Differential Photo Detector

Here are some notes, mostly from Chris Chapman, on optical displacement detectors.

6/11/2002 Chris Chapman <ChrisAtUpw aol.com>

Optical methods divide into those which depend on the wavelength of light and those which do not. Fringe counting is not too difficult for the amateur, but steps of 300 nm are not very useful in seismic work either. Techniques to measure to 1/1000 of a fringe are likely to remain beyond the capabilities of an amateur.

Optical methods which do not depend on the wavelength of the light can be easy to implement, but you may run into a relationship between the accuracy and the observation period. You can get infinite accuracy, but it takes infinitely long to do it! Practical limits are of the order of 10 to 50 nm, which is still quite respectable for amateur use. Sean said that the smallest quake he ever recorded gave a movement of about 40 nm. The ocean background may be from 500 to 15,000 nm.

The present seis design is very limited. James might try one of the very small 'grain of rice' tungsten bulbs run off 3/4 the rated voltage supply, for illumination. This gives them ~infinite life. The thing that the manufacturers don't advertise about LEDs, is that the light output changes exponentially by about x5 between 0 and 100 C. You must accurately control the temperature of a LED to prevent large drifts. The metal cased IR ones are better. The solid state lasers are incredibly noisy. Under-run tungsten is very stable and far quieter than most visible LEDs. The spectrum is well matched to a Si photocell. I use Si or GaAs photo cells in pairs in a type of bridge. This avoids a lot of problems. Si cells, like the VDT34s, come almost perfectly matched. You can't buy well matched CdS cells and they have a logR to light intensity relationship which varies in slope and zero from cell to cell. The response time is also different for increasing and decreasing illumination and varies with the illumination level. Attempting to get high precision with these devices is something which I am quite happy to leave for others!

6/12/2002 Chris Chapman <ChrisAtUpw aol.com>

In a message dated 12/06/02 152239 GMT Daylight Time, jjpub at lahr dot org writes:

The point about displacement transducers without feedback that I was making was that they are more likely to hit the end of the dynamic range of the system than would a velocity transducer. A coil/magnet system can drift quite a long ways and still provide a reasonable response to an earthquake, as long as a physical stop is not reached.

Hi John,

One way of doing this is to put a high pass filter between the preamp and the final amp. You make the gain of the preamp sufficient to give the max output range signal for the full range input. Then you add a high pass filter and the final gain amplifier / filter. This gives automatic drift compensation and high gain over the whole input range. Nuova Elettronica use this to give ~7 nm resolution over a +/- 15 mm range for their Lehman and pendulum seismometers.

I would like to experiment with an optical system. Do you have a schematic posted?

If possible I would like to stick to parts that are available from RadioShack, as they would then be easily available to a teacher as well.


I have attached a copy of the basic optical sensor circuit that I use on a SG type pendulum rig (it is self centring). I got better than 20 nm resolution out of this. The aim is to get the maximum current practicable from the photodiodes. This minimizes the signal noise. Subtracting the severely frequency limited outputs enabled me to get very high linearity over +/- 1.5 mm and low drift and noise. The gain of the INA118 could be reduced to maybe x10 and a high pass filter and a final amplifier could be added before the 10 Hz filter. I can send you a 'mod' if you are really interested. The noise tends to increase as you move away from the equal photo current centre point. You need a fairly nearly parallel light beam to prevent slight rocking motions of the carbon from being detected as signals. It is a more difficult application than the pendulum.

I had a quick look at what is available from Radio Shack and am somewhat puzzled by this suggestion. I think that you may find it really quite difficult to build precision electronic equipment with what is on offer, but that is your choice. I look for reasonable prices and precision components. Any 5% carbon resistors which are unlucky enough to fall into my hands go straight into the rubbish bin. You have several fairly good mail order firms to choose from. For less than $30, I can buy a 'starter pack' of 10 off of every 1% metal film resistor in the E24 range from 1 Ohm to 1 M Ohm, 1330 in all. (And the seller is still making a profit on this 2.3 c each deal)



Chris Chapman

PhotoDet2.gif          PhotoDet2.pdf


6/13/2002 Chris Chapman <ChrisAtUpw at aol.com>

In a message dated 13/06/02, jjpub at lahr dot org writes

Thanks for the tip on achieving great sensitivity over a wide range by high-pass filtering. On your schematic, what is REF02 with pins 2, 6, and 4 connected?


This is a low noise precision reference REF02CP made by analogue devices. The REF02AP version from Burr Brown may be cheaper. It sets a low noise voltage reference level. I choose R1 for the photodiode current to give about 0 V with the cell half exposed. The output of the OPA2604 photocell amps usually go from +5 (zero) to about -5 V (full scale)

Are the diodes on the right BZY88C5.1?


These are just for protection of the A/D converter or other output. I suggest that you check the data sheet to see the range and the max input it will stand.

For a +/- 5 V output, yes. I use +/- 9 V supply lines usually. For a +/- 10 V output from 12 V supplies, you don't need anything. For a +/- 10 V output with 15 V supplies use BZY88C10.

