SPDT analog switch for passing negative voltages

[microvia]

Hi,

I'm looking for a single-channel SPDT analog switch running at 3.3V and capable of passing negative voltages of up to -12V (-25V would be even better, yet not mandatory).
The TMUX4827 fits the needs but is a bit expensive (2.50 euros / 100pcs, which I can afford). Isn't there a cheaper similar device ? I don't mind using a 2-channel type if necessary.
However I will avoid devices like CPC1017/1117 which take too much place on my existing pcb.
Thanks.

[Benta]

Vishay/Siliconix DG419 should do it.

[voltsandjolts]

The TMUX4827 fits the needs but is a bit expensive (2.50 euros / 100pcs, which I can afford). Isn't there a cheaper similar device?

TMUX4827 is quite schmick.
1.56 euros / 100pcs at LCSC ...ah, only 74 in stock

[microvia]

Vishay/Siliconix DG419 should do it.

Interesting. A few variants are indeed cheaper, but most are significantly more expensive.
Will have to check the differences.
All I need is to let one of two negative voltages pass, both under 1mA, to drive a vaccuum tube grid.

[PCB.Wiz]

I'm looking for a single-channel SPDT analog switch running at 3.3V and capable of passing negative voltages of up to -12V (-25V would be even better, yet not mandatory).

Try TMUX4157 ,  NLHV4157, SGM3716 - those have +ve control signals and VEE is -4 to -12V, intended for depletion mode FET control I think.

All I need is to let one of two negative voltages pass, both under 1mA, to drive a vaccuum tube grid.

You could derive VEE from a couple of diodes to the -ve supplies ?

[voltsandjolts]

Maybe a couple of opto-isolators, or ssr's would do. They can be had for like 40 cents each.

[microvia]

I'm looking for a single-channel SPDT analog switch running at 3.3V and capable of passing negative voltages of up to -12V (-25V would be even better, yet not mandatory).

Try TMUX4157 ,  NLHV4157, SGM3716 - those have +ve control signals and VEE is -4 to -12V, intended for depletion mode FET control I think.

All I need is to let one of two negative voltages pass, both under 1mA, to drive a vaccuum tube grid.

You could derive VEE from a couple of diodes to the -ve supplies ?

I could, but I can only power the switch under 3.3V.

[Zero999]

I recommend the TMUX4827 in the original post.

The DG419 is no good because it won't work down to 3.3V and can't handle voltages outside the supply range.

TMUX4827 might work, but not with positive voltage. That might or might not be any good. The original post didn't mention positive.

A dual photovoltaic opto-coupler, with the output connected to depletion MOSFETs for normally closed and a enhancement MOSFETs for normally open, but that probably takes up too much room on the PCB

[microvia]

I recommend the TMUX4827 in the original post.

The DG419 is no good because it won't work down to 3.3V and can't handle voltages outside the supply range.

TMUX4827 might work, but not with positive voltage. That might or might not be any good. The original post didn't mention positive.

A dual photovoltaic opto-coupler, with the output connected to depletion MOSFETs for normally closed and a enhancement MOSFETs for normally open, but that probably takes up too much room on the PCB

It's a good option which requires additional room. Same problem if I use opto-triacs.

[voltsandjolts]

A dual photovoltaic opto-coupler, with the output connected to depletion MOSFETs for normally closed and a enhancement MOSFETs for normally open, but that probably takes up too much room on the PCB

Couldn't the optocoupler alone do the job, since it's just 1mA?
NPN output will conduct in one direction, which is all that's needed IIUC.

[PCB.Wiz]

I could, but I can only power the switch under 3.3V.

I was meaning derive the required VEE from your most negative gate voltage, the control line is positive logic, above GND.

[microvia]

How about using a NC7SB3157P6X ? Datasheet is at https://www.onsemi.com/pub/Collateral/NC7SB3157-D.PDF .
I need to connect one of two negative voltage sources to the grid of a vacuum tube. One of these two voltages won't exceed -11V. In other cases the other won't, but some circuits can deliver up to -25V.
Current passing through the grid is really negligible. I said 1mA by mistake. It's around 1µA, so much that I even maneged to drive the grid using the voltage generated by a APV1121S photovoltaic optocoupler.

[Zero999]

How about using a NC7SB3157P6X ? Datasheet is at https://www.onsemi.com/pub/Collateral/NC7SB3157-D.PDF .
Since I need to connect one of two negative voltage sources to the grid of a

(Attachment Link)

If you have a negative input voltage, then that will work, otherwise no, it won't.

I'm confused about what your requirements are.

What power supply voltages are available?

What's the voltage range of the signal being switched?

How much current does the switch have to conduct?

[microvia]

I updated my above post.
More precisely, the tube to drive is a magic eye, which grid will not pass more than 1µA with a resistor before the anode of at least 470K (which is always the case anyway).
Without resistor (for the sake of testing the worst case), the maximum current I measured didn't exceed 2mA.
In most cases the grid voltage will vary from 0V to -10V but there may be exceptions where up to -25V may have to pass through.
The device I'm considering is to be powered at 3.3V.

[Doctorandus_P]

Maybe you can use one of the "photomos" relays. These have a LED on the primary side, and a bunch of photocells and two "back to back" MOS fets on the "secondary side".

Another option could be to switch a "normal" positive voltage, and amplify it with an opamp in an inverting configuration. I guess there are a lot of ways to design around (or within) the limitations of (cheaply available) semiconductors.

I quite often see this happening. Some nice IC that replaces a handful of other parts, but when you look up the price, then that single IC is often much more (4x or more) expensive then the other / discrete solution. Just recently I bumped into the L99BM1T It provides galvanic isolation for an SPI bus. A quick check showed it costs EUR 5 in low quantities. while optocouplers are between 10ct and EUR1, depending on their speed and such.

