fast vibration measurement using lateral effect photodiode

[timeout]

Hi All!

I'm new to the forum, first post
I am a long time follower of the channel on youtube! Brilliant work! (liked the guest video concept!)

I'm just starting electronics, being a student of materials science its the perfect match in many aspects 

my goal in the lab at the moment is the vibration study of a cantilever bar having a reflective surface which can deflect a laser spot onto a lateral photodiode.
I have a (half) commercial system for exactly this task (using 1D and 2D photodiodes) at hand - this is unfortunately only specified DC up to 100kHz.
Which is fine and does the job for most of my applications.
It uses simple TL074 op amps in combination with AD633 (analog multiplier/divider) in order to do analog division on the output signals.

The laser deflection i need to measure now is of around 500 kHz plus some harmonic content which may not be important to monitor at first.
I have a rather large, Swedish made 2D Photodiode here, 10x10 mm of Sitek. http://www.sitek.se/pdf/psd/S2-0185-2L10_SU72_AA4.pdf (10x10 mm, rise time 0.8 µs max.)

here is an example of a commercial system, they seem to be using a TL034 (http://www.on-trak.com/spcpsd.html) directly attached to the Photodiode, reaching up to 400 kHz, but it seems to not supply the division signal.

As i can live with having only one axis i guess i could tie the unused axis to bias and only use one axis, perhaps again saving stray capacitance.
I guess i need something like this circuit, and tie the anodes to GND ? (http://www-cdr.stanford.edu/html/MADEFAST/catalogs/on-trak/psd/2d-circuit.gif)

I think i could also do the dividing of the SUM and DIFF signals X1,X2 in the digital domain, scope math etc.
Or else using an AD835 or AD734 ? I just dont won't to over-engineer it from the start..

I have a few LTC6244 at hand, they seem to be well suited for the job.
The datasheet (http://cds.linear.com/docs/en/datasheet/6244fb.pdf) gives an example (fig. 5a and 6a) for large area photodiodes, of which i would like to use two, because i have two ports left after biasing the two .

Here finally come some clear unclarities:

- My photodiode has a capacity of max 110 pF - so it lies somewhere between a small area <10 pF and large area ~3000 pF one - is this a problem ?
- How much gain does this circuit supply ? (my photodiode has a responsivity of ~ 0.6 A/W the laser is standard 1mW, 635 nm, reflectivity of the surface is "mirrorlike")
(http://www.osioptoelectronics.com/application-notes/AN-Photodiode-Parameters-and-Characteristics.pdf, page 4, eq. 6. I get a Quantum eff. ~ 1.2 - WHAT?!)
if i lower the gain i should get more BW out of the LTC6244, or am i missing somehting ? i wouldnt have a problem amplifying to taste in a second stage..

- Is this whole project much too optimistic, will it be possible anyway using the given combination of photodiode and amp ?

I dont require a strictly low noise design, although i have been thinking of a lead battery supply +- 12V as the photodiode is floating anyway and requires a high reverse bias voltage and i dont need to measure anything for days on end. (i know, the opamp can only take +- 6V in the HV version, which i have)

- will i need careful local voltage regulation ? or will standard 100 nF de-coupling in close vicinity to the op amp suffice ?

I appreciate any help or advice! I hope i didnt forget anything important to define the problem..
best regards and a happy weekend
Patrick

[Kleinstein]

The capacitance of the detector decides how important voltage noise is compared to current noise. To avoid extra capacitance it is worth having the initial amplification close to the detector. The relatively high frequency of interest makes detector capacity more important so the 100 pF would already be a large detector in the 500 kHz range, while is might be considered "small" when working at 10 Hz.

If you only need one dimension it could help a little to have the other side tied to a fixed potential. However usually the 1 D sensors are working better, as they are usually smaller and thus lower capacitance / lower noise.

It is OK to reduce the gain of the TIA stage quite a bit. The Laser would give up to 0. 5mA of current, so even with some loss one can expect more than 100 µA. This a high current compared to typical PD circuits.  One the voltage over the shunt is more than about 100 mV the shot noise will be higher than the resistor noise and there thus is no absolute need to use an even larger resistor. From the lower 100 mV to an upper about 5 V limit set by the supply, there is still some room to choose a suitable resistor.

It could be enough to do the division in the digital domain. However it kind of depends on how much the power fluctuates.  In theory it might be worth also monitoring the actual laser power at the same fast rate. Just a single 8 bit ADC for the total current might not be enough for the division, if relatively small movements of the laser spot are of interest.

[timeout]

Thanks!

Yes, i plan to have the detector on a small second circuit board, connected via header pins directly to the amplifier board, so all very close proximity.

What the detector is concerned, as i have this 2D one at hand it would be great to utilise it. The 1D ones are actually much cheaper that the 2D variants, which are prohibitively dear.

How did you come up with the current values ? is this experience or rough numerical estimation ?
as i pointed out, i unfortunately didnt get anywhere proper using the equation of OSI for the Quantum efficiency...

Yes, doing the division on a scope is a bit lame due to quantisation errors... but for a first shot i would be happy, especially because i can, witin practical limits, chose the distance from the reflective surface to the diode in order to max out response in the linear regime of the photoiode.

