tubes By Joe Tritschler

Borbely RIAA with Tubes Revisited

Here’s a rework of a fan favorite: a phono preamp that includes new features and improvements.

n the October 2003 aX, I Old Colony Sound Lab, www.

PHOTO 1: The completed 5842 phono preamp.
described a stereo phono pre- audioxpress.com). The first time
amplifier using 5842 triodes I saw LED bias—circa 1999—in
and an equalization scheme the first stage of a power ampli-
that was inspired by an article fier, my reaction was to make fun
written by Erno Borbely nearly 25 of it. Using a nonlinear device
years ago for The Audio Amateur. in the very circuit position from
Several readers built 5842 pream- which the input signal is derived
plifiers, including a gentleman in seemed ridiculous at the time.
the United Kingdom who called Mr. Jones made a very convinc-
it a “living, breathing instrument,” ing argument, however, so I tried
for which I’m very grateful. Since it and listened to the difference. It
that original article, I’ve made turned out to be a very significant
many changes to the unit and sonic improvement.
present them here. The preampli- LED bias has pros and cons.
fier remains the centerpiece of my LEDs tend to drop a constant
system and I use it constantly, often play- and component technology. The result is a voltage over a fairly wide range of operat-
ing several records per day. preamplifier that is a major improvement ing current. This means that their dynamic
over the original in both objective and resistance is very low. I extrapolated 137
THE CIRCUIT subjective terms. from the manufacturer’s datasheet at the
Borbely’s approach to a phono preamp in chosen operating point of 15mA and used
1985 can be described as half-active and BIASED OPINION it in my computations. The significant
half-passive; a linear gain stage up front Rather than a traditional cathode bias advantage of this is that a cathode bypass
is followed by a passive low-pass filter for resistor with bypass capacitor, each stage capacitor is no longer needed. The down-
the RIAA 75Ms rolloff and an active shelv- of the updated preamplifier (Fig. 1) uses side is that the constant-voltage behavior
ing equalizer for the 500Hz/50Hz low- a Lumex SSL-LX5093ID LED to set more or less constitutes fixed bias, losing
frequency boost. My preamplifier replaced its cathode voltage, a technique inspired the cathode resistor’s inherent advantage
Borbely’s discrete transistor circuitry with by Morgan Jones in the third edition of of providing DC feedback to stabilize the
two very high-transconductance frame- his book Valve Amplifiers (available from tube’s operating point. This is of particular
grid tubes, and the result was very good
subjective and measured performance. I’ve
elected to keep the basic topology of this
circuit intact through all the changes. A
common-cathode amplifier stage provides
voltage gain at the input, followed by an
RC low-pass filter; then a second ampli-
fier stage with anode-follower negative
feedback implements the RIAA bass boost
courtesy of a series capacitor in the feed-
back impedance. With only two active
devices, it’s indeed a very simple circuit.
Design revisions encompass the follow-
ing: a new biasing scheme and operating
point, a redesigned low-frequency section
that offers higher overall gain, a modi-
fication to the extreme high-frequency
response of the preamplifier, and special FIGURE 1: Phono preamp circuit.
attention paid to equalization accuracy

