Forums

little photo diodes and an opamp?

Started by George Herold March 18, 2020
The project seems to call for the smallest detectors.
I'm buying a few each of 'all'* the little PD's from digikey.
(I can share a list if someone wants... it's a mix of 
TH and SMD.)  
C_in between 1-3 pf when reversed biased.  
I don't see how it makes sense to bootstrap or cascode 
the PD**.  So I was looking at opamps with small C-in.  
(Big thanks to H&H of AoE as always.) 
I've now got a column of 1/f corner freq. of e_in 
(input voltage noise) for the small C_in fet opamps.
In table 4x.2 of the x-chapters.  
The best looks to be the ADA4817. 1/f ~10 kHz.
(hard to tell.. bit of a swoopy noise graph.)
(C_in = 1.5 pF, e_in = 4 nV/rthz.)  

Pure 1/f noise into a TIA must give a constant noise 
voltage.. till hit by the GBW of the opamp.  
(is that right? I'm being mathematically lazy and 
just tipping the slope of the graph.)

George H. 
  
 

*not all, but a selection.
**I'm not sure how to guesstimate the C of a cascode, 
and are there any small bjt's left?  
(The bft25a is listed as 'not for new designs' on DK)
*** 100kHz is common and higher.  
On Wed, 18 Mar 2020 17:14:32 -0700 (PDT), George Herold
<ggherold@gmail.com> wrote:

