Low drift, low noise, fast opamp

[Joerg]

Ok guys, normally I'd do this discrete and maybe I will again if there
ain't anything reasonable (<$5/1k) and available:

Photo diode receives analog signals modulated onto fiber. Modulation
frequency range up to 100MHz. The downside is that this time I must go
from full scale down to DC and that level must be accurate or at least
stable. A mV or two in offset would be ok but drift should be a lot
lower. It's going to be the usual current input thing.

What's you preference for a low noise amp in a situation like that?

TI's THS4xxx series looks good but still a bit high in terms of noise.
The ideal thing might be an AD603 without its input attenuator but these
don't come bare bones. Analog doesn't list noise in the paramatric table
which makes a search cumbersome and their photodiode design tool which
does include noise ends at 100kHz :-(

+/-12V supplies would be nice but it's all flexible at this point.

 

[Tom Bruhns]

Ok guys, normally I'd do this discrete and maybe I will again if there
ain't anything reasonable (<$5/1k) and available:

Photo diode receives analog signals modulated onto fiber. Modulation
frequency range up to 100MHz. The downside is that this time I must go
from full scale down to DC and that level must be accurate or at least
stable. A mV or two in offset would be ok but drift should be a lot
lower. It's going to be the usual current input thing.

What's you preference for a low noise amp in a situation like that?

TI's THS4xxx series looks good but still a bit high in terms of noise.
The ideal thing might be an AD603 without its input attenuator but these
don't come bare bones. Analog doesn't list noise in the paramatric table
which makes a search cumbersome and their photodiode design tool which
does include noise ends at 100kHz :-(

+/-12V supplies would be nice but it's all flexible at this point.

I haven't had to deal with high voltages like that for quite a long
time now...

OPA847 is a nice low-noise part. As is typical with high speed parts,
the noise 1/f corner is rather high, but it's still typically only
around 8nV/rtHz at 10Hz. It does need to run with a high noise gain
for stability, though. I suppose the 0.6mV input offset voltage and
1.5uV/C input offset drift, max over temperature, would do for you.
Figure 3 in the data sheet shows it connected to a 1pF photodiode. As
a bonus, it's also quite low distortion ('specially after I got them to
make a die change to fix a minor problem).

(Brian L., you owe me one for this...)

Cheers,
Tom

 

[Joerg]

Hello Tom,

I haven't had to deal with high voltages like that for quite a long
time now...

OPA847 is a nice low-noise part. As is typical with high speed parts,
the noise 1/f corner is rather high, but it's still typically only
around 8nV/rtHz at 10Hz. It does need to run with a high noise gain
for stability, though. I suppose the 0.6mV input offset voltage and
1.5uV/C input offset drift, max over temperature, would do for you.
Figure 3 in the data sheet shows it connected to a 1pF photodiode. As
a bonus, it's also quite low distortion ('specially after I got them to
make a die change to fix a minor problem).

Thanks. The OPA847 like a great candidate here. I'll have to run the
offset budget but it looks close to being acceptable. The 1/f doesn't
bother me much because noise isn't critical under a MHz or so in this
application.

(Brian L., you owe me one for this...)

Cool. I never had anyone changing the die for me except for my custom
chip designs. But there was an instant where a mfg had to correct a data
sheet after we found a gotcha.

 

[Joerg]

Hello Tom,


P.S.: The only crimp in the OPA847 is the min stable gain of 12. That's
too much in my case.

 

[Reinhard Zwirner]

Ok guys, normally I'd do this discrete and maybe I will again if there
ain't anything reasonable (<$5/1k) and available:

Photo diode receives analog signals modulated onto fiber. Modulation
frequency range up to 100MHz. The downside is that this time I must go
from full scale down to DC and that level must be accurate or at least
stable. A mV or two in offset would be ok but drift should be a lot
lower. It's going to be the usual current input thing.

What's you preference for a low noise amp in a situation like that?