audioXpress January 2010 27

 concern with high-transconductance tubes the feedback network, then every drop of
since, by definition, small changes in bias available gain could be squeezed out of the
produce large variations in plate current. circuit at low frequencies where maximum
As it turned out, I did need to plug in gain is required by the RIAA equaliza-
a few different combinations of 5842s to tion. The result is an overall gain of about
find pairs that biased similarly channel- 40dB at 1kHz, which is enough to drive
to-channel, but the improvement in sound my triode monoblock power amplifiers
was most worthwhile. Incidentally, there’s without further amplification. I performed
nothing special about these particular the computations assuming a 100k7 load
Lumex LEDs; they were simply available on the output, as presented by the stepped
locally in surplus. Morgan refers to “cheap Daven attenuators on my power amps.
red LEDs” in his book, but I tested a few With a moving-magnet cartridge, the out-
different types and they do differ slightly put is within a few dB of CD level, which
in voltage (and presumably resistance), and is close enough for my system. It allows me
seem to be very consistent within a given to switch directly between the two at the
part number. amplifiers with only a couple clicks of the
My theory is that a cathode bypass ca- volume controls.
pacitor is likely to be much more audible
than any power supply bypass capacitor. ULTRASONIC STUFF
The reason is that even though the dy- Conspicuously absent from the revised
namic plate current of a single-ended am- circuit is a shelving resistor in the low-pass
plifier stage flows through both of them, filter which was previously used to insert a
any AC voltage developed by the power 50kHz zero into the response, supposedly
supply capacitor across its impedance (in- to compensate for the deliberately limited
cluding nonlinearities and parasitic com- boost capabilities of record-cutting equal-
ponents) will be attenuated by the ampli- ization. The idea, apparently endorsed by
fier’s power supply rejection (which is high many designers, is that this compensa-
in this circuit). Whereas, any nonlinear tion audibly reduces high-frequency phase
AC signals appearing across the cathode distortion. I did a subjective comparison
bypass capacitor are placed right into the between networks with and without the
input grid-to-cathode signal and ampli- 50kHz zero and concluded that I actually
fied. This effect seems most pronounced preferred the sound without it.
in high-transconductance stages. Worth mentioning is that allowing the
Switching to LED bias in this pream- HF response to roll off seems to reduce the
plifier brought out reverb tails and other subjective effect of record surface noise and
nuances that had previously been missing, pops. Given that the interaction between
probably due to electrolytic capacitor im- cartridge inductance, cable/arm/preamp
perfections. Of all the component changes capacitance, and input resistance already
made, this was the most audible, resulting forms a second-order low-pass filter that
in a major improvement in detail and a has a much greater effect on frequency
clearer, less-fatiguing sound overall. response than a simple 50kHz first-order
low-pass filter during recording, it seems a
MORE GAIN, PLEASE moot point anyway. Golden-eared audio-
One complaint with the original 5842 pre- philes with ultrasonic hearing (or maybe
amplifier circuit was that it had only about their dogs) are entitled to their preference,
34dB of gain at 1kHz. This was barely but I no longer believe in implementing
enough for use with a moving-magnet this HF shelf.
cartridge, provided you were driving a line With regard to cartridge inductance
stage or very sensitive power amplifiers. and loading, I choppped the cable on my
Because I prefer not to use a line stage at Rega/Origin Live tonearm to about a foot
all if I can get away with it, a few more dB in length and located the preamplifier im-
overall seemed a worthwhile objective. The mediately adjacent to it, greatly reducing
original preamplifier set the second-stage shunt capacitance and audibly improving
mid-band gain to unity, rising to the re- high-end response. Ray Futrell wrote a
quired +20dB below 50Hz. good article on the subject a few years
It occurred to me that if I could set the ago in aX (“The LP Terminator,” January
mid-band gain to exactly 20dB below the 2003). I was able to increase the preampli-
open-loop gain of the stage by adjusting fier input terminating resistor to 100k7

28 audioXpress 1/10 www.audioXpress .com

 PHOTO 2: Preamp with cover off.

PHOTO 3: Inside the unit.

