Best electronic switch for audio signals-logic drive

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Account Closed
Joined 2010
I am looking for the best and cheap possibility of switching or muting audio signals with the lowest distortion effect with electronic parts driven by logic levels..Any hint?It's a cassette deck i'm working on and i need to switch very low signals(tens -hundreds of milivolts).
 
Do a search for CMOS transmission gates. CD4052 is a useful configuration. The analog signal has to be in the middle of their supply rails and they need to be in a high Z circuit so that their resistance non-linearity is less significant, and you may want to both series open and shunt for best bleed. (3x 4016/4066) If you use JFETs, add 1/2 the audio to the control voltage for best linearity. There is an old diode bridge circuit but it probable thumps badly when switched. Some circuits swap bias to multiple diff-amps, sort of bipolar linear multipliers.
 
Moderator
Joined 2011
Last edited:
Thank you!I was just reading on Doug Self Small signal amplifier book about cd4066 use.I saw the fact that cd4066 needs to be in a high z circuit for best results.

Loading a CMOS switch into a high impedance will minimize the current through the switch, and thus the resistance modulation. You can also load the switch into a virtual earth, and this minimizes the voltage variation of the switch FETs relative to the substrate, and this also minimizes resistance modulation.

The only issue with using a virtual earth load is that the switch capacitance will de-stabilize the virtual earth amplifier. You can counteract this with a feedback capacitor, but if the switch capacitance and the compensation capacitor get too large, which the amplifier's output sees as a series combination of two capacitors to ground, this capacitive load can also destabilize the amplifier. So, there is a limit to the amount of switch capacitance that can be compensated.

So, loading into a high impedance is the simplest approach, but you still have to be careful to control the switched voltage relative to the substrate voltage. Some CMOS parts can be used with wider supply voltages, well beyond 5 or 7V, and this is helpful. It should also be obvious that the CMOS switch will need to use bipolar supplies to switch a bipolar signal, since you cannot drive the switch with a signal beyond either power supply rail. This makes the logic drive signal more complex, but again, some parts let you use three voltages, two for the signal switching FETs, and one as a logic ground reference, simplifying the issue.
 
i need to switch very low signals(tens -hundreds of milivolts).
Be aware that analogue switches have some feed through. The logic control signal has some effect on the analogue signal.
This mean you might have some DC offset in the audio when the switch is ON versus OFF. You will have some spike in the audio when switching because of capacitive coupling, you better use digital control signal with slow rising and falling edges ( just clubber the digital signal with a capacitor, i guess you do not need fast switching ).
For lowest distortion, as said you need to choose switches with low Ron relative to the series impedance of the audio.
Because you deal with very low audio signals, this is good for low distortion.

In case these issues are troublesome you can better use reed relays rather than full electronic switches.
In extreme cases, one can choose mercury wetted contacts.
 
Remembered that if you slow/filter the switch voltage, it prevents clicks in the audio. After all, you don't really care if it takes 100mS to switch sources, but a 100mS transition contains no "audio" frequencies, noise.
(C)MOS transistors have no DC path between the gate and channel so there should be no DC in the audio, likewise J-FETS as long as the gate is always reverse biased.
Consider that there is no distortion in the "off" state so shunt switches are distortion free when the source is not muted. You probably don't care about a bit of distortion when muted.
My fav was P channel JFET shunt, which operates off ground (positive gate voltage) and the peak audio voltage levels are not likely an issue. But N channel may be a better shunt (with a negative gate voltage) if you have a negative supply voltage.
 
Last edited:
Administrator
Joined 2004
Paid Member
I have a couple of ancient BSS varicurves which switch high quality audio through a plethora of cmos switches and I doubt they are doing much damage to the signal.

I designed semiconductor ATE hardware somewhat over a decade ago with cmos switches in it and the cmos switches did not measurably degrade the performance.

Analog Devices makes high performance analog switches, these are what I would use.

Analog Switches Multiplexers | Analog Devices

Ones designed for bipolar supplies are what you want.

There are specific topologies that minimize distortion, and can still be made stable with high performance audio opamps - you may need to compensate for the additional capacitance at the opamp's summing input.

Look at their app notes for suggestions on how to use them effectively.
 
I have also recently looked into this matter.

Regarding getting the lowest THD from analog FET-switches you should avoid the regular 4066/4016 etc, since they add 0,5% THD :whazzat:
It looks like Analog Devices' ADG1611/1612/1613 are supreme, but they are also a bit expensive.
Then there is MAXIMs MAX4066 which offers 0,03% THD and are less expensive than AD's switches.

Next topic is the ON-resistance, which is in the 75 ohm range for regular 4066s, 0,2 ohm for ADG1612 and 45 ohms for MAX4066.