Hi Joerg

Why don't you use a wideband op amp together with a precision op amp
providing the dc performance? There are a lot of app notes available
with such a dc servo loop.

HTH

Reinhard

 

[Joerg]

Hello Reinhard,

Why don't you use a wideband op amp together with a precision op amp
providing the dc performance? There are a lot of app notes available
with such a dc servo loop.

Sewing them together later might be a challenge but that is certainly an
idea to keep in mind, especially if I can't get the desired noise
performance. Thank you.

 

[Tom Bruhns]

Hello Tom,


P.S.: The only crimp in the OPA847 is the min stable gain of 12. That's
too much in my case.

Well, I did warn you about that. ;-) But be sure to look at the whole
data sheet; there are some compensation tricks in there. Also, check
out others in the family. They are more noisy, but not outrageously
so. OPA84x...

Curious that you are so worried about input offset drift, but won't
have room for 20dB of gain...

Cheers,
Tom

 

[Joerg]

Hello Tom,

Well, I did warn you about that. ;-) But be sure to look at the whole
data sheet; there are some compensation tricks in there. Also, check
out others in the family. They are more noisy, but not outrageously
so. OPA84x...

I have done some compensation tricks with the old HA2540 amps. Hey,
anyone remember these? From the days when electrity cost next to nothing
and every chip got freaking hot?

Anyhow, at 100MHz compensation tricks can become iffy.

Curious that you are so worried about input offset drift, but won't
have room for 20dB of gain...

Well, I need precision at the low end of the spectrum but can't have too
much gain up in the VHF region because things would saturate and
intermodulate. The DC level needs to be measured very accurately even
though the actual amplitude coming in can be high.

OTOH I might just split the band up right at the photodiode. Kind of an
AC path and a DC path from there. This might even be easier than the
servo loop Reinhard had suggested. Have done it before but never with a
photodiode.

 

[John Larkin]

Ok guys, normally I'd do this discrete and maybe I will again if there
ain't anything reasonable (<$5/1k) and available:

Photo diode receives analog signals modulated onto fiber. Modulation
frequency range up to 100MHz. The downside is that this time I must go
from full scale down to DC and that level must be accurate or at least
stable. A mV or two in offset would be ok but drift should be a lot
lower. It's going to be the usual current input thing.

What's you preference for a low noise amp in a situation like that?

TI's THS4xxx series looks good but still a bit high in terms of noise.
The ideal thing might be an AD603 without its input attenuator but these
don't come bare bones. Analog doesn't list noise in the paramatric table
which makes a search cumbersome and their photodiode design tool which
does include noise ends at 100kHz :-(

+/-12V supplies would be nice but it's all flexible at this point.

What's the photodiode current level like? And how much equivalent
input noise current can you tolerate?

I like to use really fast (400 MHz and up) current-mode opamps for
DC-coupled photodiode tia's, but their noise performance probably
isn't ideal. I usually have a lot of light to work with, hundreds of
uW, whereas the telecom tia's are intended to work in the low uW
range. Most telecom tia's have AGC, which wrecks them for DC
applications.

John

 

[Joerg]

Hello John,

What's the photodiode current level like? And how much equivalent
input noise current can you tolerate?

That's still in the air at this point. Most likely a few mA max. The
noise current needs to be as low as possible because this is an analog
measuring setup.

I like to use really fast (400 MHz and up) current-mode opamps for
DC-coupled photodiode tia's, but their noise performance probably
isn't ideal. I usually have a lot of light to work with, hundreds of
uW, whereas the telecom tia's are intended to work in the low uW
range. Most telecom tia's have AGC, which wrecks them for DC
applications.

Today I am going to design the PD front end. Looks like I am going to
split it into a sub-MHZ and an RF section. The OPA847 Tom suggested
looks like a good chip for the RF part and there I'd also need more gain
so it's x12 limit won't hurt. For the sub-MHz range the AD8628 appears
to be a nice chip although I'd have to set the crossover frequency a lot
lower than a MHz.

