I forgot to put this on github, but here it is, it works, I think:
2 Likes
May I make the suggestion to plot the schematic as a PDF and include that in the repo?
2 Likes
Yes, I should do it for all my modules, but same with the BOMs, repetitive, boring work tends to be avoided 
I added it to this repo now at least 
4 Likes
This thread inadvertently sent me down a deep rabbit hole:
Quantizer â Voltage Follower Design - MOD WIGGLER
Now that Iâm on the other side, I think that the purpose of the 100R resistors must be to ensure that the behaviour of the three inputs without attenuation is as similar as possible to the input which does. Theyâre the only difference, so if thatâs not their purpose, then what is?
They would make the opamps work just a little bit harder, as would the A50k voltage divider. Any measurable imprecision introduced is compensated for by the trimpot in the output opamp feedback loop. At least thatâs what I think is happening.
Referring to the schematic here, R12 and R15 make some sense as current limiting resistors for output stage op amps. In the loop means they wonât create a voltage divider with the typically 100k input impedance of the downstream module, thereby distorting 1 V/oct signals.
(Matthew Skala argues such current limiting resistors should be larger, like 1k, because if the output is shorted to ±12 V, the voltage drop across the resistor will probably lead to power dissipation well above the 1/4 W or less the resistorâs probably rated for; itâll burn the resistor out.)
But R2, R4, R6, and R9 canât be explained the same way since those stages are followed by 100k resistors. Notice the attenuator R19 is outside the loop, and itâs 500x larger value, so Iâm skeptical the 100R resistors are supposed to be making those inputs more alike. And that doesnât explain R9 anyway.
And if 100R in the loop is good in the non-output stages, why isnât there one in the top input stage? Itâs the only stage that lacks one.
Yes, given that the resistors are on the output opamps as well, they canât also be doing what I suggested earlier. I think that itâs actually just bad design on Doepferâs part. All that any of those 100R resistors are really doing is increasing the output resistance of the opamp that they are attached to.
I think itâs a bit dodgy that there is no compensation for stray capacitance. This should be an improvement:
Precision Adder â Falstad Simulation
PS. The schematic that zedius posted is erroneous â Sebastianâs hand drawn one (and the one that he put on github) shows the correct logic.
2 Likes
I was having some difficulty calibrating my module, and realised that I made a mistake with the voltage reference. The 16k resistor chokes the current so much that the TL431 simply isnât able to function. What this means is that while the circuit does produce 1V, it does so simply by voltage-dividing the positive supply, which is not what I wanted to do at all! Sebastianâs way works. I donât know what I was thinking. 
2 Likes
got it figured out thats always nice thing !
2 Likes
Analog Devices has a fixed 1 V ± 0.35% reference:
Itâs $2.15, which maybe is more than some will want to pay, and itâs surface mount. But SOT-23-3, not hard to do.
By contrast the TL431 (adjustable > 2.5 V), which Sebastian used with a voltage divider, is 1%, though maybe if you trim your voltage divider to get 1 V ± 0.1% itâll stay there? Or you could divide down an LM4040C20 (2.048 V, 0.5%). Both are cheaper than the ADR510 and through hole (higher precision LM4040B20 and LM4040A20 seem to be SMT only), though you would need to add a trim pot and maybe a couple fixed resistors. The voltage divider impedance shouldnât matter since the outputâs going straight into an op amp buffer.
1 Like
Thanks for the suggestions. At some point Iâll probably give in and start using surface mount parts! I think that itâs the internal voltage reference of the TL431 that the 1% tolerance applies to, so once trimmed it should be fine. I like what Sebastian did with the fixed resistors and the trim pot (what youâre suggesting as well, I think) where the trim pot just needs to account for the tolerance of the resistors and the reference, and so can be a low resistance, and therefore very precise.
I think you might even get away with a diode and a trim potâŠ
Thereâs one more important thing that this module needs in order to function properly, which is a pull-down resistor on the Subtraction summing bus. Otherwise when all switches are off, the voltage being supplied to the summing buses is undefined. If you donât have that proper 0V to start off with, then any measurements you make in an effort to calibrate it will be relative to that initial error, and you wonât even be able to use it to calibrate an oscillator, let alone add sequences precisely. The op amp that you use is fairly significant. I did an A-B test between the LT1014 and a TL074. These were the results:
LT1014
0.000
1.000
2.000
3.000
TL074
0.002
0.996
1.995
2.994
This is the circuit that I built:
2 Likes