I reckon the ON-resistance - including its frequency-dependant variations - combined with the loaded and sourced impedance are the main factors for the distortion ... anyone please tell me if I am correct or not.
Anyway that leads me to my next discovery: The discreete switch JFET J105 has 3 ohms ON-resistance and is dirt-cheap.

But I have also been a bit with-holding using it for switching between MC and MM pickup signals because of the fear of DC, and audible noise. I might try this soon.
 
Administrator
Joined 2004
Paid Member
Relays have their own set of issues, and as far as I am aware they all have intrinsic thermocouples present in their construction which is mainly relevant for low level MC applications. Instrumentation reed relays are designed to have low thermal voltage offsets.

Relays really are out of scope given the title of the thread and the direction the OP said he wanted to pursue.
 
I reckon the ON-resistance - including its frequency-dependant variations - combined with the loaded and sourced impedance are the main factors for the distortion ... anyone please tell me if I am correct or not.

This is correct. The way to think about it is that FET switches distort because the channel resistance of the FET changes along with the voltage dropped across the switch FETs. This degree of change, taken into the context of the surrounding impedance, scales the nonlinearity.

For example, using a FET switch loaded into a virtual earth amplifier, assume a 10KΩ series resistor into a 100Ω FET switch into the VE amp. The signal voltage across the FET switch will be reduced by 10000/100 or 40dB by the series resistor. The resistance variation of the switching FET across this smaller voltage range may be 1Ω or less, resulting in overall distortion around -80dB.

The 'shape' of this resistance vs. voltage curve is pretty uniform, related to the geometry of the transistors, so aside from some unfortunate FET geometries, the resistance variation will directly scale along with the FET channel resistance.

This means that distortion could be reduced by 20dB by either choosing a FET that has 10Ω channel resistance, or by using a 100kΩ resistor - either will reduce the switch voltage and thus reduce the distortion of the switch circuit.

Using a higher resistance will increase circuit noise, so that can only be increased up to a point. Using a lower resistance FET will work, but the tradeoff is that a FET’s resistance is a function of the geometry of the FET, so reducing the resistance requires a larger FET geometry. Reducing the channel resistance by a factor of two will increase the FET capacitance by a factor of four - capacitance goes up as the area of the channel, whereas the resistance goes down linearly with the length of the channel. So, reducing the FET resistance by a factor of N will increase the FET capacitance by N^2. An improvement is still possible with a larger, lower resistance FET, but again, past a certain point, the capacitance will become too great to stabilize the VE amplifier, and/or the bandwidth of the switch circuit will be reduced to unacceptable levels.

The trick is therefore to optimize the FET and the circuit resistance so that noise, bandwidth and distortion are all in an optimal place relative to the circuit requirements. If you do that, it is not difficult to get distortion below -100dB, and much better if you’re clever.

A good thing to keep in mind is that traditional JFETs are designed for linear uses, which are unimportant for a switch circuit. Yes, a linear transition is useful, but the exact shape of this is unimportant - a switch FET needs to simply get through the linear region cleanly, and what’s more important is when it’s fully conducting or fully cut off. So, FET geometries optimized for switching will work far better than the traditional, nearly obsolete JFETs from the ‘70s.

The FET figure of merit is the product of its output capacitance and channel resistance - you want to minimize both, and there are FETs designed for modern switching regulators that use much more advanced vertical trench geometries, resulting in much better performance. A typical trench FET designed for a buck DC to DC converter will have a figure of merit about 100x better than a typical ancient chopper JFET like a J108. The reason is not that silicon has somehow improved, but that modern FET geometries are optimized to minimize the resistive and capacitive losses, essential to maximizing the efficiency of a switching DC to DC converter, with little regard for the linear region of these FETs.
 
Do a search for CMOS transmission gates. CD4052 is a useful configuration. The analog signal has to be in the middle of their supply rails and they need to be in a high Z circuit so that their resistance non-linearity is less significant, and you may want to both series open and shunt for best bleed. (3x 4016/4066) If you use JFETs, add 1/2 the audio to the control voltage for best linearity. There is an old diode bridge circuit but it probable thumps badly when switched. Some circuits swap bias to multiple diff-amps, sort of bipolar linear multipliers.


That's a really easy way of adding noise and distortion to the signal path. CMOS switch ICs are NOT suitable for quality audio. I've used them for PFL switching on disco consoles, but that's about it!


You're better off using cheap reed relays - a solid copper path is always preferred. It's also best to use a separate supply to operate your relays, so that the switching doesn't get into your sensitive audio paths.



I made a neat little PCB with eight reed relays and some audio sensing circuitry and some logic to select them according to which input had audio applied. This automated input selection on my hi-fi system!
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.