Composite amps

[Phil Hobbs]

Hi, all,

I'm generally prejudiced against composite amplifiers (two op amps
inside one feedback loop) because they're generally squirrelly, with
poor settling performance and weird transient response.

On the other hand, my aversion to them means that I don't have as much
experience with them as do composite-amp fans. So what do you folks say
about them?

Orchids? Onions? Actual expertise?

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 Sun, 27 May 2018 15:09:33 -0400, Phil Hobbs
<pcdhSpamM...@electrooptical.net> wrote:

>Hi, all,
>
>I'm generally prejudiced against composite amplifiers (two op amps
>inside one feedback loop) because they're generally squirrelly, with
>poor settling performance and weird transient response.
>
>On the other hand, my aversion to them means that I don't have as much
>experience with them as do composite-amp fans. So what do you folks say
>about them?
>
>Orchids? Onions? Actual expertise?
>
>Cheers
>
>Phil Hobbs

I've only done it a little, in very special cases, but if the intent
is to apply a slow DC offset correction, and the main amp and the DC
trim amp don't overlap in frequency response, it seems to work fine.

If you want to make a general -6 dB/octave amp as a composite, the
risk is probably saturating one of the amps in large-signal/slewing
cases, or at leasy doing goofy things. A composite that clips clean
would be a challenge.

A sorta similar case is where a fast signal needs to be DC coupled
across a big DC offset. A capacitor is the fast path and some slow
opamp thing does the DC part before the AC path decays. The gains have
to both be the same, about 1.00 usually, and the frequency responses
need to be matched, to get clean step response and no ISI.

Tek called this "feed-beside", a brutally fast but ugly signal path,
and slow stuff in parallel to make it clean.





--

John Larkin Highland Technology, Inc trk

jlarkin att highlandtechnology dott com
http://www.highlandtechnology.com

[Phil Hobbs]

Not so easy!

>
> Tek called this "feed-beside", a brutally fast but ugly signal path,
> and slow stuff in parallel to make it clean.

Plus a lot a lot of hand work to get them to match. I've benefited
greatly from their labours, but I have no interest in doing that myself!

What I'm mostly talking about is using a nice quiet accurate amp such as
an ADA4898 plus a faster but less accurate thing such as a THS3091 or
LM6171. The output amp is run at some fixed gain like 10, and the input
amp is run at high enough gain that the combination is stable at
quiescent conditions.

Other composite amps such as the one you mention or the common case of
using a chopamp to control the offset voltage of some fast-but-ugly
amplifier have a different set of problems.

[John Larkin]

On Sun, 27 May 2018 17:45:11 -0400, Phil Hobbs

So that would take the frequency response curve of the precision amp
and just shift it up 20 dB?

That should work, as long as the first amp keeps rolling off -6
dB/octave past its normal Ft. You could tweak a bit if not.

Make the first amp limit clean at +-2 volts or whatever, to not wind
up too much, and let the second, fast amp clip however it likes.

Or, when the second amp hits some swing limit, stuff current into the
input of the first one. Some common-base things maybe.

This would need some Spicing but shouldn't be too bad.

>
>Other composite amps such as the one you mention or the common case of
>using a chopamp to control the offset voltage of some fast-but-ugly
>amplifier have a different set of problems.
>
>Cheers
>
>Phil Hobbs

--

[bitrex]

This Wireless World article from '74 might be of interest:

<http://www.keith-snook.info/wireless-world-magazine/Wireless-World-1974/Reducing%20Amplifier%20Distortion%20-%20Sandman.pdf>

[bitrex]

On 05/27/2018 05:45 PM, Phil Hobbs wrote:

And this one from Burr-Brown:

<http://www.ti.com/lit/an/sboa002/sboa002.pdf>

[Phil Hobbs]

How so?

[Phil Hobbs]

Yeah, that's the idea. Problem is that it has horrible transient response.

[bitrex]

Guessing this stool-design problem probably only has three legs, the
labor will in be picking the appropriate length of the legs to meet the
requirements without ending up with the other kind of stool.

[bitrex]

Sorry hadn't read the follow-up yet

[Phil Hobbs]

No worries, I'm mainly interested in actual experiences. Spice we can
all do. ;)

[John Larkin]

On Sun, 27 May 2018 20:40:17 -0400, Phil Hobbs

Not to change the subject (I'd never do that) but I have made a
compound amp just to shift the power dissipation away from the
front-end diff pair, off to another chip, to avoid nanovolt thermal
hooks. I had to keep the feedback network low impedance to minimize
Johnson noise, which required a lot of feedback current.

https://www.dropbox.com/s/nieqrj2um62pdu5/L700_Shunt_Amp.jpg?raw=1

[Steve Wilson]

John Larkin <jjla...@highlandtechnology.com> wrote:

> Not to change the subject (I'd never do that) but I have made a
> compound amp just to shift the power dissipation away from the
> front-end diff pair, off to another chip, to avoid nanovolt thermal
> hooks. I had to keep the feedback network low impedance to minimize
> Johnson noise, which required a lot of feedback current.

> https://www.dropbox.com/s/nieqrj2um62pdu5/L700_Shunt_Amp.jpg?raw=1

Not to change the subject, but I have a question. There is another type of
amplifier that splits the signal into two paths - a high frequency path for
an RF amplifier with poor DC drift and small DC offset capability, and a low
frequency path for an amplifier with good DC characteristics and wide offset
capability. I thought this was a compound amplifier, and once read an article
in the HP Journal that described it.

But I can't find the article, and google is no help. Do you know the name of
this kind of amplifier?

[Phil Hobbs]

I've used emitter followers for similar things, most recently in a super
low noise laser driver. It uses MC33078 op amps and 330-ohm feedback
resistors in a 12-V system. It's basically a PNP current source with a
two-pole decoupling network between base and 10-V supply. (The +10 rail
is made by a couple of diodes from +12, so that the op amp + emitter
follower has enough output swing.)

[John Larkin]

This is usually called a compound amplifier. Tektronix called
something similar to this "feed-beside."

There are two ways to do this:

1. Split the signal with RC or bias tee circuits, amplify the AC and
DC parts with separate amps, and combine at the output.

2. Build a compound amp, with optimized AC and DC paths, but treat it
as a black-box opamp, and close a feedback loop around it.

