RF relays

No sure about their capacitance but probably really low, considering the internal distances required for a 70kV rating:

--
Regards, Joerg 

http://www.analogconsultants.com/

selling

TIA as

have

Johnson

the

the

or

Use a latching relay with both ends of the coil(s) driven by open collector drivers? Art

Can you put a guard shield around the reed relay body?

tm

Your cap figure is halfway between the 160fF Jeroen claimed and the 950fF I claimed in the the past. I wonder which is correct - I probably should go measure mine more carefully. It`s an interesting idea to guard the coil, thanks for sharing.

feedback

selling

TIA as

have

Johnson

dominate the

the

or

Good idea but those are mostly a few pF as well. Since fiberoptics guys like Phil always have high voltages for photodiode biasing available, how about taking Art's idea one step further: Drive both ends of the latching relay coil through diodes and when it has switched over reverse-bias them. That can get you down to fractions of a pF. If suitable latching relays can be found, which may not be very trivial.

--
Regards, Joerg 

http://www.analogconsultants.com/

feedback

selling

TIA as

I have

Johnson

dominate the

the

coil, or

Thanks. Fractional picofarads are always a challenge since the self-capacitance of a small but macroscopic object is down in that range--the self-capacitance of a 1 cm radius sphere is 1.12 pF. In this instance I have to bootstrap away a few picofarads worth of probe-to-shield capacitance, so another half a pF or so isn't such a big hit. It's more about trying to make it work in both the 1G and 50G modes. (Fun.) I'm tempted to use a reed switch and the guts of a dishwasher solenoid, but it has to Look Professional. (tm)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 USA 
+1 845 480 2058 

hobbs at electrooptical dot net 
http://electrooptical.net

I measured three different types. The GS6 ones were the best--the others were all above 1 pF, so I think there are real differences in the designs. A bit of iron would make a lot of difference, but there don't seem to be a lot of iron core SMT relays.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 USA 
+1 845 480 2058 

hobbs at electrooptical dot net 
http://electrooptical.net

Something like this maybe?

Maybe the bootstrapping isn't necessary. Might slow down the 50G mode,

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom timing and laser controllers 
Photonics and fiberoptic TTL data links 
VME  analog, thermocouple, LVDT, synchro, tachometer 
Multichannel arbitrary waveform generators

Thanks.

Yes, that's pretty much what I had in mind, except that maybe the coil should probably be bootstrapped from a BF862 following the SJ, I'm not sure yet. (The bandwidth/noise tradeoff is relatively subtle and layout-dependent.) My version shorts the 1G resistor to the output rather than to ground, but that doesn't matter much as long as the loose end can't couple noise current into the SJ. (It's really weird worrying about the current noise of a 1G resistor.) ;)

Putting the switches in series would help more if the coil->contact capacitance didn't dominate to such a degree, but in the present case it's about the same as using a single section.

Any experience with glass reed switches?

Thanks

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 USA 
+1 845 480 2058 

hobbs at electrooptical dot net 
http://electrooptical.net

feedback

selling

TIA as

have

Johnson

dominate the

the

or

Thanks. The output capacitance of the drivers would be several pF at least, which is more or less a dead short at these impedance levels, but I could probably bootstrap both of them from the amplifier output (as JL suggested) or from a bootstrap on the SJ.

The amplifier is a BF862 follower running straight into the - input of an ADA4898-1, so I have a low impedance copy of the input available.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 USA 
+1 845 480 2058 

hobbs at electrooptical dot net 
http://electrooptical.net

Too much. I don't like them. They twang a lot when they close (vibration in the mag field), have horrible thermoelectrics, aren't very reliable, are big and expensive. But you can buy ones with shielded/guarded contacts, and some of my complaints aren't relevant in a TIA.

They have bad magnetics, too much air gap, so need a lot of coil power. You can leverage the pickup/dropout ratio by driving the coil full blast for a few milliseconds, then back off 2:1 maybe steady-state. Their worst failure mode is probably stuck-closed.

