Photodiode Amplifier

[Guy Macon]

I am working on a circuit to detect a very faint spot of IR
light on an X-Y PSD photodiode. I can make the spot as
bright as I want, but brighter equals more IR LEDs and a higher
cost. Response time is not an major issue; I would like to
modulate my source in the 1khz to 10khz range, but will go
much lower if needed.

I made two test PWBs with short leads, ground plane,
and guard rings, using two opamp stages as follows:

My first stage is an opamp in transimpedence (current to
voltage) mode. "+" is gnd, and "-" has a photodiode to ground
and a large feedback R and small feedback C from the opamp output.
(I have no bias on the photodiode now, but will be experimenting
with using a small (0-50mV) bias later).

My second stage is a standard inverting opamp voltage gain stage.

One of my PWBs uses single opamps (+2, -3, out6, +vcc7, -vcc4)
and the other uses a quad (+1/7/8/14, -2/6/9/14, out1/7/8/14,
+vcc4, -vcc7).

I am running tests with the best opamps I know of and with the
cheapest opamp I know of. I am looking for the best tradeoff
between receiver cost and transmitter cost.

For the singles, I am using a Linear LT1792 for the first
stage and a LT1793 for the second. I chose them because
the LT1792 has (0.8fA/sqrtHz) of current noise and the LT1793
has (4.2nV/sqrtHz) of voltage noise. For the quad I am using
an LM324 because I can get them for six cents each.

That pretty much covers the high-performance end and the
cheap end. Now I need to try a few opamps that are between
the two extremes. Can anyone recommend a few parts that
have a good tradeoff between low current noise, reasonable
voltage noise, low cost, and which match one of the two
pinouts I am using?

 

[Guy Macon]

a TL080 is a good op amp, cheap.
its a JFet input with bipola out with
lots of gain over the convensional types.
TL080 = single op, TL082 a dual type.
and TL084 = quodd...

Thanks! Advice much appeciated.

 

[John Larkin]

I am working on a circuit to detect a very faint spot of IR
light on an X-Y PSD photodiode. I can make the spot as
bright as I want, but brighter equals more IR LEDs and a higher
cost. Response time is not an major issue; I would like to
modulate my source in the 1khz to 10khz range, but will go
much lower if needed.

I made two test PWBs with short leads, ground plane,
and guard rings, using two opamp stages as follows:

My first stage is an opamp in transimpedence (current to
voltage) mode. "+" is gnd, and "-" has a photodiode to ground
and a large feedback R and small feedback C from the opamp output.
(I have no bias on the photodiode now, but will be experimenting
with using a small (0-50mV) bias later).

My second stage is a standard inverting opamp voltage gain stage.

One of my PWBs uses single opamps (+2, -3, out6, +vcc7, -vcc4)
and the other uses a quad (+1/7/8/14, -2/6/9/14, out1/7/8/14,
+vcc4, -vcc7).

I am running tests with the best opamps I know of and with the
cheapest opamp I know of. I am looking for the best tradeoff
between receiver cost and transmitter cost.

For the singles, I am using a Linear LT1792 for the first
stage and a LT1793 for the second. I chose them because
the LT1792 has (0.8fA/sqrtHz) of current noise and the LT1793
has (4.2nV/sqrtHz) of voltage noise. For the quad I am using
an LM324 because I can get them for six cents each.

That pretty much covers the high-performance end and the
cheap end. Now I need to try a few opamps that are between
the two extremes. Can anyone recommend a few parts that
have a good tradeoff between low current noise, reasonable
voltage noise, low cost, and which match one of the two
pinouts I am using?

LM324 is maybe the worst opamp still in production (anybody know of a
worse one?). Lots of crossover distortion and stuff.

LF347 quad jfet amps are nice... fast and very low current noise. I
show 34 cents as our inventory price.

An LM1458 (basicly a dual 741) is very cheap.