But i am quite unsure about how to incorporate the 2D diode to the circuit in the application note of LTC6244.
here they supply the bias voltage by the opamp, as i understand it. (http://www-cdr.stanford.edu/html/MADEFAST/catalogs/on-trak/psd/2d-circuit.gif)

but isnt this quite different from the approach stated in the application note, is it ? or am i just confused with relative potentials ?
(http://cds.linear.com/docs/en/datasheet/6244fb.pdf ,fig. 5a and 6a)

i will try to post a circuit example of how i would do it..

[danadak]

Ref material for transimpedance amplifier input C reduction -


http://www.analog.com/media/en/training-seminars/tutorials/MT-059.pdf


https://electrooptical.net/static/oldsite/www/frontends/frontends.pdf


Regards, Dana.

[Kleinstein]

The current can be easy estimated from sensitivity and laser power. So the maximum current is about 1 mW times 0.6 A/W. It is only the optical loss where it takes some experience. Here the big question if an optical filter is used - this may be needed, but could reduce the intensity quite a bit.  Background light can be a problem, as it will upset the division. Especially the rather large 2 D sensor could pic up quite a bit.

The circuit from the LTC6244 data sheet, using the JFET to drive the other side is not practical to us with lateral diodes, not even the 1 D version.

The lateral diode would likely need a more conventional TIA circuit with a low noise, high speed FET based OP.

[timeout]

Thanks!

Ah, so its that easy
i dont know of any optical filter in the beam path and this 0.6 A/W is taken straight from the datasheet of the device.

i can take measures to lower stray light influx, using a long black tube in front of the diode and/or turning off the light in the lab during a measurement is a feasible option.

thats rather sad to hear....
i've compiled a circuit which i would have routed and etched tomorrow...i'll just attach it now.

As i really only need a few measurements at first a, quick and dirty approach would suit me fine...or is the task way too sophisticated for such an approach ?
anything i can set up on the bench, test, solder and whack into place would be great
i have a good stock of jellybean TH components....

if it helps i could even use a more powerful laser, which i'd of course love to avoid for the sake of lab safety...

cheers
Patrick

[nfmax]

i can take measures to lower stray light influx, using a long black tube in front of the diode and/or turning off the light in the lab during a measurement is a feasible option.

When limiting stray light, you need to use a system of baffles (i.e. thin discs with holes of carefully-calculated diameter, spaced at the correct intervals) rather than just a tube. Otherwise stray light scattered from the surface of the tube gets onto your detector anyway. There is a geometric construction for designing these - I think I still have my notes somewhere, I'll try hunting for them.

You may also need to use a narrow-band optical filter, though if you can keep your experiment in the dark, that would be best. A filter will not help you if you have stray scattering from your laser source, though.

[nfmax]

I found my notes! From June 1996. An excellent 3-day course on opto-mechanical systems design.

[timeout]

Thanks nfmax!

great Info! They should be concentric around the optical axis, i presume.

[nfmax]

Yes, concentric about the optical axis. if your aperture & detector are circular, it's simple. If they are rectangular, you should design on the basis of equivalent circle of diameter equal to the diagonal of each.

[timeout]

So, it worked

btw. i measured the photocurrent of a 1mW laser at 650 nm, having biased the cathodes biased to +12 V using a battery.
I measured between the Negative of the battery and the two anodes.
with laser illumination i measure about 300 µA combined photocurrent, of both meters.
With only the lab light on, i have about 70 µA - with lights off (and dark outside) i have about 5 µA.

So i think the value for sensitivity was overestimated with 0.6 A/W, the datasheet states that the values are valid for a wavelength of 940 nm, much more sensitive, than for 650nm, obviously.

anyhow, i bodged something together using only one axis and one TL082 yesterday  (see atachement)
I needed much less gain then expected, started out with using 680 kOhms in the feedback loop, ended using 25 kOhm - may not be optimum.
It seems to work fine at 500 kHz, the noise level is acceptable.

now i would like to do the summing and subtracting also analog, the division can be handles after digitising.
What op amp would be suitable for these jobs ? any ? TL082 or an NE5532 ?

best regards
Patrick

[Kleinstein]

The sensitivity is lower at 650 nm than at 950 nm, about by the ratio of wavelengths. So the 300 µA sound reasonable, so do the 25 KOhms. 25 K times 300 µA are already 8 V - thus well above the shot noise limit. So there would be room to reduce the resistors even further.

For the OP for the TIA I would look at a faster and low noise JFET based OPs.  A OPA2134 would be good low noise candidate and also a bit faster than the TL082. The OPAx604 would be considerable faster, but still manageably.

For the subtraction the NE5532 might work. The TL082 is a bit slower but likely still OK.

To reduce the light from the environment it might be a good idea to use a kind of apertures to reduce light from the wrong direction.
70 µA from the lab light could be a problem as this might include AC components and would also cause an error in division.

Using an optical filter is kind of tricky, as much of the intensity will come from red light anyway and a narrow band (e.g. interference type) filter can also reduce the laser light quite a bit.

[nfmax]

There is a good article by Philip Hobbs, on interfacing to high capacitance photodiode with a wide bandwidth & low noise, available at https://electrooptical.net/static/oldsite/www/frontends/frontends.pdf.