+ of 280V; this one has 274V B+ after the 2.5K
voltage dropping resistor, separate for the two channels.
Power supply:
150-0-150V mil. surplus transformer (same that we use in the Darling
amplifiers). Full wave bridge, using ultrafast diodes, paralleled with
either a 5ar4 / 5r4 / 5u4 / 5t4 rectifier, or a 80 / 5z3 / 5y3 type
rectifier tube. Installed both 4 pin and octal sockets to accomodate
any of these tubes. The voltage drop of the rectifier tubes is of no
consequence, as the 1.7V voltage drop of the ultrafast diodes will
dominate. The diodes are bypassed by 50nF @ 630V orange drop caps to
filter the diode noise.
The power supply filtering sonsists of a CLC sction, with 47uF oil
caps rated 550VAC, and a 20H (150DCR; 150mA?) choke. Duncan amps
simulation yields a not too bad result for this supply, with a few
nanoamps and 2 microvolts ripple. Nanoamps is OK, but 2 microvolts is
too much for a preamp, would be insufficient in itself.
A pair of 0D3 gas regulator tubes in series follow the last C section,
with a 2.7K @ 5W resistor in between, to protect the gas tubes. The
450VDC voltage after the last C section is dropped to 300V regulated
after the gas tubes. The gas regulators are bypassed by a 50nF orange
drop @ 630V to filter out the gas tube whooshing noise, and the top
gas tube is bypassed by a 30K @ 5W resistor to ensure the two tubes do
not strike at once.
Filament supply
6.3Vct secondary mil. surplus filament transformers are run with full
wave bridge Schottky diodes; one for the 76 tubes, and a separate for
the 7193 tubes. The reason for two separate fl supplies is the big
voltage difference between the two stages. The 76 can take 90V fl –
cathode difference, but the 7193 datasheet recommends to tie the fl as
closely to cathode as possible. The 166V difference detween the two
cathodes calls for two separated supplies, each referenced to the
appropriate cathodes, at one channel only. Neither Esmilla nor Berman
address this issue; they use one fl supply, floating between the two
cathodes, derived by dividing resistors from the B+ supply.
An 1uF @ 630V polyethylene cap is the first to receive the voltage to
filter out diode noise. The 76 tubes fl filtering after the 1uF cap:
-1R-10mF @ 50V - .33R – 15mF @ 16V; on the ground there are also
resistors, .27R ; and .22R. Tightly twisted AWG14 solid copper wires
carry the FL supply to the type 76, where an additional 1uF @ 630V PE
cap takes care of HF fluctations and HF noise pickup. The tubes are
running in starved cathode mode, at 5.87VDC instead 6.3VDC designer
center value. The cooler cathode has much lower tube noise, and tube
life is extended significantly.
The filtering of the cathode follower's fl supply would have been the
same, but the electrolytics caps ther did not reform, so I put in
instead a 0.27R followed by a 2mF @ 16V Sprague Atom. Tightly twisted
AWG14 copper, and the 1uF @ 630V caps at the tube base, similar to the
76 stage.
Wireing
Tightly twisted pair of AWG 12 solid core copper wires power cord,
AWG 8 solid core copper ground bus. RCA input jacks wired with: AWG
26 silver to the ground bus, and AWG 30 silver to the selector switch
& grids of the 76. The cathode of the tubes is also wired with an AWG
26 silver wire. All the cathodes and the input grounds connect to the
ground bus at one single point. The plate voltage supply of the
cathode followers is AWG 22 silver, and the wire connecting the 76
plate to the 7193 grid is AWG 26 silver, both in teflon tubing, as
they are exposed on the outside. Except these teflon tubed wires, the
filament wires, and the power cord, every wire (silver wires and
ground bus) is uninsulated. The 7193 has as grid and plate stoppers
85R carbon comp resistors. (The 76 also has 85R grid stoppers).
The output coupling cap
… is a Sprague Vitamin Q oil cap, .22uF @1kV. Chopped off the copper
leads at the base, and added silver leads, with lead-free silver
solder. AWG 26 silver on the input side, and AWG 30 on the output
side.
The cap is encased in a bamboo case, and the void space is filled with
wax, to dampen the capacitor. The caps are mounted on a wooden block,
to increase the degree of isolation from common mode mechanical
vibrations with the chassis.
The RCA jacks are rhodium plated, input impedance is 380K, output
impedance is 440K.
Mounting of the transformers and choke
Every iron is mounted with dampening, to the wooden frame, to avoid
building up transformer resonances. .
