Driven switch body. (what about greater than one bootstrap)

G

George Herold

Guest
So I copper clad proto-ed the grayhill rotary switch into my
PD thing. I can post pics.. at 1 MHz an inch of bus wire is
not a big deal. Anyway with the switch body grounded the
step response/ BW went to crap. (100 k ohm FB) When
I floated the switch with my dremel, and I got a nice improvement
in BW ~400 kHz to 800 kHz
(I wasn't looking at the step response at the time..
and BW measurements are squishier than the time domain.)

Anyway then I tried driving the switch body from the input signal,
I got a little improvement. (In step response.)
My thinking was that the driving the switch body, would pull all the
un-connceted switch pins along with it. And now I'm thinking
if I drive it a bit over unity...? or is that the road to misery?

I could add a shell connected copper plane under the switch
and add a circular copper fence.

George H.
 
George Herold wrote...
So I copper clad proto-ed the grayhill rotary switch into my
PD thing. I can post pics.. at 1 MHz an inch of bus wire is
not a big deal. Anyway with the switch body grounded the
step response/ BW went to crap. (100 k ohm FB) When
I floated the switch with my dremel, and I got a nice improvement
in BW ~400 kHz to 800 kHz
(I wasn't looking at the step response at the time..
and BW measurements are squishier than the time domain.)

Anyway then I tried driving the switch body from the input signal,
I got a little improvement. (In step response.)
My thinking was that the driving the switch body, would pull all the
un-connceted switch pins along with it. And now I'm thinking
if I drive it a bit over unity...? or is that the road to misery?

I could add a shell connected copper plane under the switch
and add a circular copper fence.
Click to expand...

You have to check all the gain positions.
With optical signals.


--
Thanks,
- Win
 
On Monday, September 16, 2019 at 10:19:14 PM UTC-4, Winfield Hill wrote:
George Herold wrote...

So I copper clad proto-ed the grayhill rotary switch into my
PD thing. I can post pics.. at 1 MHz an inch of bus wire is
not a big deal. Anyway with the switch body grounded the
step response/ BW went to crap. (100 k ohm FB) When
I floated the switch with my dremel, and I got a nice improvement
in BW ~400 kHz to 800 kHz
(I wasn't looking at the step response at the time..
and BW measurements are squishier than the time domain.)

Anyway then I tried driving the switch body from the input signal,
I got a little improvement. (In step response.)
My thinking was that the driving the switch body, would pull all the
un-connceted switch pins along with it. And now I'm thinking
if I drive it a bit over unity...? or is that the road to misery?

I could add a shell connected copper plane under the switch
and add a circular copper fence.

You have to check all the gain positions.
With optical signals.


--
Thanks,
- Win
Click to expand...

Yeah well the driven switch idea is a total fail at this point.
(I'm not sure why I saw an improvement when I dremeled the copper
around my switch body... maybe a poor solder junction in the first
circuit?) Anyway that reported result is not true.
I think I can have a 5 pF switch and fast,
or a 5pf switch and low noise,
but not both*. And the back panel (~1"x2" Al sheet) could be
changed... For the diode laser experiments the full range can be
totally useful.. that's a fast + switch app. (I think?)

George H.

*Or low noise and fast but no switch.. or lower C switch...
some jumper thing? (it seems crazy to invite people to re-solder
your circuits...)
 
George Herold wrote...
Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)
Click to expand...

It's low, but worls horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?


--
Thanks,
- Win
 
On Tuesday, September 17, 2019 at 7:35:29 PM UTC-4, George Herold wrote:
On Monday, September 16, 2019 at 10:19:14 PM UTC-4, Winfield Hill wrote:
George Herold wrote...

So I copper clad proto-ed the grayhill rotary switch into my
PD thing. I can post pics.. at 1 MHz an inch of bus wire is
not a big deal. Anyway with the switch body grounded the
step response/ BW went to crap. (100 k ohm FB) When
I floated the switch with my dremel, and I got a nice improvement
in BW ~400 kHz to 800 kHz
(I wasn't looking at the step response at the time..
and BW measurements are squishier than the time domain.)

