Switching with a switch vs with a FET

[fryingpan]

FETs used as switches are convenient, but are there any drawbacks in the signal path? (eg. noise)

Say I wanted to redirect a signal, for example switching in a presence boost, or something more complex (for example rerouting a signal from one pre-amp stage to another, or switching between different values of resistors and capacitors in a tonestack), what would the real-life consequences be, as opposed to getting a switch of the right kind (maybe a three or four pole kind, and running cables from and to the board)? I know that FETs have some resistance, but compared to the usual values seen in filters and stuff, this doesn't seem to be consequential.

[bluebunny]

Seems to work OK for Boss. 

[teemuk]

Audio signal path wiring tends to become longer with mechanic switches because signal path must go to switch and back. This might introduce extra sources of interference. With FET switching you can keep signal paths to switch short and only paths of switches control signal remain moderately long, interference there is less of an issue than in audio signal path.

[ElectricDruid]

FET switching is more complicated. You need to make sure the signal is biased correctly, you need the right FETs, you need a circuit that can provide the right control signal, etc etc. It makes for a lot of circuit.

*But* it can switch silently, it's more reliable than a clunky footswitch, and you don't have signal wires running all over the place.

It's a balance. You have to decide whether the costs are worth the benefits.

[vigilante397]

One of the big ones, as part of ElectricDruid's comment on the right FETs, there's inherent resistance across the FET (RDS-on) that isn't present in mechanical switches, so you're basically putting a resistor in series with your signal, you just have to choose a FET with low enough resistance that it isn't noticeable.

[amptramp]

True bypass has some advantages and disadvantages as listed below:

Advantages

1. Negligible signal distortion in bypassed mode
2. Exact unity gain in bypassed mode
3. Negligible frequency response deterioration, connected or bypassed
4. Minimal noise addition
5. Minimal interaction with effect circuit
6. No power required for switch function
7. No other circuitry required except switch pop reduction resistors

Disadvantages

1. Expensive 3PDT alternate-action switch required
2. Introduction of switch pops unless circuit modified
3. Different behaviour depending on order of contact transfer
4. Low reliability and life of switch
5. Difficulty in adding remote switching capability
6. Cannot be initialized on turn-on
7. Advantages not necessarily applicable to delay functions

Let's take a closer look at this.  True bypass connects the output to the input in bypass mode or to the circuit output in operational mode via switch contacts that are less than one ohm in resistance.  Even the worst pedals have an input impedance of 47KΩ, so the gain would be 47000/47001 ohms or as close to unity as whoremonger is to swearing.  This resistance does not change with input or output voltage and the total capacitance within the switch and the entire effect would be equivalent to less than the capacitance of a foot of cable.  The only noise addition may come from capacitive coupling to noise sources such as LFO's and switching regulators.  Switch voltage and current limits are above anything we would use in a pedal.  If you have only a few pedals, true bypass is not that much of an advantage.  But if you have a dozen pedals in a row, the lack of distortion, low noise, unity gain and retained frequency response is worthwhile.

[fryingpan]

You do ensure me that a FET used in saturation mode does not distort the incoming signal, right? (I suppose it would be Vds, in this case).

[fryingpan]

Would two of these parts be a good choice for a DPDT switch that can also be controlled remotely?

https://www.mouser.it/ProductDetail/Vishay-Siliconix/DG419DJ-E3?qs=3bFTlaDNFuV0juGT9hJcaQ%3D%3D

Distortion stats are not provided, but the MAX4659 which is pin-compatible (and I suppose a "clone") claims 0.005% distortion or so.

Bear in mind that this is not for a guitar pedal, but for an amplifier (with bipolar signal levels over 1Vrms).

[ElectricDruid]

Would two of these parts be a good choice for a DPDT switch that can also be controlled remotely?

https://www.mouser.it/ProductDetail/Vishay-Siliconix/DG419DJ-E3?qs=3bFTlaDNFuV0juGT9hJcaQ%3D%3D

Yes, the DG switches are good parts, and it's easy to control them since they have level shifters on the control inputs.

Distortion stats are not provided, but the MAX4659 which is pin-compatible (and I suppose a "clone") claims 0.005% distortion or so.

0.005% sounds like it's a hundred times below anything we might even need to worry about.

Bear in mind that this is not for a guitar pedal, but for an amplifier (with bipolar signal levels over 1Vrms).

What power supply are you going to run the chip on? the typical level is probably +/-15V, so a 30Vpp signal level max. Hence your 1Vrms signal (while not being any problem) isn't going to take best advantage of the headroom and signal/noise ratio you have available. I'd feed it something hotter if possible.

Secondly, these switches switch as close to instantly as makes no odds (nsecs) so if there is any signal present at the point the switch switches, you'll get an abrupt transient - a click. That's not the switch's fault, exactly. But the "silent" FET switch on things like the old boss pedals deliberately slows down the switching process to a few tens of milliseconds to "hide the join" and avoid any click. You can't do those kind of tricks with these chips.

Still, the DG switches are good. Very easy to use, switch full range analog signals without any messing about with biasing, don't need level shifters, etcetc - useful chips all round!

[iainpunk]

Distortion stats are not provided, but the MAX4659 which is pin-compatible (and I suppose a "clone") claims 0.005% distortion or so.