Tube sockets
The gas tubes and the 76 use DIY tube sockets: a material similar in
properties to teflon, I believe it's called dacron. Holes are punched,
where the tubes fit in very tightly, but there are no metalic parts
for contacts. Silver leads are directly soldered into the tube bases,
after sucking out the original solder from the tube bases, connecting
the tube pins to the tube socket pins. This way the number of contacts
and different metals is drastically reduced – the electrons have a
free way to go, from tube pin directly to the silver lead, with only a
small amount silver-bearing lead-free solder to jump. This could be
the best tweak that can be applied to a tube to improve its sound, but
it makes tube-rolling very painful: you have to solder and desolder
every time you roll tubes…. The analogy of the regular tube socket
way of using tubes (with soldered-on bases) and this way is the race
cars hauling a heap of junk compared to the unburdened race cars…..
huge difference, but can't tell exactly how much unless you compare
them.
Measurements
Voltages
450VDC unreg high voltage, 300V reg,, 274V B+ after the 2.5K resistors
protecting the tubes and separating the two channels. 165V plate at
the 76, 167V at the cathode of the 7193, 274V 7193 plate. 76 running
at 3.6mA, 7193 at 6.8mA.
Frequency range
Setup: Before I connected the pre to the system, hooked it up to the
Tek oscilloscope, and a Heatkit signal generator. Used a plain, ugly
black PVC IC, that comes with junk equipment connecting the signal
generator to the 76 with a RS BNC / RCA adapter, and connected the
scope probe to the RCA output jack through a questionable looking RCA
to aligator clips adaptor.
Specs: square wave signals were generated. The gain is roughly 5.2; a
2V in signal gives a 10.4V out signal. The gain is constant over a
wide range, from mV levels to 20V on the output side, indicating a
very large headroom…. At 20Hz (and 1V out) the amplitude of the signal
was yet unchanged compared to 1kHz level. However, the rise time and
decay was slightly slow – the square wave was not exactly square at
the output. I have to find out if this was caused by the bad cabling
used for the test, or the preamp needs a hand at this respect.
However, it looked much-much better than the square wave output of my
souped-up PAS at any frequency…. I have not checked lower frequencies
– it shall be done with better cabling, and the signal generator goes
down to 0.1Hz, so I shall see…
At 20kHz (also 1V out) the square wave looks like at 1kHz. At 200kHz
just the same, though there is a slight lag in the rise time. Even at
600kHz there is but 1dB or less decrease in the amplitude…. I could
say, extended frequency range?! Probably the best high frequency
extension for any preamp so far…
However, at 600kHz the square wave looked more like a sawtooth wave. I
have to figure out if it was tube distortion or the weak interconnect.
I suspect the IC had a major contribution to this fact, as it is not
supposed to do much at these frequencies. The signal was very stable,
though, the line stage handled this frequency with ease.
Mode of operation
As you have noticed, there is no volume control. The volume control
will go after the preamp. The rationale is that at low signal levels
the volume control introduces a greater amount of distortion, and
greater tube noise to signal levels, impairing amplification. After
the line stage, the signal is higher level (in this case – after a 5.2
gain) and at a significantly greater current, driven by the sturdy
cathode follower.
TVC will be responsible for the volume control, between the line stage
and the amplifier. It will be of a great benefit, increasing the
current drive and low frequency performance as the volume is turned
down.
As some of you know, I have been designing a TVC autoformer, in
collaboration with Akos (who came up with the paper version of
Charlie's CCSSWP IC) and Denes Ba, a master-transformer winder in
Hungary, with Stu's advices. The design process and the continuous
exchange of ideas was the most intense, and fruitful DIY experience I
ever had.
I'll use this TVC for volume control, that Denes Ba winds – I rather
have him wind a high quality pair for me, than venture in such a
complex task as my first adventure into transformer-winding. I rather
start with much simpler tasks, such as a filament transformer or a
choke.
First listening impressions
Very audiophile sound. Lots'a base, dynamic, extended highs, brittle
mid-range. The instruments jump out, kind-of under a magnifying glass,
and the vocals are sharp and shrill. Both Esmilla and Berman cautioned
against regulated supplies, saying that these result in an uninvolving
hifi-sound.
Now, after 15 hours or so of breaking in, this shrill quality has
disappeared. The high end is not unnatural either, but would need some
extension according to Nelli. I think the high extension is better
than the modded Dynaco PASs, but it could use a touch at the very
high.
