Opinions on analog switch ICs vs Opto-Isolators to controls effects chain.

[PeterPan]

Simple story... I won't even include a diagram. I want to be able to use an MCU to control individual pedals being "in-chain" or bypassed. I've noted that inside most "vintage" stomp box's that contain mechanical SPDT switches, the input always feeds the pedal even when the effect is off. So the switch simply connects the output jack to either the effect's output or the input. So to re-cap, I just want the pedal user to be able to just keep all the pedals "on", while I handle the "in-chain" or bypass of each one externally.

So... I first thought to do the analog switching with pairs of opto-isolators, the kind with an encapsulated LED and Cadmium resistive cells. I figured it would be easy to make one OPTO handle the bypass while the other handles the in-chain/enable, as long as I drive them with inverted logic (only one on at a time). then at the end of the chain a simple resistor to ground would keep the operation stable, while the input and the output of the whole chain (of hopfully many pedals) would be buffered by OP-AMPs.

I should say that I THOUGHT the optos would be a nice approach because of their slow response time of opto isplators. I reasoned that the slow response would enable a kind of "smooth" transition for each pedal controlled, so if someone hits a pedal in the middle of a chord, it would be less obtrusive a change. Well that may have been true, but I foolishly forgot the feedback problem during the short time where BOTH the bypass and the enable Opto's would be in transition. Sure enough, during that transition, many pedals that will "chirp" for a split second, because its input and output are momentarily bridged. Not good!

So now I'm back to my original plan of using some real analog audio switching ICs. Since my circuit already has 5V available, with rail-rail op-amps biased to 2.5V, I thought the Analog devices ADG884 http://www.analog.com/en/switchesmultiplexers/analog-switches/adg884/products/product.html might be a good choice.  Each chip has two analog SPDT switches, have much better "ON" resistance than any onto-isolator, and they even have "break before make" switching properties. These chips are $2.70 each (Qty 1) at Mouser.com, which isn't bad considering each one will replace 4 separate OPTOs, and I won't need that "inverter" logic I previously needed to turn one OPTO on while the other went OFF. So it seems this will be an easier way, except that by using these chips I will lose that "smooth transition" feature I thought I'd gain with opto isolators.

Anyway, before I order some of these ADG-884s I thought I'd check with you folks, as I'm sure I'm not the only who's done this sort of thing. Does anyone have any other suggestions for better choices in analog switch ICs? Any similar chips like this I may have missed?

[merlinb]

Sure enough, during that transition, many pedals that will "chirp" for a split second, because its input and output are momentarily bridged. Not good!

But if you're using a uC, why diodn't you just arrange for the LEDs to turn on/off  at slightly different times (a few milliseconds) to prevent the feedback? Also, why do you need inverting logic if you have a uC?

[samhay]

I've noted that inside most "vintage" stomp box's that contain mechanical SPDT switches, the input always feeds the pedal even when the effect is off. So the switch simply connects the output jack to either the effect's output or the input. So to re-cap, I just want the pedal user to be able to just keep all the pedals "on", while I handle the "in-chain" or bypass of each one externally.

I don't think this is a great strategy, as many of these vintage pedals were also known to tone-suck pretty badly.

If you are willing (and capable) to play with uC's, then it should be trivial to replace each 2/3PDT true-bypass stomp switch with a DPDT relay (Edit, or CMOS switch, etc). You can even leave your existing pedals alone and place each of them in a loop...

Edit2 - suspect this pre-coffee post is not very clear. Point being that, if you use a Dpdt approach then you can't bridge the input/output.

[amptramp]

Check out the first application circuit of Figure 1 on page 2 of this datasheet:

http://www.onsemi.com/pub_link/Collateral/MC14007UB-D.PDF

You can add RC networks to the select lines on pins 3 and 10 to slow down switching so that there is no pop from changing inputs.  Easy to connect to a microcontroller.

[PeterPan]

Sure enough, during that transition, many pedals that will "chirp" for a split second, because its input and output are momentarily bridged. Not good!

