Using modern analog switches for bypass?

[somefolkslikewater]

Hi Everyone,

After a long hiatus rotating through my various hobbies I've found myself back on the wagon building pedals.

For both fun (to use different switches) and reliability I'm looking to move away from 3pdt switches for future builds. I've tried to get a feel for the options out there and the microcontroller + relay solution looks to be a widely adopted and relatively flexible route to explore, but by no means a free ride.

The reason for my post however is that I'm curious about the current thinking regarding analog switches for bypass in pedals. Having repaired my own builds a few times over the years I'm more conscious about longevity and can see the benefits of electronic switching over moving parts. While analog switch specs have presumably improved (I'm guessing both CMOS and solid state relays) I don't know enough about the intricacies of applying them to audio bypass applications to judge whether this has been a help or a hindrance. There also doesn't seem to be a huge amount of recent discussion about using them and I've struggled to find examples in commercial pedals more exotic than discrete FETs, but please correct me if I'm wrong!

To summarise my bypass design requirements:

• 
  Use momentary SPST switches
• More reliable (in theory) than 3pdt
• Avoid clicks, pops, noise etc. due to switching
• 
  Avoid buffers in bypassed signal chain

Questions:

• Have advances in analog switch technology made them worth exploring beyond what's already been done with things like 4066 / 4053 CMOS Switches?
• Is their any change in how they would be applied for use in guitar pedal bypass? I'm thinking challenges, limitations, relevant specs and whether existing circuits are all still relevant?
• Is anyone actively experimenting with this? If so do you have any pointers, resources or perhaps IC recommendations to share?
• Is this worth spending time worrying about or am I facing diminishing returns vs something like a microcontroller + relay?

Thanks in advance for your help!

[iainpunk]

honestly, i think using a 3pdt is the easiest way to go for a DIY build on any type of build. with most "digital switches" there needs to be a buffer before and after because of the less than ideal throughput impedances, because they actually do the same thing as Boss style Jfet switching, but inside a plastic case, instead of discrete. i personally despise relays, because they are big, need lots of power, they aren't as reliable as an actual switch...,

granted, my bias against relays comes from the place i used to be an intern at, where they just let relays run until the unit fails and then, stop the whole production line to replace the little bugger, recalibrate the whole machine, take out and clean up failed bottles/spills/broken bottles... and i was tasked with doing that. being the lazy MF i am, i introduced them to solid state relays that fit in to the relay sockets, and they haven't had a line stop ever since they overhauled all mechanicals in to solid states

cheers,
Iain

[somefolkslikewater]

Thanks Iain. Completely agree with you about 3pdt being the easiest to implement and relays have their own set of problems but ultimately every switching method entails some compromise. My intention wasn't to compare the various switching methods though.

Analog switching would be my preference if it came fairly close to mechanical alternatives in performance without the need for buffers, hence my interest. I'm really wondering if there have been or could be any advancements using this technology for pedal bypass with more modern chips or if all the same problems/caveats/challenges of using them remain?

[Fancy Lime]

The only two problems with using JFET or CMOS switches without buffers are:

1. There is an on-resistance. Typically a few hundred Ω but may be as low as single digit Ω.

2.You need to bias the input and output of the switch.

Problem 1 is really not a problem in most circumstances but it has to be considered. If problem 2 really is a problem depends on the circuit. I have never seen it done but it should be possible to design a switching circuit with no buffer using low-on-resistance JFETs. The biggest challenge is getting the right JFETs. For a negative ground circuit, we would need a P-channel device. The lowest resistance one I know is the J175 with 125Ω. Not great but no biggy if whatever comes after has a high enough input impedance and you don't stack a ton of those bypasses. On positive ground, we could use an N-channel like the J105 with just 3Ω. All in all a feasible concept for a looper but a lot of hassle in a single pedal without any obvious benefits.

Anyway, why the aversion to buffers?

Andy

[antonis]

So, do you guys feel comfortable to stomp on a CMOS delicate cofiguration..??

[Fancy Lime]

So, do you guys feel comfortable to stomp on a CMOS delicate cofiguration..??

Hence the buffer, so that the CMOS does not get crushed underfoot 

[amptramp]

I wrote up a blurb on silent switching using CMOS and diode switching here:

https://www.diystompboxes.com/smfforum/index.php?topic=120006.msg1122270#msg1122270

One design uses a 4007 as a switch with adjustable time constants.

Another uses diode bridges and while not classically silent, it would work for some fuzzes where the noise level would be below the fuzz level.

Another switches when the input and output voltages match for a minimum of popping.

[ElectricDruid]

I've used the modern DG-series switches and for what they're designed for they're a big improvement over the old 4066/4053/4051 switches. The problems with those were:

1) Limited power supply range
You couldn't run them on a +/-15V supply, so they often finished up run on +15V/0V, and then had to have the input biased. Wot a faff.

2) No Logic input level shifting
This was another good reason why they finished up being run on +15/0V. You could run 4000-series CMOS logic on the same supply and it would control the switches no problem.

The DG series fixes both of these things, and lets you run the chips on +/-15V or +/-12V and feed them full level audio, while switching the chip with a 5V logic signal. Much simpler, and handy for controlling audio with a uP.

HOWEVER...none of this is really what we want for a stompbox. So although I think the chips are much improved over the old ones, they're not intended for the use you want to use them for, and won't be ideal for that as a result.

I think the on resistance is a red herring, frankly. I usually stick a 560R resistor on the outputs of my pedals anyway to protect the output op-amp from shorts. An extra hundred ohms here or there won't make any odds if your pedal has a modern concept of what its input impedance should be and many modern switches are far lower than that anyway.