The sonic character is similar to that of the souped-up PAS (that is
similar to the original PAS), but it projects the sound further, and
has better center-fill. With the 76 I have the same level of detail
and resolution at the far end of the room that I have with near-field
listening with the PAS. The PASs sound is also more up-front, the 76
images deeper, and with a wider sound-stage. The sound is more
detached from the speakers. The modded PAS (or rather the
single-driver speakers, by default) did quite a good job in this
aspect, as the sound was always detached, but some of the music
material focused aroud the drivers. With the 76 there is no feeling
that the drivers are limiting the stereospace.
My conclusion after the initial break-in is positive. Excellent
stereospace, nice top and bottom extension, very dynamic, and lots of
headroom. There is some woolyness in the base, and some trouble with
the mids/highs. I think most of that is volume control issue (see
below), but some might be the result of using wirewound resistors as
cathode resistors. This is yet a prototype, an infant preamp. I have a
very long way to tweak it out, find the optimal input / output
impedances, bypass cap values, the resistor-rolling…
Regulate or not-to-regulate?
The gas tubes are not evil, and involving sound can be achieved using
them. Both Esmilla and Berman recommended using plastic output
coupling caps, and I am using oil caps. Maybe the gas regulator tubes
need oil output caps to warm their sound? Would it sound clinical with
plastic caps? The other thing I have noticed is the whoosing noise of
the gas tubes. If the 76 output arrives to the amp without
attenuation, I can hear a quite loud whoosing noise – similar to a
white noise or the noise of a TV set when there's no reception, but
unique, has its own sound. This sound is plainly audible from the
listening position. However, this volume is 14dB over the max input
sensitivity, that is around 40-50dB over normal listening levels. At
normal listening levels the whooshing noise is only audible if the
ears are close to the driver.
However, the ultimate test of the benefits of regulation is to listen
to the 76 without it. This would require mods to the power supply…..
Volume control issues
I am not able to give a correct evaluation of sonics yet, as the
current volume control mechanism is extremely compromised. I am using
the amps shunt volume pot to decrease the volume – and it needs to be
used a lot. Three-quarter turns for volumes with the DVD or cassette
tape as the source, and almost down to zero level with the CD as the
source. This means that at these points the input impedance of the amp
is super-low, guzzling up all the driving current, and virtually all
the low frequencies. This effect of the shunt pot is very apparent:
from DVD / tape there is a decent amount of bass coming out with 3/4
turn of the pot, but maxing out the pot for CD results in very weak
bass. Even the cathode follower cannot drive the resulting
ridiculously low input impedance.
The TVC will reverse the situation, as turning the volume down will
increase drive current – resulting in better and better base as I go
down with the volume. We have already observed this peculiarity of the
TVC with Charlie's S&B TVC – the softer the volume gets, the more
emphatic and controlled the base gets, the more body and tone appears.
The only difficulty for the TVC is the recordings with low volume
levels, and playing it at loud volumes, where the source just does not
have enough drive. This is not the shortcoming of the TVC, it is the
shortcoming of the lack of current drive from the source. The 76
preamp is designed to get around this exact shortcoming, and combine
the best of the two worlds.
Whet else?
The 76 and 7193 give a very faint glow, but the gas tubes are awesome.
At daylight their color is more purple on the side, and orange, when
viewed from the top, and at night they look white, with a shade of
purple on the side. Very interesting phenomenon. As soon as I hit the
power switch, the gas tubes light up. They start regulating
immediately as the voltage goes up. Very nice feature, and I know
immediately that there is a problem, and a need to troubleshoot when
they do not light up right away…
Gas Tube Lore
Yesrerday at 1AM I got saturated with the PhD dissertation, and picked
up the pdf printout of the Navy's Electronics Textbook on electron
tubes, that was sitting on my shelf unread for a year or more. It has
a nice section on gas tubes. I have read many different interpretation
on how to use these obscure tubes correctly, and people seem to place
the current limiting resistor either in front of the tube, or after,
and there is no consensus on the value of resistor either. The
Radiotron Designers Handbook is the best source on this matter, and
states that the resistor should be before the tube, and be designed to
limit the current to the maximum current permissible by the gas tube.
This led me to believe that gas tubes are of very limited usage, as
the 0D3's max permissible current is 40mA.
The Navy manual had more detail on how gas tubes regulate, and there
is no limit to the current the load can draw. The only catch is that
the higher the load current, the narrower the window of regulation
offered by the gas tubes.