with only a slight response peak a little 1% silver micas, which is a very controver- of copper-clad PC board material as the
over 20kHz with my particular cartridge. sial choice. Many folks seem to find them original with the same phenolic tube sock-
metallic sounding and some engineers ets. I have changed nearly every other com-
PARTS IS PARTS have reported hysteresis distortion with ponent. All connections on the board are
In my original article, I stated that I pre- them. I’ve found that placing a large DC component-to-component, with the copper
ferred the sound of carbon composition re- bias across them (in this case, the plate cladding forming an excellent ground plane.
sistors despite their relatively poor objective voltage of the stage, which should negate
noise performance. Several readers took any hysteresis effects) makes a world of POWER SUPPLY
exception to this, mostly on the grounds difference, and they sound truly transpar- At the moment, I’m using a tube-regulated
that they also suffer from poor tolerance ent to my ears. I use the old types which Gen-Rad laboratory power supply from
and drift. Well, you guys were right: Car- are reportedly made of cleaved sheets of the 1960s, which is especially nice because
bon comps really aren’t stable enough for natural mica, as opposed to modern ones it’s not mine (thanks, Chris). It provides
complex equalization networks, particu- which are a ground-up slurry. But again, very well regulated 300V B+ and 6.3V fil-
larly those that must be both accurate to don’t sweat the small stuff. The precision ament supplies, both of which are essential
a reference standard and precise channel- polypropylene film-and-foil types available
to-channel. I now reserve them solely for now would probably work and sound just
restoration of older equipment and occa- fine if you don’t choose to use micas.
sionally in grid-stopper applications.
I prefer non-inductive precision wire- TUBES AND WIRE
wound resistors by Shallcross, Ultraohm, The 1980s JAN Raytheon 5842Qs that
Mepco, and Daven for their sound and ac- were very plentiful a few years ago work
curacy, but if you don’t have access to them, fine in the second stage of the amplifier, but
you can use regular metal films with per- I had a little trouble with them in the front-
fectly acceptable results. For plate-load re- end due to microphonics. I ended up scav-
sistors, I used 7W half-percent non-induc- enging some late-60s Amperexes, which
tives made by Tepro, but any quality wire- were much quieter. Strangely enough, these
wound of at least 5W would work, provided seem to have much more subjective bass
it’s within the specified tolerance of 1%. response than the Raytheons, but only in
The blocking capacitors make a subtly the first stage; probably because the second
audible difference, but I wouldn’t lose sleep amplifier stage has a negative feedback loop.
over it. I happen to really enjoy the sound The extra bass is lovely, but this type of
of extended-film/paper-in-oil Sprague Vi- thing can drive an electrical engineer crazy.
tamin-Qs (196P-series) and also use them The short IERC heatsink/shields are nice if
in my power amps (and even the crossover you can find them.
capacitors on my horn tweeters). But, like Does the type of wire used really make
the non-inductive resistors, if you don’t that much of a sonic difference? The engi-
have a good supply of them, it’s certainly neer in me doubts it at audio frequencies,
not the end of the world. I had some SBE but, then again, it seems to. I used Teflon-
716P polypropylene film-and-foil “orange insulated silver-plated stranded hookup
drop” capacitors in there for a while and wire because I had access to it. No ques-
they sounded very good; maybe a little “ar- tion Teflon is much more forgiving toward
tificially detailed” for my tastes, but it’s an soldering technique.
entirely subjective call. Photo 3 shows my preamplifier in its cur-
The equalization capacitors are old-style rent modified state, still on the same piece