I don't know of they have specific names. As Phil noted at the start
here, it's tricky to manage the overlap with precision.

There is an RF power amp configuration that has a high-power amp with
some distortion, and a paralleled low-power amp with correcting
distortion behavior. That probably has a name. I think cell towers use
that.

[Phil Hobbs]

Dunno. I cordially dislike architectures like that on account of their
weird settling behaviour. It's super hard to get the nonlinearities to
match, so it's generally much better to put bandaids on the RF amp to
make it behave properly. I've only done that once or twice, so I don't
have too many specifics to contribute.

[Phil Hobbs]

Feedforward.

[John Larkin]

On Mon, 28 May 2018 11:16:24 -0400, Phil Hobbs

Opamp data sheets and appnotes are pretty good at avoiding any mention
of transient thermal effects, or static effects of output stage
dissipation on input offset. I know it is a serious consideration in
opamp chip layout.

They can measure their DC specs and AC/noise specs and large-signal
swing specs with different setups. Sneaky devils.

A compound amp can avoid some of these problems.

[John Larkin]

Not to be confused with predistortion, I guess.

https://en.wikipedia.org/wiki/Predistortion

I'd play with the compound amp thing, but I need to force myself to do
less interesting grunt work. Like revising proposals and replacing
faucets. Hard to decide which is less appealing.

[Phil Hobbs]

For temperature controllers you normally want the front end amp on the
cold plate, but you don't want its dissipation to vary. I often put a
lead-lag integrator inside the loop but off the cold plate. That forces
the input amp's bias to stay very very stable, so there's no change in
dissipation.

[Phil Hobbs]

Do the proposal, then you can pay somebody to do the faucet. ;)

[Steve Wilson]

John Larkin <jjla...@highlandtechnology.com> wrote:

> This is usually called a compound amplifier. Tektronix called
> something similar to this "feed-beside."

> There are two ways to do this:

> 1. Split the signal with RC or bias tee circuits, amplify the AC and
> DC parts with separate amps, and combine at the output.

> 2. Build a compound amp, with optimized AC and DC paths, but treat it
> as a black-box opamp, and close a feedback loop around it.

> I don't know of they have specific names. As Phil noted at the start
> here, it's tricky to manage the overlap with precision.

> There is an RF power amp configuration that has a high-power amp with
> some distortion, and a paralleled low-power amp with correcting
> distortion behavior. That probably has a name. I think cell towers use
> that.

Compund, not composite. Thanks.

[bitrex]

What is this "predistortion" discussed in the article but itself a form
of "regular" negative feedback?

It says "In essence, 'inverse distortion' is introduced into the input
of the amplifier, thereby cancelling any non-linearity the amplifier
might have." Yes, that's what negative feedback is.

[John Larkin]

Except that predistortion is not negative feedback; all the signals
are going in the same direction. It's hard to do feedback on a GHz
power amp.

[George Herold]

I broke the feedback loop with an RC lowpass. (int-amp reading the R,
there was at least one gain adjustment in each leg.)
But I don't think that's what Phil is asking about.

George H.

[George Herold]

Right, I've never done that, but as long as one can be tuned
down to match the slowest (time-wise).

George H.

[George Herold]

Have you peaked at the in between opamp signal?
GH

[George Herold]

Oh you've got some time delay front to back,
maybe feed back around the 1st opamp...
an RC in series? at some intermediate freq./time.
George H.
Inside cooling off, I'm grilling more meat later.
Hey a shoutout to any vet's on Memorial Day.
Thanks!
(I should give some buddies a call.)

GH

[Jim Thompson]

On Sun, 27 May 2018 15:09:33 -0400, Phil Hobbs
<pcdhSpamM...@electrooptical.net> wrote:

>Hi, all,
>
>I'm generally prejudiced against composite amplifiers (two op amps
>inside one feedback loop) because they're generally squirrelly, with
>poor settling performance and weird transient response.
>
>On the other hand, my aversion to them means that I don't have as much
>experience with them as do composite-amp fans. So what do you folks say
>about them?
>
>Orchids? Onions? Actual expertise?
>
>Cheers
>
>Phil Hobbs

I don't have time (*) to join the discussion, but read this...

<http://www.analog-innovations.com/SED/CompositeAmplifiers.pdf>

(*) Pursuing a new trick for behavioral modeling ;-)

...Jim Thompson
--
| James E.Thompson | mens |
| Analog Innovations | et |
| Analog/Mixed-Signal ASIC's and Discrete Systems | manus |
| STV, Queen Creek, AZ 85142 Skype: skypeanalog | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |

Thinking outside the box... producing elegant solutions,
by understanding what nature is hiding.

"It is not in doing what you like, but in liking what you do that
is the secret of happiness." -James Barrie

[Phil Hobbs]

On 05/28/18 14:56, Jim Thompson wrote:
> On Sun, 27 May 2018 15:09:33 -0400, Phil Hobbs
> <pcdhSpamM...@electrooptical.net> wrote:
>
>> Hi, all,
>>
>> I'm generally prejudiced against composite amplifiers (two op amps
>> inside one feedback loop) because they're generally squirrelly, with
>> poor settling performance and weird transient response.
>>
>> On the other hand, my aversion to them means that I don't have as much
>> experience with them as do composite-amp fans. So what do you folks say
>> about them?
>>
>> Orchids? Onions? Actual expertise?
>>
>> Cheers
>>
>> Phil Hobbs
>
> I don't have time (*) to join the discussion, but read this...
>
> <http://www.analog-innovations.com/SED/CompositeAmplifiers.pdf>

I have that one, thanks. Lots of the circuits are summing-junction
snoopers, i.e. circuits wih a slow-but-accurate amp looking at the
time-averaged input error of a fast-but-cruder amp, and nulling it out.
That's a useful trick sometimes, and are examples of "putting a bandaid
on the fast circuit", which I was talking about upthread.

There are other sorts of bandaids, e.g. the White cathode follower and
many sorts of local feedback. Often the key is to figure out a way that
the bandaid can be much slower than the main amplifier, as in the
snooper circuits.

I often use op amps to force JFETs to run at exactly I_DSS, for
instance--the problem then is to keep the low-frequency noise from going
nuts.

[Phil Hobbs]

It's generally well behaved until the output amp rails, the reason being
that the input amp is the slow one.