But if you're using a uC, why diodn't you just arrange for the LEDs to turn on/off  at slightly different times (a few milliseconds) to prevent the feedback? Also, why do you need inverting logic if you have a uC?

That's true, and thanks for replying. I have 2 reasons for avoiding that route. first, the MCU module I'm using has a limited number of IO pins. Requiring 2 outputs for each pedal would indeed give me maximum control over this problem, but it also means cutting the number or pins available for pedal control in half. Hence, I was using some simple inversion logic to allow each single bit to operate as an SPDT switch. But even if that were not an issue, now that I've seen how slow the transitions are, I'm not sure I'd want to double it to totally cure the momentary feedback. This is why I'm seriously considering the real analog switching/mux chips.

[PeterPan]

I've noted that inside most "vintage" stomp box's that contain mechanical SPDT switches, the input always feeds the pedal even when the effect is off. So the switch simply connects the output jack to either the effect's output or the input. So to re-cap, I just want the pedal user to be able to just keep all the pedals "on", while I handle the "in-chain" or bypass of each one externally.

I don't think this is a great strategy, as many of these vintage pedals were also known to tone-suck pretty badly.

If you are willing (and capable) to play with uC's, then it should be trivial to replace each 2/3PDT true-bypass stomp switch with a DPDT relay (Edit, or CMOS switch, etc). You can even leave your existing pedals alone and place each of them in a loop...

Edit2 - suspect this pre-coffee post is not very clear. Point being that, if you use a Dpdt approach then you can't bridge the input/output.

Yeah thanks. Thats kind of where Iw as heading with those Analog Device's chips. Each has 2 analog SPDT switches with break before make logic built in. I should pint out that I had done something like this before and got away with it, but at that time I was using much faster OPTO isolators, (5C3 Vactrols I think), but they are much too expensive IMHO.

[PeterPan]

Check out the first application circuit of Figure 1 on page 2 of this datasheet:

http://www.onsemi.com/pub_link/Collateral/MC14007UB-D.PDF

You can add RC networks to the select lines on pins 3 and 10 to slow down switching so that there is no pop from changing inputs.  Easy to connect to a microcontroller.

Thanks AmpTramp. I'd only get one SPDT operation per chip with that one, and it would take up more board space to do the 8 pedal/channels I want to control.  But I'll give it some thought because its definitely cheaper than the AD chip, even with an added RC delay parts. I will say that I once tried some audio switching with a really cheap  "Quad bi-lateral switch" chip (CMOS 4066), and I did note that any slowness in the logic transition caused a pretty nasty current spike when the transition voltage was reached, which translated into an unwanted "pop" in the audio. This 14007 is probably much more sophisticated (maybe it has some built in hysteresis or Schmidt triggers on the inputs?),   but because of bad experiences in the past, I'd be leery of doing a simple RC delay on the logic inputs. Unless you can tell me you've done it with this chip can gotten away with it?

[PeterPan]

Check out the first application circuit of Figure 1 on page 2 of this datasheet:

http://www.onsemi.com/pub_link/Collateral/MC14007UB-D.PDF

You can add RC networks to the select lines on pins 3 and 10 to slow down switching so that there is no pop from changing inputs.  Easy to connect to a microcontroller.

Hey AmpTramp! I don't think that last chip was exactly what I need, but thanks for pointing out "On SemiConductor". Outside of a slight variation in the pin-out, their NLAS4684MR2 seems to be nearly identical to that AD884 I mentioned, but is less than 1/2 the price! Good deal!

[R.G.]

Have you read the material at geofex on signal switching and programmable effects switchers?

[PeterPan]

Have you read the material at geofex on signal switching and programmable effects switchers?

No. I've found their site and tried to search, but haven't found that. Link?

[R.G.]

No. I've found their site and tried to search, but haven't found that. Link?

