differential signal detector...

[server]

I probably won\'t use this one, but it is pleasingly weird.

Watch for line wrap on the BAT15 model.

Version 4
SHEET 1 1724 680
WIRE -416 -352 -480 -352
WIRE -368 -352 -416 -352
WIRE -208 -352 -288 -352
WIRE -128 -352 -208 -352
WIRE 32 -352 -128 -352
WIRE 192 -352 112 -352
WIRE -208 -304 -208 -352
WIRE 192 -304 192 -352
WIRE -480 -224 -480 -352
WIRE -576 -208 -608 -208
WIRE -528 -208 -576 -208
WIRE -208 -192 -208 -240
WIRE -160 -192 -208 -192
WIRE -32 -192 -96 -192
WIRE 80 -192 -32 -192
WIRE 240 -192 160 -192
WIRE 320 -192 240 -192
WIRE 384 -192 320 -192
WIRE 240 -176 240 -192
WIRE 384 -176 384 -192
WIRE -528 -160 -560 -160
WIRE -32 -160 -32 -192
WIRE -560 -96 -560 -160
WIRE -480 -96 -480 -144
WIRE -480 -96 -560 -96
WIRE -208 -80 -208 -192
WIRE -160 -80 -208 -80
WIRE -32 -80 -32 -96
WIRE -32 -80 -96 -80
WIRE 80 -80 -32 -80
WIRE 240 -80 240 -112
WIRE 240 -80 160 -80
WIRE 384 -80 384 -96
WIRE 384 -80 240 -80
WIRE -480 -64 -480 -96
WIRE -32 -16 -32 -80
WIRE 384 0 384 -80
WIRE -928 96 -992 96
WIRE -896 96 -928 96
WIRE -656 96 -752 96
WIRE -608 96 -656 96
WIRE -448 96 -480 96
WIRE -368 96 -448 96
WIRE -128 96 -288 96
WIRE -32 96 -32 48
WIRE -32 96 -128 96
WIRE 32 96 -32 96
WIRE 192 96 112 96
WIRE -992 112 -992 96
WIRE -752 112 -752 96
WIRE -480 128 -480 96
WIRE 192 144 192 96
WIRE -992 240 -992 192
WIRE -752 240 -752 192
WIRE -480 240 -480 208
FLAG -752 240 0
FLAG -480 240 0
FLAG -656 96 sig
FLAG -992 240 0
FLAG -928 96 gate
FLAG -448 96 GEN+
FLAG -480 -64 0
FLAG -576 -208 GEN+
FLAG 192 -304 0
FLAG 192 144 0
FLAG 384 0 0
FLAG -416 -352 GEN-
FLAG -128 96 SFP+
FLAG -128 -352 SFP-
FLAG 320 -192 DET
SYMBOL voltage -752 96 R0
WINDOW 0 31 93 Left 2
WINDOW 3 12 126 Left 2
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName Vclk
SYMATTR Value SINE(0 1 {F} 10u)
SYMBOL bv -480 112 R0
WINDOW 0 17 108 Left 2
WINDOW 3 38 157 Left 2
SYMATTR InstName B1
SYMATTR Value V=V(gate) * 0.25*(tanh(1e6*(V(sig))))
SYMBOL voltage -992 96 R0
WINDOW 0 28 124 Left 2