I've been playing with a CPH3910/ADA4899/THS3091 combination that is
starting to look reasonable in the spherical-cow universe. It's a
little gain-of-50 amp with about 120 MHz bandwidth, Zin of 10M // <<1pF,
and flatband 1-Hz noise of about 1.5 nV. (The capacitance of the
connector will dominate.)

I seriously doubt it'll work that well in real life.

[Phil Hobbs]

> Oh you've got some time delay front to back,
> maybe feed back around the 1st opamp...
> an RC in series? at some intermediate freq./time.
> George H.
> Inside cooling off, I'm grilling more meat later.

Overcast here--has been all weekend. :(

> Hey a shoutout to any vet's on Memorial Day.
> Thanks!

+1

[Phil Hobbs]

On 05/27/18 20:05, Phil Hobbs wrote:

>> This Wireless World article from '74 might be of interest:
>>
>> <http://www.keith-snook.info/wireless-world-magazine/Wireless-World-1974/Reducing%20Amplifier%20Distortion%20-%20Sandman.pdf>
>
>
> How so?

I had a look at the article, and it does have some points of interest,
thanks. The author distinguishes between his technique and feedforward,
but I'd argue that for circuit purposes it's a species of feedforward,
whose basic principle is to sense just the error of an amplifier and
apply the appropriate correction to the _output_ and not the _input_.

I didn't know that FF was invented by Harold Black.

[bitrex]

Ah, helpfully the Wiki seems to never mention that it's a feed-forward
and not feedback topology.

[bitrex]

Have you tried designing a composite-amp oscillator instead? If my
theory is correct it should make a wonderful amplifier.

[bitrex]

I'm taking today to remember my Dad, who passed about two and a half
weeks ago just shy of 92. S/Sgt 10th Mountain, served in Italy late 1944
to 1946.

Thanks, Dad!

And also sort through all his stuff, he was of the generation that never
really liked to throw anything away that looked like it would come in
handy someday. Sigh...

[George Herold]

I'm sorry for your loss. My dad was part of the 'greatest' generation too,
but he passed away in his 70's. Years ago. He served in the South Pacific.

George H.

[bitrex]

I'm the youngest of four kids, I got more time with him than most guys
whose fathers were in their early 50s when they were born. He remained
pretty much independent until he died but required a few hours a week of
help from me and a nurse the past few years that I was glad to assist
with, we were still going out for lunch about once a week until last month.

About a week before he died he'd got some test results back that didn't
look too good. He was rarely a pessimist, about the closest he came was
that day when he looked very tired and said that he missed his friends,
all the people who he used to know who by now were all gone.

I could understand that, one son who visits regularly is nice but can
hardly make up for all the rest. He became very ill the next week and
passed away about 48 hours after arrival at the hospital, just enough
time for the rest of the family to make it to be there with us.

It went about as "well" as it could and at that age and I'd known for
quite a while the day would come sooner or later, but I'm still thinking
it's him when the phone rings. It's all still very strange. :)

[bitrex]

On 05/29/2018 08:53 AM, George Herold wrote:

And thank you! Dad was kind of hit-or-miss as a father (like most
fathers I suppose) but it was remarkable hearing the stories about the
time he came from which was both so different and so much the same as th
time I grew up in, I am happy to say he was a man I was proud to have
had a chance to know.

[George Herold]

Sure, my dad died one night, massive heart attack, he was set to go
hunting the next day with buddies. It was totally unexpected from my
point of view, but we learned after the fact (from his girlfriend)
that he wasn't taking his high blood pressure meds. Some type of
advanced warning is mostly a good thing, let's you sorta prepare...
emotionally.

George H.

[Jim Thompson]

On Mon, 28 May 2018 15:26:58 -0400, Phil Hobbs

I have a paper somewhere that discusses GBW-product and optimization
of the two-OpAmp composite. Quite nicely done, as I recall...
unfortunately I haven't re-found it :-(

[bitrex]

No warning feels extremely unfair, but having a couple last
conversations is extremely hard too. There's so much you want to say but
gosh you don't want to be too emotional or look like you're coming
unglued yourself! It feels "unfair" any way you slice it I suppose.

We mostly talked about his grandkids, some old movies he wanted to see
again, girls, and boats. Seems as reasonable an assortment of topics as
any other :)

[George Herold]

Yeah, life's not fair. :^) I keep telling my kids this, but I think
you only learn it through hard knocks. Me mum passed away about 1 year ago
now. Alzheimer's, which is a particularly icky way to go. When I heard,
after shock, my next emotion was relief.

George H.

[k...@notreal.com]

My dad died suddenly at age 52, shoveling snow. I was 12 at the time.

[k...@notreal.com]

Ouch! That is probably the hardest way to go, at least for the
family. My mother was getting pretty bad at the end but it wasn't
Alzheimer's. She was 95, with heart and kidney problems that were
taking their toll on the brain. I understand your comment about
"relief".

[glen walpert]

That sounds like something I would have saved if I had seen it, so I
searched my SED notes folder without luck but found some old app notes
and a post from Win Hill that might be of interest to someone. (Probably
old hat to Phil):

Analyzing feedback loops containing secondary amplifiers:
www.ti.com/lit/an/slyt103/slyt103.pdf

Burr Brown Application Bulletin
FEEDBACK PLOTS DEFINE OP AMP AC PERFORMANCE
www.ti.com/lit/an/sboa015/sboa015.pdf
(composite amp discussion pp 9-13)


From: Winfield Hill <Winfiel...@newsguy.com>

Newsgroups: sci.electronics.design

Subject: Re: Photodiode wich is fast enough to detect +50Mhz analog
(sinus) signal??

Date: 11 Aug 2004 03:31:18 -0700


Yannick wrote...

>

> Winfield wrote...

>> Careful, using a sufficiently-wideband opamp can insure the summing-

>> junction impedance will be low compared to the total shunt
capacitance.

>> Resistors have 0.05pF to 0.1pF of self capacitance, this should be
your

>> total feedback capacitance. With 3k resistor you'd have a -3dB
rolloff

>> at 530MHz. You want high R for low noise, so we'll try 100k, yielding

>> a 16MHz rolloff. Then we can apply the standard R-C-R trick (this is

>> more than 30 years old) to get a flat frequency response to 75MHz, or

>> whatever you decide your bandwidth should be.

