[sdiy] Signals leaking into the PSU?

Tue Feb 28 03:50:49 CET 2023

 Actually, crappy oscillators prone to pitch variations are a feature...
don't fuck with it. It's very musically useful and people pay crazy money
for stuff like that when it's vintage and has "the sound". I'll be blunt,
the only reason you think this doesn't sound cool is because you're not
listening to it in a mix. I know two different songs off the top of my head
that specifically create that sound that you have in the mp3 and are
absolute bangers that were played incessantly and constantly by everyone
for a good decade or so. For a musician, listening to VCOs alone in a
sterile recording is audio necrophilia, it's very similar to those people
who go to hi fi tradeshows, walk up to a $100k hi fi, and put in a cd with
full-range sweeps from 0 to 22.050 kHz (the 050 is so the bats can enjoy it
too).

Instead, buy/build a different VCO, which is properly decoupled.

The value of a drifty VCO becomes obvious only if you have a stable VCO
playing together with it.

You'll want the drifty VCO on bass lines and bass drums as well as on pads.

I haven't read the whole thread, but when decoupling, there's something
that a lot of people forget.

A decoupling capacitor works because it essentially forms a first-order low
pass filter between the rails and its circuits. So what you have is:

V+ ----> series R/L 1 ----> cap to ground ----> circuit ---->  series R/L 2
----> ground

If there's no physical resistor in position 1, what you get is just a
parasitic. That's a stupid design because it means the low pass filter
isn't very effective.

Or otherwise you can think of it this way: when the node "circuit" is
sourcing current, i.e. becomes a short circuit, you want it to be sourcing
the current as much as possible from "cap to ground". To do this, it has to
go through ESR:

cap to ground = positive terminal ----> ESR ----> capacitance ----> ESR
----> negative terminal

That ESR of a single electrolytic is usually on the order of 1-2 ohms.

If your parasitic resistance is 1-2 ohms too, then from the point of view
of "circuit", looking towards "V+", you have two ways you can source
current, and they are equivalent in difficulty, because they have the same
impedance between where you are, and the charge you want to get. So
"circuit" will source current equally from both. That's a mere 50%
reduction in noise, meaning merely 6dB! That's shitty. What you want is for
the current to be coming from "cap to ground" almost exclusively. That's
what you do:

1. make "cap to ground" actually two capacitors, paralleled. One larger
electrolytic, and one smaller ceramic. The ceramic will have low capacity,
but what it will also have is very low ESR, on the order of 0.1 or less
ohms, at higher frequencies than the electrolytic. This way you get a
"compound capacitor" of sorts that has very low ESR on spikes, which are
usually where most of the noise comes from anyways. The electrolytic
handles lower frequencies which need higher currents. Every capacitor has
its lowest ESR at a different frequency, and you want those ranges to be
different so they complement each other, like the frequency bands of an
equalizer. This is called a "bypass capacitor bank".

2. Add a 50-200 ohm resistor in "series R/L 1". Adjust to taste.

Now with a 200 ohm "series R/L 1" and a 0.1 ohm ESR "cap to ground" you
have a 200+0.1 : 0.1 or 2001 : 1 reduction. That's a 66 decibel reduction,
basically below noise floor, using parts that cost $0.01 total. The
capacitor will have to charge up before "circuit" sees a voltage rail at
full voltage, so don't play your synth within the first 0.1 seconds or so,
unless you specifically want to go for that experimental sound, that is.
There's value in that too. There should be a collaboration album of people
playing synths within the first minute of turning them on, during winter.

I don't know why no one's telling you to use series resistors.

As regarding the potentiometers being referenced to the voltage rail
directly, instead you can build a voltage reference... but this here should
be plenty stable for you anyways, at least as a start.

But again. Just get a different VCO. You'll be missing the shitty one
sooner than you think.

On Mon, Feb 20, 2023 at 6:15 PM A.M. Barrio via Synth-diy <
synth-diy at synth-diy.org> wrote:

   - To Mike Beauchamp, Chris McDowell, Roman Sowa: The VCO design comes
   from Moritz Klein's one. I'm attaching a schematic of the exact circuit I'm
   implementing. After reading all the responses it is quite obvious to me
   that it is VERY prone to changes in output frequency: the CV pots are tied
   directly to the rails, as well as the CD40106 IC in charge of generating
   the oscillation. Roman, I am taking absolutely no offense from your
   comment: it is indeed a bad design. I am not willing to improve *this*
   particular VCO design as it was just a test and a bad example of me being
   too excited to have my first panel mounted on my case. Please ignore the
   wasteful use of the 40106 and the crappy expo converter, much more care
   will be put into the next VCO I put together. About this: * If you need
   help deciding which ones those are, post the VCO schematics.* Would you
   mind pointing them out? I mentioned a couple spots above, but I'd like to
   make sure I cover them all. That being said, I'd love to hear about the
   reference voltages. Chris, thanks for the IC examples, I'll check them out.
   Besides, I have some spare 7805 regulators at home. Although they are quite
   big, would they do the job for testing before I go ahead and grab some of
   the examples you mentioned?