WINDOW 3 -33 180 Left 2
WINDOW 123 0 0 Left 0
WINDOW 39 0 0 Left 0
SYMATTR InstName Vgate
SYMATTR Value PULSE(0 1 10u 0 0 100u)
SYMBOL res -272 80 R90
WINDOW 0 67 56 VBottom 2
WINDOW 3 82 55 VTop 2
SYMATTR InstName R1
SYMATTR Value 50
SYMBOL res 128 80 R90
WINDOW 0 68 59 VBottom 2
WINDOW 3 72 58 VTop 2
SYMATTR InstName R2
SYMATTR Value 50
SYMBOL e -480 -240 R0
SYMATTR InstName E1
SYMATTR Value -1
SYMBOL res -272 -368 R90
WINDOW 0 70 61 VBottom 2
WINDOW 3 79 61 VTop 2
SYMATTR InstName R3
SYMATTR Value 50
SYMBOL res 128 -368 R90
WINDOW 0 71 83 VBottom 2
WINDOW 3 79 84 VTop 2
SYMATTR InstName R4
SYMATTR Value 50
SYMBOL cap -224 -304 R0
WINDOW 0 -46 24 Left 2
WINDOW 3 -43 51 Left 2
SYMATTR InstName C3
SYMATTR Value 5p
SYMBOL cap -48 -16 R0
WINDOW 0 -46 24 Left 2
WINDOW 3 -43 51 Left 2
SYMATTR InstName C4
SYMATTR Value 5p
SYMBOL diode -160 -176 R270
WINDOW 0 32 32 VTop 2
WINDOW 3 0 32 VBottom 2
SYMATTR InstName D3
SYMATTR Value BAT15
SYMBOL res 176 -96 R90
WINDOW 0 72 59 VBottom 2
WINDOW 3 76 59 VTop 2
SYMATTR InstName R5
SYMATTR Value 10K
SYMBOL res 368 -192 R0
WINDOW 0 50 39 Left 2
WINDOW 3 48 66 Left 2
SYMATTR InstName R7
SYMATTR Value 1e6
SYMBOL cap -48 -160 R0
WINDOW 0 60 8 Left 2
WINDOW 3 54 40 Left 2
SYMATTR InstName C5
SYMATTR Value 100p
SYMBOL diode -96 -64 M270
WINDOW 0 32 32 VTop 2
WINDOW 3 0 32 VBottom 2
SYMATTR InstName D1
SYMATTR Value BAT15
SYMBOL cap 224 -176 R0
WINDOW 0 55 29 Left 2
WINDOW 3 46 60 Left 2
SYMATTR InstName C1
SYMATTR Value 100p
SYMBOL res 176 -208 R90
WINDOW 0 -13 61 VBottom 2
WINDOW 3 47 59 VTop 2
SYMATTR InstName R6
SYMATTR Value 10K
TEXT -1016 -16 Left 2 !.MODEL BAT15 D(IS=130n RS=4.5 N=1.08 XTI=1.8
EG=.68\\n+ CJO=260f M=.047 VJ=.11 FC=.5 BV=4 IBV=10U TT=25p)
TEXT -1016 -144 Left 2 !.tran 200u
TEXT -976 -312 Left 2 ;K420 Differential Signal Detector
TEXT -888 -216 Left 2 ;JL Oct 17 2020
TEXT -1016 -112 Left 2 !.param F = 5e6
TEXT -1016 -80 Left 2 !.step param F list 5e6 500e6 2e9
TEXT -952 -264 Left 2 ;ECL or CML into SFP Module