>

> R-C-R trick to get a flat frequency responce, i never heard of this

> before, can you explain how it works?



I posted the circuit previously in this thread (20 July) and described

in detail how it works. Here's the circuit again:



| Rf R2 adjustable

| ,---/\/\---+---/\/\--/\/\----,

| | '--||--' | C2 R3 | nA-sensitivity wideband

| | Cf '--||--/\/\-- gnd | transresistance amplifier

| | |

| | __ ,-||--/\/\--+ correction network details

| input O--+---|+ \ | __ | R2 C2 = Rf Cf

| | >-+-/\/\-+-|- \ | R3 C2 sets bandwidth

| ,-|-_/ | | >-----+---

| | | gnd --|+_/

| gnd --/\/\--+-/\/\--' composite amplifier



The undesired Rf capacitance Cf is canceled by the R2 C2 network.

R3 is used to limit the upper frequency of this cancellation.



R2 C2 and R3 constitute the standard R-C-R trick. I thought of this

about 18 years ago, and have used it with great success since then.

Later I learned that it had been described in an old Keithley manual,

and probably in many other places years before that.



BTW, the circuit above will outperform (sensitivity, bandwidth, SNR,

phase accuracy) any of the resonant schemes you've been contemplating

here. It's not true that feedback makes things more noisy. In this

circuit feedback (and a high-performance composite amplifier) insures

that all of the signal current is used by the amplifier, rather than

becoming uselessly drained away by the input-node capacitance. Thus

feedback actually improves the SNR. Using a resonant input doesn't

solve the capacitance problem because if a high enough Q is used for

a solution, it simply creates insurmountable phase-error problems.







--

Thanks,

- Win

[John Larkin]

On Mon, 28 May 2018 11:58:52 -0400, Phil Hobbs
<pcdhSpamM...@electrooptical.net> wrote:

>On 05/28/18 11:42, John Larkin wrote:

>Do the proposal, then you can pay somebody to do the faucet. ;)
>
>Cheers
>
>Phil Hobbs

Well, it's refreshing to, once in a while, swap a screen for the
underside of a bathroom sink.

But really, most consumer products, including plumbing, are such
trash.

--

John Larkin Highland Technology, Inc

picosecond timing precision measurement

[plastco...@gmail.com]

Zero DC offset low noise composite TIA for high capacitance photodides
http://ixbt.photo/?id=photo:1329265

2200pF input capacitance
http://ixbt.photo/?id=photo:1329266

Zero-drift OPA189 composite TIA
http://ixbt.photo/?id=photo:1329269

Inductance LB3218T102K (1207 smd)

[John Larkin]

On Fri, 21 Feb 2020 12:37:03 -0800 (PST), plastco...@gmail.com
wrote:

Just a note: a photodiode has two leads.

[plastco...@gmail.com]

Capacitor 2200pF is my photodiode.

[John Larkin]

On Fri, 21 Feb 2020 13:11:52 -0800 (PST), plastco...@gmail.com
wrote:

>Capacitor 2200pF is my photodiode.

That's a lot. You could drive a cascode emitter to keep the load
impedance way down on the photodiode. The transistor collector, much
lower capacitance, can drive the TIA.

The problem there is that you should have a standing current in the
cascode transistor, to keep the emitter impedance low. That makes an
offset and maybe messes up the TIA.

Hence the notion that a photodiode has two leads. Take the fast stuff
from one end and the slow signal from the other. It's the Tektronix
"feed-beside" idea.

I'm expecting the first article of my GHz o/e converter any day now.
It uses the feed-beside idea.

[plastco...@gmail.com]

Internal input capacitance OPA140 :
differential Cd = 10 pF
common mode Ccm = 7 pF

Cascode transistor or voltage follower useless in this case.

Zero bias voltage photodiode boosted TIA :

http://ixbt.photo/?id=photo:1329322

OPA140 with internal compensation cap - amplifier like integrator.

[plastco...@gmail.com]

Metamorphoses :

http://ixbt.photo/?id=photo:1329419

Bootstrapping TIA grounded photodiode

[George Herold]

Dmitriy, What's the advantage of this? At one point in your
circuit evolution I thought you were going to move the non-inverting
input away from ground to bias the photodiode (PD).
(You can reduce PD capacitance ~factor of 3 or so with reverse bias.)
Have you read Phil Hobb's book? Most of his ideas can be found
for free online... check his website.

George H.

[plastco...@gmail.com]

1) grounded detector, grounded source/emitter HF transistor
2) true zero-bias operation of detector
3) my circuit is simpler
I like to read books))

[George Herold]

OK, what do you find better about zero bias operation?
I should admit that for many years I ran all my PD's at
zero bias. I thought this gave me better 'zero' light detection.
(No DC offset with no light... but the dark current from
PDs is generally pretty low.)
Running with some bias has two main advantages.
1.) reduced C.. faster
2.) Higher saturation current (light intensity) without bias the
electrons build up in the junction and it saturates.. more light
gives no more electrons.

George H.
(who is addicted to reading... I need to find a few new fiction writers)

[Phil Hobbs]

Zero bias is better in one respect: you can get zero leakage current.
For jobs such as very wide range, very slow photometers, that's a win.
Garry Epeldauer et al. wrote a beautiful paper about getting 14 orders
of magnitude in photocurrent, if you don't mind being stuck with
millihertz bandwidths:

<https://electrooptical.net/www/optics/eppeldauer14decadephotocurrent.pdf>

Crappy PN photodiodes and solar cells don't respond well to large
reverse bias either.

For just about anything else, zero bias is a complete crock.

With almost any PIN diode, APD, MPPC, (etc) zero bias is a disaster.
Applying reverse bias to a PIN diode can reduce its capacitance by a
factor of 7 or so, which reduces the high frequency noise by the same
factor.

[John Larkin]

I know of a large organization that has wasted about a million dollars
a year, since 2002, by running a lot of very expensive Hamamatsu
photodiodes at zero bias.

https://www.dropbox.com/s/wm3a3cpxa8tcarg/S8551_1.JPG?raw=1

[Phil Hobbs]

I suspect I know the organization. ;)
They have some very good folks though.