Anyway then I tried driving the switch body from the input signal,
I got a little improvement. (In step response.)
My thinking was that the driving the switch body, would pull all the
un-connceted switch pins along with it. And now I'm thinking
if I drive it a bit over unity...? or is that the road to misery?

I could add a shell connected copper plane under the switch
and add a circular copper fence.

You have to check all the gain positions.
With optical signals.


--
Thanks,
- Win

Yeah well the driven switch idea is a total fail at this point.
(I'm not sure why I saw an improvement when I dremeled the copper
around my switch body... maybe a poor solder junction in the first
circuit?) Anyway that reported result is not true.
I think I can have a 5 pF switch and fast,
or a 5pf switch and low noise,
but not both*. And the back panel (~1"x2" Al sheet) could be
changed... For the diode laser experiments the full range can be
totally useful.. that's a fast + switch app. (I think?)

George H.

*Or low noise and fast but no switch.. or lower C switch...
some jumper thing? (it seems crazy to invite people to re-solder
your circuits...)
Click to expand...

Oh or JL's idea of switched cascodes... what's the C_ce capacitance
when switched off? (collector to emitter)

GH
 
On Tuesday, September 17, 2019 at 8:03:19 PM UTC-4, Winfield Hill wrote:
George Herold wrote...

Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)

It's low, but worls horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?
Click to expand...

I think the cascode gives me lower noise. (But I haven't done the
calculation nor any noise measurements... I find it useful to do those
together as I can use the electronics to check my math :^)

I've got this hair-brained idea of looking at diode laser noise just below threshold.
My guess is it probably won't work, but I need a low noise, fast PD,
to have a chance.

George H.
--
Thanks,
- Win
Click to expand...
 
On 17 Sep 2019 17:03:06 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

George Herold wrote...

Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)

It's low, but worls horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?
Click to expand...

A cascode needs a keep-alive current. Cascode really helps to get good
noise and bandwidth results for a high-capacitance photodiode.
 
On Wednesday, September 18, 2019 at 3:04:09 PM UTC-4, John Larkin wrote:
On 17 Sep 2019 17:03:06 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:

George Herold wrote...

Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)

It's low, but worls horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?

A cascode needs a keep-alive current. Cascode really helps to get good
noise and bandwidth results for a high-capacitance photodiode.
Click to expand...

Many of the signals I look at are small, 0-10%, changes on some background
light level.. so the cascode should be fine.

What I didn't appreciate, until today.. (even though it says as much
in Phil's book.) Is that with the cascode the speed of the TIA opamp is not
that important anymore. You can't speed it up with a faster opamp.
The time cosntant is mostly the PD capacitance times R_e (25mV/I_c)

I keep bashing my head into this 5pF of switch C. (Even with no cascode,
just a bootstrap.) I've ordered a few of these faster THS4631 opamps
http://www.ti.com/lit/ds/symlink/ths4631.pdf

Four times the BW should mean I need twice as big a compensating feedback cap.
I'll see.
 
On Wednesday, 18 September 2019 20:35:19 UTC+1, George Herold wrote:

> I keep bashing my head into this 5pF of switch C.

could always reduce it with a knife switch :)

What's the c of a coaxial relay?


NT
 
On Wednesday, September 18, 2019 at 7:35:07 PM UTC-4, George Herold wrote:
On Wednesday, September 18, 2019 at 3:35:19 PM UTC-4, George Herold wrote:
On Wednesday, September 18, 2019 at 3:04:09 PM UTC-4, John Larkin wrote:
On 17 Sep 2019 17:03:06 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:

George Herold wrote...

Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)

It's low, but worls horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?

A cascode needs a keep-alive current. Cascode really helps to get good
noise and bandwidth results for a high-capacitance photodiode.

Many of the signals I look at are small, 0-10%, changes on some background
light level.. so the cascode should be fine.