--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Phil Hobbs]

On 2020-10-17 15:59, jlarkin@highlandsniptechnology.com wrote:

I probably won\'t use this one, but it is pleasingly weird.

Watch for line wrap on the BAT15 model.

snip

Fun--a floating one-stage Cockroft-Walton.

Who says you need a transformer?

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

 

[server]

On Sat, 17 Oct 2020 17:27:23 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-17 15:59, jlarkin@highlandsniptechnology.com wrote:

I probably won\'t use this one, but it is pleasingly weird.

Watch for line wrap on the BAT15 model.

snip

Fun--a floating one-stage Cockroft-Walton.

Who says you need a transformer?

Cheers

Phil Hobbs

I was considering using a Mini-Circuits wideband transformer to pick
off the diff signal, but this circuit occurred to me and is suitably
strange.

But I think I\'ll use a 1K resistive pickoff from just the high side,
into a 5 GHz MMIC, then a fairly normal diode doubler detector. We
might allow the user to apply a wide range of inputs, including
single-ended.

An SAV551 source follower would be fun as the pickoff too, but might
get risky. Have you tried that?

I\'ve been playing with 10G SFP modules. They seem to accept
single-ended or diff inputs, from about 50 mV p-p to full 5 volt CMOS,
and just work. They are amazing for $20.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Phil Hobbs]

On 2020-10-17 18:08, jlarkin@highlandsniptechnology.com wrote:

On Sat, 17 Oct 2020 17:27:23 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-17 15:59, jlarkin@highlandsniptechnology.com wrote:

I probably won\'t use this one, but it is pleasingly weird.

Watch for line wrap on the BAT15 model.

snip

Fun--a floating one-stage Cockroft-Walton.

Who says you need a transformer?

I was considering using a Mini-Circuits wideband transformer to pick
off the diff signal, but this circuit occurred to me and is suitably
strange.

But I think I\'ll use a 1K resistive pickoff from just the high side,
into a 5 GHz MMIC, then a fairly normal diode doubler detector. We
might allow the user to apply a wide range of inputs, including
single-ended.
An SAV551 source follower would be fun as the pickoff too, but might
get risky. Have you tried that?

SAV551s (like other 3-GHz class pHEMTS) are amazingly stable. I\'ve run
one (without beads) bootstrapping a 100-pF Hamamatsu MPPC at the wrong
end of a 50-mm flex. The bootstrap bandwidth wasn\'t nearly as wide as
if it had been on the right end, but it worked very well.

I\'ve used SKY65050s as followers--they work about as well as JFETS, only
10 times faster. (The Avago parts were the pits for this because their
drain impedance was so low--I measured an ATF38143 as having a voltage
gain of less than 0.7 as a follower.)

I\'ve been playing with 10G SFP modules. They seem to accept
single-ended or diff inputs, from about 50 mV p-p to full 5 volt CMOS,
and just work. They are amazing for $20.

Yup. Funny how that happens when there\'s a huge development effort with
a multi-year roadmap.

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

 

[server]

On Sat, 17 Oct 2020 19:52:27 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-17 18:08, jlarkin@highlandsniptechnology.com wrote:
On Sat, 17 Oct 2020 17:27:23 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-17 15:59, jlarkin@highlandsniptechnology.com wrote:

I probably won\'t use this one, but it is pleasingly weird.

Watch for line wrap on the BAT15 model.

snip

Fun--a floating one-stage Cockroft-Walton.

Who says you need a transformer?

I was considering using a Mini-Circuits wideband transformer to pick
off the diff signal, but this circuit occurred to me and is suitably
strange.

But I think I\'ll use a 1K resistive pickoff from just the high side,
into a 5 GHz MMIC, then a fairly normal diode doubler detector. We
might allow the user to apply a wide range of inputs, including
single-ended.
An SAV551 source follower would be fun as the pickoff too, but might
get risky. Have you tried that?

SAV551s (like other 3-GHz class pHEMTS) are amazingly stable. I\'ve run
one (without beads) bootstrapping a 100-pF Hamamatsu MPPC at the wrong
end of a 50-mm flex. The bootstrap bandwidth wasn\'t nearly as wide as
if it had been on the right end, but it worked very well.

I\'ve used SKY65050s as followers--they work about as well as JFETS, only
10 times faster. (The Avago parts were the pits for this because their
drain impedance was so low--I measured an ATF38143 as having a voltage
gain of less than 0.7 as a follower.)


I\'ve been playing with 10G SFP modules. They seem to accept
single-ended or diff inputs, from about 50 mV p-p to full 5 volt CMOS,
and just work. They are amazing for $20.

Yup. Funny how that happens when there\'s a huge development effort with
a multi-year roadmap.

Cheers

Phil Hobbs

Did you play with my SFP board? I have more.

The 10G Cisco that I like is hard to get, but we\'ve found some cheap
equivalents that look good. Cisco\'s list price is $1200 or something
silly now.

We have a new 1 GHz analog o/e converter, something I did for fun
during lockdown. Jonathan will test it with an SFP as the light
source. Directly pulsing vcsels gets weird.

My o/e was crazy slow, and he discovered that the pin photodiode has
20 pF from the cathode lead to the TO18 can. The spec is 1p max.




