By=20
Arthur Grannell
The Dynaco ST-35 Power Amplifier has been =
the center=20
of my music system since I discovered it in 1982. I used one to =
replace my=20
pair of Dynaco Mark IVs after performing lengthy comparative =
listening=20
tests, and deciding that the smaller, cheaper ST-35 provided =
inherently=20
superior sound. Being a =93tinkerer,=94 I have always looked for =
ways to=20
improve the circuitry further, but have found little, other than =
the well=20
known power supply and individual component =
improvements.
Dave=20
Gillespie=92s EFB bias regulator is the first =93true=94 circuit =
modification=20
that promised to provide genuine, measurable results, that I=92ve =
found,=20
since taking an interest in this amp. I incorporated this simple =
circuit=20
in a completely refurbished and upgraded ST-35 amp project that I =
recently=20
completed, and performed power output testing with one and both =
channels=20
driven. The results were very similar to Dave=92s published =
results, except=20
that this amp has a considerably increased power supply =
=93reserve=94, so the=20
difference in power output was not quite as dramatic. My next =
project was=20
then to add the EFB bias regulator to a completely =93stock=94 =
ST-35. The=20
amplifier described below is a =93factory built=94 ST-35 with no =
circuit=20
changes, other than a rectifier diode upgrade to UF4007s, the =
elimination=20
of input capacitor C1, and the changing of capacitor C2 from a =
ceramic to=20
a polystyrene type of the same value. Here is a photo of =
the bottom=20
side, before modifications:
The SCA35 that Dave describes and the ST-35 =
share=20
similar circuitry in the output and power stages, but have =
numerous=20
physical differences. Adding additional components to the ST-35 is =
challenging due to the lack of available =93real estate.=94 The chassis is less =
than 3/4=94=20
deep, so component size is limited. I started by adding the =
necessary grid=20
resistors as shown in the photo at right. The 1K resistors =
were=20
installed in place of the jumper wire from eyelets 8 and 9 to pin =
2 of=20
each output tube. The 100 ohm resistors were bridged across =
terminals 8=20
and 9, and the output transformer screen lead attachment was moved =
to pin=20
8 of each output tube.
I fabricated a small board =
(at=20
left) from 1/16=94 phenolic punchboard that I had lying =
around. The grid=20
spacing on this board is 0.1875=94, so it=92s a product that =
seems to be=20
no longer available. An alterative would be to drill your own =
holes in a=20
piece of phenolic or FR4 board. The cut board dimensions are =
1-13/16=94 x=20
1-13/16=94, with a 7/16=94 x 13/16=94 cutout to clear the output =
transformer=20
leads. A 3/16=94 mounting hole was drilled to allow support with an additional #8 nut =
on the output=20
transformer mounting screw as shown.
I favor the use of =
turret=20
terminals for construction of small boards. The ones that I used =
are=20
Keystone #152, which are available from major electronics =
suppliers.=20
However, the staking tools are expensive and difficult to find. =
Therefore,=20
I attached the turrets into the board holes with a dab of =93JB =
Weld=94 (JB=20
Quick), an epoxy that will withstand the heat from soldering. An=20
alternative to turrets could be small pieces of 14 AWG solid wire =
cut to=20
=BD=94 length, placed into appropriate sized holes drilled in the =
board, and=20
also held in place with JB Weld. JB Weld is electrically =
conductive,=20
however, so watch for =93bridging=94 between components. Of =
course,=20
fabricating an etched pc board would work equally as well, keeping =
in mind=20
that attachment points such as eyelets or terminals would still be =
needed.=20
I used 5 ohm, 0.1% cathode resistors, =
rather than=20
paralleled 10 ohm resistors and mounted them directly on the =
board. A=20
Bourns 5K trimmer was mounted in a position to allow adjustment =
from an=20
area away from high voltage. A dab of common yellow wood glue was =
placed=20
under the trimmer to help secure it to the board. I decided to =
also mount=20
the output capacitor for the LM337 on the board, rather than use =
the aging=20
one enclosed in the original multi-section can. The value I had on =
hand is=20
33mfd, 25VDC, which will work fine in this circuit. One needs to =
watch the=20
polarity, here. The negative lead is grounded.