John

 

[Jamie]

a TL080 is a good op amp, cheap.
its a JFet input with bipola out with
lots of gain over the convensional types.
TL080 = single op, TL082 a dual type.
and TL084 = quodd...

 

[Rene Tschaggelar]

I am working on a circuit to detect a very faint spot of IR
light on an X-Y PSD photodiode. I can make the spot as
bright as I want, but brighter equals more IR LEDs and a higher
cost. Response time is not an major issue; I would like to
modulate my source in the 1khz to 10khz range, but will go
much lower if needed.

Modulating the light is an excellent idea.
Do a lock-in then. A few AC coupled amplifiers
with sufficient high gain plus a narrowband
filter. I can recommend the LT1128, with 2 stages
you can get a gain of 10k..100k (80..100dB)

Rene

 

[Guy Macon]

LM324 is maybe the worst opamp still in production (anybody know of a
worse one?). Lots of crossover distortion and stuff.

Worst for what purpose? Low-light / low-speed pPhotodiode amplifiers,
to take one particularly germane example, never cross over, and thus
don't care about crossover distortion.

LF347 quad jfet amps are nice... fast and very low current noise. I
show 34 cents as our inventory price.

That's a good one from a price/performance standpoint. Thanks!
The cheapest LM324D costs 6.7 cents (Digikey) while the cheapest
LF347 is 19.6 cents. My current project will be using 500,000
quad opamps per year (or two million singles) so that extra 12.9
cents will cost us $64,500 (12.9 cents times 20 opamps per unit
times 100,000 units). (I can get both parts cheaper, but the
difference won't change much.) That's not a bad price if it
turns out that the LF347 allows me to take a dollar of the cost
of my 100,000 light emitter assemblies.

You can see by these numbers why I care so much about price vs.
performance. At it's peak, my largest quantity design was selling
well over 100,000 units per day at a retail price of $24.99.

 

[Guy Macon]

Modulating the light is an excellent idea.

It's essential when dealing with low level of light from
the emitter and high levels of DC and 120Hz modulated light
from the environment. Another thing that helps a lot is a
narrowband optical filter at the color of the emitter.

A few AC coupled amplifiers with sufficient high gain
plus a narrowband filter.

I am doing that. I can do a good analog filter for a
dollar or two, depending on how good a filter I want
to make, or I can use a four dollar DSP and do a FFT
filter, which will allow me to have a cheaper emitter.
I am looking into both options, and will choose whichever
gives me the lowest total cost.

I can recommend the LT1128, with 2 stages
you can get a gain of 10k..100k (80..100dB)

The LT1128 is a fine part, but at roughly $4 each times
four opamps (using the LT1128 for the critical first
stages only and cheaper opamps for the other 16) times
100,000 units it will cost me about a million an a half
dollars to get (4.7pA/sqrtHz) of current noise. The
LT1792 has (0.8fA/sqrtHz) and is a buck or two cheaper
per part.

 

[Rene Tschaggelar]

It's essential when dealing with low level of light from
the emitter and high levels of DC and 120Hz modulated light
from the environment. Another thing that helps a lot is a
narrowband optical filter at the color of the emitter.

The frequency could be higher, and the filters become
more manageable in terms of capacitor size.

I am doing that. I can do a good analog filter for a
dollar or two, depending on how good a filter I want
to make, or I can use a four dollar DSP and do a FFT
filter, which will allow me to have a cheaper emitter.
I am looking into both options, and will choose whichever
gives me the lowest total cost.

The two are not comparable.
While you can have an analog filter operating at
signal levels below 1mV, that is not the case for a
digital filter.

The LT1128 is a fine part, but at roughly $4 each times
four opamps (using the LT1128 for the critical first
stages only and cheaper opamps for the other 16) times
100,000 units it will cost me about a million an a half
dollars to get (4.7pA/sqrtHz) of current noise. The
LT1792 has (0.8fA/sqrtHz) and is a buck or two cheaper
per part.

Ok, current noise appears to be the spec then. I was
more thinking in terms of voltage noise.