What I didn't appreciate, until today.. (even though it says as much
in Phil's book.) Is that with the cascode the speed of the TIA opamp is not
that important anymore. You can't speed it up with a faster opamp.
The time cosntant is mostly the PD capacitance times R_e (25mV/I_c)

I keep bashing my head into this 5pF of switch C. (Even with no cascode,
just a bootstrap.) I've ordered a few of these faster THS4631 opamps
http://www.ti.com/lit/ds/symlink/ths4631.pdf

Four times the BW should mean I need twice as big a compensating feedback cap.
I'll see.

OK, f the rotary switch. (I've thought about some internal
jumpers or something... but if you ask students to open up your stuff..
well be prepared.)
Anyway how about little toggle switches?
Maybe two or three of them? there could be lots of
combinations... but some series thing would look to
keep the capacitance down. I bought a bunch of little
plastic toggles... let's say 1 k ohm to 10 Meg.

(I ordered some little toggle switches,
I think I know which box they are in. :^)

one toggle switch is easy, what about if you have two
toggle switches, two or three position, (single pole)
I never know what to call three position toggles.
on-open-on?
The one thing nice about the rotary switch, is that at
DC, 10k ohm was a 10k resistor.

George H.



George H.
Click to expand...

(Oh dear, can you tell I've had too much beer?
please excuse me. I'm taking tomorrow off, and it's
beautiful around here. GH)
 
On Wednesday, September 18, 2019 at 3:35:19 PM UTC-4, George Herold wrote:
On Wednesday, September 18, 2019 at 3:04:09 PM UTC-4, John Larkin wrote:
On 17 Sep 2019 17:03:06 -0700, Winfield Hill <winfieldhill@yahoo.com
wrote:

George Herold wrote...

Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)

It's low, but worls horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?

A cascode needs a keep-alive current. Cascode really helps to get good
noise and bandwidth results for a high-capacitance photodiode.

Many of the signals I look at are small, 0-10%, changes on some background
light level.. so the cascode should be fine.

What I didn't appreciate, until today.. (even though it says as much
in Phil's book.) Is that with the cascode the speed of the TIA opamp is not
that important anymore. You can't speed it up with a faster opamp.
The time cosntant is mostly the PD capacitance times R_e (25mV/I_c)

I keep bashing my head into this 5pF of switch C. (Even with no cascode,
just a bootstrap.) I've ordered a few of these faster THS4631 opamps
http://www.ti.com/lit/ds/symlink/ths4631.pdf

Four times the BW should mean I need twice as big a compensating feedback cap.
I'll see.
Click to expand...

OK, f the rotary switch. (I've thought about some internal
jumpers or something... but if you ask students to open up your stuff..
well be prepared.)
Anyway how about little toggle switches?
Maybe two or three of them? there could be lots of
combinations... but some series thing would look to
keep the capacitance down. I bought a bunch of little
plastic toggles... let's say 1 k ohm to 10 Meg.

(I ordered some little toggle switches,
I think I know which box they are in. :^)

one toggle switch is easy, what about if you have two
toggle switches, two or three position, (single pole)
I never know what to call three position toggles.
on-open-on?
The one thing nice about the rotary switch, is that at
DC, 10k ohm was a 10k resistor.

George H.



George H.
 
On Wed, 18 Sep 2019 15:56:21 -0700 (PDT), tabbypurr@gmail.com wrote:

On Wednesday, 18 September 2019 20:35:19 UTC+1, George Herold wrote:

I keep bashing my head into this 5pF of switch C.

could always reduce it with a knife switch :)

What's the c of a coaxial relay?


NT
Click to expand...

We use a lot of little Fujitsu FTR-B3GA relays. Capacitance is tiny
and they work nicely to about 3 GHz.

George could use them to do the actual switching.

https://www.dropbox.com/s/se162xpw86hpmzs/DSC06884.JPG?raw=1
 
John Larkin wrote...
On 17 Sep 2019, Winfield Hill wrote:
George Herold wrote...

Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)

It's low, but works horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?

A cascode needs a keep-alive current. Cascode really helps
to get good noise and bandwidth results for a high-capacitance
photodiode.
Click to expand...

So does the bootstrap, without the bias-current penalty.
We're talking lots of degradation below from 1 to 100nA.