Try these:
http://www.geofex.com/article_folders/cd4053/cd4053.htm
http://www.geofex.com/article_folders/fxswitchr/fxswitchr.htm
http://www.geofex.com/article_folders/rmtswtch/rmtsw.htm
http://geofex.com/Article_Folders/remoteftsw.pdf
http://geofex.com/Article_Folders/pffootsw.pdf
http://geofex.com/Article_Folders/relays/relays_for_switching_audio_signa.htm
http://geofex.com/Article_Folders/ASMOP/ASMOP.htm
No particular order there.

I wrote most of this long ago, but it's still good background.

[PeterPan]

No. I've found their site and tried to search, but haven't found that. Link?

Try these:
http://www.geofex.com/article_folders/cd4053/cd4053.htm
http://www.geofex.com/article_folders/fxswitchr/fxswitchr.htm
http://www.geofex.com/article_folders/rmtswtch/rmtsw.htm
http://geofex.com/Article_Folders/remoteftsw.pdf
http://geofex.com/Article_Folders/pffootsw.pdf
http://geofex.com/Article_Folders/relays/relays_for_switching_audio_signa.htm
http://geofex.com/Article_Folders/ASMOP/ASMOP.htm
No particular order there.

I wrote most of this long ago, but it's still good background.

Thanks RG. Its very good of you to have taken the time to write all this up. For all my work on things like this, I have seldom taken the time to share the things I'd learned over the years in such friendly tutorials. Anyway, I'm pretty familiar with most of the methodology you've written about here. I guess my main issue (for lack of a better word) was what various switching elements were possible or available these days, and the merits of each. The analog opto's seemed to offer a nice "crossfade" option when switching from bypass to effect, which i thought would be a unique approach. But now I see that to make use of such a scheme, I'd need another judiciously placed op-amp for each effect controlled, to make sure signal only traveled in the direction I want it to. But seeing that added complexity, and not wanting to add more expense to the ultimate project (and hopefully product) that this will be part of, I think I'm going to just use one of those newer analog switch chips, with the built in break-before-make functionality, with circuits similar to the ones you've posed. My only complaint is that they make a lot of these wonderful chips to fit inside very small mobil devices without taking up much board space, so many of them have pin spacing of like 1/2 mm apart. Well its nice to conserve board space, but for what I'm doing these "itsy bitsy" chip sizes are going to be a hindrance to do by hand. But their specs sure are nice!

[R.G.]

One I didn't list there is the Technology of Boss and Ibanez Bypasses. What's interesting about that for your purposes is that they use JFETs for a slow and controlled fade on signals. JFETs let you pick how fast the signal fades in and out with a single RC constant per JFET switch.

I've always been fascinated by "morphers", these being really slow crossfade switchers, and I think this may be where you're trying to go. All of the things done up as switches will have problems with this application because they really can't stand too much time spent dwelling in the middle of the fade, at least for signals bigger than maybe 10-50mV.

I built a couple of prototype morphers using VCAs a while back. It was kind of an N-way crossfade. The enabled channel faded up at a rate set by a "fade-in speed" pot and the previously enabled channel faded down at a speed controlled by a "fade-out speed" pot.  Varying the speed pots got some really interesting crossfade effects. You could set both fade-in and fade-out long for the kind of crossfade you expect, but things like one long, one very short was interesting.

I used a uC for the control on the fade, and always intended to do a 2-of-N controller, letting both the current and one-past channels stay active and panning in/out one as footswitches were hit. Never got the chance to build that.

I saw it as more of a crossfader for small chains of effects in a playing situation.

[PeterPan]

One I didn't list there is the Technology of Boss and Ibanez Bypasses. What's interesting about that for your purposes is that they use JFETs for a slow and controlled fade on signals. JFETs let you pick how fast the signal fades in and out with a single RC constant per JFET switch.

I've always been fascinated by "morphers", these being really slow crossfade switchers, and I think this may be where you're trying to go. All of the things done up as switches will have problems with this application because they really can't stand too much time spent dwelling in the middle of the fade, at least for signals bigger than maybe 10-50mV.