>The project seems to call for the smallest detectors. >I'm buying a few each of 'all'* the little PD's from digikey. >(I can share a list if someone wants... it's a mix of >TH and SMD.) >C_in between 1-3 pf when reversed biased. >I don't see how it makes sense to bootstrap or cascode >the PD**. So I was looking at opamps with small C-in. >(Big thanks to H&H of AoE as always.) >I've now got a column of 1/f corner freq. of e_in >(input voltage noise) for the small C_in fet opamps. >In table 4x.2 of the x-chapters. >The best looks to be the ADA4817. 1/f ~10 kHz. >(hard to tell.. bit of a swoopy noise graph.) >(C_in = 1.5 pF, e_in = 4 nV/rthz.) > >Pure 1/f noise into a TIA must give a constant noise >voltage.. till hit by the GBW of the opamp. >(is that right? I'm being mathematically lazy and >just tipping the slope of the graph.) > >George H. > > > >*not all, but a selection. >**I'm not sure how to guesstimate the C of a cascode, >and are there any small bjt's left? >(The bft25a is listed as 'not for new designs' on DK) >*** 100kHz is common and higher.
There are still lots of fast (5 GHz and up to 100G) NPNs around, and lots of BFT25 equivalents. Try a Digikey search. I'm currently designing an oscillator with two BFT25s! Spice is showing 1.2 volts Vbe at a reasonable collector current. It's a small chip, but that's ridiculous. When Da City allows me to legally go back to work, I might test one. Fast PNPs are pretty much history. We use a fiber-coupled, very fast photodiode that's 0.7 pF. -- John Larkin Highland Technology, Inc The cork popped merrily, and Lord Peter rose to his feet. "Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"
On 19.03.20 12:14 am, George Herold wrote:
> The best looks to be the ADA4817.
Also check out the new OPA818 (haven't used it yet). > I've now got a column of 1/f corner freq. How much bandwidth and gain do you need? The 1/f noise contribution is often drowned out by the feedback impedance Johnson noise. &mdash; David
On 2020-03-18 20:14, George Herold wrote:
> The project seems to call for the smallest detectors. I'm buying a > few each of 'all'* the little PD's from digikey. (I can share a list > if someone wants... it's a mix of TH and SMD.) C_in between 1-3 pf > when reversed biased. I don't see how it makes sense to bootstrap or > cascode the PD**.
You can get about 15 dB improvement at high frequency by using a SAV-551+ bootstrap, if you keep the stray capacitance low enough. So I was looking at opamps with small C-in.
> (Big thanks to H&H of AoE as always.) I've now got a column of 1/f > corner freq. of e_in (input voltage noise) for the small C_in fet > opamps. In table 4x.2 of the x-chapters. The best looks to be the > ADA4817. 1/f ~10 kHz. (hard to tell.. bit of a swoopy noise graph.) > (C_in = 1.5 pF, e_in = 4 nV/rthz.) > > Pure 1/f noise into a TIA must give a constant noise voltage.. till > hit by the GBW of the opamp. (is that right? I'm being mathematically > lazy and just tipping the slope of the graph.)
No. The eN*C noise power density goes as f**2, whereas 1/f noise power goes as (drum roll) 1/f. That's why pHEMTs make very good bootstraps despite having 1/f corners of 10 MHz or so--the f**2 from the admittance of the capacitor knocks out the 1/f from flicker noise.
> **I'm not sure how to guesstimate the C of a cascode, and are there > any small bjt's left?
For a common-base stage it's C_CE, but you generally don't care because it appears across r_E. For a FET bootstrap it's basically C_DG + (1-Av) C_GS. For a pHEMT that can be way below 1 pF, so you're limited by the strays. Cheers Phil Hobbs
> > George H. > > > > *not all, but a selection. > > (The bft25a is listed as 'not for new designs' on DK) *** 100kHz is > common and higher. >
-- 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
jlarkin@highlandsniptechnology.com wrote...
> > When Da City allows me to legally go back > to work, I might test one.
Your entire company is shutdown? -- Thanks, - Win
On 2020-03-18 21:04, jlarkin@highlandsniptechnology.com wrote:
> On Wed, 18 Mar 2020 17:14:32 -0700 (PDT), George Herold > <ggherold@gmail.com> wrote: > >> The project seems to call for the smallest detectors. >> I'm buying a few each of 'all'* the little PD's from digikey. >> (I can share a list if someone wants... it's a mix of >> TH and SMD.) >> C_in between 1-3 pf when reversed biased. >> I don't see how it makes sense to bootstrap or cascode >> the PD**. So I was looking at opamps with small C-in. >> (Big thanks to H&H of AoE as always.) >> I've now got a column of 1/f corner freq. of e_in >> (input voltage noise) for the small C_in fet opamps. >> In table 4x.2 of the x-chapters. >> The best looks to be the ADA4817. 1/f ~10 kHz. >> (hard to tell.. bit of a swoopy noise graph.) >> (C_in = 1.5 pF, e_in = 4 nV/rthz.) >> >> Pure 1/f noise into a TIA must give a constant noise >> voltage.. till hit by the GBW of the opamp. >> (is that right? I'm being mathematically lazy and >> just tipping the slope of the graph.) >> >> George H. >> >> >> >> *not all, but a selection. >> **I'm not sure how to guesstimate the C of a cascode, >> and are there any small bjt's left? >> (The bft25a is listed as 'not for new designs' on DK) >> *** 100kHz is common and higher. > > There are still lots of fast (5 GHz and up to 100G) NPNs around, and > lots of BFT25 equivalents. Try a Digikey search.
Have you found any with the super low C-B leakage? 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 2020-03-19 13:03, Phil Hobbs wrote:
> On 2020-03-18 20:14, George Herold wrote: >> The project seems to call for the smallest detectors. I'm buying a >> few each of 'all'* the little PD's from digikey. (I can share a list >> if someone wants... it's a mix of TH and SMD.) C_in between 1-3 pf >> when reversed biased. I don't see how it makes sense to bootstrap or >> cascode the PD**. > > You can get about 15 dB improvement at high frequency by using a > SAV-551+ bootstrap, if you keep the stray capacitance low enough. > > So I was looking at opamps with small C-in. >> (Big thanks to H&H of AoE as always.) I've now got a column of 1/f >> corner freq. of e_in (input voltage noise) for the small C_in fet >> opamps. In table 4x.2 of the x-chapters. The best looks to be the >> ADA4817. 1/f ~10 kHz. (hard to tell.. bit of a swoopy noise graph.) >> (C_in = 1.5 pF, e_in = 4 nV/rthz.) >> >> Pure 1/f noise into a TIA must give a constant noise voltage.. till >> hit by the GBW of the opamp. (is that right? I'm being mathematically >> lazy and just tipping the slope of the graph.) > > No.&nbsp; The eN*C noise power density goes as f**2, whereas 1/f noise power > goes as&nbsp; (drum roll) 1/f. > > That's why pHEMTs make very good bootstraps despite having 1/f corners > of 10 MHz or so--the f**2 from the admittance of the capacitor knocks > out the 1/f from flicker noise. > >> **I'm not sure how to guesstimate the C of a cascode, and are there >> any small bjt's left? > > For a common-base stage it's
C_BE, but you generally don't care because
> it appears across r_E. > > For a FET bootstrap it's basically C_DG + (1-Av) C_GS.&nbsp; For a pHEMT that > can be way below 1 pF, so you're limited by the strays. > > 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 19 Mar 2020 10:55:33 -0700, Winfield Hill <winfieldhill@yahoo.com>
wrote:

>jlarkin@highlandsniptechnology.com wrote... >> >> When Da City allows me to legally go back >> to work, I might test one. > > Your entire company is shutdown?
Yes. I go in part-time to check on things, which is technically legal for fundamental company functions and to enable other people to work from home. Most engineers are working at home, but of course production people aren't. It's absurd. We are currently still paying everyone. Imagine Apple, Google, Intel, Salesforce, Twitter, all of SF and Silicon Valley and Napa Valley and Tahoe ski areas shut down. Glen Canyon is full of unemployed people walking kids and dogs. The crush at Safeway is over: how much rice and toilet paper will people buy until they run out of space? Still can't find flour for some reason, so the St Patrick's Day Guinness Cake is on hold. Musk doesn't want to shut down the plant in Fremont. I'm sure Tesla fans approve. -- John Larkin Highland Technology, Inc The cork popped merrily, and Lord Peter rose to his feet. "Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"
On Thu, 19 Mar 2020 13:57:27 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