When the current drawn by the load from the power supply is low (let's
say 3mA) compared to the gas tube's average current ((5mA + 40mA) / 2
= 22.5mA for 0D3), then the gas tube will offer steady regulated
voltage even if there are very big variations in the line voltage –
the 120VAC line voltage fluctuations form 70V to 180V do not affect
the regulated B+! If the current is large, let's say 100mA, then the
range is narrower, 105V…143V in the line AC, which is still pretty
awesome. (Note: the 100mA data is from the textbook, the 3mA data is
but an (un?)educated guess…)
The formula to calculate the value of the protective resistor in front
of the gas tube is:
R = ( source voltage – regulated voltage ) / ( gas tube average
current + average load current )
This resistor sets the current for the gas tube. The considerations of
choosing a specific value are: if you run it at average gas tube
current, you have the widest range of regulation in case of line AC
fluctuations, and the harder you run a gas tube, the effects of
changing load will be smaller. So, you can optimize for either best
sonics or widest protection for poor line AC areas. (Are there such?)
Janos
_______________________________________
Test setup: my system, consisting of the usual tweaked Micromega Stage
3 / Décor input autoformer (steps down to Uin max: 0.2V) / 76 pre /
Sweetheart (Darling) amp / Voigt pipes.
The preamp's power supply
The 76 pre has a hybrid power supply: a full wave bridge, with
ultrafast diodes (1A max 800V max Uinv). The top two diodes are
bypassed by the rectifier tube. 50nF orange drops @ 600VDC bypass each
diode to suppress diode switching noise.
A CLC filter (47uF / 20H / 47uF) is followed by a current limiting
resistor (3.5K @ 25W dale, limits current to 40mA, dissipates 5W). Two
OD3 gas tubes in series drop the voltage to 297V regulated, and they
burn 18mA current, the 76 & 7193 tubes use 10.5mA per channel. One
100nF @ 600VDC orange drop filters out the gas tube broadband noise.
The gas tube whooshing noise is barely audible on the Voigt pipes, when
the ears are placed close to the driver.
The diodes allow a lot of current, and a mere 1.7V voltage drop. If a
rectifier tube is inserted, the voltage drop will be determined by the
diode, regardless of the rectifier tube's inherent voltage drop. The
gas tube regulation theoretically should filter out most of the power
supply's irregularities. The expectation is that a rectifier tube
should not change sonics, as the diodes and gas tubes should level out
any differences. However, a rectifier tube makes for a noticeable
change, and there are perceptible differences between different
rectifier tubes. I thought that the changes between the rectifier tubes
are not as great as observed in unregulated and ultrafast-diode-less
amps, yet significant enough to make the listening experience from
average, through good to outstanding.
Dec 8, 2005: 11AM - 2PM, electricity quality: quite good.
Test material: Cantate Domino, 1st track (on repeat). Begins with an
organ-solo, followed by a choir, trumpets, then choir with trumpets and
organ... very demanding test material, many aspect of reproduction
tested to the limits. Excellent test track, that sheds light on how a
system handles vocals, harsh sounds, soft sounds, dynamics, texture,
soft passages, and extreme demands.
Each tube had a few minutes of warmup before critical listening. Tube
brand, type, and state of break-in shown.
In the order of testing:
Marconi 5Z3, ST14, 5 hrs:
Musical, coherent, organ is rich in harmonics, but the high notes are
not clear. Choir is not full sounding. Trumpets lack dynamics and
impact. Peak at the end is mushed in stereospace.
No rectifier, diodes only (100hrs break-in):
Bass notes of the organ are clearer, easier to get, yet squeaky
sounding, like mice being squeezed. The 5Z3 in comparison was much
warmer, fuller, yet smeared. Choir is much more transparent, more
information, detail, and precision. Organ has more attack, impact,
drama, and contrast. Wider soundstage, and colder presentation; trumpet
shines, but at the cost of becoming narrow and squeaky. Slight blur at
the peak, making the sound harsh and the ears aching slightly.
5T4, RCA 0 hrs, NOS, metal can T tube:
Clear, wide soundstaging organ, wider than diodes only. Highs soar to
the sky. Trumpets get 5 stars: great tone, shine, body - they sound
perfect. Very clear choir, more voices are identifiable. Strongest
fundamental on the organ as yet. The choir remains clear even when
organ plays. Dynamics is not restricted at peak, individual sounds
remain clear. !CAUTION: I got shocked when inserting the tube, and
touching both metal can and chassis. After that I checked if there is
voltage difference between tube can and chassis and there was not...
however, that shock felt like 120VAC line voltage...
5R4WGB, potato masher, JAN, 1000hrs:
Similar to 5T4, with narrower soundstage and more harmonic richess.
Trumpets pretty nice, yet a bit muffled compared to 5T4. Choir is very
natural, "human" sounding, best so far. Best balanced organ so far:
has bite on the high pipes, and the bass pipes sound like true
counterpoints to them - indicating correct harmonic structure
throughout the entire frequency range. This can be the reason why
vocals sound so natural.... The choir and the organ are clearly
distinguishable, each can be followed effortlessly, yet the music does
not fall apart, forms a single musical unit. The best handling of
complex peaks so far. The pedal notes of the organ were a tad clearer
and stronger on the 5T4.