[George Herold]

(Well thanks for trashing the first ~20 years of my PD career. :^)

In defense of zero bias it's got great simplicity, doesn't
break if put in backwards*. And if your DMM has a 200uA scale
a ~10mm^2 PD makes an easy light detector. Lots of applications
don't care so much about speed.
(a bow to the lm324, I never used one, but enjoyed the schematics.)

>
> With almost any PIN diode, APD, MPPC, (etc) zero bias is a disaster.
> Applying reverse bias to a PIN diode can reduce its capacitance by a
> factor of 7 or so, which reduces the high frequency noise by the same
> factor.

George H.
*I've never tried breaking a PD by forward biasing hard.
my bpw34 looks about the same volume as a 1/4 watt resistor,
so ~1/4 W, 0.6V and 0.4A

[jla...@highlandsniptechnology.com]

On Mon, 24 Feb 2020 15:46:09 -0500, Phil Hobbs

Price's law. They have a few brilliant people and an army of duds.

But if the brilliant people retire, or otherwise go away, it might
take a while for the world to notice.

IBM, Xerox, Kodak, Polaroid, Boeing, Motorola, RCA, DEC, Nokia.

--

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"

[plastco...@gmail.com]

Photodiodes from Russia for similar applications -
http://www.technoexan.ru/pdf/silicon_detector/spd100uv.pdf
https://www.researchgate.net/publication/266227378_Silicon_precision_detectors_for_near_IR_visible_UV_XUV_and_soft_X-ray_spectral_range
price less than $300
http://ixbt.photo/?id=photo:1183963
Quantum noise from fatty X-ray photons is much larger than the shot noise of the dark current at the same value))

[Phil Hobbs]

$300 for a garden variety PN photodiode? Nice work if you can get it.
You can get an MPPC that size with a gain of a million for about the
same money.

[George Herold]

Seems spendy, I think the big PD's from osi-optoelectronics are ~$50.

The extra noise from x-rays is easy to understand. Each x-ray is going
to come in and make a number of e-h pairs. (let's say on average
there are ten electrons made per photon.) So the 'pieces' of charge
causing the noise are ten times bigger and you get ten times the
noise density.

George H.

[George Herold]

Years back I bought PD's from advanced-photonix (I think they had a
different name to start.. bought out or something.)
Then the prices more than doubled and I went elsewhere.
They still seem to be in business.

https://www.digikey.com/product-detail/en/advanced-photonix/290-12-22-241/209-290-12-22-241-ND/1012520

George H.

[plastco...@gmail.com]

Russian photodiode for 13 nm, is expensive.
Example: 4keV photon, 4000/3.6 ~1100e, square root = 33
quantum noise 33 times greater than dark current

Demo version e-PHEMT 40-140 mA (or SiGe BFU910F 2mA) boosted TIA :
http://ixbt.photo/?id=photo:1329848

Gate current ATF-53189 ~85nA

[Phil Hobbs]

That's sort of silly, because it relies on the OPA140 to complete the
loop, and that grossly limits the speed. You can do a _lot_ better than
2 us rise time (massively underdamped) with 2200 pF.

It works a lot better if you run the pHEMT as a follower and return the
anode of the PD to the pHEMT source via a capacitor. You'll need a good
ferrite bead in the gate to prevent oscillation--try a BLM18BA100SN1.

The ATF53189 is obsolete, like all the 3-GHz HP/Avago/Broadcom pHEMTs.

A gizmo of mine that works that way is at
<https://electrooptical.net/News/silicon-photomultiplier-sipm-mppc-system-for-cathodoluminescence/>.
The detector capacitance is between 350 and 1400 pF, so the comparison
is reasonable.

If you used a PIN diode and reverse biased it, you'd be around 100 pF
rather than 8 nF for that 1 cm**2 device. That would get you a noise
reduction of about 38 dB essentially free. That sort of improvement
matters to some of us. ;)

Of course it might or might not work with X-rays.

[plastco...@gmail.com]

>
>It works a lot better if you run the pHEMT as a follower and return the
>anode of the PD to the pHEMT source via a capacitor.
>

IMHO, the follower or my wiring diagram does not make any difference - only the location of the symbol GND !
From the point of view of the photodiode, one circuit pulls on the lower terminal, the other circuit operates on the upper pin.
All properties are similar.
My JFET BF862 is installed more conveniently - single-supply compatible.
It is very interesting to observe the voltage at the input of the amplifier.

http://ixbt.photo/?id=album:30781

Why no one broke the taboo - disconnect the non-inverting input Op Amp from GND ?

[George Herold]

On Tuesday, February 25, 2020 at 2:13:52 PM UTC-5, plastco...@gmail.com wrote:
> Russian photodiode for 13 nm, is expensive.

OK sure... just run with no window?

> Example: 4keV photon, 4000/3.6 ~1100e, square root = 33
> quantum noise 33 times greater than dark current

Yeah, I know little of x-rays. Is the depth of the depletion
region such that the x-ray loses all it's energy there?
You could use the noise to measure the QE of the detector.
(If you knew the energy.)
(maybe you already do that.)
(or measure x-ray energy from the noise..
OK that's probably a silly idea.. easier to measure x-ray
energy from a pulse height or total current or something.)

>
> Demo version e-PHEMT 40-140 mA (or SiGe BFU910F 2mA) boosted TIA :
> http://ixbt.photo/?id=photo:1329848
>
> Gate current ATF-53189 ~85nA

OK that's fun.

Dmitriy, (SED's own Russian troll* :^) 2 us rise time
with 1 meg and 2200pF is nice. With a smaller PD and
both bootstrap and cascode I think I got about a
1 us rise time. (Phil and Win do better.)
(I had slower opamp and transitors.. )
Is the yellow trace, chan. 1, the output? That looks mostly like a
single pole. Which means (to me) that there is a single capacitance slowing
things down. It could be C in parallel with the 1 meg FB resistor.
(what does step response with 100k ohms look like?)

If not that you should consider setting aside an afternoon to try
biasing the PD.
Have you read the TIA section in AoE3? (Win's book)

George H.
*I say that lovingly.

