Archive of the former Yahoo!Groups mailing list: tube synth diy

I recently completed construction and initial testing of the VCA
using a 1SH24B tube. I built it in MOTM-format using a panel and
bracket from Bridechamber.

The version I built is different than the original circuit. In order
to avoid damaging the PCB from excessive unsoldering, I did not build
a stock unmodified version for comparison purposes.

From what others had said, there were three issues: the output was
noisy, there was a "high-pass effect" or "loss of bass", and there
was a "thump" during the attack of the EG when used as a VCA. So I
modified the circuit in order to reduce noise and increase bandwidth.

LED's are very noisy. Since the cathode is in the signal path, I
eliminated the two signal diodes, the two LED's, and their resistors.
I connected a 1.2K 1/2W resistor between Pin 1 of the cathode-heater
and ground.

I was out of TL072's. In order to use a fast op-amp (which would not
be an LM358 or LT1013) to drive the grid capacitor that was better
suited for audio, I replaced it with an OP275. The OP275 requires
power bypassing to operate within specification, and the stock 100nF
bypass caps are too far away to be effective. Also, the stock 100nF
cap on the positive looks like it forms ~15kHz LPF on the tube
output. So I eliminated those caps, and underneath the PCB, added
two .1uF ceramic caps with heat shrink tubing going from ground, with
the capacitors as close as possible to the power pins on the socket.
The OP275 can drive capacitive loads, and works better as a
non-inverting amp if the parallel value of the gain resistors are
less than 2K. So eliminated the 47pF cap on the output, and replaced
it with a resistor lead. I did not add the gain resistor. While 330
Ohm is the minimum specified resistor for output protection, I
replaced it with 1K (outside the feedback path) to have the same
output impedance as MOTM.

The grid capacitor forms a high-pass filter with the (330K) grid
resistor in parallel with the grid resistance of the tube. The grid
resistance of the 1SH24B is specified as 100K. So the resulting
corner frequency with a 220nF cap would be 9.4Hz -- not high enough
to cause a problem, imho. However that specification is when the
cathode-heater is run off 1.2V, which is not what we are doing here,
so therefore the grid resistance is unknown. So I increased the two
220nF caps to 470nF (both of them so the tube would still sit level :)

At audio frequencies, the effective impedance of the plate is the
plate resistor (8K2) in parallel with the load resistor (100K).
Imho, coupling caps sound better driving lower impedances. So I
changed the plate resistor to 20K and the load resistor (connected to
pin 5 of the OP275) to 33K. 20K||33K = 12.5K yielding a bit more
gain than the stock circuit.

However, I do not think the "high-pass effect" noticed by others was
largely due to the coupling caps. I think it was due to cathode
depletion -- negative feedback that increases with lower frequencies.
So I added a cathode bypass capacitor. I could not figure out a way
to derive an appropriate value mathematically. So I used a 47uF
axial electrolytic. Which may be way larger than necessary, and also
might not sound as good as a film cap, YMMV. I connected it between
pin 2 of the tube (connected to the negative rail) and ground. Do
not connect it to the ground side of the cathode or the in-rush
current could burn out the heater.

I wired up the bias switch -- using properly shielded cable -- to two
settings: 330K towards the negative rail, and 1M towards ground.
They sound a bit different from each other, but more on that below.

I hooked it up and started testing it. I didn't notice any noise on
the output. While distortion increases the high frequency content, I
did not notice a "loss of bass". As one would expect, the output
does roll off at low frequencies eventually, but this seems to occur
below the audio range. While distortion increases with input level,
it does not reach the point where it squares up completely.

With the CV INIT trimmed, and no CV input, the output of pin 7 of the
LM358 is +6V with the gain control potted all the way down, and
-14.38V with the gain potted all the way up, the output of pin 1 is
-12.1V and 13.77V respectively. The Gain knob (I used a linear 50K
pot) has no effect within the last tick or so as it runs into the
rails. The output level seems fine.

With the stock 100nF cap installed, I did not notice a click or thump
when driving it with an MOTM-800 EG. So far, with a minimum attack,
it seems better than the MOTM-190 in that regard. However, there is
a noticeable peak or boost in output at the beginning of the
envelope. Which seems like a nice feature, and it it is easily
reduced by increasing the length of the attack. While the resulting
curve seems neither perfectly linear or exponential, it is easily
adjusted by ear. Although considering the range of the gain pot,
I'll try decreasing the CV input resistor to 50K. So besides that,
and given the distortion, it works fine as a "regular" VCA.

With a roughly 10Vpp source, distortion becomes obvious when the
input is more than half-way up (I used linear 50K pots for both audio
inputs). While the sound is quite different when using either
positive or negative feedback (and imho, changing Input 2 to a
reversing attenuator would be a nice mod), between the two bias
settings, the difference without feedback is subtle. There is also a
slight effect on the CV response between the two settings.

I'm still trying to figure out the best way to wire the bias switch.
I would like one setting to have the least distortion possible, and
another to have more distortion available.

With no signal, the voltage on the grid measures -14.66V connected to
the negative rail with a 330K resistor. With the signal grid
connected to ground with a 1M resistor, when first powered up when
already warm, it reads -13.50V then slowly climbs to -14.16V and
stays there. Switching between the two switches between -14.66V and
-14.16V.

I checked all my connections. My conclusion is that the grid is
capacitively coupled to the -15V cathode. Keep in mind that this
tube was designed to run the cathode near ground, and the grid biased
a bit below that.

With a signal, the negative voltage on the grid increases. If I
measure 3.15 VAC rms (about 9Vpp) on pin 1 of the OP275, with the
grid connected to the negative rail with a 330K resistor, the grid
measures -17.85 VDC. Which is roughly equal to 14.66 plus 3.15.
Well, anyway, at least the grid is negative to the cathode, which I
guess is kind of normal :)

Luckily, it doesn't become so negative it that it cuts off the plate
current. Still, I would like to figure out the optimal bias
settings. Although I was planning the resistor towards ground as the
clean setting, the slightly more negative (.5V) setting has less
distortion. So while there isn't much wiggle room, I'm going to try
reducing the 330K resistor for less distortion. Beyond that, my only
other idea is to make the cathode less negative. I'm also going to
try decreasing the 1M for more distortion.

If some of you who have this module could measure the cathode and
grid voltages that would be appreciated.


Thanks :)