Referring =
now to the=20
photo at the top of article, I removed the original 95 ohm, =
5 watt=20
resistor from the =93can capacitor=94 as well as the two wires =
from the #4=20
terminal leading to pin 3 of the output tubes. I also temporarily =
removed=20
the black wire from the output transformer to the common terminal =
of the=20
right speaker terminal strip to facilitate board installation. I =
also=20
clipped off the ground wire from the can capacitor tab at the same =
speaker=20
terminal.
After temporary installation of the circuit =
board, I=20
located the LM337 on the chassis support and marked the mounting =
hole,=20
drilling it with a 1/16 inch bit. For convenience, I decided =
to mount a=20
ring terminal to the tab for the input signal (ground, in =
this=20
situation), rather than the center pin on the bottom which I =
clipped off.=20
Two wires were soldered into the ring terminal. One lead goes to =
the bias=20
board ground, the other to the can tab ground. I felt this was =
necessary=20
as the chassis support is spot welded to the chassis, and could be =
a=20
marginal grounding point.
After soldering an extension wire and heat =
shrink=20
tubing on the LM337=92s output terminal, I then installed it on =
the support=20
along with the ring terminal, using thermal compound, but without =
a mica=20
insulator and hole spacer. I then shortened the previously removed =
speaker=20
terminal ground wire and attached it to the =93side=94 can =
capacitor ground=20
lug. This will, in theory, improve the grounding scheme, rather =
than=20
relying on grounding through the chassis or aluminum capacitor =
can. I ran=20
a new speaker ground wire to the output strip on the other side of =
the=20
chassis support. I then soldered the LM337 leads to the board =
terminals,=20
and shortened and attached the wires from pin 3 of the output =
tubes to=20
each 5 ohm resistor=97one from each channel.
The installation was completed by attaching =
a wire=20
from the can capacitor (marked with a square) to the 360K =
resistor. This=20
resistor was mounted slightly above the board as it is a 2 =
=93watter=94, and gets=20
warm. Finally, I added a #8 screw with two nylon nuts, in a =
=93sandwich=94=20
arrangement, one above and one below the board in the corner of =
the notch=20
to stabilize the board. This screw, plus the connection to the =
LM337 on=20
the other corner, makes the board physically stable without any =
additional=20
chassis drilling. For extra insurance, I placed a small piece of =
tubing=20
over the black output transformer wire, as it now contacts this=20
stabilizing screw. The installation was complete with the drilling =
of only=20
one 1/16=94 hole as the only physical chassis alteration. =
Some=20
of the essential parts that I purchased from Mouser=97
Mouser # 511-LM337SP Linear =
Regulators -=20
Standard 1.2-37V Adj Negative
Mouser # =
71-CPF15R0000BEE14 Metal=20
Film Resistors - Through Hole 1watt 5ohms 0.1%
Mouser #=20
652-3296Y-1-502LF Trimmer Resistors - Multi Turn 3/8 5Kohms =
Sealed=20
Vertical Adjust
Addendum
Since the above article appeared, I=92ve =
been asked a=20
number of times for my opinion of any sound changes apparent in =
the=20
amplifier due to the inclusion of the EFB Bias regulator. This is =
a valid=20
question, since it=92s well known that better =93specs=94 in an =
audio circuit=20
don=92t necessarily equate to better sound. A few months ago I did =
a=20
personal evaluation, comparing Dynaco cathode bias vs. EFB Bias. =
For those=20
who haven=92t seen this report before, it is included here.
Listening Observations Dynaco ST-35 =
Cathode Bias=20
compared with EFB Bias
Music =
Source:=20
Yamaha DVD-C940 DVD-SACD-CD Changer. CD=92s upsampled at 176Kb/s. =
Preamp:=20
Dynaco PAS-2 with modifications to the heater supply, power supply =
and=20
line stage.
Amp:=20
Essentially stock ST-35 with EFB mod incorporating an external =
switch that=20
changes between cathode and EFB bias. Tubes: Outputs: JJ EL84 =
(matched=20
Quad) Voltage Amp/phase inverter: 7247s=96one marked =93Dynaco=94, =
one marked=20
=93United Electron=94 both are marked =93Made in Gt. Britain=94 =
and both believed=20
to be =93Mullard=94.
