Unity resistance analog switch

[AccountRemovedPerUsersRequest]

hello.

I have a circuit that needs to have switching between analog signals. In a prototype I am using 4051 and it works but ON-Resistance is depended on analog signal level. I assume that this is a feature that every analog switch has more or less. However, maybe there are better alternatives or tricks I could use. Something like ADG841 (that looks almost too good to be true).

Analog signal is between 500mV and 3V. Frequency 1kHz. Supply voltage 5V.

A

[TimFox]

IIRC, with the 4051 switch, the ON resistance is more constant in one range of analog voltage vs. supply voltages than another.  Of course, all FET and CMOS switches have some variation in ON resistance vs. the channel voltage (analog input/output) vs. supply voltages.  You may be able to improve your performance by tweaking the supply voltages to make the "sweet spot" agree with your analog range.

[Zero999]

The on resistance changes because the gate-source voltages of the MOSFETs, which compose the switch, varies with the signal voltage.


http://dangerousprototypes.com/blog/2012/08/18/app-note-13/
https://www.analog.com/media/en/training-seminars/tutorials/MT-088.pdf
https://www.analog.com/media/en/news-marketing-collateral/product-selection-guide/choosing_switches_or_muxes.pdf

Use a an analogue switch with as lower on resistance as possible.

Keep the load impedance high.

And the smaller, the signal amplitude, the better.

[T3sl4co1l]

4051 what, the old school CD4000 version?  It'll be pretty awful at 5V, yeah (what is it 600 ohms peak, something like that?).  Basically it's made with MOSFETs with Vgs(th) around 2V, so, for intermediate voltages (like 2V above VSS and 2V below VDD), it's just barely beginning to turn on.  And at 5V say, there's that 1V inbetween where neither N/P is on very much, and that's basically why you get the lumpy Rds(on) vs. Vin curve.

Solution is to use bigger/better FETs with lower threshold; the 74HC version is good for 3-6V so should do quite nicely here.

And, yes indeedy, purpose-made analog switches have very good specs indeed; you can get them in quite low resistances (with nearly enough capacity for actual power switching(!)), just be aware of capacitance, both to GND and thru (off state, between terminals).

Or, if your signal is based on some kind of impedance (as most high frequency analog and digital signals are), then the on/off attenuation may be more meaningful here than Rds(on) and C(off).  The on/off ratio can be quite substantial even at high frequencies, for purpose made switches; they may use a 'tee' design internally (two series switches turned off, one shunt switch turned on inbetween them -- shunting crosstalk).

Lots of fine parts with DGxxx prefixes, or ADG from Analog Devices, like that one, yeah (note the supply rating though).  Some of them can get fairly pricey, so watch out for that; specialties include wide supply range (30V or more?), low leakage, integrated ESD protection, etc.  Common variants include SPST, SPDT, ganged, and SPI (or shift register) interface.

You mention frequency, but not current or impedance, so we can't tell what exactly you need, but at least the low frequency suggests you may not be very concerned with capacitance, so you can opt for quite low resistance if you wish.

Tim

[Cerebus]

You'll have to show us more of where and how you're using an analogue switch, plus all the stuff that people perennially miss out of these kinds of questions: type of signal, range of frequencies involved, impedances, voltage ranges, etc. before anyone can be more than handwavingly helpful.

There are all kinds of tricks and topologies to get around the non-linearities and other limitations of analogue switches and you could narrow down the list of ones to explain/suggest by filling the question out a bit as suggested above.

[moffy]

There is a technique used by multislope ADC's that might help. As long as you keep the input resistor values high compared to the switch ON resistance, the voltage at the switch will always be around 0v when on due to the virtual ground of the op amp.

[TimFox]

The point of the diagram above is that the resistance of the ON channel depends on the voltage between the channel and the power supplies to the switch, and the virtual ground keeps that voltage constant.

[Cerebus]

Er, it reduces the voltage swing (in the proportion of the R_DSon to the input resistor) but it doesn't keep it constant.

There's actually a better trick, same as that but put another matched switch in the feedback loop. At unity gain the input and feedback switches come close to cancelling each other's non-linearities. (A plot of RDSon versus voltage is roughly symmetrical and U shaped.)

[Terry Bites]

All mosfet swiches have the fwd/ reverse charcteristic. But by choosing a very low Ron Switch it becomes less of an issue.
There are many .25R swiches. This means better than 0.1% gain error in 600ohm system. ADG841 bring one of the be

Analog switches the movie:https://www.analog.com/en/products.html#

Of course relays still rule. Klictsach!. I'm sittin' here smoking wetted reed right now.

[Kleinstein]

Switching with a JFET has a constant on resistance (if a suitable gate signal is available), but it needs extra effort to control the gate signals. For 8:1  mux like the 4051 this can be a bit cumbersome.

Lowe R_on really helps. There are also a few with a lower Ron and the same footprint als the 4051, though often with low maximum supply.
74LV4051 from would be a relatively afordable option. The Ti version is around 30 Ohms - the Phillips/ons version is much higher !
Lower resistance often also comes with more leakage.

[Cerebus]

Switching with a JFET has a constant on resistance (if a suitable gate signal is available), but it needs extra effort to control the gate signals.

One has to qualify that with "if you keep V_ds low enough". As soon as V_ds becomes significant the channel begins to look like a constant current source with anything but a constant on resistance.

With the kind of scheme that Kleinstein is talking about, where (for an n-channel JFET) an on condition is V_GS = 0 (i.e. the gate is fed with a buffered copy of the signal you're switiching) and off V_GS = -lots (the gate is pulled down to the rails)  you need to keep V_ds significantly below V_pinch-off. i.e. you must keep your signal well into the ohmic region.

Random graph of JFET characteristic curves that shows all the relevant variables:



In the example device above, the maximum V_ds you could tolerate and still keep a constant R_on of ~200Ω would be about 1V.

[mawyatt]

Here's a NMOS switch that uses a charge pump to raise the gate voltage above the supply for lower Ron. Used these for for some special applications.

https://www.ti.com/lit/ds/symlink/ts5n412.pdf?HQS=dis-mous-null-mousermode-dsf-pf-null-wwe&ts=1645196249386

Best

[AccountRemovedPerUsersRequest]

Thank you! I identified two new ideas to improve the design.

A

[Zoli]

Thank you! I identified two new ideas to improve the design.

A

A third one:
https://www.ixysic.com/home/pdfs.nsf/www/CPC1018N.pdf/$file/CPC1018N.pdf