sending a signal through alot of electro caps?!?

[Quackzed]

i'm trying to do up a remote bypass using a 4053 as in r.g.'s 'Bypassing and Switching with the CD4053 CMOS Analog MUX'. found here...
http://www.geofex.com/article_folders/cd4053/cd4053.htm
Basically I want to string up 10 of these in series and am starting to see that, due to the requirements of the 4053 needing to be decoupled and biased to VR (1/2V) at the chip, the signal will be going through 2x 2.2uf electro's each 4053 i use, for 10 in series thats 20 electro caps! when all effects are bypassed and if all effects are engaged it's 40!!   
i'm thinking that this is really pushing the 'can you hear a difference' argument well past the breaking point!
signal goes in - 1M to Ground - 2.2uf - 1M to VR - 4053 - 1M to VR - 2.2uf - 1M to ground- out
for each 4053 switch in bypass
and goes   in - 1M to Ground - 2.2uf - 1M to VR - 4053 - - 1M to VR - 2.2uf - 1M to ground- send (fx box)
return - 1M to Ground - 2.2uf - 1M to VR - 4053 - 1M to VR - 2.2uf - 1M to ground- out
for each 4053 when 'effect engaged'...
what's the best way to avoid all these electros in the signal path, or do I need to ?

[R.G.]

There are a couple of ways of going about this.

First, counting electrolytics in an audio chain gets futile. They're everywhere. It's only in recent years that it has been measured that electrolytics do have a distortion mechanism built in. It's small enough that it took a while to measure it. But it is there; it's related to the AC voltage across the cap. If this is small, the distortion gets even smaller and is almost undetectable. So the trick in using electros undetectably in the audio is to make the values about 10x what you need for passing the lowest frequencies they need to pass, and never, ever using an electro for filtering or cutoff duties. Doing this runs the distortion back down to very, very small.

Second: us being guitarists and searching for distortion, who's to say that electrolytic distortion is a bad thing?  The hifi tweako crowd thinks that everything that's affordable is junk, and only incredibly expensive stuff is any good, but we guitarists like some distortions. Probably worth TRYING to get electro distortion. It might be good.

Third: Be sneaky. The CD4053 has a power option available that lets you run it from bipolar supplies up to +/-7.5V (or other combinations up to a total of 15V). You do this by supplying a negative voltage to the Vee pin. So if you ran your 4053 switches from +/- 5V to +/-7V, you could leave the signals at ground for DC, and have NO electrolytics in the path, other than the two that are probably in every pedal. All you'd need would be pull-downs, and if the pedals have proper pull-downs and the amp is also pulled down with a 1M or so, you wouldn't need that.

Several ways...

[Quackzed]

Thank you for that R.G. As always , your experience and explanation of these mechanisms offers some clarity and perspective that is indeed a comfort. I'm not going to worry about the 'distortion' from the electrolytics unless i hear anything objectionable, which i doubt i will. ill check out running it with a bipolar supply as well, save some parts etc... though i recall you mentioning it can be a bit tricky... time will tell if i can keep the chip happy.
Once i get this project together i'll post a proper write up with modular type layouts to make it easier for others down the road to build a single or multiple remote bypass thing using the 4053 like i'm planning. good stuff!
Thanks Again.     

[R.G.]

The real PITA is that the CD405x family will not run from +9 and -8.5 so you could get your bipolar from a charge pump converter. It's really got to be no bigger than +7.5, -7.5V for operations inside the datasheet limits.

[PRR]

> the CD405x family will not run from +9 and -8.5 ...It's really got to be no bigger than +7.5, -7.5V for operations inside the datasheet limits.

I remembered it as 18V. So +/-9.0000, or +9/-8.5 with a hair to spare.

However today's TI CD4051 datasheet shows (page 5) 20V Ab Max, and

Features
. Wide Range of Digital and Analog Signal Levels
- Digital . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3V to 20V
- Analog. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20VP-P

I'd think a good 9V (battery or good-brand spot-box) and a flip would be safe. (A GoodWill "9V" wart probably would not.)

And then again, if ONLY powering the CMOS, some very simple techniques could knock it down. I would consider Zeners if I could find 7V with specs for very low current. There are very low-power pass and shunt regulators, though nothing I'm likely to find in my shoeboxes. If power is not too precious, 10K:33K voltage-dividers would power a lot of very slow CMOS at +/-6V.

[PRR]

> avoid all these electros in the signal path

Do not forget the switches!

One CMOS switch with ample supplies into 1Meg is low-low THD.

It sounds like you may have 20 switches. And possible they terminate in a medium impedance.

The PDF that I linked above, page 7, shows numbers like 0.1%-0.2% into 10K. For 20 switches cascaded into 10K that would be several percent. Part-percent into 100K. Far greater than any e-cap chain that I've met.

Also there's 100-200 ohms in each switch. Twenty is several K Ohms series resistance. Into 10K load you have audible loss.

