Basic advance question - design of photodiode amplifiers

[Charlie E.]

Hi All,
Just posing a little question on the design of photodiode amplifiers.
It is interesting how, so often, when you look into doing something
you haven't done before, you run into interesting bits of 'common
knowledge' that you don't know, and have a hard time finding!

In this case, I am designing a simple color sensor. I originally was
going to cook book it with a pre-made sensor IC that I2C communicated,
but ran into two months of programming difficulties, non-documented
features and bugs, and have finally decided to 'simplify' things, and
just roll my own.

Now, National has Sensor Webench, that will basically let you take a
sensor, and they will build the amplifiers you need for you.
Interesting, but you don't LEARN anything. Also, if you need a sensor
not on their 'list,' you are pretty muich SOL. I need a visible light
photodiode, and I didn't find one that I liked on their list.

So, I decided to try rolling my own, and looking at both their
examples, and in AoE, I see one interesting thing - a negative bias on
the photodiode. So, my first question - how is this negative bias
usually generated? And at what voltage? It is just 'assumed' that
you know these things already! I am trying to build a very portable
device, and wanted to use a single 3.3V supply for everything.

Thanks for any advice and good sources for more research.

Charlie

 

[Tim Williams]

So, I decided to try rolling my own, and looking at both their
examples, and in AoE, I see one interesting thing - a negative bias on
the photodiode.  So, my first question - how is this negative bias
usually generated?  And at what voltage?  It is just 'assumed' that
you know these things already!  I am trying to build a very portable
device, and wanted to use a single 3.3V supply for everything.  

Well, part common knowledge, part can be inferred. For instance, from
the physical phenomena at work: photons generate electron-hole pairs.
They're just kind of loose in the silicon, wandering around randomly.
Some may recombine, generating a photon (in suitable semiconductors,
e.g. GaAs, which is to say, LEDs and photodiodes are the same damn
thing, which is true to a degree) or heat (silicon doesn't produce
photons very well), but most are retained in their respective region
(N or P). This action generates a photocurrent (at Vf = 0V), with
some maximum voltage behind it (If = 0 at whatever Vf, maybe 0.5V, the
current being returned through recombination or whatever).

Now, even if the material is nonlinear (which it is, being a diode
junction), it stands to reason that you'll get more current at 0V than
at 0.5V. It then stands to reason that, if you put a back bias on the
thing, you can get even more current out. Physically, you can imagine
applying an electric field, it pulls the electrons and holes apart,
minimizing recombination and therefore maximising photocurrent. There
is also a secondary effect, when the electric field is strong enough,
electrons or holes will be accelerated to ionization potential,
resulting in inelastic collisions, knocking out more electrons and
holes, resulting in even more current. This is how very high gain
photodiodes are operated.

There is one more advantage to back bias: junctions capacitance drops
dramatically. Even if you aren't doing it to maximize photocurrent or
generate avalanche currents, you can still improve speed noticably for
this reason.

3.3V might be enough (although you'll probably manage more like 1.5V,
needing some headroom for op-amps?), but the best way to tell is to
read datasheets. If nothing else, you can put in a charge pump
(minding that it may be noisy!).

Tim

 

[GregS]

Well, part common knowledge, part can be inferred. For instance, from
the physical phenomena at work: photons generate electron-hole pairs.
They're just kind of loose in the silicon, wandering around randomly.
Some may recombine, generating a photon (in suitable semiconductors,
e.g. GaAs, which is to say, LEDs and photodiodes are the same damn
thing, which is true to a degree) or heat (silicon doesn't produce
photons very well), but most are retained in their respective region
(N or P). This action generates a photocurrent (at Vf =3D 0V), with
some maximum voltage behind it (If =3D 0 at whatever Vf, maybe 0.5V, the
current being returned through recombination or whatever).

Now, even if the material is nonlinear (which it is, being a diode
junction), it stands to reason that you'll get more current at 0V than
at 0.5V. It then stands to reason that, if you put a back bias on the
thing, you can get even more current out. Physically, you can imagine
applying an electric field, it pulls the electrons and holes apart,
minimizing recombination and therefore maximising photocurrent. There
is also a secondary effect, when the electric field is strong enough,
electrons or holes will be accelerated to ionization potential,
resulting in inelastic collisions, knocking out more electrons and
holes, resulting in even more current. This is how very high gain
photodiodes are operated.