Philips Miniwatt 80, Australia, T bulb, 20hrs:
This tube is in a different league. Feels like we switched to analog
source from digital. Church acoustics could be heard the first time.
The organ is amazing. Ten times more information, harmonics, texture,
and micro detail. A real choir with real people singing. Soundstage is
as wide as 5T4, but the sound fills the entire room, and beyond... the
sound echoes back from the very skies, shining with the light above....
like a real cathedral. (Well, not as clear, and grasping, less dynamic
and impactful, but does it!)
5R4GY, RCA, ST, brown base, 10 hrs:
Same league as the PM80, but not as dark - it's lighter sounding.
Better separation of voices, clearer organ. Larger cathedral, yet the
chathedral echoes are not so apparent - the cathedral size & echo
delay & volume matched better on the PM80. Trumpets are cleaner, yet
sound less emotional, but more real - PM80 made trumpets sound as if
blown by angels, superseding human talent and real-world insturmets.
Choir sounds cleaner, but more edgy, slight digital taste. At dynamic
peaks choir separates from music. I have a sense that mixing errors on
the disc are ruthlessly revealed.
Arcturus 80, ST14, 1967, 10 hrs (tube A):
Soundstage shrunk, does not fill the room like the 5R4GY / PM80; sounds
softer and much less detail and harmonic richess. Quite nice trumpet,
though blown with diminished force. Organ is dry, like an organ with
wooden pipes, that has been neglected during the last 300 years...
(heard one such in Transylvania;). Overall: nice tone, but blurred and
small in dynamics and detail. Gets hard to follow at peak.
Arcturus 80, ST14, 1967, 0 hrs (tube B):
At first similar to the previous tube... but cleaner sound, and more
detail. The tube starts warming up (and breaking in...) after a few
minutes transforms completely. One of the best so far at dynamic peaks!
High notes of the organ are still dry; excellent trumpets, and a very
transparent choir. Sounds like midway between PM80 and 5R4GY.
Philips Miniwatt 80, Australia, T bulb, 0hrs, tube B:
Back in the audio heavens... Like PM80, tube A, yet somewhat brighter,
and the choir has expanded...
Brimar 80, T bulb, 0hrs:
Tone is like that of the Arcturus 80. Strongest on the organ
fundamentals, and with the most texture. The highs of the organ come
almost from the ceiling... best 3D imaging. Best dynamics, in tutti can
hear specific individual the singers making mistakes. Perhaps the best
tube so far. The PM80 was more emotional, though. This tube gets the
detail level of the 5R4GY with richer texture, and even better detail.
Mullard 5AR4/GZ34, 0 hrs:
Loudest tube, very precise fundamentals, but not as textured as the
Brimar/PM 80. Excellent, yet uninvolving choir. The choir is too
sybillant - way more than any previous tubes. At peak the soundstage
shrinks, room fill also shrinks, and the sound gets rough and
unsophisticated, and messed up.
Emerson 80, ST14, 0 hrs:
Better tone and texture than 5AR4. More refined. The choir is
unbelievable. You fall in love with the female singers... Organ is
excellent. The pedals feel like counterpoints, lots of detail with
musical coherence. Takes peak with ease; no compression, no loss in
emotional content or detail. Similar in power to Brimar 80, emotional
like the PM 80, and bright like the 5R4GY.
I just got a tip from Stu that rectifier testing is not this easy: we
have to try out combinations with different power supply caps, as the
effect of the rectifier tube is modulated by its interaction with the
filtering capacitor... this time I have tested the interaction with
gigantic 47uF oil caps rated at 1kV. These caps seem to be very
satisfying: very relaxed sound, wide stereospace, great high frequency
extension, precise base with plenty of texture.
Probably even more was going on, as the gas tubes also have to be taken
into account. I have a hunch that the gas tubes sound, in turn, depends
on the current they are burning up - or on the current limiting
resistor, or on both.
The 5W rated 2.73K IRC (dissipating 6.7W of heat) was running the OD3s
at 30mA. It was got hot as crazy in a few seconds, but added a copper
heatsink to it, and it worked without failing, even 5-6hrs at one
stretch. When disassembled after 80hrs or so, it had no burn-marks or
scorches, and kept original resistor value. (Did not look stressed
out.) The 3D imaging was much sharper with it, yet the gas tube
whooshing noise was way louder, compared to the current 25W-er 3.50K
Dale.
Janos