CD4053 as an Audio Switch

[@p_mouse8676]

I have actually seen a pretty nice paper by Texas Instruments, where they put these kinds of switches inside the feedback loop of an opamp. This drastically improves linearity and distortion, which often isn't that great with these switches.

[@GadgetReboot]

there’s also some amount of signal isolation bleed through when a switch is off for example if I put a 1 V sine wave in and have a switch disconnected I still might see 100 mV getting through on the scope but in this application it didn’t seem to interfere. Plus adding distortion may actually enhance the overall guitar sound!

[@p_mouse8676]

@GadgetReboot it may or it may not. Without (or at least not a significant amount) we know it will be the same as the input.

[@DrexProjects]

Very cool project.

[@martingerken7094]

Thanks for the thorough explanation. Will the DGxxx switches do better?

[@GadgetReboot]

i’ve never used the DG parts, the ones I looked at seemed very expensive although they had better on resistance and higher voltage and better channel isolation specs. another one I’m looking at is TMUX4053 with slightly more supply voltage capability than CD and better channel isolation again and improved on resistance but it looks like a regular price as long as it’s a surface mount design. I don’t think it comes as through hole.

[@TheEmbeddedHobbyist]

Just an opinion here but the circuit would look nicer with the inputs on the left and the outputs on the right of the circuit diagram. this is the normal way a circuit flows and makes it easer in my opinion to read. but I'm just old and set in my ways. 🙂 edit: forgot to add nice idea.

[@GadgetReboot]

In this case it would be hard because the audio headers are both inputs and outputs on the same headers, but generally I draw left to right with the signal. I also wasn't sure how I felt about directly wiring everything, causing lines all over the place instead of just having net labels and a choose your own adventure signal path hunt.

[@TheEmbeddedHobbyist]

@@GadgetReboot Don't get me started on the lots of boxes and net labels, KZbin university style of circuit design. Each track is a line and if there is more than one in the function like an address buss then you use a line defined as a bus. it makes the circuit so much easer to read and IMHO less likely to give you connectivity errors. Again just my opinion as an old design engineer. 🙂

[@GadgetReboot]

when I was working in a fancy place where they did those group design reviews all the time and there were 5 to 10 designers all taking turns in different meetings going through their schematic for various boards in the whole system, all of the different artistic ways of doing schematics and resulting discussions would always happen. One person would pull up their schematic and it would be directly drawn individual wires all over the place, the next person would pull up a schematic where all of the circuit blocks are separated by dashed lines and there are net labels to get in and out of each section and things like that. and the oldest timer in the group would complain that all of the schematics should be drawn with the circuit blocks separated in dashed lines with net labels and have text saying what each little box function does and then a 10 minute debate would go on, nothing would change and on the next design review everyone would be back to doing it the way they want.

[@TheEmbeddedHobbyist]

@@GadgetReboot We had it easy, we had design guides which you had to use. This ensured that all drawings looked the same. But as a engineer i had it easy, I just drew the circuit on A3 or A0 and put comments on how much current a track was expected to draw. These went in to the drawing office to be draw and given there final drawing number. They then came back to be checked and signed by the engineer before review and then signed off by quality. We had a quality system for going to have a pee 🙂

[@HazeAnderson]

The signal flow is still primarily left to right, very normal, very easy to read ... instead of headers, GR could use TRS jacks (tip for input and ring for output) and some breakout cables. 👍

[@AnalogDude_]

You don't need R5, you because you used R10, R12, R14 to tie in the reference voltage unless you used a cap before R1. You could use "LT1054" to generate the negative voltage for the Vee pin on the CD4053 and skip the reference voltage all together, maybe your problem is than solved. the Vss pin on the CD4053 is to operate the internal logic.

[@GadgetReboot]

I was also trying to keep the same reference voltage on both sides of the switches in case there may be any clicks or pops when they are switching if voltages are drifting around on one side but it’s still experimental and more testing is being done. If I did only want to use one I would probably keep R5 and get rid of the one on the other side of the switch because R5 is a lower value resistor which I found worked better with that buffer. i’m still considering using a negative voltage generator for more headroom as well as keeping ground reference at 0V and making sure the switching frequency is out of the audio range of course.

[@AnalogDude_]

@@GadgetReboot You don't have too, R5 is redundant and maybe a error, the voltage you put in the cd4053 is also coming out, they are fets in a certain configuration. you could place a double low pass filter after the negative voltage generation, just use a online calculator to cut off above 1 or 2 Hz, the cap will be large if you choose small resistors to allow current to pass.

[@AnalogDude_]

@@GadgetReboot it's not necessary that the negative voltage is equal to the positive voltage. it works just fine if you use +9V and -8V (for example).

[@AnalogDude_]

@@GadgetReboot I did a simulation with 2 low pass filters in series, using 180 Ohm and 820uF, the cutoff is 1.08 Hz and still have caps smaller than 1000uF, but if you draw 7.x milli amp, the voltage drops from 5V to 2.4V. Maybe it needs inductors.

[@AnalogDude_]

@@GadgetReboot reading some more, LM2776 is made for audio (and to be operated from batteries if needed), 200 mA instead of 100 mA for the LT1054. no filters needed.