[George Herold]

OK the layout of the circuit is of utmost importance. You could post
a pic or something.
You do realize that you've biased the photodiode by 1/2 your supply
voltage. (Doesn't that also bias your DC output?..
checks trace, right ~1.3 volts below ground.)
I've seen other schematics that bias the opamp rather than the photodiode.
I've never liked them because I like DC =0 at zero light. Better for
me is a lop-sided supply for the opamp. (and bias the PD more. At least 10V
unless the manufacturer says not to.)

George H.

[Phil Hobbs]

Because you've got a 10 MHz op amp inside the feedback loop of a 12-GHz
f_max pHEMT, which is just plain silly.

There's nothing special about ground anyway.

[plastco...@gmail.com]

The capacitance of my detectors is below 300 pF and i use negative bias.
cap 2200 pF for picture only))

Very thin dead layer is important for photons 13 nm.
Responsivity of the detector for photons 1keV - 5keV is always the same = 0.27 A/W
Very funny detector this is a CCD camera without a window.
Only you need to connect it not to the TV, but to the oscilloscope and you can immediately analyze the materials by XRF.

Phil Hobbs wrote:
>That's sort of silly, because it relies on the OPA140 to complete the
>loop, and that grossly limits the speed.

>Because you've got a 10 MHz op amp inside the feedback loop of
> a 12-GHz f_max pHEMT, which is just plain silly.

From bootstrapping photodiode to bootsrapping non-inverting input Op Amp.
The task of the bootstrap circuit is to keep the voltage at the photodiode constant by directing the photocurrent to the amplifier.
STOP ! Error detected))
The current should not go to the amplifier, but to the feedback resistor Rf !
Now you guess what needs to be changed in the circuit so that the amplifier OPA140 turns from a snail into a cheetah ?
Slew Rate OPA140 20 V/μs

http://ixbt.photo/?id=photo:1329866

[George Herold]

On Monday, February 24, 2020 at 2:47:45 PM UTC-5, Phil Hobbs wrote:

Hi Phil, I downloaded the above and was chewing through it last night.
Great stuff!
1.) Rs (PD shunt resistance... I've always just treated this as
infinite. Can I measure leakage current and get Rs?
2.) pg 3094 has a nice discussion of 1/f noise.
3.) Are there even better low current opamps these days?
4.) Nice effective BW calcs in App A.
I would add to that, the ENBW for a two pole filter, f_3dB and Q,
is
ENBW = f_3dB * Q *pi/2 = ~1.11 f_3dB (Q=0.707.. Butterworth)

George H.

[Phil Hobbs]

Re: noise BW
Yeah, it's like 1.22x for two noninteracting RC poles, so 1.11 for
Butterworth sounds roughly right.

Re: shunt resistance

For bias voltages << kT/e, both the forward and reverse diffusion
currents are contributing to the conductance--it's just dI/dV, and so is
fairly far from zero for a large-area diode run at zero bias.

Ideally the effective shunt resistance goes up by a factor of 2 or so
with 50 mV of reverse bias, because you shut off the reverse diffusion
current without introducing significant additional leakage. That's a
super useful trick with InGaAs diodes in dim light.

re: 1/f noise
Haven't read it recently, but in photodiodes you actually can get
significant 1/f noise at zero bias, unlike in the case of resistors.

re: low current op amps

BITD I used to really like the OPA111. Its performance was easy to
remember: 1 MHz bandwidth, 1 mV offset, 1 uV/K drift, 1 pA input bias.
(See? I haven't used one in 30 years and I still remember.) ;)

It was one of the primo op amps used in early tunnelling and atomic
force microscopy. Of course it's noisy, but not nearly as bad as the
other popular super-high-Z op amp of the day, namely the LM11.

Nowadays there are much better choices, e.g. JL's fave OPA197.

[George Herold]

Yeah well the 1.11 number is in AoE so I figured everyone knew it. :^)

>
> Re: shunt resistance
>
> For bias voltages << kT/e, both the forward and reverse diffusion
> currents are contributing to the conductance--it's just dI/dV, and so is
> fairly far from zero for a large-area diode run at zero bias.

Huh OK... I did this measurement on bpw34 today. At modest voltages
(6 to 21V I got ~10 mV from a TIA with 100 meg FB ~0.1 nA..
at 10V rev. that's ~100 G ohm..!? OK I was also seeing ~10 mVrms of
60 Hz crude.. so I'm not sure of these numbers.

>
> Ideally the effective shunt resistance goes up by a factor of 2 or so
> with 50 mV of reverse bias, because you shut off the reverse diffusion
> current without introducing significant additional leakage. That's a
> super useful trick with InGaAs diodes in dim light.
>
> re: 1/f noise
> Haven't read it recently, but in photodiodes you actually can get
> significant 1/f noise at zero bias, unlike in the case of resistors.

Oh, probably simple stuff you know, but it resonated with me.
1/f noise density (V^2) ~ Log(f_high/f_low)
And then if you are recording a number (P) of separate measurements.
The low freq BW ~1/P and the noise density goes as Log(P) the number
of measurements...
"Oh", I said to myself.

>
> re: low current op amps
>
> BITD I used to really like the OPA111. Its performance was easy to
> remember: 1 MHz bandwidth, 1 mV offset, 1 uV/K drift, 1 pA input bias.
> (See? I haven't used one in 30 years and I still remember.) ;)
>
> It was one of the primo op amps used in early tunnelling and atomic
> force microscopy. Of course it's noisy, but not nearly as bad as the
> other popular super-high-Z op amp of the day, namely the LM11.
>
> Nowadays there are much better choices, e.g. JL's fave OPA197.

Thanks, I saw the opa111 was the last time buy on DK.. for
~$50 you can relive past. :^)

George H.
Re opa197: at least spec wise doesn't the opa192 knock it away?
(maybe I'm missing something?) (5uV offset and 0.2 uV/C drift)

[jla...@highlandsniptechnology.com]

I don't use that as a low noise high-performance amp, but as a
general-purpose gumdrop. It's stable with a big output cap, 3.3u film
or 100u polymer.

[George Herold]

OK Thanks. Is there some nice opamp like this (opa197 or 192) that
has a little more GBW? I'm looking at table 4X.2 (AoEx) High spped
VFB's... there are a lot to choose from. 50-100 MHz would be nice.

opa1611 looks OK

George H.