I built a couple of prototype morphers using VCAs a while back. It was kind of an N-way crossfade. The enabled channel faded up at a rate set by a "fade-in speed" pot and the previously enabled channel faded down at a speed controlled by a "fade-out speed" pot.  Varying the speed pots got some really interesting crossfade effects. You could set both fade-in and fade-out long for the kind of crossfade you expect, but things like one long, one very short was interesting.

I used a uC for the control on the fade, and always intended to do a 2-of-N controller, letting both the current and one-past channels stay active and panning in/out one as footswitches were hit. Never got the chance to build that.

I saw it as more of a crossfader for small chains of effects in a playing situation.

Interesting! Well I should clarify that the "crossfade" feature was kind of an afterthought for me. Its part of a much bigger project (Its called "SoundMan", and I'll explain it better on another thread sometime), involving wireless remote controllers for both audio levels and multi effects stompboxes. As "project creep" always happens, i soon realized it would be an important add on to have a "pedal adapter", for interfacing with "stand-alone" vintage pedals. I had done something like this before with opto isolators and I "got away" with it, at least for the particular pedals I was interfacing to. But of course to be universal, my overlapped-on/feedback issue would have to be eliminated for all cases. Well i found these VERY cheap OPTOs at Electronic Goldmine" (http://www.goldmine-elec-products.com/prodinfo.asp?number=G15396B ). And by cheap, you have to understand that they had a sale, offering the bags of 50 for $15 (not the current $49). So I thought I'd consider doing the crossfade idea instead of analog switches. And of course even with the optos, you can control the response speed to some extent if you want to go to the extra trouble to control the rate of increase, of current applied to the internal LED.

Well I put in some sample orders for a lot of ordinary analog switch ICs yesterday, so I figured I'd do some bench work Sure enough I found a few configurations that kind of solved the problem using an op-amp. If you bring the bypass and effect outputs together at the summing junction of an op-amp inverter, for example, it will prevent the feedback problem, because the summing junction always remains ZERO. Not surprisingly, I really had to watch my wiring configuration to make that work, and there were issues with unwanted gain or gain reduction in an op-amp inverter. But another way that seemed hopeful was to put non inverting unity gain buffer configured op-amps, in both the bypass and effect outputs, and then bring both these outputs into the summing junction of an inverting op amp. That gave me the best avoidance of unwanted (r unpredictable) gain shifts, but at the expense of 3 op-amps. So although it was a useful exercise, this is getting to be too much of an added expense. I'm supporting up to 8 pedals, and its just too much extra expense for what was supposed to be a relatively inexpensive "pedal adapter. Especially when I consider that I wasn't specifically after "crossfade", but rather just a less obtrusive transition. I have another solution to the transition problem because the overall project has digital potentiometers available for audio level control  (I'm using Dallas 1881s for that... http://www.maximintegrated.com/en/products/analog/data-converters/digital-potentiometers/DS1881.html if you're interested). Those chips are controlled by I2C interface from the MCU.  So anyway, under program control I've already added what i call a "volume swell" effect, kind of like pedal-steel guitar. It basically drops the volume to about -64db quickly (but not instantly) and then restores it again to its original level at a controlled speed (I use this for playing Alman Borthers songs like "Melissa" without a volume pedal). So now in my system, along with any "patch" being able to enable or bypass selected pedals, any patch can also trigger a "volume swell" effect. And after trying it out I realized that the volume swell did a very nice job, not exactly of crossfading, but of accomplishing the original goal, which I kind of lost sight of. The goal was to simply allows for less obtrusive changes between effects patches. It turns out that at least in my opinion, the quick level fade to near silence and controlled return serves this purpose just as well.