--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Phil Hobbs]

On 2020-10-17 20:06, jlarkin@highlandsniptechnology.com wrote:

On Sat, 17 Oct 2020 19:52:27 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-17 18:08, jlarkin@highlandsniptechnology.com wrote:
On Sat, 17 Oct 2020 17:27:23 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-17 15:59, jlarkin@highlandsniptechnology.com wrote:

I probably won\'t use this one, but it is pleasingly weird.

Watch for line wrap on the BAT15 model.

snip

Fun--a floating one-stage Cockroft-Walton.

Who says you need a transformer?

I was considering using a Mini-Circuits wideband transformer to pick
off the diff signal, but this circuit occurred to me and is suitably
strange.

But I think I\'ll use a 1K resistive pickoff from just the high side,
into a 5 GHz MMIC, then a fairly normal diode doubler detector. We
might allow the user to apply a wide range of inputs, including
single-ended.
An SAV551 source follower would be fun as the pickoff too, but might
get risky. Have you tried that?

SAV551s (like other 3-GHz class pHEMTS) are amazingly stable. I\'ve run
one (without beads) bootstrapping a 100-pF Hamamatsu MPPC at the wrong
end of a 50-mm flex. The bootstrap bandwidth wasn\'t nearly as wide as
if it had been on the right end, but it worked very well.

I\'ve used SKY65050s as followers--they work about as well as JFETS, only
10 times faster. (The Avago parts were the pits for this because their
drain impedance was so low--I measured an ATF38143 as having a voltage
gain of less than 0.7 as a follower.)


I\'ve been playing with 10G SFP modules. They seem to accept
single-ended or diff inputs, from about 50 mV p-p to full 5 volt CMOS,
and just work. They are amazing for $20.

Yup. Funny how that happens when there\'s a huge development effort with
a multi-year roadmap.

Cheers

Phil Hobbs

Did you play with my SFP board? I have more.

Not yet. Last week I built a test setup for my new laser noise
canceller idea.

Since there are no really good fast matched BJT chips(*), trying to
extend the original MAT04 versions
<https://electrooptical.net/www/canceller/WithoutTears.pdf> to much
higher bandwidth is a problem.

There are two basic approaches: use the DC precision of the slowish
MAT04/MAT14 and put a bandaid on to reach higher frequency, or take some
amazingly good disctete transistors, e.g. the BFP640H (45 GHz, VAF > 1
kV beta ~700) and make them act as though they were monolithic. (Mainly
that means getting their die temperatures to track accurately from DC
out to highish speeds, like 50 kHz or so.)

The first approach might put fast cascode transistors on the MAT14 diff
pair to get rid of their output capacitance, and use Darlington base
drive to recycle the base current to the collector circuit, so that (in
the required bandwidth) nothing was lost to base current.

Turns out, of course, that the collector-base time constant puts a
fairly sharp limit on how far you can go with the bandaid approach--the
extrinsic base resistance and BC capacitance prevent the recycling at
high frequencies.

So this gizmo uses a BFP640 diff pair, exploiting the very high VAF to
allow equalization of the power dissipation on the two sides by dorking
the V_CE of one side to match the ratio of I_C. We\'ll see how it works
this week.

Cheers

Phil Hobbs

(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

--
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

 

[Gerhard Hoffmann]

Am 18.10.20 um 03:47 schrieb Phil Hobbs:

(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these? Renesas?

< https://www.digikey.de/products/de?keywords=HFA3134 >
< https://www.digikey.de/products/de?keywords=HFA3135 >
< https://www.digikey.de/products/de?keywords=hfa3096 >

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

cheers, Gerhard

 

[Phil Hobbs]

On 2020-10-18 04:09, Gerhard Hoffmann wrote:

Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.

    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

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

 

[server]

On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

Cheers

Phil Hobbs

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.

I wonder if using a balun could help in situations like this.





--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Phil Hobbs]

On 2020-10-18 09:27, jlarkin@highlandsniptechnology.com wrote:

On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.
I wonder if using a balun could help in situations like this.