There is one more advantage to back bias: junctions capacitance drops
dramatically. Even if you aren't doing it to maximize photocurrent or
generate avalanche currents, you can still improve speed noticably for
this reason.

I have never tried to bias for more sensitivity. I should. If the bias voltage has noise at the
recorded frequency, your liable to run into more problems. I also don't know how
biasing affects overall s/n ration assuming the bias is noise free.

greg

 

[John Devereux]

Hi All,
Just posing a little question on the design of photodiode amplifiers.
It is interesting how, so often, when you look into doing something
you haven't done before, you run into interesting bits of 'common
knowledge' that you don't know, and have a hard time finding!

In this case, I am designing a simple color sensor. I originally was
going to cook book it with a pre-made sensor IC that I2C communicated,
but ran into two months of programming difficulties, non-documented
features and bugs, and have finally decided to 'simplify' things, and
just roll my own.

Now, National has Sensor Webench, that will basically let you take a
sensor, and they will build the amplifiers you need for you.
Interesting, but you don't LEARN anything. Also, if you need a sensor
not on their 'list,' you are pretty muich SOL. I need a visible light
photodiode, and I didn't find one that I liked on their list.

So, I decided to try rolling my own, and looking at both their
examples, and in AoE, I see one interesting thing - a negative bias on
the photodiode. So, my first question - how is this negative bias
usually generated? And at what voltage? It is just 'assumed' that
you know these things already! I am trying to build a very portable
device, and wanted to use a single 3.3V supply for everything.

Thanks for any advice and good sources for more research.

The negative bias is to reduce the capacitance of the photodiode. You
only need it if you are trying for a high speed circuit. You may well be
able to use a single supply configuration. It depends on the photodiode
and the speed required.

 

[Ecnerwal]

The negative bias is to reduce the capacitance of the photodiode. You
only need it if you are trying for a high speed circuit. You may well be
able to use a single supply configuration. It depends on the photodiode
and the speed required.

....and most photodiode circuits tend to be in the "high speed" category.
Otherwise a phototransistor is often a better choice - except where a
very non-rohs compliant CdS resistive sensor is better...photodiodes are
usually only chosen when speed is of the essence (in my limited
experience - we used them for looking at laser pulse risetimes...)

 

[Charlie E.]

http://www.national.com/onlineseminar/2004/photodiode/photodiode.html

Download the presenttion.

Bob

Hi Bob,
Thanks! I found that one this morning. Nice general data, but no
math...

Charlie

 

[Charlie E.]

Full of misinformation, unfortunately. They perpetuate the silly notion
that there's nothing you can do about the eN*Cd*Rf noise peak, whereas
changing the circuit topology can help a lot. See
http://electrooptical.net/www/frontends/frontends.pdf.

Cheers,

Phil Hobbs

Hi Phil,
Thanks, looks good, but is overkill for my application. Only going to
need to do a few ADCs (maybe 6-8) in 500ms, so don't need it fast.
Need it cheap and accurate! ;-)

Charlie

 

[MooseFET]

Hi All,
Just posing a little question on the design of photodiode amplifiers.
It is interesting how, so often, when you look into doing something
you haven't done before, you run into interesting bits of 'common
knowledge' that you don't know, and have a hard time finding!

In this case, I am designing a simple color sensor.  I originally was
going to cook book it with a pre-made sensor IC that I2C communicated,
but ran into two months of programming difficulties, non-documented
features and bugs, and have finally decided to 'simplify' things, and
just roll my own.

Now, National has Sensor Webench, that will basically let you take a
sensor, and they will build the amplifiers you need for you.
Interesting, but you don't LEARN anything.  Also, if you need a sensor
not on their 'list,' you are pretty muich SOL.  I need a visible light
photodiode, and I didn't find one that I liked on their list.

So, I decided to try rolling my own, and looking at both their
examples, and in AoE, I see one interesting thing - a negative bias on
the photodiode.  So, my first question - how is this negative bias
usually generated?  And at what voltage?  It is just 'assumed' that
you know these things already!  I am trying to build a very portable
device, and wanted to use a single 3.3V supply for everything.  