Speakers:=20
Altec-Lansing 890C (Bolero) LF: 10=92 active (approximately a 604 =
with 10=94=20
cone), 10=94passive. HF: Small horn made by =93Foster=94. =
Estimated system=20
efficiency: 95-98 dB (1W, 1m).
I selected =
three specific=20
CD=92s and one SACD for the tests. I played the tracks listed =
below using=20
one amplifier bias setting, then switching the amp off, changing =
only the=20
bias switch position, and immediately relistening to the tracks, =
going=20
back and forth between the two settings in no particular order. =
1/ Stereophile Test CD2 I =
concentrated on=20
the first two tracks for evaluation. These were recorded by =
connecting a=20
Fender Bass guitar directly into the mixing console. Anyone who =
has heard=20
this, may agree that the effect is rather startling! My opinion is =
that=20
the sound has even more realism and =93punch=94 with EFB bias. =
2/ George Wright Playing the Mighty =
Wurlitzer Pipe=20
Organ, Vol. 3 I chose this CD figuring that if anything =
would=20
demonstrate an increase in amplifier power, this would be it! =
Anyone who=20
has listened to theater pipe organ, live or recorded, knows the =
floor=20
shaking, window rattling sound that emanates from the 16 and 32 =
foot=20
pipes. The lowest frequency produced is 16 Hz, and I tend to think =
that=20
was Mr. Wright=92s favorite note!
Oddly, after =
listening=20
extensively to this CD, and switching back and forth, I could =
detect no=20
difference at all between the two bias settings. I could only =
guess the=20
reason for this, and that guess is that it has something to do =
with the=20
fact that my speakers do a poor job at 16 Hz. Altec rated them =
down to=20
40Hz, and I suspect response falls off quickly below that point. =
The test=20
tones down to 20Hz on the =93Stereophile=94 CD confirms this =
suspicion. I have=20
heard this (organ) CD on a system that does reproduce 16 Hz fairly =
well,=20
and the sound is definitely different. Nevertheless, I=92m still a =
little=20
puzzled by my lack of ability to hear a difference between the two =
settings.
3/ Telarc Classical SACD =
Sampler 2=20
Track 2 and one or two other tracks were probably included on =
this=20
recording for the reason that the closely =93miked=94 orchestral =
side drum=20
literally lifts you out of your seat, sounding much like thunder, =
at=20
certain points. Here, the difference between the two bias settings =
was=20
definitely noticeable, as I came further off the couch when the =
EFB bias=20
setting was used. I also think that other differences were =
apparent using=20
EFB bias including an improvement in the sound of cymbals and some =
brass=20
instruments. I tend to enjoy the music more with the latter bias =
setting=20
in ways that I find a little difficult to define.
4/ Bela Fleck=96Double =
Time I=20
consider the acoustic quality of this CD to be the highest of any=20
recording that I have in my collection; better than any SACD or =
DVD I=92ve=20
heard. Consisting mostly of music played on banjo, mandolin, =
guitar and=20
bass, it defines the meaning of =93transient response=94.
To my surprise, =
the=20
difference between the two bias settings was more noticeable with =
this=20
one, than on any of the other disks. I asked another person to =
listen to a=20
track played both ways, and to tell me if she heard any =
differences, and=20
if so, if she could describe what she heard. She stated that on =
position 2=20
(EFB) the very subtle finger and pick sounds on the strings were=20
noticeably more defined and apparent. She preferred this setting, =
overall.=20
Her opinions matched mine, but I found another generally =
noninvolved,=20
non-critical listener=92s opinion to be quite valuable, since she =
didn=92t=20
know in advance what she was going to hear, and has no knowledge =
of the=20
technical difference.
In=20
summation, I=92d say that, yes, there is a =93sound=94 difference =
between Dynaco=20
cathode bias output, and Dave Gillespie=92s EFB bias. The =
improvement is=20
measurable on the bench as increased power output and reduced =
distortion.=20
But, the effect on the final sound is apparent in ways that I for =
one,=20
would not have expected. The increase in power output in improved =
bass=20
would be expected, but it=92s interesting that the modification =
shows up=20
with improved transient response and superior =93musicality=94,=20
overall. |