Or plastic-packaged PDs. We use a 300 um Hamamatsu one in a TMB package
that costs $5.

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

 

[server]

On Sun, 18 Oct 2020 12:39:36 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 09:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.
I wonder if using a balun could help in situations like this.

Or plastic-packaged PDs. We use a 300 um Hamamatsu one in a TMB package
that costs $5.

Cheers

Phil Hobbs

We buy this one in a metal sugar-cube package, ST or FC fiber
connector, all aligned and stuff. 20 pF from cathode to TO18 can, then
xx pF from the can to the metal housing. Yuk.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Phil Hobbs]

On 2020-10-18 14:26, jlarkin@highlandsniptechnology.com wrote:

On Sun, 18 Oct 2020 12:39:36 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 09:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.
I wonder if using a balun could help in situations like this.

Or plastic-packaged PDs. We use a 300 um Hamamatsu one in a TMB package
that costs $5.

We buy this one in a metal sugar-cube package, ST or FC fiber
connector, all aligned and stuff. 20 pF from cathode to TO18 can, then
xx pF from the can to the metal housing. Yuk.

One gathers that they expect you to use a positive bias supply. That\'s
a pain.

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

 

[server]

On Sun, 18 Oct 2020 20:01:56 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 14:26, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 12:39:36 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 09:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.
I wonder if using a balun could help in situations like this.

Or plastic-packaged PDs. We use a 300 um Hamamatsu one in a TMB package
that costs $5.

We buy this one in a metal sugar-cube package, ST or FC fiber
connector, all aligned and stuff. 20 pF from cathode to TO18 can, then
xx pF from the can to the metal housing. Yuk.

One gathers that they expect you to use a positive bias supply. That\'s
a pain.

Cheers

Phil Hobbs

Given that my circuit takes signals out of both ends, I just used the
wrong ends.



--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard

 

[Phil Hobbs]

On 2020-10-18 20:27, jlarkin@highlandsniptechnology.com wrote:

On Sun, 18 Oct 2020 20:01:56 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 14:26, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 12:39:36 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 09:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.
I wonder if using a balun could help in situations like this.

Or plastic-packaged PDs. We use a 300 um Hamamatsu one in a TMB package
that costs $5.

We buy this one in a metal sugar-cube package, ST or FC fiber
connector, all aligned and stuff. 20 pF from cathode to TO18 can, then
xx pF from the can to the metal housing. Yuk.

One gathers that they expect you to use a positive bias supply. That\'s
a pain.


Given that my circuit takes signals out of both ends, I just used the
wrong ends.

Using a negative bias supply makes TIA design easier, because you can
easily make zero volts = zero photocurrent, simply by grounding the +
input, which means coming out the cathode. For single-ended photodiode
circuits, you normally don\'t care how much capacitance they want to hang
on the anode, because it\'s connected to the bias supply anyhow.

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

 

[John Larkin]

On Mon, 19 Oct 2020 13:13:05 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 20:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 20:01:56 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 14:26, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 12:39:36 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 09:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.
I wonder if using a balun could help in situations like this.

Or plastic-packaged PDs. We use a 300 um Hamamatsu one in a TMB package
that costs $5.

We buy this one in a metal sugar-cube package, ST or FC fiber
connector, all aligned and stuff. 20 pF from cathode to TO18 can, then
xx pF from the can to the metal housing. Yuk.

One gathers that they expect you to use a positive bias supply. That\'s
a pain.


Given that my circuit takes signals out of both ends, I just used the
wrong ends.

Using a negative bias supply makes TIA design easier, because you can
easily make zero volts = zero photocurrent, simply by grounding the +
input, which means coming out the cathode. For single-ended photodiode
circuits, you normally don\'t care how much capacitance they want to hang
on the anode, because it\'s connected to the bias supply anyhow.

Cheers

Phil Hobbs

I\'m hanging the PD between a -5 supply and the input of a MMIC which
biases itself up about +2 volts, 7 volts net. Seemed all nice until
the capacitance glitch. We can probably find another supplier for the
850 nm connectorized photodiode. Lots of people make them.

