discrete JFETs vs CMOS analog switch

I need a switchable 0 /40dB attenuator for a 10Hz-20MHz 1Vp-p signal into a buffer amplifier (10k|3pF input loading). First I thought of using a SPDT analog switch and resistor divider, but even 'low capacitance' devices come with 15~20pF capacitance on the common pin and have poor off isolation for my purpose.

in --.--SW1--.--> to buffer | | 1kR | |--SW2--' 10R | gnd

When I simulated a 2N5484 J-FET, which have been around for at lease 30 yea rs, they beat the pants of any modern CMOS analog switch for off isolation and shunt capacitance. They only downside is more complex driving circuit because of negative gate drive and should have good AC grounding for best o ff isolation.

Is the small signal JFET best suited for this application ? There seems ver y little information on comparing the JEFTs analog switch to the CMOS switc h, especially AC performance.

Adam Seychell

Reply to
Adam Seychell
Loading thread data ...

Thanks for the part umber, I might get some. I got JFETs but not for that f requency. (I got complimentaries !)

Anyway, the only reason to use real JFETs is to use the linear range and th eir unique capability to deal with AC. The other thing about them is they c an do it in their linear range. In other words they can be easily used for gain control, there is no need for DC bias on he drain. No other device can do that except for those old light bulb things, opto something but they we re a Cds and a light bulb inside, my HP 339 has a couple. But of course the thing dos about dead turtle speed.

If all you need to do is switch, not modulate, mix or any of that other shi t you are probably not best off using the JFET. There are so many other opt ions. And if the signal is HF you can bias anything.

The 40dB drop you want should be accomplished with resistors, and the turni ng on or off of devices. And if you have to match impedance, damn. But assu ming you don't, figure just WHY you want those JFETs. There are other devic es that can easily do what you want, if I read this right. You said nothing about gain control or any of that. Just attenuation and, well whatever.

If all you want is to switch in attenuation, maybe a 2N7000 ? But I am not sure of the frequency on those.

Still, don't burn up my JFETs, I got a real use for them. }:-)>

Reply to
jurb6006

Another thing about using a MOSFET for switching signals is that the integral diode prevents blocking of signals more than a few hundred millivolts in the off state.

Reply to
Pimpom

Use two MOSFETs in series, gates tied together and driven, source tied to source and using the drains as the input and output. This prevents body diode conduction.

Reply to
Steve Goldstein

Achieving a flat spectral response in an attenuator or amplifier simply needs to understand this is done by impedance ratios. So when you have a load impedance of 10k 2pF ,@ 10 MHz realize that any ratio must use a ratio that makes the component capacitance negligible. This can be done by lowering the R impedance load so that 2 pF = 8kOhm at 10Mhz is negligible.

Reply to
Tony Stewart

There is a diode is from the drain and source to the body/backgate. In discrete MOSFETs the backgate is tied to the source so the diode is from drain to source, but on a chip (e.g. analog switch chip) the designers could connect the backgate to something else, (e.g. to the supply rails in the old 4016), so there is no diode from drain to source. Provided the signals are with the supply rails, the switch can block conduction between its terminals in either direction. More recent analog switch chips try to do things like reduce the on-resistance, in the 4066 this is done by switching the backgate of one FET to sometimes tie it to the signal pins, but this has disadvantages for charge injection, if that matters. Yet more recent ones probably do much more fancy stuff.

Reply to
Chris Jones

If the signals are small, it may be worth evaluating a NMOS-only switch like the FST3125, or one of the similar ones that are specified for analog operation. NMOS has a better on-resistance vs. capacitance trade-off than PMOS.

Also, I would suggest considering not trying to achieve >>40dB isolation in one stage of switching, and for higher frequencies, perhaps not even in one chip package. How about:

in --.--SW1a--.--SW1c--.--> to buffer | | | | S | | W | | 1 | | b | | | | | gnd | | | | | | | 1kR | |--SW2------------' 10R | gnd

Reply to
Chris Jones

Yes, that should work - as long as the switching bias voltage is substantially higher than the signal peak amplitude. That is, higher by at least the full 'on' gate voltage.

Reply to
Pimpom

a buffer amplifier (10k|3pF input loading). First I thought of using a SPD T analog switch and resistor divider, but even 'low capacitance' devices co me with 15~20pF capacitance on the common pin and have poor off isolation f or my purpose.

ears, they beat the pants of any modern CMOS analog switch for off isolati on and shunt capacitance. They only downside is more complex driving circui t because of negative gate drive and should have good AC grounding for best off isolation.

ery little information on comparing the JEFTs analog switch to the CMOS swi tch, especially AC performance.

Mechanical switches have ~0.5 - 1.5 pF (NKK are lowest I know) or would one of those telecom relays that John L. likes work for you?

George H.

Reply to
George Herold

20 MHz at 1 v p2p isn't that fast, why not use an op-amp with switchable gain resistors that can do less-than-unity gain:
Reply to
bitrex

The telecom relays are miles better than most any semiconductor in any spec except switching speed.

Jfets are OK switches, but capacitances can be high and gate drive is usually a nuisance.

A 2-stage attenuator might be better for a 100:1 ratio. Maybe use one DPDT cmos switch.

--
John Larkin         Highland Technology, Inc   trk 

The cork popped merrily, and Lord Peter rose to his feet.   
 Click to see the full signature
Reply to
John Larkin

JFETs work fine for that, especially if your buffer has a high-impedance input.

There are lots better muxes than the ones you cite--check out the TMUX1511--it's almost as good as a relay, and 100k times faster.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

There aren't any small-signal discrete MOSFETs left, except for a few dual-gate ones such as the late lamented BF998. Looks as though Toshiba is the only remaining maker.

Dual-gate FETs make really good medium-speed sampling gates.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Don't understand why to use an analog switch in the direct signal path here at all, in isolation from un-stated requirements. The signal is AC coupled as far as I can tell, and isn't that fast or large amplitude. A switchable VGA eliminates the isolation and loading problems by not having them. only advantage I can see is budget

Reply to
bitrex

Yep, that's the ticket. Amplifiers for good linearity in this range are available for analog video, and are generally 75 ohm items. The good news: there's lots of em. The bad news: they don't generally use or need 40 dB of attenuation.

Reply to
whit3rd

Fortunately there are a number of small N-MOSFETs with ~3pF output capacita nce and few ohms on resistance. A quick simulation of back to back N-FETs with schematic in link below. This circuit will have horrendous charge inje ction and slow switch speed, but ~3pF capacitance loading in off state, eve n less with on state thanks to low collector capacitance of $0.05 bipolars. The 1nF across the gates add AC grounding when in off state, or the NPN is in full saturation.

Small dual MOSFETs in 6 pin packages NX3020NAKS NXP, VT6K1 Rohm. The NPN/PN P are also readily available as a 6pin SOT363, SOT666 et.

formatting link
ng?dl=0

Reply to
Adam Seychell

So where do you buy them?

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
 Click to see the full signature
Reply to
Phil Hobbs

Digikey claims to have the NX3020NAKS - 23,000 in stock

CH

Reply to
Clifford Heath

Am 02.07.20 um 21:12 schrieb Phil Hobbs:

what about these?

<
formatting link
>

Gerhard

Reply to
Gerhard Hoffmann

A BFT25 has about 0.25 pF of collector capacitance and a reverse beta around 4. Something could be done with that.

--
John Larkin         Highland Technology, Inc   trk 

The cork popped merrily, and Lord Peter rose to his feet.   
 Click to see the full signature
Reply to
John Larkin

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.