--
Thanks,
- Win
 
John Larkin wrote...
On Wed, 18 Sep 2019, tabbypurr@gmail.com wrote:
On Wednesday, 18 September 2019, George Herold wrote:

I keep bashing my head into this 5pF of switch C.

What's the c of a coaxial relay?

We use a lot of little Fujitsu FTR-B3GA relays.
Capacitance is tiny and they work nicely to about 3 GHz.

George could use them to do the actual switching.

https://www.dropbox.com/s/se162xpw86hpmzs/DSC06884.JPG?raw=1
Click to expand...

0.3pF. Use them with bootstrap. Forget cascode, get
a full current range. Have your cake and eat it too.


--
Thanks,
- Win
 
On 9/18/19 7:50 PM, John Larkin wrote:
On Wed, 18 Sep 2019 15:56:21 -0700 (PDT), tabbypurr@gmail.com wrote:

On Wednesday, 18 September 2019 20:35:19 UTC+1, George Herold wrote:

I keep bashing my head into this 5pF of switch C.

could always reduce it with a knife switch :)

What's the c of a coaxial relay?


NT

We use a lot of little Fujitsu FTR-B3GA relays. Capacitance is tiny
and they work nicely to about 3 GHz.

George could use them to do the actual switching.

https://www.dropbox.com/s/se162xpw86hpmzs/DSC06884.JPG?raw=1
Click to expand...

Typically about 0.2 pF contact-to-contact, and a bit more from contact
to coil.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com
 
On 9/18/19 9:00 PM, Winfield Hill wrote:
John Larkin wrote...

On 17 Sep 2019, Winfield Hill wrote:
George Herold wrote...

Oh or JL's idea of switched cascodes... what's the C_ce
capacitance when switched off? (collector to emitter)

It's low, but works horribly below 1uA. Bootstrap works
well over the whole range, why are you fixated on cascode?

A cascode needs a keep-alive current. Cascode really helps
to get good noise and bandwidth results for a high-capacitance
photodiode.

So does the bootstrap, without the bias-current penalty.
We're talking lots of degradation below from 1 to 100nA.
Click to expand...

Agreed. However, the bootstrap and cascode are sort of orthogonal, so
with equally good devices you wind up several dB ahead by using both.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com
 
On Friday, September 20, 2019 at 7:21:21 PM UTC-4, Phil Hobbs wrote:
On 9/18/19 7:50 PM, John Larkin wrote:
On Wed, 18 Sep 2019 15:56:21 -0700 (PDT), tabbypurr@gmail.com wrote:

On Wednesday, 18 September 2019 20:35:19 UTC+1, George Herold wrote:

I keep bashing my head into this 5pF of switch C.

could always reduce it with a knife switch :)

What's the c of a coaxial relay?


NT

We use a lot of little Fujitsu FTR-B3GA relays. Capacitance is tiny
and they work nicely to about 3 GHz.

George could use them to do the actual switching.

https://www.dropbox.com/s/se162xpw86hpmzs/DSC06884.JPG?raw=1



Typically about 0.2 pF contact-to-contact, and a bit more from contact
to coil.
Click to expand...
I was measuring toggle switch capacitance Friday.
Three position, single pole. measured with SRS RCL
thing. I added about 1" of bus wire to the switches
to plug 'em into the contacts... The srs contact interface
is meant for production. But I got it to read a 1pF cap correctly
and then two 1pFs in series...

Panel mount C&K toggles were a bit less than 0.5 pF
between pins and in the center 'no contact' position
about ~1pF to both. (7xxx series)

I have a little NKK plastic toggle it was about 0.35 pF pin to pin
and 0.8 both, but all these numbers are good to ~10% if I'm lucky. :^)

This,
https://www.digikey.com/product-detail/en/nkk-switches/B13AP/360-1013-ND/379100

George H.


Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com
Click to expand...
 
On Fri, 20 Sep 2019 19:21:13 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 9/18/19 7:50 PM, John Larkin wrote:
On Wed, 18 Sep 2019 15:56:21 -0700 (PDT), tabbypurr@gmail.com wrote:

On Wednesday, 18 September 2019 20:35:19 UTC+1, George Herold wrote:

I keep bashing my head into this 5pF of switch C.

could always reduce it with a knife switch :)

What's the c of a coaxial relay?