More voltage makes the photodiodes faster. Not all capacitive... seems
like more voltage sweeps charges out faster. Does that keep happening,
or does it level off like capacitance?

 

[Phil Hobbs]

On 2020-10-19 13:29, John Larkin wrote:

On Mon, 19 Oct 2020 13:13:05 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 20:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 20:01:56 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 14:26, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 12:39:36 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 09:27, jlarkin@highlandsniptechnology.com wrote:
On Sun, 18 Oct 2020 06:28:02 -0400, Phil Hobbs
pcdhSpamMeSenseless@electrooptical.net> wrote:

On 2020-10-18 04:09, Gerhard Hoffmann wrote:
Am 18.10.20 um 03:47 schrieb Phil Hobbs:


(*) The Infineon HFA series has very disappointing R_ee\' and R_bb\'.

Do you mean these?   Renesas?

Right. I meant Intersil, not Infineon.


    https://www.digikey.de/products/de?keywords=HFA3134   
    https://www.digikey.de/products/de?keywords=HFA3135   
    https://www.digikey.de/products/de?keywords=hfa3096   

I\'ve used the HFA3096 in a pretty good 1:5000 time stretcher.
It was available in a hermetic flat pack for space apps.
Might have been still Intersil.
The 3096 family seems to be bond out options of the same
SOI chip. The last fast-ish PNPs alive?

Yeah. Unfortunately they have ridiculously high R_bb\' and R_ee\', which
makes them unsuitable for noise cancellers. By a combination of
bootstrapping and cascoding, with various tricks for bias current
cancellation, I managed to get the whole thing going with no PNPs at
all. (Noise cancellers need to use both ends of the photodiodes, which
makes it trickier.) Those SiGe:C things really are amazing, all for 20
cents in reels.

In the spherical cow universe, that gets me >60 dB cancellation out to
30 MHz or thereabouts, over a pretty wide range of photocurrents. We\'ll
find out what the real ruminants can do in the next few days.

My new o/e converter uses both ends of the photodiode, one for a fast
AC path and the other end for a slow DC path. I merge them back
together later. That\'s how I got into trouble with that crazy pd, the
one with 20 pF from the cathode to the can. Only the 850 pd does
that... it works great with the longer wavelength parts.
I wonder if using a balun could help in situations like this.

Or plastic-packaged PDs. We use a 300 um Hamamatsu one in a TMB package
that costs $5.

We buy this one in a metal sugar-cube package, ST or FC fiber
connector, all aligned and stuff. 20 pF from cathode to TO18 can, then
xx pF from the can to the metal housing. Yuk.

One gathers that they expect you to use a positive bias supply. That\'s
a pain.


Given that my circuit takes signals out of both ends, I just used the
wrong ends.

Using a negative bias supply makes TIA design easier, because you can
easily make zero volts = zero photocurrent, simply by grounding the +
input, which means coming out the cathode. For single-ended photodiode
circuits, you normally don\'t care how much capacitance they want to hang
on the anode, because it\'s connected to the bias supply anyhow.

Cheers

Phil Hobbs

I\'m hanging the PD between a -5 supply and the input of a MMIC which
biases itself up about +2 volts, 7 volts net. Seemed all nice until
the capacitance glitch. We can probably find another supplier for the
850 nm connectorized photodiode. Lots of people make them.

More voltage makes the photodiodes faster. Not all capacitive... seems
like more voltage sweeps charges out faster. Does that keep happening,
or does it level off like capacitance?

Depends on the device. If the high-doped regions are thin enough, you
can deplete them before the device breaks down. That sometimes leads to
quite startling speedups.

The physics of that is similar to forward recovery in PN diodes--due to
Debye shielding, the carriers nearest the junction don\'t get the message
until it\'s had a chance to get there by diffusion.

Schottky photodiodes are very fast too, but the interdigitated metal
fingers block a lot of the light, and you have to bias them above
reachthrough because there are two barriers back-to-back.


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