[whit3rd]

On Tuesday, February 25, 2020 at 3:55:22 PM UTC-8, George Herold wrote:
> On Tuesday, February 25, 2020 at 2:13:52 PM UTC-5, plastco...@gmail.com wrote:

> > Russian photodiode for 13 nm, is expensive.

> > Example: 4keV photon,...

>.. easier to measure x-ray
> energy from a pulse height or total current or something.)

At 4 keV, I'd want a proportional counter with (Xenon?) gas,
or a fairly large volume ion chamber (leakage current in biased
parallel-plate capacitor). Neither is a tiny solid detector.

A phosphor, of course, can generate light flashes when hit by X-rays,
and photodiodes can be efficient at detecting the secondary radiation.

[plastco...@gmail.com]

For gas sensor :
4 keV / 30 = 133ē
For scintillator - SiPM !

http://ixbt.photo/?id=photo:1330019

[Lasse Langwadt Christensen]

years ago I did X-ray fluorescence spectrometry with an Fe55 source (5.9keV)
and a CCD as detector, image processed to find all pixels with empty neighbours

[tabb...@gmail.com]

eh?

[plastco...@gmail.com]

Best result OPA140+BF862 (L=1mH+1mH drain load):
http://ixbt.photo/?id=photo:1329969

Rf=91k and Rf=1M
http://ixbt.photo/?id=photo:1330021

Pushmi-Pullyu
dual accelerator - input JFET, output pnp BJT
DC precision
power supply +-3.8V
active probe - bootstrapped dual JFET half-bridge CPH6904 (out=in+0.86V)
(mediocre, better jfet+ADA4860 or FVF (Flipped Voltage Follower) with BF998)
http://ixbt.photo/?id=photo:1330027

Tortoise outrun Achilles !
time: 1.25us
Сircuit speed indifferent to input capacitance !
http://ixbt.photo/?id=photo:1330025

Сheetah enters the arena:
http://ixbt.photo/?id=photo:1330022
only 1M + 2200pF + OPA656
poor substitute, OPA140+jfet better

The final, a curtain.
OPA656 + BF862
http://ixbt.photo/?id=photo:1330023
GBW=31 GHz ? ((

[George Herold]

On Thursday, February 27, 2020 at 6:10:34 PM UTC-5, plastco...@gmail.com wrote:
> Best result OPA140+BF862 (L=1mH+1mH drain load):
> http://ixbt.photo/?id=photo:1329969

I'm confused about the two traces are here.
The BF862 is unfortunately no more...
You can only surf so long on the trailing edge of technology.

>
> Rf=91k and Rf=1M
> http://ixbt.photo/?id=photo:1330021

Is this for me? Hard to tell the shape of the response
for the 91k. But the 1M looks more like a two pole response
now anyway.

>
> Pushmi-Pullyu
> dual accelerator - input JFET, output pnp BJT
> DC precision
> power supply +-3.8V
> active probe - bootstrapped dual JFET half-bridge CPH6904 (out=in+0.86V)
> (mediocre, better jfet+ADA4860 or FVF (Flipped Voltage Follower) with BF998)
> http://ixbt.photo/?id=photo:1330027

So all the stuff on the left is an active probe?
To look at the inverting node without loading down with capacitance?
So is this right? The jfet is taking the fast part of the
current signal and sending it through to the non-inverting input?

I don't know what the pnp is doing on the output. Unless
it's a driver for some coax cable output?
It looks the the 'output' (your arrow between collector
of pnp and 1k R to -V) should be near -V... are you
now biasing PD via the output?

>
> Tortoise outrun Achilles !
> time: 1.25us
> Сircuit speed indifferent to input capacitance !
> http://ixbt.photo/?id=photo:1330025

Well at least to C on the inverting node.

>
> Сheetah enters the arena:
> http://ixbt.photo/?id=photo:1330022
> only 1M + 2200pF + OPA656
> poor substitute, OPA140+jfet better
>
> The final, a curtain.
> OPA656 + BF862
> http://ixbt.photo/?id=photo:1330023
> GBW=31 GHz ? ((

Huh the opa656 looks nice.. +/- 5V.
Thanks

Have you been lurking here long... I'm just reminded of
Jan Pantelje (sp) Which is a good thing. I like Jan.

George H.

[plastco...@gmail.com]

> Rf=91k and Rf=1M
> http://ixbt.photo/?id=photo:1330021
>Is this for me? Hard to tell the shape of the response
>for the 91k. But the 1M looks more like a two pole response
>now anyway.

With Rf=91k now Cf>0pF must be present

>
> Pushmi-Pullyu
> dual accelerator - input JFET, output pnp BJT
> DC precision
> power supply +-3.8V
> active probe - bootstrapped dual JFET half-bridge CPH6904 (out=in+0.86V)
> (mediocre, better jfet+ADA4860 or FVF (Flipped Voltage Follower) with BF998)
> http://ixbt.photo/?id=photo:1330027

>So all the stuff on the left is an active probe?
>To look at the inverting node without loading down with capacitance?

Yes, the circuit is very sensitive to the probe capacitance.
Non-inverting node !

>So is this right? The jfet is taking the fast part of the
>current signal and sending it through to the non-inverting input?

JFET is input AC amplifier with feedback through Op Amp TIA))

>I don't know what the pnp is doing on the output. Unless
>it's a driver for some coax cable output?
>It looks the the 'output' (your arrow between collector
>of pnp and 1k R to -V) should be near -V... are you
>now biasing PD via the output?

MMBTH81 is amplifier with gain~10
look in the upper right corner of the book AoE))
no photodiode offset !
slew rate only doubles((
overall gain increases

Capacitance SiPM 60035 = 3400pF
Dmitriy P.

[tabb...@gmail.com]

On Friday, 28 February 2020 00:44:11 UTC, George Herold wrote:
> On Thursday, February 27, 2020 at 6:10:34 PM UTC-5, plastco...@gmail.com wrote:

> > Best result OPA140+BF862 (L=1mH+1mH drain load):
> > http://ixbt.photo/?id=photo:1329969

> I'm confused about the two traces are here.
> The BF862 is unfortunately no more...
> You can only surf so long on the trailing edge of technology.

Russia is expert at that game.