audioXpress January 2010 29

 if you don’t plan any further power-sup- even at very high volume levels, I hear
ply decoupling. The preamp draws 60mA nothing when switching from CD to LP.
from the B+ and 1.2A for the filaments. I
included a 1nF silver-mica capacitor inside LISTENING & CONCLUSION
the preamplifier as a precaution to shunt How does it sound? Well, if a meteor were
any rising impedance at RF due to stray to crash through my roof and wipe out
inductance in the B+ power supply umbili- my entire system, I’d spend the insurance
cal, but it worked just fine without it. money on parts to build another preamp
exactly like this one. But seriously, folks. . .
MEASURED SPECS I believe this is about the best I can do.
Photo 4 shows the actual measured fre- My friends really seem to like it, too—and
quency response of the preamp using a some of them have very expensive tastes in
Tektronix SG 505 MOD WQ oscillator audio gear. For those who can’t afford to
and a Tektronix AA 501 distortion ana- build something quite this exotic, there’s
lyzer as a precision AC voltmeter. The re- a much cheaper version of the circuit that
sponse is indeed very flat; I didn’t need to sounds almost as good. But that’ll have to
do any empirical tweaking of the calculated wait until next time.
circuit values. The low frequency response
is about 1dB underdamped from around 20 THANKS
to 50Hz, which is likely due to the presence Many thanks to my father Tony Tritschler,
of three capacitors within the second stage an exceptional audio engineer and loud-
feedback loop. My rationale for this design speaker expert without whom I wouldn’t
choice is that I preferred the sound with have a system at all; to Duke Fecteau, Bob
this configuration; maybe it’s because both Jones, Larry Mehal, Curtis Welch, Fred
blocking capacitors are subject to negative Garber, and all my audio buddies for good
feedback. Stability is ensured regardless be- times and invaluable feedback; to Roger
cause the amp runs open loop at VLF. Hughes at Midwest Surplus Electronics
At 20kHz the response is down a little in Fairborn, Ohio, who has been taking
over 1dB, but it starts to rise again, becom- IOUs and giving me good-natured grief
ing underdamped by 4dB at 100kHz. This for almost 15 years now; and to Chris Ivan,
is probably due to the negative feedback a first-rate tech and tremendous source
loop as well; in this case, an HF pole at of knowledge from whom I’ve stolen al-
the grid of the second stage interacts with most all of my electronic components, and
an output pole caused mostly by stray ca- whose hi-fi system I’ve shamelessly imi-
pacitance imposed by the test equipment tated almost entirely. aX
and interconnecting cables,
which should be similar to 5842 Preamplifier Parts List
conditions encountered in R1.....................................................100k7 ¼W 5%
R2.....................................................1007 ¼W 5%
actual use. R3, R8 .............................................10k7 5W 1%
In the middle of the pass- R4.....................................................39.2k7 ¼W 1%
R5.....................................................53.6k7 ¼W 1%
band, where it really counts R6.....................................................320k7 ¼W 1%
(my use of Edgar midrange R7.....................................................1M7 ¼W 5%
R9.....................................................10M7 ¼W 5%
horns reinforces my strong C1 .....................................................3000pF 300V 1%
C2 .....................................................0.1MF 400V 10%
belief in this philosophy), the C3 .....................................................1000pF 300V 1%
response is flat from 200Hz C4 .....................................................0.47MF 400V 10%
V1, V2 ..............................................5842
to 5kHz, with only a 0.1dB D1, D2..............................................Lumex SSL-LX5093ID LED
crest at the 75μs RIAA turn- RIAA Equalization Measurement chart
over point. This is certainly (75Ms, 318Ms, 3180Ms)
flatter than my (pretty darn Freq. Ref. gain L gain L error R gain R error
10Hz +19.7dB +19.8dB +0.1dB +20.0dB +0.3dB
flat) speakers. Gain mea- 20Hz +19.3dB +20.1dB +0.8dB +20.3dB +1.0dB
sured 41.2dB in one channel 50Hz (3180Ms) +16.9dB +17.5dB +0.6dB +17.7dB +0.8dB
100Hz +13.1dB +13.2dB +0.1dB +13.3dB +0.2dB
and 41.3dB in the other, and 200Hz +8.2dB +8.2dB 0.0dB +8.2dB 0.0dB
the “A”-weighted noise with 500Hz (318Ms) +2.6dB +2.6dB 0.0dB +2.6dB 0.0dB
1kHz 0.0dB 0 0 0 0
inputs shorted is -76 and 2.12kHz (75Ms) –2.9dB -2.8dB +0.1dB -2.8dB +0.1dB
-75dB, respectively, referred 5kHz -8.2dB -8.2dB 0.0 -8.2dB 0.0
10kHz -13.7dB -14.1dB -0.4dB -14.1dB -0.4dB
to a typical moving-magnet 20kHz -19.6dB -20.8dB -1.2dB -20.8dB -1.2dB
input level of 4mV. Subjec- 50kHz -27.5dB -28.2dB -0.7dB -28.4dB -0.9dB
100kHz -33.6dB -29.4dB +4.2dB -29.9dB +3.7dB
tive noise level is negligible;

30 audioXpress 1/10 www.audioXpress .com