4093 - Strange Behaviour

[ianbax]

Hello all,

I'll built a bank of slow 'LFO' 4093 oscillators (100k pot, 100uf capacitor) to use as clocks and another bank in audio range (100k pot, 100nf cap).

As per Nicholas Collins and many others, the idea is that the output of the first can be plugged into one side of the 4093 and gate the audio ones.

They're on two separate chips built as 'modules' and I'm using the tip of 3.5mm jacks to make the connections.

The gate input (one side of the 4093) is normaled to 5v (the power supply) so unless a gate is plugged in the audio range oscillator drones away. When I plug in the gate, it works, but the pitch drops quite significantly I think I can even hear a bit of a ramp up rather than a clean switch on and off.

One other thing - on the LFO module there are indicator LEDs - so the output is about a 3v square wave because of the LED voltage drop (it's just wired straight to the output, no separate driver). I thought perhaps I was running too close to the switching threshold of the 4093 but I tried upping the voltage and it was the same.

Now from a bit of reading *I think* I know what's going on - it's some sort of 'loading' and impedance issue - The reason I say that is I seem to have solved the problem by plugging a buffer (either two stages of a 4049 or recently discovered 4050 non-inverting) between the stages i.e. on the gate input of the 4093. I'm just curious what's going on - I didn't think there was an issue of CMOS gates driving other gates - like you'd have to get to trying to drive 20 other gates before fanout limits became an issue (according to my R.A. Penfold book).

Can someone with more advanced electronic knowledge shed any light? I'm kinda hungry for a proper explanation. As I build more modules do I need to think about buffering all gate inputs - say, adding a clock divider from a 4040 or using the 'LFO' to clock a 4017 sequencer. That is potentially an extra IC each time and more connections. Perhaps this only effects oscillator circuits because of the external components (big caps?) involved?

[PHOBoS]

If the LFOs are at their highest speeds it requires the chip to charge/dicharge the 100uF caps with a higher current, combined with the voltage
drop caused by the LEDs this might be an problem. Also with a 5V supply a 3V logic level is pushing it a bit but since they are schmitt trigger
inputs it should be less of an issue.

A drop in pitch is interesting if you look at what could affect the pitch in the first place. With these kind of oscillators the first thing to
suspect would be the supply voltage and a drop in pitch would suggest a rise in the voltage. Unless there is something funny going on
with the normalling that doesn't seem very likely. Whatever you do to the other input of the NAND gate shouldn't really affect its
frequency (unless you are modulating it with a high frequency). The only other way that I see that could affect it is through the output
of the gate as that is part of the circuit that makes it oscillate. If you load it too much it could affect the frequency.

[ianbax]

Thanks for taking the time Phobos.

Advice taken on 5v supplies - I as thinking ahead to interfacing with an Arduino which is why I plumped for 5v.

As it was within tolerance of the resistors I chose for the LEDs I bumped it up to 9v but got the same result.

It's difficult to take the LED out of the equation with this soldered up version (I only noticed this behaviour after I'd finished the LFO module with a face plate and everything). I'll probably breadboard something up to satisfy my curiosity.

I did breadboard up a quick experiment dividing these LFO clocks with a 4040 and all was well - as you say 3v pushing it but it worked.

I did a schematic (it's a bit crappy - I wasn't sure how to represent the separation of the circuits and their interconnection with a patch lead). As the buffer between the two works the only conclusion I come to is that somehow they interact once connected. I twigged this as in the logic noise series and the article on filters:

https://hackaday.com/2015/03/25/logic-noise-filters-and-drums/#:~:text=Logic%20Noise%20is%20an%20exploration,filter%20with%20a%20single%20part.

in the video (not in the text) Elliot notices his filter 'loading' the oscillator and altering it's pitch - he adds a buffer between the two to solve the problem.

So I've got a solution at least. I've rigged up a 4050 between the input and the gate.

---

nand_circuit.png
Description:
Filesize:	34.21 KB
Viewed:	158 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

---

nand_circuit.png
Description:
Filesize:	34.21 KB
Viewed:	145 Time(s)
This image has been reduced to fit the page. Click on it to enlarge.

---

[ianbax]

Reading back - to be super clear. The pitch change is noticed when you switch from the constant 5v (the normalled connection) to the external input (square wave from the 'lfo').

So at 5v constant into one side of the 4093, the oscillation is at a higher pitch than when the square wave is applied.

It could be as simple as the gated square wave is roughly 3v peak to peak and the constant source is 5v but I thought that as long as one side of the 4093 is higher than it's threshold it'll read high and that's that.

[ianbax]

As a follow up to this - just in case anyone searches for the same problem - after getting p'd off with the complexity of the circuit needing extra ICs to buffer I went 'back to basics' and looked again at oscillator options.

I breadboarded up the 40106 oscillator with a diode on the input to achieve gating which sound wise (and perhaps electrically) is much the same as the 4093 circuit. For whatever reason (internals in the 4093?) the pitch issue is corrected and a much more compact circuit is achievable - a N4148 and resistor on each input rather than an IC.

Diodes as gates - there's schematics everywhere I followed this one:

https://hackaday.com/2015/02/04/logic-noise-sweet-sweet-oscillator-sounds/

No normalling is needed to +5v either as the default of the oscillator is continuous if no signal is connected to the diode. So a bit of wiring saved.

Now I have another issue with switching noise at high frequencies but I think that's an unavoidable consequence of switching a square wave on and off quickly in this fashion. [

[PHOBoS]

[ianbax]

[PHOBoS]

ok good, so you have resistors.
Now for the value, I'd say 330 is too low @5V.
If we take some averages and presume the voltage drop across the LED is 2.5V that gives you about 7.5mA.
Should be fine for the LED but a quick look at one of the datasheets for CD4093 gives me a typical output current of about 0.88mA @5V
You can push this a bit, I assume you are not building a medical grade device, but 7.5ma is quite a bit more, hence the voltage drop
on the output and it could cause other unwanted behaviour (though it doesn't really explain the change in pitch to me). For a 1mA current
the resistor should be around 2K5 so maybe use 2K2 as the lowest. Of course it also depends on how bright you want the LED to be but
with todays LEDs you usually don't need a lot of current to get blinded.

[piedwagtail]

In my experience - NANDulators with 12-15V supplies can be hard to wake if the charge capacitors are in the xxxuf range.

R

[ianbax]

[zaphod betamax]

Even at 5volts I find 1k too blinding on a RED LED.
I try for 2.2k or even 4.7k