We also have a lot of light to work with (thousands of uW...) but must
detect extremely minute changes.

Then comes a whole 'nother little problem. This client doesn't do
electronics much so they have no tech who could build a neat lab
prototype on a large Veroboard with ground plane. I can do it but that
wouldn't be cost effective. Oh well, got to cross that bridge in a
couple of weeks.

 

[John Larkin]

Hello John,


That's still in the air at this point. Most likely a few mA max. The
noise current needs to be as low as possible because this is an analog
measuring setup.



Today I am going to design the PD front end. Looks like I am going to
split it into a sub-MHZ and an RF section. The OPA847 Tom suggested
looks like a good chip for the RF part and there I'd also need more gain
so it's x12 limit won't hurt. For the sub-MHz range the AD8628 appears
to be a nice chip although I'd have to set the crossover frequency a lot
lower than a MHz.

We also have a lot of light to work with (thousands of uW...) but must
detect extremely minute changes.

Then comes a whole 'nother little problem. This client doesn't do
electronics much so they have no tech who could build a neat lab
prototype on a large Veroboard with ground plane. I can do it but that
wouldn't be cost effective. Oh well, got to cross that bridge in a
couple of weeks.

I could let you use one of my J730 boards for experimenting...

http://www.highlandtechnology.com/DSS/J720-730DS.html

This has a DC-180 MHz tia/amp chain, currently outputting 1 v/mW at
the analog output connector, using a fairly pokey silicon PIN diode.
It would be easy to replace the silicon pin with something else; I
recently spun a 1310 nm version with a GaAs pin diode, just dropped it
in and changed one gain-set resistor. Email me if interested.

John

 

[Joerg]

Hello John,

I could let you use one of my J730 boards for experimenting...

http://www.highlandtechnology.com/DSS/J720-730DS.html

This has a DC-180 MHz tia/amp chain, currently outputting 1 v/mW at
the analog output connector, using a fairly pokey silicon PIN diode.
It would be easy to replace the silicon pin with something else; I
recently spun a 1310 nm version with a GaAs pin diode, just dropped it
in and changed one gain-set resistor. Email me if interested.

Thanks for the offer. That is very kind. However, this app is analog.
Got a discrete solution on paper, still have to compare it with the
OPA847 from a noise point of view. But my wife just put our favorite
casserole in the oven and I can almost smell it so I don't know if I'll
get that done tonight.

The BFR505 looks really nice for the discrete fiber amp and is fast
enough for this case. Of course these are European transistors and that
leaves a bit of worry. Got a black I with EU semiconductors before when
they sudenly became unobtanium in the US for the prototype phase.
Somehow my favorite hotrod RF BJT from the days of Motorola have
vanished. They offered stunning noise figures.

 

[Joerg]

Hello John,

I could let you use one of my J730 boards for experimenting...

http://www.highlandtechnology.com/DSS/J720-730DS.html

This has a DC-180 MHz tia/amp chain, currently outputting 1 v/mW at
the analog output connector, using a fairly pokey silicon PIN diode.
It would be easy to replace the silicon pin with something else; I
recently spun a 1310 nm version with a GaAs pin diode, just dropped it
in and changed one gain-set resistor. Email me if interested.

Just re-read your post and the spec and realized it does provide analog
out. Unless it's a trade secret which device did you use in the TIA?

 

[John Larkin]

Hello John,


Just re-read your post and the spec and realized it does provide analog
out. Unless it's a trade secret which device did you use in the TIA?

AD8014, current-mode opamp.

John

 

[Ian]

Hello Tom,


I have done some compensation tricks with the old HA2540 amps. Hey, anyone
remember these? From the days when electrity cost next to nothing and
every chip got freaking hot?

Anyhow, at 100MHz compensation tricks can become iffy.



Well, I need precision at the low end of the spectrum but can't have too
much gain up in the VHF region because things would saturate and
intermodulate. The DC level needs to be measured very accurately even
though the actual amplitude coming in can be high.

