Summing Amplifier Channel Isolation Problem

Thread Starter

violapwr

Joined Apr 8, 2021
3
I'm designing a quad-preamp circuit (for an experimental musical instrument). Each pre-amp channel is read by the analog input of a microcontroller before being summed together.

The issue is that there is a ton of crosstalk being reported by the microcontroller between the four channels. I'm wondering if there is a problem with the values of the summing amplifier's input resistors? Any other ideas?

Thank you!
 

Attachments

WBahn

Joined Mar 31, 2012
29,489
Your opamp can require as much as 500 nA of input bias current. With all of the resistors connected to the inverting input being 10 MΩ, you are very possibly starving it of the bias current it needs.

Are you trying to convert the input signal to some kind of PWM or other digital signal in the pre-amps? If not, why are you using a digital inverter? IIRC, the 4069 is not a Schmitt trigger device, so when you operate it near the transition point with a slowly changing signal, it can oscillate -- which is going to be reflected in every device on the chip.
 

Ian0

Joined Aug 7, 2020
8,940
Yes - 10MΩ is not a good choice, it will pick up a fly sneezing in the next room. Keep it below 100k.
CMOS unbuffered inverters do make reasonable amplifiers, but when they are biassed into the linear mode they will take a lot of current, which means that they can easily superimpose their signal on the power supply, where it will be picked up by other inverters in the same IC. It gets worse as the power supply voltage increases.
Why didn't you just use op-amps? That's what they are meant for!
 

AnalogKid

Joined Aug 1, 2013
10,777
Are you trying to convert the input signal to some kind of PWM or other digital signal in the pre-amps? If not, why are you using a digital inverter? IIRC, the 4069 is not a Schmitt trigger device, so when you operate it near the transition point with a slowly changing signal, it can oscillate -- which is going to be reflected in every device on the chip.
Nope.

He is using an unbuffered CMOS inverter as a linear inverting amplifier. This really does work, but not very well. Waaaay back, as a lab project in school, I built one channel of a home stereo amplifier using nothing but these gates. Phono preamp with RIAA compensation, tone controls and a low current class D "power" amplifier. Gain issues, bandwidth issues, etc., but the little puppy actually made audio.

But I digress . . .

The schematic has some issues. There is no decoupling on the V+ (Vcc/2) source, which makes it a target for crosstalk problems. There also is no decoupling at the device power pins, another crosstalk source. And some of the resistances are waaaaay too high. The thing to remember is that a CMOS gate in linear amp mode has a most zero power supply and common mode rejection.

If this circuit approach is the TS's original thought, gold star. But there are a ton of cheap quad opamps that will improve the fidelity 1000%.

ak
 
Last edited:

Thread Starter

violapwr

Joined Apr 8, 2021
3
Thank you all for your quick responses. I should add, one confusing thing about this circuit is that it worked great when I breadboarded it. But these crosstalk issues arose once I made a PCB version. I tried to separate the leads on the PCB so the four channels were isolated, however that hasn't seemed to help.

The sense I'm getting from @AnalogKid and @Ian0 is that op-amps would work much better (in terms of my crosstalk issue) than the CD4049's. If I were to go this route, are there any specific op-amps recommended?

As an aside, I had gotten the idea to use the CD4049's from Nic Collin's book Handmade Electronic Music, and liked the sound of them when overdriven (and to clarify, there are no digital signals here @WBahn, all audio).

Also, if I were try and make this current CD4049 approach work, how might you all tackle it? (@AnalogKid I had included two 100nF decoupling capacitors, however, I had mistakenly removed them from the 1st schematic I posted. New attached schematic has them.)

@hrs that for sure sounds promising. Do you think any other schematic modifications would be needed? Or just swap the CD4049 for the CD4069?

Thanks again!
 

Attachments

hrs

Joined Jun 13, 2014
390
You may still want to add a capacitor between V+ and ground.
@hrs that for sure sounds promising. Do you think any other schematic modifications would be needed? Or just swap the CD4049 for the CD4069?
You can swap the CD4049 for the CD4069. The CD4069 has a symetric output so will be more linear. The CD4049 has an asymetric output and may give you more overdrive if thats what you're looking for. From my experiments with the CD4049 up to 5 volts would be ok, above 6 volts I ran the risk of burning the chip. Your CD4049 at 6 volts may draw a different current than mine.
 

AnalogKid

Joined Aug 1, 2013
10,777
You added two decoupling caps, but you need at least two more.

One cap for each IC. These go as close to each IC.s power and GND pins as possible, with the shortest possible leads.
One for the V+ node. Actually, two. 1 - 100 uF electrolytic and 1 - 0.1 uf ceramic.

Opamps - more of the 4558's should work fine in this application. Consider dropping all of those 10 M resistors to 110 K, and scaling the other gain-setting parts everything else to match.

ak
 

MisterBill2

Joined Jan 23, 2018
16,495
I would not attempt to get useful performance from a scheme using inverters as amplifiers .
Use a quad op-amp IC .Then each channel will be linear. And standard mixer circuits will perform in expected manners.
 

Thread Starter

violapwr

Joined Apr 8, 2021
3
Solid advice - thank you all! I'll try do a few versions of the design based on this feedback. I'll get back if anything -- I appreciate the assistance.
 
Top