NT

We use a lot of little Fujitsu FTR-B3GA relays. Capacitance is tiny
and they work nicely to about 3 GHz.

George could use them to do the actual switching.

https://www.dropbox.com/s/se162xpw86hpmzs/DSC06884.JPG?raw=1



Typically about 0.2 pF contact-to-contact, and a bit more from contact
to coil.

Cheers

Phil Hobbs
Click to expand...

Being DPDT, tricks can be played in a TIA topology.

I guess one could booststrap unused contacts, or even the coil. But
I'm working at 400 Hz and hundreds of watts lately.
 
On Saturday, September 21, 2019 at 5:06:08 PM UTC-4, jla...@highlandsniptechnology.com wrote:
On Fri, 20 Sep 2019 19:21:13 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 9/18/19 7:50 PM, John Larkin wrote:
On Wed, 18 Sep 2019 15:56:21 -0700 (PDT), tabbypurr@gmail.com wrote:

On Wednesday, 18 September 2019 20:35:19 UTC+1, George Herold wrote:

I keep bashing my head into this 5pF of switch C.

could always reduce it with a knife switch :)

What's the c of a coaxial relay?


NT

We use a lot of little Fujitsu FTR-B3GA relays. Capacitance is tiny
and they work nicely to about 3 GHz.

George could use them to do the actual switching.

https://www.dropbox.com/s/se162xpw86hpmzs/DSC06884.JPG?raw=1



Typically about 0.2 pF contact-to-contact, and a bit more from contact
to coil.

Cheers

Phil Hobbs

Being DPDT, tricks can be played in a TIA topology.
Click to expand...
Huh... OK let me think about it.
I've got a layout with three toggles which I think...
discounting stray stuff, gives me just one switch C.
(~0.5 pF) for my one 'sweet spot' TIA resistor.
(100 k ohm.) I'll post it. three toggles and
six resistor settings.. brute force.
George H.
I guess one could booststrap unused contacts, or even the coil. But
I'm working at 400 Hz and hundreds of watts lately.
Click to expand...
 
On 9/21/19 5:05 PM, jlarkin@highlandsniptechnology.com wrote:
On Fri, 20 Sep 2019 19:21:13 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 9/18/19 7:50 PM, John Larkin wrote:
On Wed, 18 Sep 2019 15:56:21 -0700 (PDT), tabbypurr@gmail.com wrote:

On Wednesday, 18 September 2019 20:35:19 UTC+1, George Herold wrote:

I keep bashing my head into this 5pF of switch C.

could always reduce it with a knife switch :)

What's the c of a coaxial relay?


NT

We use a lot of little Fujitsu FTR-B3GA relays. Capacitance is tiny
and they work nicely to about 3 GHz.

George could use them to do the actual switching.

https://www.dropbox.com/s/se162xpw86hpmzs/DSC06884.JPG?raw=1



Typically about 0.2 pF contact-to-contact, and a bit more from contact
to coil.

Cheers

Phil Hobbs

Being DPDT, tricks can be played in a TIA topology.

I guess one could booststrap unused contacts, or even the coil. But
I'm working at 400 Hz and hundreds of watts lately.
Click to expand...

Yup, I did both in a switchable 1G/50G TIA for a scanning surface
voltage tool for a now-defunct company called Qcept. Plus I had to
short out the 1G resistor when on the 50G range, because otherwise its
Johnson noise across the 0.2 pF of the open contacts would have
dominated the noise floor. :)

That's one of those times that not doing a careful error budget will
bite you in the posterior.

Since the company went mammaries-topmost, here's part of the schematic:
<https://electrooptical.net/www/sed/QceptTransimpedanceAmpRev1.pdf>


Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC / Hobbs ElectroOptics
Optics, Electro-optics, Photonics, Analog Electronics
Briarcliff Manor NY 10510

http://electrooptical.net
http://hobbs-eo.com
 
You must log in or register to reply here.

Welcome to EDABoard.com

Sponsor

Back
Top