Boy, do I remember those Harris amps, and the way they did a die shrink
every couple of years which made everything go unstable.....

Another candidate might be the AD8099.

Regards
Ian

 

[colin]

Ok guys, normally I'd do this discrete and maybe I will again if there
ain't anything reasonable (<$5/1k) and available:

Photo diode receives analog signals modulated onto fiber. Modulation
frequency range up to 100MHz. The downside is that this time I must go
from full scale down to DC and that level must be accurate or at least
stable. A mV or two in offset would be ok but drift should be a lot
lower. It's going to be the usual current input thing.

What's you preference for a low noise amp in a situation like that?

TI's THS4xxx series looks good but still a bit high in terms of noise.
The ideal thing might be an AD603 without its input attenuator but these
don't come bare bones. Analog doesn't list noise in the paramatric table
which makes a search cumbersome and their photodiode design tool which
does include noise ends at 100kHz :-(

+/-12V supplies would be nice but it's all flexible at this point.

Is this 100mhz just a carrier wave,
or is the light itself just modulated upto 100mhz ?
If not could you modulate it with a higher frequency ?

What sort of current do you get out of the PD?
for real low drift and low freq noise floor maybe you need to use a zero
drift op amp as a servo,
then you have wider choice for the main device.

Jfets types look like they promise good speed/noise.
or phemt would be good, if/when they make them.

Colin =^.^=

 

[Joerg]

Hello John,

AD8014, current-mode opamp.

Maybe you could try out the lower noise OPA847 that Tom suggested if the
board in there ever needs a re-spin. In the past I always found discrete
solutions better in terms of noise but this was mostly for all kinds of
transducers and not photodiodes.

BTW, it's interesting that you use kilometers in the advertising text.
The only other places I saw doing that in the US were NASA and the IEEE.

 

[Joerg]

Hello Ian,

Boy, do I remember those Harris amps, and the way they did a die shrink
every couple of years which made everything go unstable.....

Another candidate might be the AD8099.

Thanks. At around 1nv/rtHz that one comes quite close to what's possible
with RF transistors. I'll give the AD8099 a good look.

 

[Joerg]

Hello Colin,

Is this 100mhz just a carrier wave,
or is the light itself just modulated upto 100mhz ?

It's light modulated at up to 100MHz. Plain old AM style.

If not could you modulate it with a higher frequency ?

Well, that's not really my call. I just have to make the receive end
work as good as possible.

What sort of current do you get out of the PD?

Probably around a mA or so. The challenge is to detect wee changes in that.

for real low drift and low freq noise floor maybe you need to use a zero
drift op amp as a servo,
then you have wider choice for the main device.

Drift isn't an issue by now anymore since I already split out the lower
MHz range down to DC. That part works ok (on paper...).

Jfets types look like they promise good speed/noise.
or phemt would be good, if/when they make them.

I am also looking at a discrete amp for the higher frequency portion.
That would help the design longevity because with chips there is always
a chance that the last order bell clangeth one day.

 

[John Larkin]

Hello John,


Maybe you could try out the lower noise OPA847 that Tom suggested if the
board in there ever needs a re-spin. In the past I always found discrete
solutions better in terms of noise but this was mostly for all kinds of
transducers and not photodiodes.

I'd like to do a faster version some day, so a better amp would help.
Noise isn't a big deal in my apps, as I usually have a lot of
fiber-coupled light... my lasers easily launch a milliwatt. Maybe a
common-base transistor stage would decouple the pd capacitance from
the opamp input, but by the time you put all that junk in the emitter,
it's gonna want to oscillate, and by the time you damp that the bw may
go away. Life sucks.

BTW, it's interesting that you use kilometers in the advertising text.
The only other places I saw doing that in the US were NASA and the IEEE.

We use SI units for anything that matters. Fiber attenuation is always
spec'd in dB/km and dispersion in ps/km. Dumb stuff, like box dims and
patch cable lengths, we still tend to do in inches and feet.

Gotta go make bread pudding. Popular demand.


John