But, since you mentioned those Boss bypasses (I'm curious what their method is too), I will say that having worked with those DS1881s a bit, I'd consider using them for an ultimate cross-fade! Each chip has 2 analog taper pots that are controlled (by I2C interface) in 64 1dB steps. (63 actually, the last one past -63 is a mute, at about -84dB). You can pretty much write to all its registers with one short command chain, and you can write those registers pretty fast (the serial equivalent of about 500Kbps!). So with an MCU its reasonably easy to orchestrate a gradual change in volume (or a cross fade by writing to both its channels) at any rate you need, from what seems like instantaneous to as slow as necessary. And happily, 1dB steps are pretty hard to discern unless you are REALLY listening carefully with headphones.

But anyway, I'm just sharing there. Like I said, I'm going to go back to trying actual IC based analog switches, and let my "volume swell effect" handle the slow transition, on patches where it seems bothersome. My next issue,(at least for me its an issue) is that many of the better analog SPDT switches chips are too fregin' tiny for prototyping. Even to control 8 channels, the jacks are already the biggest parts, and I'd much rather work with .050" pitch SOICs, then these itsy bitsy MSOPs, with their stupid 0.5mm pin spacing! One hopeful thing I found yesterday is a TI chip, TS3A5018 http://www.ti.com/product/TS3A5018/samplebuy. It's a QUAD (!) SPDT analog switch, and guess what... its available in a 16 pin SOIC!!! (easy on the eyes!). Now their "on" resistance is higher than the others, 10 ohms. But then again I'm not switching speaker / power loads here. Still beats the hell out of the on resistance of those OPTOs. I'm a LITTLE concerned because these chips don't have that "break before make" feature. The specs seem to be telling me the transition to "off" time is generally a few nS faster than the turn on time, which might help. But even there was some unwanted overlap, I'm NOT switching high current loads where there would be a burn-out risk of short circuit. I just need to make sure any overlap happens so fast that the whole transition is no worse then the mechanical SPDT switches the pedals originally had. Well we'll see... One thing about TI is that they are ridiculously fast. I'm in Tampa, Florida, a good distance away from Texas. Yet, my sample order I placed yesterday at 5PM just arrived here this morning! So it looks like I'd better head to my garage and get to work, seeing how these TS3A5018s perform. Most of my workshop is in the garage area, which in Florida gets pretty uncomfortable by mid afternoon!

Anyway, thanks again for the helpful conversation, and for listening. It really is nice to be able to bounce ideas of someone that is both technically knowledgeable, AND a musician that understand the challenges of the applications! :-)

[R.G.]

Interesting! Well I should clarify that the "crossfade" feature was kind of an afterthought for me. Its part of a much bigger project ...even with the optos, you can control the response speed to some extent if you want to go to the extra trouble to control the rate of increase, of current applied to the internal LED.

If you're already at a microcontroller, why not mute the output or multiple points while the switching is taking place? JFETs would do that very well.

... this is getting to be too much of an added expense. I'm supporting up to 8 pedals, and its just too much extra expense for what was supposed to be a relatively inexpensive "pedal adapter.

To a first approximation, the electronics inside a widget are free compared to the box it goes in, the controls and the power supply. It helps to take the costs of all the parts in a spreadsheet and sort by price. The expensive stuff all clusters together a the top. Mostly you'll find the majority of the cost in the top 10-20% of the items.

And happily, 1dB steps are pretty hard to discern unless you are REALLY listening carefully with headphones.

In my experience, you have to dither the step timing or the regularity of the steps comes out as a "zipper" noise if it's happening at an audio frequency. Maybe for very slow fades. Or maybe I'm just too picky. Even making the volume steps happen at signal zero crossings can be noticed sometimes. But probably not in all the racket on a stage.

My next issue,(at least for me its an issue) is that many of the better analog SPDT switches chips are too fregin' tiny for prototyping. Even to control 8 channels, the jacks are already the biggest parts, and I'd much rather work with .050" pitch SOICs, then these itsy bitsy MSOPs, with their stupid 0.5mm pin spacing!

Go beat the used-goods sales for a Bell and Howell stereo zoom microscope. You'll thank yourself in the future. I got mine for under US$100. Worth every penny.