Can you just build two TIAs and switch between them with a relay? Guess not, if you need super low input capacitance.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom timing and laser controllers 
Photonics and fiberoptic TTL data links 
VME  analog, thermocouple, LVDT, synchro, tachometer 
Multichannel arbitrary waveform generators

Use a reed switch, and a magnet stuck on the end of a car door lock actuator. The door lock actuator runs on 12 V and some (most?) have a plastic business end. Plus it runs on the low voltage you probably already have.

Throw it all in an opaque box and call it shielding. :)

Matt Roberds

OK, more like this?

I've got to get out of the silly habit of assuming that the voltage at a summing point is zero.

--

John Larkin                  Highland Technology Inc 
www.highlandtechnology.com   jlarkin at highlandtechnology dot com    

Precision electronic instrumentation 
Picosecond-resolution Digital Delay and Pulse generators 
Custom timing and laser controllers 
Photonics and fiberoptic TTL data links 
VME  analog, thermocouple, LVDT, synchro, tachometer 
Multichannel arbitrary waveform generators

Phil Hobbs schrieb:

Hello,

may be a relay with more distance between the coil and the contacts would be better. I think of a distance of some inches between them and a plastic rod to transfer the force from the magnet to the contacts. I don't know if such relays exist.

Bye

Uwe Hercksen schrieb:

Hello again,

if you take a microswitch and a solenoid coil, you may build such a relay by yourself. You only need a push bar and a sort of guidance to fix the push bar at the right place. I hope there are some microswitches suitable for your problem.

Bye

r

IJust out of curiosity.... Why did you call your thread RF relay? I am missing the RF relay part in your text. I looked up the Omron G6K which you referenced and it does not strike me as an RF relay.

This is what I think of when I think RF relay:

l,

far

there a different kinds of g6k

-Lasse

summing

Yes, thanks. I think I win by using only one section of the relay and grounding the cold end of the relay coil, making sure that it's the outer layer that's grounded. There's a few pF from the SJ to ground anyway, so an extra half pF won't change the noise much. The BF862 + ADA4898 circuit is 20-odd dB quieter than their production model, so nobody will complain about half a dB, I shouldn't think. ;)

I get more and more fond of the BF862 + BFT92 PNP wraparound circuit. Its simplest form is like this:

+2.7V 0 | +----------+ | | | | R | R 60 | R | R | | |< +--------| BFT92 | |\ | | In | | |-+ | 0-->| BF862 | |-+ *----0 Out | | *----------+ | | | | Isink | V

It gives a loop gain of V_BE/26 mV, so since 1/g_m of the FET is about

30 ohms, you get a 1-ohm output impedance follower *with the same noise as a barefoot BF862*. You also get a whole bunch of bandwidth so it doesn't destabilize the op amp loop. You bootstrap it just the way you show, with something like a BFR505 follower. The FET should run with about 1.8-2.2V D-S to get the full transconductance with low gate leakage. Because of the PNP's lowish beta, it works better with most of the current going through the FET--10 mA for the FET and 2 mA for the PNP is about right.

You can get more loop gain, and hence better high-frequency linearity, by using a second BF862 (D-G shorted) as the drain load, but that costs

3 dB in SNR. In this instance I'm making a copy of the second FET's I_DSS using an instrumentation amp and a BJT, and using that for the source load.

I usually tart it up a bit to reduce drift and get a bit more loop gain--an NPN follower between the drain load and the PNP can be a win, but you have to make it _fast_.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 USA 
+1 845 480 2058 

hobbs at electrooptical dot net 
http://electrooptical.net

They do, but you have to worry about the self-capacitance of their much larger contacts. I'd really like to have the mechanism of the G6K ones with a bit of iron in the magnetic loop to allow the coil to be further away.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC 
Optics, Electro-optics, Photonics, Analog Electronics 

160 North State Road #203 
Briarcliff Manor NY 10510 USA 
+1 845 480 2058 

hobbs at electrooptical dot net 
http://electrooptical.net