3V is enough bias to hugely improve the capacitance of the
photodiode. You didn't say a lot about the speed the sensor needs to
work at.

www.interfet.com makes some giant JFETs for making photopreamps on
large photodiodes. Unfortunately you need a couple of car batteries
to provide the Idss of them

You may only need a fairly simple design if the bandwidth is narrow.
You pick the op-amp on the basis of its input noise voltage and input
noise current. Match the v(noise)/I(noise) to the Xc of the
photodiode at your working frequency.

 

[John Devereux]

...and most photodiode circuits tend to be in the "high speed" category.
Otherwise a phototransistor is often a better choice - except where a
very non-rohs compliant CdS resistive sensor is better...photodiodes are
usually only chosen when speed is of the essence (in my limited
experience - we used them for looking at laser pulse risetimes...)

I disagree here, there are plenty of slower applications where you need
photodiodes rather than phototransistors. In fact I don't think I've
ever used a phototransistor in a real design, do people still use them?
Last one I had was ~30 years ago as a kid.

Photodiodes can give you large active areas when needed and are
extremely linear. They can still be pretty fast - even unbiased - if run
into e.g. an inverting opamp circuit to cancel the capacitance.

 

[[email protected]]

Phototransistors...eeeeeewwwwwwwww.  Friends don't let friends, and all
that.

They're horrible.  Okay for night lights and slow IRDA, maybe, but
always a very poor choice for anything requiring decent SNR, due to (a)
small active area, (b) crummy packaging so you can't concentrate the
light on the small active area, (c) poor linearity, (d) poorly
controlled gain, (e) excessive noise.

Photo transistiors are actually quite nice.
You will find them in optocouplers for example.
It al depends on the requirements.
Manufacturers optocouplers seem to have no problem making
the light coupling, and, as the world is digital now, who cares
about linearity. And noise ;-)
Lots of gain is what we need, saturation.. so who
cares if it varies a bit.

 

[Jan Panteltje]

Okay, if you're building night lights or the moral equivalent thereof,
go for it. Confusing high output with high sensitivity is unfortunately
one of the most common mistakes in electro-optics. Assuming that noise
floors are flat is another.

Cheers,

Phil Hobbs

Silly, 'nightlights' seems to be on your mind all the time.
Have you any idea what optocouplers (and photo transistors) are used for?

Dr Philip C D Hobbs (Just for you, Jan)

Thank you, unfortunately it seems that anybody who looked at one aspect
of something and writes some paper on it, can be Dr.
Well OK that is the way it is.
Slow IRDA?
http://scitation.aip.org/getabs/ser...00079000006000773000001&idtype=cvips&gifs=yes
Maye you should google 'fast phototransistor ns'

 

[John Devereux]

[...]

Somewhat more than a paper and somewhat more than one aspect:

http://www.amazon.com/Building-Elec...=sr_1_1?ie=UTF8&s=books&qid=1240586529&sr=1-1

If you're doing anything electro-optical and non-trivial, you need
this book.

Seconded.

There is also a lot of good general analog and low noise electronics
stuff in there, as well as "electro-optical". It has an unusual
combination of hands-on practical advice, optical and electronic theory,
with advanced (to me anyway) mathematical techniques.

 

[Rich Webb]

I've been having terrible trouble with the index--they stopped accepting
WordPerfect, so I had to change to LaTeX because I wasn't going to use
Word at any price, and their macro sets make it really hard to get
accurate pagination for the index. I still owe it to them--probably get
done this coming week, so the second edition will be out in June or July.

WordPerfect, indeed - gets me feeling all Charlton Heston-ish; "Cold,
dead hands" an all that. Word is for the HR types ...

New edition, it's a race between you and Win then is it? ;-)

 

[Tim Williams]

WordPerfect, indeed - gets me feeling all Charlton Heston-ish; "Cold,
dead hands" an all that. Word is for the HR types ...

New edition, it's a race between you and Win then is it?   ;-)

Ah, but Win's had two editions under his belt for almost three
decades. Phil will have to roll out a quick third edition sometime
between June and "When It's Done" to catch up. Which I guess means
he's in no rush at all.

Tim