Now their "on" resistance is higher than the others, 10 ohms. But then again I'm not switching speaker / power loads here. Still beats the hell out of the on resistance of those OPTOs. I'm a LITTLE concerned because these chips don't have that "break before make" feature.

Switch resistance up to maybe 500 ohms or so is not generally an issue for audio signal level switching.

Good luck with the design.

[PeterPan]

Interesting! Well I should clarify that the "crossfade" feature was kind of an afterthought for me. Its part of a much bigger project ...even with the optos, you can control the response speed to some extent if you want to go to the extra trouble to control the rate of increase, of current applied to the internal LED.

If you're already at a microcontroller, why not mute the output or multiple points while the switching is taking place? JFETs would do that very well.

... this is getting to be too much of an added expense. I'm supporting up to 8 pedals, and its just too much extra expense for what was supposed to be a relatively inexpensive "pedal adapter.

To a first approximation, the electronics inside a widget are free compared to the box it goes in, the controls and the power supply. It helps to take the costs of all the parts in a spreadsheet and sort by price. The expensive stuff all clusters together a the top. Mostly you'll find the majority of the cost in the top 10-20% of the items.

And happily, 1dB steps are pretty hard to discern unless you are REALLY listening carefully with headphones.

In my experience, you have to dither the step timing or the regularity of the steps comes out as a "zipper" noise if it's happening at an audio frequency. Maybe for very slow fades. Or maybe I'm just too picky. Even making the volume steps happen at signal zero crossings can be noticed sometimes. But probably not in all the racket on a stage.

My next issue,(at least for me its an issue) is that many of the better analog SPDT switches chips are too fregin' tiny for prototyping. Even to control 8 channels, the jacks are already the biggest parts, and I'd much rather work with .050" pitch SOICs, then these itsy bitsy MSOPs, with their stupid 0.5mm pin spacing!

Go beat the used-goods sales for a Bell and Howell stereo zoom microscope. You'll thank yourself in the future. I got mine for under US$100. Worth every penny.

Now their "on" resistance is higher than the others, 10 ohms. But then again I'm not switching speaker / power loads here. Still beats the hell out of the on resistance of those OPTOs. I'm a LITTLE concerned because these chips don't have that "break before make" feature.

Switch resistance up to maybe 500 ohms or so is not generally an issue for audio signal level switching.

Good luck with the design.

Well thanks again for all that. One of the very nice features of that particular digital pot is that you can tell it to perform each transition at zero crossing points on your signal, That goes a LONG way to quieting that "zipper" effect you're talking about. You definitely owe it to yourself  to playing with those chips!

As far as the visual magnification, that's not as big an issue for me as the fact that my hands just are not as steady as they once were. I don't think I have "parkinsons" or anything like that, but some other meds I'm on for asthma have some involuntary twitching as a side effect. When i have to solder down anything smaller than 50 mill pitch, I usually just give up and solder a whole side of the chip, and come back later with solder wick. But its still an unnecessary pain in the but on something where maximum miniaturization  isn't an issue.

As far as costs go, you know.. it does start to add up, no matter how you slice it. Larger 4 layer PC boards can get costly if you're not totally sure the design is final, and have to make a short order. And if you really want a nice enclosure for some things, that can cost too... as does preparing at least a modicum of pro looking labels. I used to balk at the prices some people change for their own home brewed designs, but now I understand. So now i find myself scrutinized each part to consider where there might be a lower cost substitute that would accomplish the same thing. Take that quad analog switch. I was happy to find that cheap TI part, but then I realized the so called "quad" SPDT switch was really all controlled together, by one input. Well, Max makes a really nice and supposedly very quiet quad SPDT, the Max394. But try and get one for less than $6! Intersil has a drop in replacement (ISL8394) which you can get at a few places for a couple of bucks less. But bottom line, it adds up.

I don't know if its allowed, but if so I'll post more about it when its closer to "final". One problem with anything containing an MCU is that everytime you go to bed satisfied, the next morning you wake up with 4 new features in your head you could add. :-)