Rene

 

[Guy Macon]

(Does anyone else have a favorite quad opamp? I need to pick
another 4 or 5 from the near-infinite choices, and I figure I
might as well start with ones that others here like. My main
tradeoff is cost vs. current noise, but I will look into any
part that anyone suggests.)

The frequency could be higher, and the filters become
more manageable in terms of capacitor size.

The limiting factor is the response time of the photodiode.
Low light levels call for large area photodiodes (slower)
running in photovoltaic mode (slower). This puts your upper
modulation limit in the 1khz to 10khz range. Within that
range you want to be as far above the 120hz from fluorescent
lights as possible.

I am not sure that I understand your comment about capacitor
size. You can make any kind of opamp filter you want in
the 100hz to 10khz range with capacitors in the 0.01uF to
0.1uF range and still have reasonable resistor values.

The two are not comparable.
While you can have an analog filter operating at
signal levels below 1mV, that is not the case for a
digital filter.

I know of very few applications where a 1mV signal stays at 1mV.
Usually it gets amplified to a reasonable level. In my design,
it ends up in an ADC on a microcontroller or DSP. You still
need minimal analog filtering with the DSP approach, of course;
you need a low pass RC for antialiasing, and a high pass RC
filter insures that the later stages don't saturate because
of the low-level 1-10khz AC modulation signal riding on the
high level DC and 120hz background signal. The tradeoff is
in what I do after the signal is goes through the minimal HPF
and LPF and is amplified to a few volts; do I run the signal
through a multipole analog bandpass filter or do I run the
signal through a digital filter using a DSP? I can do a good
analog filter for a dollar or two, depending on how good a
filter I want to make, or I can use a four dollar DSP and do
a FFT filter, which will allow me to have a cheaper (less
bright) emitter. I am looking into both options, and will
choose whichever gives me the lowest total cost.

 

[John Larkin]

(Does anyone else have a favorite quad opamp? I need to pick
another 4 or 5 from the near-infinite choices, and I figure I
might as well start with ones that others here like. My main
tradeoff is cost vs. current noise, but I will look into any
part that anyone suggests.)



The limiting factor is the response time of the photodiode.
Low light levels call for large area photodiodes (slower)
running in photovoltaic mode (slower). This puts your upper
modulation limit in the 1khz to 10khz range. Within that
range you want to be as far above the 120hz from fluorescent
lights as possible.

I am not sure that I understand your comment about capacitor
size. You can make any kind of opamp filter you want in
the 100hz to 10khz range with capacitors in the 0.01uF to
0.1uF range and still have reasonable resistor values.


I know of very few applications where a 1mV signal stays at 1mV.
Usually it gets amplified to a reasonable level. In my design,
it ends up in an ADC on a microcontroller or DSP. You still
need minimal analog filtering with the DSP approach, of course;
you need a low pass RC for antialiasing, and a high pass RC
filter insures that the later stages don't saturate because
of the low-level 1-10khz AC modulation signal riding on the
high level DC and 120hz background signal. The tradeoff is
in what I do after the signal is goes through the minimal HPF
and LPF and is amplified to a few volts; do I run the signal
through a multipole analog bandpass filter or do I run the
signal through a digital filter using a DSP? I can do a good
analog filter for a dollar or two, depending on how good a
filter I want to make, or I can use a four dollar DSP and do
a FFT filter, which will allow me to have a cheaper (less
bright) emitter. I am looking into both options, and will
choose whichever gives me the lowest total cost.

Will this thing have an ADC and a uP to control it? If so, why not
just turn the light source on/off under software control and digitize
the optical sigs, making a synchronous lockin amplifier out of the
whole thing? The on/off cycles wouldn't even need to be strictly
periodic... you could, accidentally or by intent, wind up with a
synchronously detected pseudo-random-period-spread-spectrum lockin
thing. You can do some amazing things with one of those 75-cent uPs
with built-in adc/mux. You only have to pay for code once.

John

 

[Guy Macon]

Will this thing have an ADC and a uP to control it?

