Paia EK-7 Digitizer (4 Bit ADC)

[Ruebezahl]

Ever since i came across the ingenious Electro Harmonix Clockworks, an kind of analogue clock divider link
, i was trying to take this concept of using a potentiometer as a voltage divider and kind of quantize the output to control digital chips, like for example the 4089, to make cool beats or other things in the spirit of lunetta. basically what i need is a low resolution ADC. While an DAC is kind of simple with the classic R2R ladder, the go to way to make an ADC with discrete components seems to be a comparator stage for every digital value followed by an En5coder. For a 4 bit word that already needs 15 comparators plus the encoder, so at least 5 chips, so not as simple as i hoped. But in this thread here somebody posted an article that was originally published in the polyphony magazine, complimenting a kit by Paia. I think the kit doesn't exist anymore, but it looked much simpler than the classic way. The Digitizer, how they call it, achieves 4 bit ADC with only four comparators and a couple resistors (almost 30), so one chip! Phobos said he tried it and it didn't work, which was a bit discouraging, since i trust his skills a lot

article:
https://paia.com/manuals/docs/EK-7_200.pdf

original thread:
https://electro-music.com/forum/viewtopic.php?highlight=adc&t=73314

Anyway i am a big boy and i found i actually have all the resistor values needed lying around so i gave it a try and... it kind of worked. Well i had to fix some obvious mistakes, but after i double triple checked, there was one last bug i just couldn't find an error: the last step was missing, it always froze at 1110 and never jumped to 1111, even if i connected the reference voltage right to the positive rail. so i figured there must be something wrong with the least significant digit and randomly swapped resistors around there. And it turns out if i swap the resistor R6 that is pulling the output D3 up for a much bigger number (100k worked for me), everything works now! i don't really know why the value is soo off the schematic, but to be honest i am also not 100% sure about a few things in the schematic, like why there are all this 39k resistors before the inputs, or those pullup resistors in general. in the 2 bit version they are not there. Actually i might try a version omitting them, might work just as well

Anyway, sorry for the wall of text, i am a bit excited. So long story short: I can confirm this schematic works, but i had to swap the R6 for 100k. Also i used an LM339 as a comparator, if anyone was wondering.

but: i only tested it with LEDs so far and it is obvious it's not a clear cut digital robot, but rather a wonky wannabee. the digital states represented by the LEDs were actually fading in and out each other, which is very interesting, especially since i buffered all the ouputs. Definitely looking forward to do more experiments. I can quite appreciate this remaining analogue ambiguiness!

---

paia 4 bit digitizer.png
Description:	Paia 4 Bit "Digitizer" aka Analog to Digital Converter
Filesize:	104.56 KB
Viewed:	70 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[PHOBoS]

ah yes that quirky thing, with it's magical, somewhat fractal like, resistor network.

[Ruebezahl]

Ahaha, you just got me with the diodes! I just straight ignored them, i thought it is just for reverse polarity protection, never even thought about why they stuck two in there

So it turns out they actually have a function: After studying the PCB layout i realized that the Comparators are powered by the 5V and the rest of the circuit takes its positive rail from the (diode dropped) voltage (8.8V but my supply voltage is 10V, i just tried 5V just in case and it works pretty much the same).

I do think it is not so clear in the schematic, since it doesn't say where the comparators power supply is coming from, but still, i could have thought about that earlier...

And moreover, it works now with the proper R6=2.2k, so basically all my pondering and experimenting was "for the cat", as we would say in German.

Speaking of cat, i shot a video of the operation, including my cat jumping in and stopping my operations for one hour, taking apart my breadboard to find the error. And in the end realizing i just turned the pot to one of the gaps between logic states and kept it there for the whole time i was trouble shooting!!!

So yeah, that is definitely a little annoying, between a few states (but only at two points actually) instead of giving out a clear cut 1 or 0, the outputs start to wildly oscillate (several megahertz sometimes), which is also audible in the sound in the end. Maybe i could just filter it out? But would be even nicer i could find a way to make the transition smoother, i mean it works for most of them!


So i have to correct myself:

[Ruebezahl]

[blue hell]

[Ruebezahl]

[Ruebezahl]

A little update!

I did the following revisions:

-Omitting all resistors on the inverting inputs (so far i couldn't see any bad influence from that), just to lower the parts count

-Changing some values of the resistors at the non inverting inputs (R4 and R9 in my new schematic), in order to make it more stable around the transition points

A bit about why i changed those: First of all, there where two transitions that were especially prone to self oscillation of the comparators:

0011-->0100
0111-->1000

so basically the two times D2 and D3 turned on respectively, so i was trying to change something about those two stages. In the datasheet of the LM339 is actually a nice graphic explaining how the feedback and input resistor set the hysteresis. So making the input resistor bigger results in a bigger hysteresis. Of course it shouldn't be too big, otherwise the jumping will be too off when the counting goes up or down respectively. But i found those values give me the best results in killing the oscillation without influencing other factors. Feel free to experiment with other values. Alternatively i guess one could also raise the feedback resistor values, but not many people have a varied assortment of >Megaohm Resistors, so i guess its easier that way.

Long story short: I got it working pretty smoothly, and even managed to lower the parts count a little, if you feel anxious about that you can also put them back in. The response might not be 100% linear, but for my purposes it is good enough and i think this circuit is pretty neat for what it is!

---

lm339 datasheet example.jpg
Description:
Filesize:	38.85 KB
Viewed:	14443 Time(s)

---

---

4 bit adc.jpg
Description:
Filesize:	71.97 KB
Viewed:	74 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[PHOBoS]

nice work!


I honestly have no idea why I used a zener or if I even used it to separate the power. I probably did test with ordinary diodes first though.

[Grumble]

Why don’t you use a TLC0820, that is an 8 bit flash ADC.
Sample time max. 2.5uSec, 5 volt supply.
Use the 4 highest bits…

---

tlc0820a.pdf
Description:	8bit flash adc	Download (listen)
Filename:	tlc0820a.pdf
Filesize:	1.16 MB
Downloaded:	65 Time(s)

---

[Ruebezahl]

[elektrouwe]