8
\$\begingroup\$

I have always thought that MOSFETs have four pins [see wikipedia], one of them connected to the body (i.e. substrate). However, when trying to order some, I have been completely unable to find any with a separate connection for the body; all the countless ones I checked had it already internally connected. It would be really nice if someone knew some way (ie. nomenclature or something) to differenciate those! In other words: I am asking for an efficient way to search for MOSFETs where the body is not internally connected.

The generally applicable question above is more interesting of course but in case I have missed any possible solutions I will also describe my specific problem: I am designing a quite accurate analogue amplifier and a sample and hold subcircuit. The sample and hold circuit needs an analogue switch with low leakage (the next stage has high input impedance and the usual leakage of a few nanoamps gives unaceptably high voltage errors and drift). I had planned to either use an IC or build a similar cicuit to that shown in the lower right corner of the attached schematic. The trick with that circuit is that Q1 has almost no voltage drop across it when the switch is closed, so it also has almost no leakage. Now the problem I’ve encountered is that the brute force approach of checking datasheets (on the farnell.com webshop) of either MOSFETs with 4 or more pins or analogue switches has not yielded any results after more that 2 hours. The problem with MOSFETs is that they have several pins connected to the same thing (e.g. drain) and all analogue switches I’ve seen have too much leakage current and generally bad documentation.

Thank you for your time

[English is not my native language; please forgive my blunders or — much better — edit them]

EDIT: Whilst the part suggestions are helpful for my specific case, I would also be interested in general terminology as this is more likely to be helpful to others.

similar circuit (my supply is ±5V and uC logic levels 0–3.3V)

\$\endgroup\$
3

6 Answers 6

4
\$\begingroup\$

The parts on that schematic are very low capacitance (and, unfortunately, that generally means relatively high 'on' resistance) compared to the otherwise nice ones that Ignacio mentions in his comment.

The switch is designed to have low charge injection since that directly affects the accuracy of the nulling.

You might consider something like the ADG5236 in a modern design, which has 0.6pC of charge injection, which is only 60nV on a 10uF capacitor.

Or use an amplifier that has the autozero circuitry internally, but I see you're actually working on a sample-and-hold circuit.

\$\endgroup\$
3
  • \$\begingroup\$ This part seems suitable, thank you. Yes I want a sample and hold circuit, not a chopper-stabilized amp. I have now noticed that googling “low charge injection” does yield good results; I had apparently used the wrong keywords and was drowned in noise. \$\endgroup\$
    – user82975
    Commented Aug 11, 2015 at 16:39
  • \$\begingroup\$ But how did you find that part? What did you search for, did you use any references etc.? This is likely to seem very stupid for someone experienced, but for some reason narrowing down search results has been very hard for me. \$\endgroup\$
    – user82975
    Commented Aug 11, 2015 at 16:46
  • \$\begingroup\$ I gave you a bit of a hint- I searched with a parametric engine for a switch that has high voltage capability and relatively high switch 'on' resistance. You might try searching directly for 'low charge injection' too. \$\endgroup\$ Commented Aug 11, 2015 at 17:07
3
\$\begingroup\$

The 4 pins shown inside the MOSFET symbol are a representation of the physical structure of the MOSFET. However, for the vast majority of MOSFETs, the substrate is an integral part of the MOSFET and can't be separated from the source.

I have seen low-current MOSFETs that DO have the substrate come out as a separate lead but I haven't seen any of those available for several decades. Doesn't mean that they don't currently exist - I just have never needed to find any.

Most people use J-FETs such as J112 (N-channel) or J172 (P-channel) for doing the kind of switching that you are trying to do.

\$\endgroup\$
1
  • \$\begingroup\$ Perhaps I am confused, but wouldn’t J-FETs be forward biased sometimes, messing up the charge on the capacitor? \$\endgroup\$
    – user82975
    Commented Aug 11, 2015 at 16:34
2
\$\begingroup\$

4 terminal MOSFETs are getting hard to find.

A former Micrel P-channel family MIC94030/31/50/51 is now made by Microchip and seems to be in production. https://www.microchip.com/wwwproducts/en/MIC94030#additional-features .

Also check out the 4007 FET array IC e.g. https://datasheet.octopart.com/MC14007UBCPG-ON-Semiconductor-datasheet-531527.pdf It has 2 N channel MOSFETs with a shared substrate pin and 2 P channel MOSFETs with a separate shared substrate pin. The P and N gate connections are tied together in pairs.

\$\endgroup\$
2
\$\begingroup\$

Take a look at the ALD1115 or ALD1105 from ADVANCED LINEAR DEVICES, INC. It'a a complemantary MOSFET N/P pair. V+ is connected to the substrate, which is the most positive voltage potential. V- is connected to the most negative voltage potential

\$\endgroup\$
1
\$\begingroup\$

These are not strictly discrete NMOSFETS, but actually NMOS analog switches arranged to form quad 2:1 MUXs: ADG774A. I guess it is still relevant for this topic, as NMOS switches are a prime example of the usefulness of 4-terminal MOSFETs.

I recently became aware of this part when looking into low leakage switches. Their specs are so far beyond the conventional CMOS switches, that it is ridiculous. Ridiculous that there don't seem to be more of them in production.

If you use Analog's parametric search for switches, you realize that this part has about 1 order of magnitude better on-resistance and 1 order of magnitude better leakage and 1-2 orders of magnitude better switch speed than comparable CMOS switches...

Their sole drawback seems to be that you loose about 1-2 Volt of signal handling range below VDD.

\$\endgroup\$
0
\$\begingroup\$

The big picture here is that the picture itself is old. The schematic you show was how it may have had to be done in the early 1980s. Today you can get op-amps and integrated switches that will perform much better than the now-40-year-old semi-discrete solution. Orders of magnitude better, in fact. In a way, we live in somewhat easier times when it comes to such designs. And, most importantly, it’s not so obvious that you need the sample-and hold. An ADC followed by a DAC will perform better than most analog sample-and holds would for the slow signals you’re working with. In fact, the suitable ADC and DAC may cost less than high-performance switches and op-amps needed for a good sample-and hold. It sounds crazy but it’s not as crazy as it sounds in fact.

\$\endgroup\$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.