But of course!

If so, why not just turn the light source on/off under software
control and digitize the optical sigs, making a synchronous
lockin amplifier out of the whole thing?

Alas, my emitter is fixed, my detector is mobile, and my detector
sees a spot that the emitter put on a ceiling. I can't say more
without revealing too much about what I am designing, so I apologize
for only describing the subsystem rather than giving the big
picture.

The on/off cycles wouldn't even need to be strictly periodic...
you could, accidentally or by intent, wind up with a synchronously
detected pseudo-random-period-spread-spectrum lockin thing.

I think I can still do this. I just need to detect the on/off
cycles and lock in to them.

You can do some amazing things with one of those 75-cent uPs
with built-in adc/mux. You only have to pay for code once.

SEVENTY FIVE CENTS!?!?! I don't need a *supercomputer*!!! <grin>.
At Mattel, we paid between five cents and thirty cents for a uP,
plus a few thousand for a ROM mask.

 

[John Larkin]

SEVENTY FIVE CENTS!?!?! I don't need a *supercomputer*!!! <grin>.
At Mattel, we paid between five cents and thirty cents for a uP,
plus a few thousand for a ROM mask.

Then why use a $4 DSP?

John

 

[Guy Macon]

John Larkin says...

Then why use a $4 DSP?

Because it may be cheaper.

With a DSP I can do sophisticated FFT-based filtering.

With better filtering the system works with a worse SNR.

With a lower SNR requirement the light emitters can be dimmer.

Dimmer light emitters are cheaper.

(But are they enough cheaper to pay for the DSP?)

With a cheap uP I need to do analog opamp-based filtering.

With analog filtering the system needs a better SNR.

With a higher SNR requirement the light emitters must be brighter.

Brighter light emitters cost more.

(But does the cheaper uP save enough to pay for them?)

Any suggestions for the cheapest way to project a modulated
infrared dot (size doesn't matter with a PSD) on a wall are
welcome. Right now I am using the IR LED with the most light
for the lowest price, and using multiple IR LEDs to get a
brighter dot. How many? I don't know how many yet.

 

[John Popelish]

Any suggestions for the cheapest way to project a modulated
infrared dot (size doesn't matter with a PSD) on a wall are
welcome. Right now I am using the IR LED with the most light
for the lowest price, and using multiple IR LEDs to get a
brighter dot. How many? I don't know how many yet.

Vishay makes some narrow angle IRLEDs that have a pretty high power
output for very little money (TSAL6100). I an interested on your
system optics. Have you considered a plastic fresnel lens that is
made of visible blocking material? You can get a good light
collection efficiency and eliminate an extra filter layer.

 

[Winfield Hill]

John Popelish wrote...

Vishay makes some narrow angle IRLEDs that have a pretty
high power output for very little money (TSAL6100). ...

The LED's 10-degree angle isn't so unusual. But its
1A pulse rating is rather nice.

Thanks,
- Win

whill_at_picovolt-dot-com

 

[Guy Macon]

Vishay makes some narrow angle IRLEDs that have a pretty high power
output for very little money (TSAL6100).

Thanks! mouser has them for 10.6 cents in Qty 2000 (I emailed for a
Qty 100,000 quote). That part is much brighter than what I was
thinking of using.

I an interested on your system optics. Have you considered a
plastic fresnel lens that is made of visible blocking material?
You can get a good light collection efficiency and eliminate
an extra filter layer.

I was figuring on seeing if combining the lens and filter would
lower my cost, but hadn't considered a fresnel. Good idea.
Avoids large filter material thickness variations. With a nice
wide spot (PSDs care about total light, not how bright one part
is) optical distortion at the ridges shouldn't be a problem.
Thanks!

 

[John Popelish]

John Popelish wrote...

The LED's 10-degree angle isn't so unusual. But its
1A pulse rating is rather nice.