NT

[Phil Hobbs]

I didn't mean that you thought it was the best one, but it's sure better
than an OPA111. ;)

I'm also partial to the OPA140, which is just about the perfect JFET op
amp: 0.5 pA typical Ibias at 25C, 5 nV 1-Hz flatband noise, 30 Hz 1/f
corner, 11 MHz BW, 220 uV max offset over temperature, 0.25 uV/K typical
drift, $2. Not horrible at all.

[plastco...@gmail.com]

I remembered where the zero offset of the photodiodes is still important !
Excellent mid-infrared photodiodes from Russia -
http://www.mirdog.spb.ru/products.htm
Ro = 5 Ω ... 10 k

Dmitriy P.

[George Herold]

OK got it, thanks. I'm so-so at single transistor design.

George H.

[plastco...@gmail.com]

Composite voltage follower
CFA ADA4860 + BF862
http://ixbt.photo/?id=photo:1330880

Output signals
yellow - circuit A
white - circuit B (bootstrapped)
http://ixbt.photo/?id=photo:1330881

Rise time circuit B (bootstrapped)
http://ixbt.photo/?id=photo:1330882

[George Herold]

On Saturday, March 7, 2020 at 3:15:17 AM UTC-5, plastco...@gmail.com wrote:
> Composite voltage follower
> CFA ADA4860 + BF862
> http://ixbt.photo/?id=photo:1330880

Huh? scratch scratch. Dmitriy, I have no idea what
you are doing? What is the input? (for one)
You'd have to describe what you are doing if you wanted some
sort of input/ critic/ ideas. Which would most likely come not
from me, I mostly just ask questions.

George H.

[plastco...@gmail.com]

very simple - imagine that a capacitor 2200pF is a photodiode with pulse photocurrent ~ 1 µA

[plastco...@gmail.com]

A unity-gain buffer removes the photodiode's parasitic capacitance :
circuit A - 2200pF >> 15pF
circuit B - 2200pF >> 0.8pF
Bootstrap photodiodes known since 1984, may have been used before.
The operation of the circuit "A" is limited by the capacitance gate-drain of the JFET transistor.

[plastco...@gmail.com]

Versatile composite amplifier configuration

https://www.researchgate.net/publication/271747584_Versatile_composite_amplifier_configuration

[plastco...@gmail.com]

"Shunt bootstrapping technique to improve bandwidth of transimpedance amplifiers"
https://www.researchgate.net/publication/3380589_Shunt_bootstrapping_technique_to_improve_bandwidth_of_transimpedance_amplifiers

fig.3 Output of bootstrap amplifier

my composite voltage follower -
http://ixbt.photo/?id=photo:1332241

output signal from ADA4860 with photocurrent 1.23µA
http://ixbt.photo/?id=photo:1332242

output signal from OPA656 with photocurrent 1.23µA
http://ixbt.photo/?id=photo:1332243

rise time 70ns, noise 150mVp-p
Channel 2 = output ADA4860

I wish you all good health.

[plastco...@gmail.com]

Composite amplifier EPHEMT + RGC + OPA656
Cdet=2200pF Rf=1M
ATF-54143 Igate= 18 nA
Bead on gate blm18bb050sn1d
http://ixbt.photo/?id=photo:1339025

Without e-phemt :
http://ixbt.photo/?id=photo:1339026

Rise time :
http://ixbt.photo/?id=photo:1339027
Dmitriy P.

[plastco...@gmail.com]

Retrospective "composite amplifier" challenge))
Question for experts and authors of the AoE3.
What is the function of capacitors 10nF between base and emitter MMBT6429 ?
https://electronix.ru/forum/uploads/monthly_06_2016/post-21169-1466350930.png
Why is there only one capacitor in the service manual between base and emitter ?
https://electronix.ru/forum/uploads/monthly_06_2016/post-21169-1466350949.png
Dmitriy P.

[Winfield Hill]

plastco...@gmail.com wrote...

You're right, my copy of the service manual
only shows one capacitor. The symmetrical
extra one came in, either when I made the
pencil drawing for the draftsman, or they
made a drafting error. As for the purpose
of the capacitors, we can only guess. It'd
be nice to be able to ask the HP engineer(s).

Did we point out in the text that those are
cascode transistors, whose only purpose is to
isolate the op-amp active nodes from the very
high capacitance of the huge IF3602 JFETs?
The Darlington is to insure all of the emitter
current ends up in the collector, and almost
none in the base.

The performance specs for their G=10,000
amplifier are pretty amazing. We have one of
those HP 34420A multimeters in my lab, but
it's rarely used, it's far too good! And the
special nanovolt input connector is awkward.


--
Thanks,
- Win

[Michael Terrell]

On Sunday, May 3, 2020 at 8:09:24 AM UTC-4, Winfield Hill wrote:
> plastcontrol wrote...

Some former HP engineers are on:

https://groups.io/g/HP-Agilent-Keysight-equipment

This replaced the old Yahoo HP group.

[Cursitor Doom]

On Sun, 03 May 2020 09:15:17 -0700, Michael Terrell wrote:

> Some former HP engineers are on:
>
> https://groups.io/g/HP-Agilent-Keysight-equipment

Yeah, but so is Gardner lately.

[jla...@highlandsniptechnology.com]

On Sun, 22 Mar 2020 02:17:10 -0700 (PDT), plastco...@gmail.com
wrote:

>"Shunt bootstrapping technique to improve bandwidth of transimpedance amplifiers"
>https://www.researchgate.net/publication/3380589_Shunt_bootstrapping_technique_to_improve_bandwidth_of_transimpedance_amplifiers
>
>fig.3 Output of bootstrap amplifier
>
>my composite voltage follower -
>http://ixbt.photo/?id=photo:1332241
>

>output signal from ADA4860 with photocurrent 1.23渙
>http://ixbt.photo/?id=photo:1332242
>
>output signal from OPA656 with photocurrent 1.23渙

>http://ixbt.photo/?id=photo:1332243
>
>rise time 70ns, noise 150mVp-p
>Channel 2 = output ADA4860
>
>I wish you all good health.

I have a new design, a little thing to keep busy with during the
shutdown. It's a GHz o/e converter. The trick is to use both ends of
the photodiode, use the current twice. One end goes into a fast
AC-coupled amp, and one end into a slow, precise DC gain path, and
combine them later.

--

John Larkin Highland Technology, Inc

Science teaches us to doubt.

Claude Bernard