So, to summarize: I must look for a way to improve my bus board (uh
consider spending some money on a proper PSU/bus board). I must plug 100uF
to 470uF caps between each rail and GND on every module as well as the bus
board. 100nF caps between each IC and GND are recommended. Sensitive parts
of circuits should never be tied directly to the power rails, but to a
voltage regulator instead. Would these measures remove (or at least notably
reduce) the problem I'm facing?

Thanks a lot beforehand,

A.M. Barrio.

 VCO.png
<https://drive.google.com/file/d/1EBAwnrlCLTUUEoa8b8sTuSTGzpBY5-kL/view?usp=drive_web>

>
> On Sun, Feb 19, 2023 at 5:21 PM A.M. Barrio <albertomunozbarrio at gmail.com>
> wrote:
>
>> This is the complete email I wanted to send. There's another thread where
>> the attached files and additional info is missing. My bad, pressed CTRL +
>> Enter by mistake whoops. Please ignore that one! pata at ieee.org and
>> mbryant at futurehorizons.com, thank you for your replies, I have taken
>> them into account :-)
>>
>> ------------------------------------
>>
>> Hello,
>> I'm having an issue where the frequency of my VCO is being affected by
>> the operation of other modules. I'll give a broader explanation:
>>
>> I'm a beginner on synth DIY, right now I have my home made case, a PSU,
>> two VCOs, an LFO, a clock module and an amplifier/speaker (built in the
>> case, not externally). Everything has been arranged by me, I haven't
>> purchased any module yet. The clock and the LFO are the latest modules I
>> have built (pretty much both at the same time), and when I tested them on
>> their own, they worked just fine. However I noticed that when they are
>> powered, the frequency of the VCO would stutter following the operation of
>> the clock and the LFO.
>>
>> That would be: *without connecting the clock or the LFO anywhere*, just
>> having them powered up, whenever the clock is up or down, or the waves
>> generated by the LFO reset, the frequency of the VCO varies a tiny bit. It
>> gets more noticeable the higher the pitch of the VCO.
>>
>> After some head scratching I've realised that the signals generated by
>> any of the modules I have (both VCOs, the clock and the LFO) are somehow
>> leaking into the +12 and -12 rails of my PSU. Measuring any of the rails on
>> my scope (on AC mode, so only the noise is shown), I can see that the waves
>> generated by the modules I mentioned above are there as well in the form of
>> noise (around 20mVpp each of them). That explains the little variations in
>> frequency of the VCO. However I have no idea why this is happening or how
>> to fix it. I have tried with different PSUs I have around:
>>
>>
>>    - 11V 0.750A SMPS into DCDC to get +12 and -12
>>    - Two 12V 1A SMPS together to get +12 and -12
>>    - Linear PSU with 7812 and 7912
>>
>> I have the same problem with all of them. I'm positive there's an issue
>> somewhere with a lack of filtering, but I don't know where, or why.
>>
>> I'm attaching a sample of the VCO output where the stutter can be heard
>> and the schematic of the LFO. The way the LFO is set up in terms of caps
>> and isolation can be extrapolated to every other circuit I have made, I
>> always place those 100nF caps between each rail and ground.
>>
>>
>>  LFO (square).jpg
>> <https://drive.google.com/file/d/1ts1WGcyl2y1Ypp5eP79Ptb5fHXawetZh/view?usp=drive_web>
>>
>>  VCO stutter.mp3
>> <https://drive.google.com/file/d/1RrUNX_Of_RGLAwn2CyV7I5fxRrlHvfHL/view?usp=drive_web>
>>
>>
>> In case you have any idea what could be wrong, your reply will be greatly
>> appreciated. Thanks beforehand!
>>
>> Kind regards,
>>
>> A.M. Barrio.
>>
>> ------------------------------------
>>
>> In response to pata and Mike Bryant:
>>
>>    - Yes, the current PSU I'm using can provide 8W in total. I'm only
>>    powering one VCO and one LFO with some LEDs, power shouldn't be the issue
>>    in this case.
>>    - I always place 100nF caps between each rail and GND on every
>>    circuit/board. However I hadn't heard of placing caps on each IC. I suppose
>>    it should be between their power pins and GND? I will have a look at
>>    capacitance multipliers as well as I haven't heard about them before.
>>
>> _______________________________________________
> Synth-diy mailing list
> Synth-diy at synth-diy.org
> http://synth-diy.org/mailman/listinfo/synth-diy
> Selling or trading? Use marketplace at synth-diy.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://synth-diy.org/pipermail/synth-diy/attachments/20230228/9b53deda/attachment.htm>