0.005% sounds like it's a hundred times below anything we might even need to worry about.

how about 1000 times, up to about 5% you don't hear distortion in a guitar, you hear 'colouration', 5% to 15% is ''lightly overdriven/distorted'' and a big muff is in the 40% to 85% range depending on sustain and tone settings. (84% is what i measured with max sustain, volume and full brightness)

cheers

[fryingpan]

I wasn't really concerned with distortion (that was just to say that it is probably negligible) and signal levels would be approx. 1Vrms on one switch and possibly 4-5Vrms on the other one, all feeding into high impedance inputs (a JFET or a non-inverting opamp).

[PRR]

....What power supply are you going to run the chip on? the typical level is probably +/-15V, so a 30Vpp signal level max. Hence your 1Vrms signal (while not being any problem) isn't going to take best advantage of the headroom and signal/noise ratio you have available. I'd feed it something hotter if possible.......

No. The curve is canoe shape. Stay near the middle and curvature is small. Come up to 10V peaks and you have non-zero 3rd harmonic. I strongly suspect DG were not too dumb and did pick the implementation for around 1V signal, a popular level in industry. (The audio market does not support switch chips, we have to crib from other fields.) 

[fryingpan]

....What power supply are you going to run the chip on? the typical level is probably +/-15V, so a 30Vpp signal level max. Hence your 1Vrms signal (while not being any problem) isn't going to take best advantage of the headroom and signal/noise ratio you have available. I'd feed it something hotter if possible.......

No. The curve is canoe shape. Stay near the middle and curvature is small. Come up to 10V peaks and you have non-zero 3rd harmonic. I strongly suspect DG were not too dumb and did pick the implementation for around 1V signal, a popular level in industry. (The audio market does not support switch chips, we have to crib from other fields.) 

In fact the MAX4659 part provides THD stats measured at 5vp-p. Which is equivalent to about 1.7Vrms. Still, being so low, I guess you could push it to 10-12Vp-p without significant distortion (after all, anything under 0.1% is enough, and in a guitar amplifier with lots of distortion, practically zero).

[amptramp]

I did a thread on noiseless switching here:

https://www.diystompboxes.com/smfforum/index.php?topic=120006.msg1122294#msg1122294

The drawings are in the final post as postimage had a colossal wipeout in the spring of 2018.  The idea of noiseless switching with 4066 or 4053 switches is to delay switching until the input and output match which may give a sudden change in the direction of voltage but no step function of voltage.

[ElectricDruid]

....What power supply are you going to run the chip on? the typical level is probably +/-15V, so a 30Vpp signal level max. Hence your 1Vrms signal (while not being any problem) isn't going to take best advantage of the headroom and signal/noise ratio you have available. I'd feed it something hotter if possible.......

No. The curve is canoe shape. Stay near the middle and curvature is small. Come up to 10V peaks and you have non-zero 3rd harmonic. I strongly suspect DG were not too dumb and did pick the implementation for around 1V signal, a popular level in industry. (The audio market does not support switch chips, we have to crib from other fields.) 

In fact the MAX4659 part provides THD stats measured at 5vp-p. Which is equivalent to about 1.7Vrms. Still, being so low, I guess you could push it to 10-12Vp-p without significant distortion (after all, anything under 0.1% is enough, and in a guitar amplifier with lots of distortion, practically zero).

+1 agree. That graph looks pretty good to me up to 10Vpp or so. But ultimately, you decide how much distortion you can live with. There's always trade-offs.

Thanks for the graph though, Paul. It's definitely not a simple linear situation, so just boosting the level to get better S/N (as I originally suggested) isn't going to work.

[PRR]

Adjusting signal level only helps so much. The switch designer already did as much as possible for a price.

At home the next most useful design factor is load impedance. That switch shows about 4 Ohms resistance change. If we loaded with 400 Ohms then nonlinearity is near 1% (actually not bad for stage).  If we loaded with 4 Megs (hard) then nonlinearity is obviously near PPM, 0.000,1%.

[fryingpan]

In my case, the load would be 500k. As I said, distortion will never be an issue.

[ElectricDruid]

Adjusting signal level only helps so much. The switch designer already did as much as possible for a price.

So I guess what we need to know is what signal level they were designing for!

On a +/-15V supply, a 10Vpp or even 20Vpp signal is not unusual (after all, the supply is 30V, so you're well clear of any clipping). If that's not what the switches are intending, they really need to say so.

The series resistance of the switches is a issue that needs to be considered, but for the modern switches like we're discussing, it's a order of magnitude better than the older 4000 series stuff, and if you feed a decently high impedance input, you don't need to worry. 100K input impedance is easy to achieve.

Tom

[choklitlove]

With a switch you can say "True Bypass".

[amptramp]

FET switch distortion is not that bad if the switching is done near an inverting op amp node.  If the non-inverting node is held at Vcc/2, the inverting node will operate at the same voltage.  The only reason to avoid being right at the inverting input is the capacitance to ground of the FET switch may cause a rise in high-frequency closed loop response with the possibility of instability.  The signal is nearly zero at the node and can be made small by using resistance between the switch and the inverting input.