I ran a bunch with a half amp pulse of 100 us repeated every
millisecond for several days and saw very little change in the output
power. I have an array of 12 of them aimed at a single spot about an
inch from the LEDs that warms my finger. At 5 for a dollar they are
quite a deal.

 

[John Popelish]

Thanks! mouser has them for 10.6 cents in Qty 2000 (I emailed for a
Qty 100,000 quote). That part is much brighter than what I was
thinking of using.


I was figuring on seeing if combining the lens and filter would
lower my cost, but hadn't considered a fresnel. Good idea.
Avoids large filter material thickness variations. With a nice
wide spot (PSDs care about total light, not how bright one part
is) optical distortion at the ridges shouldn't be a problem.
Thanks!

You might also want to check out some of the other Vishay choices,
like this one that has an 870 nm output that may produce a little
higher response (especially speed) from your PSD.

http://www.vishay.com/docs/81060/81060.pdf

All choices:
http://www.vishay.com/ir-emitting-diodes/

With quantities above well 1000, you may be able to get a good price
directly from Vishay.

 

[Phil Hobbs]

Worst for what purpose? Low-light / low-speed pPhotodiode amplifiers,
to take one particularly germane example, never cross over, and thus
don't care about crossover distortion.




That's a good one from a price/performance standpoint. Thanks!
The cheapest LM324D costs 6.7 cents (Digikey) while the cheapest
LF347 is 19.6 cents. My current project will be using 500,000
quad opamps per year (or two million singles) so that extra 12.9
cents will cost us $64,500 (12.9 cents times 20 opamps per unit
times 100,000 units). (I can get both parts cheaper, but the
difference won't change much.) That's not a bad price if it
turns out that the LF347 allows me to take a dollar of the cost
of my 100,000 light emitter assemblies.

You can see by these numbers why I care so much about price vs.
performance. At it's peak, my largest quantity design was selling
well over 100,000 units per day at a retail price of $24.99.

I would have thought that the price of the position-sensing diode would dwarf
any of these cost savings. Are you using a quad cell or a lateral-effect
device? In photoconductive or photovoltaic mode?

Replacing an expensive position-sensing diode with a couple of small solar
cells plus a shadow mask ought to save much more money than a dime a circuit,
no? There's a discussion of this exact thing on P. 364 of my book, which you
can read free on Amazon if you like--start at
http://www.amazon.com/exec/obidos/asin/0471246816
and search for "364" and then on "94" for a lot more details on p. 94.

(I tried pasting the links directly, which used to work, but Amazon seems to
have changed their "search inside the book" service to be session-specific.)

Cheers,

Phil Hobbs

 

[Guy Macon]

I would have thought that the price of the position-sensing diode

Roughly $3.40. (waiting on quote of exact price)

would dwarf any of these cost savings.

When you are designing high-volume consumer items, the concept of
dwarfing other cost savings doesn't apply. If I can shave 2 bucks
off of the PSD, I will *still* want to shave off a fraction of
a penny anywhere else that I can.

Are you using a quad cell or a lateral-effect
device?

Hamamatsu S5991-01 tetra-lateral.
http://usa.hamamatsu.com/hcpdf/parts_S/S5991-01.pdf

In photoconductive or photovoltaic mode?

Photvoltaic. My light source is dim and I don't need fast response.
I also have my technician setting up a "near photovoltaic" test
where you apply a small (1 to 50mV) bias.

Replacing an expensive position-sensing diode with a couple of
small solar cells plus a shadow mask ought to save much more
money than a dime a circuit, no?

As you can imagine, i would very much like to replacing my
expensive PSD with a couple of small solar cells plus a shadow
mask!

There's a discussion of this exact thing on P. 364 of my book, which you
can read free on Amazon if you like--start at
http://www.amazon.com/exec/obidos/asin/0471246816
and search for "364" and then on "94" for a lot more details on p. 94.

You just sold another copy of your book! I will definitely be
looking into this.

In your opinion, is your technique suitable for detecting the positions
of three modulated spots projected on a ceiling by IR LEDs?