Maxim has become notorious for this. They made a lot of nice IC's, often started with somewhat reasonable prices, and after a few years prices become significantly more expensive and the parts get discontinued altogether.

Another example is the TPIC6B595. It's a nice IC to get some (higher current) drivers on a uC. but it's much more expensive then a 74hc595 (20ct) with a bunch of "digital transistors" (with built in resistors, these cost around 2ct each). I am thinking about designing in an STP16CP05 as a relay driver. It costs around EUR1.5, but you get 16 outputs for that. BOM cost is still 3x that of the solution with the 2xHC595 + 16 BJT's but it saves PCB area and assembly costs.

[microvia]

Maybe you can use one of the "photomos" relays. These have a LED on the primary side, and a bunch of photocells and two "back to back" MOS fets on the "secondary side".

Another option could be to switch a "normal" positive voltage, and amplify it with an opamp in an inverting configuration. I guess there are a lot of ways to design around (or within) the limitations of (cheaply available) semiconductors.
(...)

Photomos is too expensive. OPAs will take too much space on the pcb which is already tight and exclusively powered at 3.3V.
Hence why I considered the NC7SB3157P6X as a SPDT switch to pass negative voltages as described above.

[PCB.Wiz]

Hence why I considered the NC7SB3157P6X as a SPDT switch to pass negative voltages as described above.

Not even close, that is a +1.65 to 5.5V part !
If you really need to tolerate -25v the -13v parts I gave above could have clamp zeners.
Otherwise, a pair of digital PNP parts driving P-JFET x2 could switch -25v

[microvia]

As there's no IC that can handle the requirements, and since I hardly use mosfet transistors, I took the opportunity of asking flux.ai to design it.
Well, I'm quite impressed by the output even though it took quite some time to obtain a readable schematic, which in turn had to be manually entered in KiCAD.
From what I understood, bootstraps / charge pumps had to be added in order to let negative voltages pass using a GPIO command.
I don't yet understand how this works but here's the overall schematic:


What do you think ? Thanks !

[PCB.Wiz]

I took the opportunity of asking flux.ai to design it....
I don't yet understand how this works but here's the overall schematic:
What do you think ? Thanks !

 
That's because it is AI ideas-in-a-blender, and simply will not work.
You can sort of see what they were trying to do, create a series FET SW, a la SSR's  but for that FETS need to be common source, not diodes in series as drawn.
FETS also need level shifters, but negative rails are not common so AI blender missed that detail.
 

You can do this with a single fet to each rail, and a level shifter to each gate. (PNP or PMOS)
To avoid cross conduction spikes during handover, you can either use 2 diodes, or 2 resistors may be enough at low currents.

What range of actual grid voltages and currents does this need to handle ?
If you need to go to 0V, a bit more care is needed on the FET gate level shifters.

[microvia]

Thanks for your feedback.
The grid current is in the µA range. In all cases, magic eye tubes have a series resistor at the anode of its triode section (which grid receives the negative voltage from the SPDT).
Removing this resistor would cause the grid current to reach almost 2mA.
Grid voltage can be anything between -12V and 0V.

[voltsandjolts]

With such low currents, simply two opto-isolators is worth trying.

[PCB.Wiz]

Thanks for your feedback.
The grid current is in the µA range. In all cases, magic eye tubes have a series resistor at the anode of its triode section (which grid receives the negative voltage from the SPDT).
Removing this resistor would cause the grid current to reach almost 2mA.
Grid voltage can be anything between -12V and 0V.

To get to 0V, you need to drive the FET gate above 0V with the level shifter, choose a low threshold FET - generic 7002's are a bit high on vgs.

Something like this can drive the gates all the way up to close to +3V from a 3.3V CMOS out MUX signal, while allowing pull down to -25V when off.
L = NGB and H = NGA
I just picked a round number, ~ 100uA could even be high for a MUX FET gate driver.
Or use a mute transistor for switch,  they avoid the S-D diode of a nmos part.
eg ROHM IMX25 gets a dual mute transistor.

[microvia]

It looks like the additional parts will take quite some space and I'm not familiar with FET design (hence the flux.ai experiment).
I guess I'd rather use opto-couplers. Photovoltaic devices aren't suitable: the grid current of the magic eye is so low that they can directly drive it ! I sent a square wave of variable frequency to the APV1121S such that the magic eye could deflect from min to max ! But I had to opt for a diode/cap charge pump for cost reasons. So... I'm a bit stuck for now.

Edit: I found the CPC2317, a dual optocoupler with one SPST in NO and the other SPST in NC position by default, which allows making a SPDT driven by a single GPIO without using an inverter.
However, this part takes quite some place, and I don't like enlarging the whole pcb. The image below looks like if some space could be easily saved to include the optocoupler, but that space is either taken by silkscreen detailing the pads, or is simply required to protect nearby components from being too close to soldering pads.

[PCB.Wiz]

It looks like the additional parts will take quite some space and I'm not familiar with FET design (hence the flux.ai experiment).
I guess I'd rather use opto-couplers.

The parts come in sot363, but if you only need -12v, why not use one of the VEE switches I already mentioned?
Ie Ti TMUX4157N ,  or onsemi NLHV4157, or SGM3716
Addit: maybe also TI TMUX4053, has 1v8 control, slightly out of spec at 3.3v,-12v  but that may affect only RON above 0V


It’s not clear what your signals are, but another approach is to mux in the 0~3v3 domain, and use a single sot25 opamp to amplify and invert to 0~-12v ?

[azone]

Good answer was the first reply....DG-series. DG419 etc... many of them have VL for logic (3.3v) and can do +/-15v