The extremely fast switching of most CMOS switches will lead to audible clicks if the input/output aren't at the same level. The advantage of "raw" JFETs or basic switches like the 4066 is that you can slow down the control signals to ramp briefly from on to off and limit the click. But then there are modern switches that wait for a zero-crossing automatically, so maybe that's not an issue either. The choice is *far* wider than it used to be, and the range of options has given us chips that are much more specialised for a specific use.


EDIT: I thought I remembered the 4066's switching being able to be slowed down but checking it, that doesn't seem to be the case. Was it the 4016 instead? I was almost sure one of those two could do it?

[Fancy Lime]

The on resistance may become an issue when and if you have three dozen pedals all turned off and run straight from a guitar and into a fairly low impedance amp input. I agree with Tom that if that ever happens, lots of things have gone wrong and that it is not really a realistic problem scenario. However, I have also learned that a lot (and I mean a LOT) of people obsess about "tone suck". So my suggestion that on resistance is a problem should be read as "some people will think of it as a problem but it very rarely really is one".

I'm going to try and sketch the J175 bufferless switching scheme as soon as I have a few minutes.

Andy

[Fancy Lime]

I think this ought to work (untested):



The opamp buffer is used as a stand-in for whatever the actual effect may be. Ground is the negative rail, VR is half supply voltage, ctrl is a 0V/+9V signal from the source of your choice. When ctrl is at 0V, the effect is bypassed. When ctrl is at 9V, bypass is closed and effect is on.

Cheers,
Andy

[PRR]

I think it has multiple problems. One JFET references zero, the other V/2. The "bypass" seems to short the dry signal. Maybe I'm reading it wrong.

[Fancy Lime]

I think it has multiple problems. One JFET references zero, the other V/2. The "bypass" seems to short the dry signal. Maybe I'm reading it wrong.

Why is the JFET reference voltage a problem? As long as the "open signal" at ctrl makes Vgs ≤ 0V (for a p-channel) and the "close signal" makes Vgs ≥ Vgs(off) it should work, no? "Open" means ctrl is high, "close" means ctrl is low. I think that that is the case anyway, p-channels are a bit of a mindf*** for me.

The shorting: yes, I forgot a 100k resistor in series with C4. I'll update the drawing later today. Thanks for catching it!

I had another look at the datasheet and realized that Q1 should be a J176 instead of J175 (different Vgs(off)).

Also, I alluded to this in an earlier post but maybe it needs stating outright here: This thing needs to be fed with a fairly low impedance input signal (<< 100kΩ) and there is really no real benefit of using a scheme like that over a buffered bypass that I can see, which is probably why you never see it done like that. The only reason to ever use this is if you realized how unreliable mechanical switches are but oppose buffers on religious grounds.

Andy

[composition4]

FWIW I've never had ANY problems with 4066 or 4051/2/3 switches. Bias both sides of the switch to VREF... And that's it. Never (not once) had any clicking or popping or distortion/tone altering problems whatsoever. If you're yet to try them, don't dismiss them because "you heard" people have dramas with them - find out for yourself!

[merlinb]

CMOS switches are OK, but you just cannot get around the click problem. Often it passes unnoticed, after all, we tolerate mechanical switches which are even clickier. But in situations where silence really matters, you just can't beat a discrete FET for that 'fast fade' capability.

Why is the JFET reference voltage a problem? As long as the "open signal" at ctrl makes Vgs ≤ 0V (for a p-channel) and the "close signal" makes Vgs ≥ Vgs(off) it should work, no?

If your control signal can swing negative as well as positive, sure. But if not, both JFETs need to be bised up to Vref, kinda like:


(reverse the diodes if using P type JEFTs)

[Fancy Lime]

@Merlin
Why is pulling the gate to the same voltage (0V in this case) as source and drain not enough? Oh, yes, now that I type it out like that I realize that there is a signal swing. Never mind my stupid questions
OK, so we need to have a +9V to -9V converter to drive the control voltage. It's getting less practical by the minute... A "soft un-buffered bypass" Klone is just about the only thing where the "extra mojo" vs. "extra effort" equation may make sense to some people. 

Or we can just do as Merlin showed, which is much easier and then we can use J107 N-channel JFETs with only 8 Ohms of on-resistance and advertise "now with even less tone suck".

Andy

[R.G.]

If you haven't yet, you might enjoy reading "The Techmology of Bypasses" at geofex.com.

The CD4051 works very well indeed. The company I design for has sold literally tens of thousands of pedals using that for bypass, and has had exactly one complaint. That one CD4051 was faulty.

As Merlin says, there will always be a slight transient in the CD4051. This transient can be heard faintly if you're running into a very high gain amp at full volume if you put your ear in front of the speakers. But you do have to bias the ins and outs to the same voltage. JFETs need that too.

Hard contact switches have about the same transient as CMOS unless you're quite careful about how you use them. Sometimes worse.

JFETs actually have switching transients, you just get to slow them down so they're below the hearing threshold.

Re JFET control voltages. JFETs are on when you let their gate and source float to whatever voltage they want to be at. That's what that series diode at the gate is for. To turn them off, you have to pull the gate to Vgsoff. Notice that this is relative to the source. You bias the source at some reference voltage but the signal on the source subtracts from the reference voltage on every other half cycle. That's no problem with 100mV signals from a guitar, but it is an issue if your Belchfire Megablaster pedal in front of the JFET switching is cramming in +/- 4.5V of signal to it. In this case the signal itself can pull the JFET out of "offness". It's a design issue with all JFETs.

Having tried the newer LED-Photo MOSFET "relays",  I like them. No transients, just switching. The TLP222 series seem to work very well. They come in a 4 pin dip single and an 8 pin dip double. Mouser lists them for about $1.10 per switch.