>On 2020-03-18 21:04, jlarkin@highlandsniptechnology.com wrote: >> On Wed, 18 Mar 2020 17:14:32 -0700 (PDT), George Herold >> <ggherold@gmail.com> wrote: >> >>> The project seems to call for the smallest detectors. >>> I'm buying a few each of 'all'* the little PD's from digikey. >>> (I can share a list if someone wants... it's a mix of >>> TH and SMD.) >>> C_in between 1-3 pf when reversed biased. >>> I don't see how it makes sense to bootstrap or cascode >>> the PD**. So I was looking at opamps with small C-in. >>> (Big thanks to H&H of AoE as always.) >>> I've now got a column of 1/f corner freq. of e_in >>> (input voltage noise) for the small C_in fet opamps. >>> In table 4x.2 of the x-chapters. >>> The best looks to be the ADA4817. 1/f ~10 kHz. >>> (hard to tell.. bit of a swoopy noise graph.) >>> (C_in = 1.5 pF, e_in = 4 nV/rthz.) >>> >>> Pure 1/f noise into a TIA must give a constant noise >>> voltage.. till hit by the GBW of the opamp. >>> (is that right? I'm being mathematically lazy and >>> just tipping the slope of the graph.) >>> >>> George H. >>> >>> >>> >>> *not all, but a selection. >>> **I'm not sure how to guesstimate the C of a cascode, >>> and are there any small bjt's left? >>> (The bft25a is listed as 'not for new designs' on DK) >>> *** 100kHz is common and higher. >> >> There are still lots of fast (5 GHz and up to 100G) NPNs around, and >> lots of BFT25 equivalents. Try a Digikey search. > >Have you found any with the super low C-B leakage? > >Cheers > >Phil Hobbs
Haven't tried, but that's a good idea. I'll get some and fire up the old Boonton. I wonder what SiGe leakage is like. Well, I'll get some when UPS can deliver them. BFT25 seems to leak single digits of fA as a diode. 0.25 pF too. -- John Larkin Highland Technology, Inc The cork popped merrily, and Lord Peter rose to his feet. "Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"
On Thu, 19 Mar 2020 13:03:48 -0400, Phil Hobbs
<pcdhSpamMeSenseless@electrooptical.net> wrote:

>On 2020-03-18 20:14, George Herold wrote: >> The project seems to call for the smallest detectors. I'm buying a >> few each of 'all'* the little PD's from digikey. (I can share a list >> if someone wants... it's a mix of TH and SMD.) C_in between 1-3 pf >> when reversed biased. I don't see how it makes sense to bootstrap or >> cascode the PD**. > >You can get about 15 dB improvement at high frequency by using a >SAV-551+ bootstrap, if you keep the stray capacitance low enough. > >So I was looking at opamps with small C-in. >> (Big thanks to H&H of AoE as always.) I've now got a column of 1/f >> corner freq. of e_in (input voltage noise) for the small C_in fet >> opamps. In table 4x.2 of the x-chapters. The best looks to be the >> ADA4817. 1/f ~10 kHz. (hard to tell.. bit of a swoopy noise graph.) >> (C_in = 1.5 pF, e_in = 4 nV/rthz.) >> >> Pure 1/f noise into a TIA must give a constant noise voltage.. till >> hit by the GBW of the opamp. (is that right? I'm being mathematically >> lazy and just tipping the slope of the graph.) > >No. The eN*C noise power density goes as f**2, whereas 1/f noise power >goes as (drum roll) 1/f. > >That's why pHEMTs make very good bootstraps despite having 1/f corners >of 10 MHz or so--the f**2 from the admittance of the capacitor knocks >out the 1/f from flicker noise. > >> **I'm not sure how to guesstimate the C of a cascode, and are there >> any small bjt's left? > >For a common-base stage it's C_CE, but you generally don't care because >it appears across r_E. > >For a FET bootstrap it's basically C_DG + (1-Av) C_GS. For a pHEMT that >can be way below 1 pF, so you're limited by the strays.
I'm just finishing up the layout of my Colpitts test oscillator. It's a 4-layer prototype. I punched a hole in the L2 ground plane and inserted a copper patch below the critical oscillator nodes. I'll compare tempco with the patch grounded, or connected to the emitter as a bootstrap/guard. https://www.dropbox.com/s/n69fqw8xef46l3n/Z482_Guard.jpg?raw=1 PADS is being obtuse. I can't flood the copper guard over connection point TP5 without it throwing a clearance error. Too bad. -- John Larkin Highland Technology, Inc The cork popped merrily, and Lord Peter rose to his feet. "Bunter", he said, "I give you a toast. The triumph of Instinct over Reason"