You can buffer much of this out. Lots more pins.

Have you considered relays?

Or just pretend that both e-cap and CMOS THD tends to be inoffensive, and it's only rock-n-roll (reggae, waltz, whatever), and like it.

[bool]

relays..

[R.G.]

I remembered it as 18V. So +/-9.0000, or +9/-8.5 with a hair to spare.
...
- Digital . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3V to 20V
- Analog. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .20VP-P

I'd think a good 9V (battery or good-brand spot-box) and a flip would be safe. (A GoodWill "9V" wart probably would not.)
[/quote]
Fresh alkaline batteries are 9.4-9.6V. Still, with a 20 V rating, it's OK. But you have to be sure you have a TI date code after the changes.

And then again, if ONLY powering the CMOS, some very simple techniques could knock it down. I would consider Zeners if I could find 7V with specs for very low current. There are very low-power pass and shunt regulators, though nothing I'm likely to find in my shoeboxes. If power is not too precious, 10K:33K voltage-dividers would power a lot of very slow CMOS at +/-6V.

Yep.

Do not forget the switches!
One CMOS switch with ample supplies into 1Meg is low-low THD.
It sounds like you may have 20 switches. And possible they terminate in a medium impedance.
The PDF that I linked above, page 7, shows numbers like 0.1%-0.2% into 10K. For 20 switches cascaded into 10K that would be several percent. Part-percent into 100K. Far greater than any e-cap chain that I've met.
Also there's 100-200 ohms in each switch. Twenty is several K Ohms series resistance. Into 10K load you have audible loss.
You can buffer much of this out. Lots more pins.

An extremely good solution is to put one buffer before the string of switches and one at the end. Eight more pins, good performance.

Relays have their own set of challenges.

Frankly, a string of P-channel JFETs might be a better solution.

[Mark Hammer]

My understanding, from the dim recesses of when I used to be an avid reader of audio magazines, was that having a lot of electros in series with the signal, would incur a certain amount of "group delay".  Indeed, differential group phase delay was the very reason that BBE Sonic Maximizers were developed.

But of course, having a misalignment between the higher and lower frequency content, such that harmonics become harder to perceptually "assign" to fundamentals, is really more problematic when a) the bandwidth of the signal is noticeable broader than what your typical guitar and guitar amp will present, and b) the audio signal being processed involves multiple sources mixed down to mono, such as when dealing with music reproduction.

One way of dealing with the group delay issue, though, as I understand it, should it actually be a problem, is to use smaller value caps in parallel with the electros.  So (and I'm pulling a number out of my backside here, so don't treat this as gospel), a decent 100nf cap in parallel with the 2u2.  But note that this is ONLY for purposes of helping all harmonic content "arrive" at the same time as the fundamentals they are married to, and is completely different than any distortion issues RG addressed.

[Quackzed]

Some good points! I had been just looking at one section, working out the layout, high pass, led driver etc... all seemed well.
But once i started to look at the whole shebang I started to realize some of the little issues were getting BIG.
And yes, 20 switches basically.
so, for the sake of argument, lets keep the 4053s in the mix and see what i'm looking at.
1. the electros
   -two in the path each bypass/effect. i could do bi-polar power and try to get rid of 'em but that leaves the real possibility of plugging in something that'll pop without 'em. With them in thats 20 cascaded caps! And its 20 cascaded high pass filters as well ( 2.2uf - 1M to ground) steep! but as long as its low enough, sure! Electro distortion? rock and roll baby!!! 
2. switch resistance.
    -say 200ohms per switch, 400ohms per bypass/effect block (though ill need to look hard at the datasheet, see if i can get that as low as possible). 4k ouch! buffer up front (my fuzzes are complaining) maybee buffer after the first block? (ok they're happy again) and less problems with volume/high end fluctuations between pedals that can drive it and passive guitars that can't.
still, not too bad with a buffered signal. or is it? 200ohms/500k (parallel 1M's) fine, BUT the pedals may be alot lower input imp. for one percent loss it'd need to be @20k? unlikelly, but 100k might be reasonable and that'd be okish till you stack 20 of em up...
hmmm....

[Quackzed]

the chips are cd4053be 's got 'em from an ebay seller a few weeks ago, but is there any benifit to going bi-polar other than losing the vr resistors? as losing the caps seems like looking for trouble down the line.  no pun intended.
and looks like the further down the line you get placed, the more volume loss/ loading you get. maybee a second buffer after the first 5 ? and with a buffer at the end to compensate?
i hadn't really thought about phase shift, only in the sense of a bunch of cascaded high pass filters/ low end rolloff...
and yeah, if 2 cascaded hpf's do 180 degrees then 20 will do 1800 degrees?!? even kevin bacon could probably hear that as a time delay , out of my depth there...
but a few buffers wouldn't be too much more hassle really, and offers some peace of mind?