You wrote " Under-run tungsten is very stable and far quieter than most visible LEDs. The spectrum is well matched to a Si photocell. I use Si or GaAs photo cells in pairs in a type of bridge. " What type are photo diodes SLSD-71N2 and VTD34? What model bulb do you use and at what voltage?

VTD34 is from Newark at http://www.newark.com 

SLSD is from Allied at http://www.alliedelec.com/catalog/pf.asp?FN=505.pdf  

We have had all sorts of trouble getting suitable bulbs. The line filament lamps I use do not seem to be readily available in the US. I try to get a reasonably straight filament of about 3 W 12 V and run it on 9 V. Some festoon car bulbs with a straight tube and end contacts may be suitable. You may be able to get lamps to fit in standard open 1.25" fuseholders. The other alternative is to go to a metal can IR LED with a flat face. I use a small x14 two element eye glass to collimate the beam of the filament bulbs.

By 3 Hz BW do you mean that the output covers DC to 3 Hz?

In this case, yes. I can plug in values for a 10 Hz filter. The 180 Ks ---> 69 K and the 227.5 K ---> 78 K. It only takes a minute on the computer.

For a high pass filter I would use a 4.7 micro F polyester capacitor onto +in on a LF112 dual opamp with 3.3 or 4.7 Meg Ohm input resistor to ground. Put -in to 0 V using a 4.7 K resistor and connect to Vout with a 47 K potentiometer to give gains of up to 11. Use the second opamp in place of the TL071.

If your first stage is set up so 5 V out = 1 mm, you will get a max output of 5mV x 11 = 55 mV / micron. With a +/- 10 V in 16 bit A/D, this gives 0.305 mV / count and a nominal resolution of 5.5 nm. Unless you have a very good A/D, the bottom two counts are noise, so the electronic steps are still a bit smaller than the optical resolution. It should be roughly right.


9/20/2002 Bob McClure <Bobhelenmcclure at aol.com> 

The following description may help. The LED above the pendulum illuminates a square mirror on the top of the pendulum. The light reflected back is a square sided cone. Each edge of this cone half-illuminates a photodiode facing down from the emitting plane of the LED. There are four photodiodes in all. If the pendulum moves, the cone of light shift in angle, and the light reaching the photodetectors becomes unbalanced.

Here is George Harris's original description~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~

Subject Re DRIP (diamagnetically restored inverted pendulum)

From "George Harris" <gjharris at earthlink.net>
Date Wed, 14 Aug 2002 222816 -0700

A suggestion from George Harris

A very sensitive and inexpensive two axis sensor can be made as follows

Above the top surface of the pendulum, place an LED with the front surface flattened so that it become a point source about 1/4 inch above the surface.

On the top surface cement a 6mm square mirror (stock H43866 from Edmund).

MIRROR 4-6 WAVE AL 6 X 6MM 2 Float Glass 60-40 4-6wave per inch Protected Aluminum NT43-866 $12.10 http://www.edmundoptics.com 

On each of the four sides of the LED place small solar cells (stock 980-0150 from allied) in such a position that each will be half illuminated by the LED. VTD34 $1.59


The photocells can be connected in parallel, plus to minus, on the opposite sides across the inputs to a low noise operational amplifier (like an OP27 from Precision monolithics) with about a megohm in the feedback. The resulting output is very sensitive (nanometers) and linear.

To check the sensitivity, I mounted the assembly on a three point platform with a dial indicator on one end of the platform. Tilting the platform by small motions of a support screw permits the accurate measurement of the angular (or acceleration) sensitivity.

8/16/2004 George Writes and Chris Responds

While leafing through the current issue of Electronic Design I read an ad  from National Semiconductor (p. 34-35, 08/09/04 EDN) the might be of interest  to amateurs who are building their own seismometers and need a way to  determine the absolute position of the boom and output a signal proportional to the  boom's displacement. The ad is a design idea for an application of their  LMV2011 precision op amp. they are suggesting that it be used to control the  output of a led and stabilise it. They are doing this by using a photo diode to  look at the LED and through the op amp, feedback the signal to the LED and  stabilise it's output.

My thought would be to place the occulting vane on the seismometer boom in  the light path and as the boom moves to obscure the light, the amp will  increase the light level to make up. We should get a usable signal across the resistor that the design has in series with the LED. I do not know if this will work or if it may be too noisy or temperamental, or some other "gotchas," but I am going to try and collect the parts and play with it. You can read the application note if you are interested at:..

Hi George,

To measure seismometer movements, you usually allow over +/-0.5mm total movement (or your seismometer becomes very difficult to set up) and you require to measure the arm position to maybe 20 nano metres for 'amateur' use. To do this you have to reduce / compensate / design for low drifts and minimum noise, both with time and with temperature - you are considering 1 part in 25,000.

You need two large area photo diodes (~7sq mm - VTD34?) connected to the inverting input of two low noise opamps TLC2201? with a suitable value of feedback resistors. You then subtract the opamp outputs with a differential opamp eg INA118 and apply bandwidth filtering. This reduces the effects of temperature on the photo sensitivity and on the leakage currents. Noise considerations in photodiodes require you to use a photo current of the order of 0.5 mA, which implies a high intensity light source. The photodiodes should be fixed to a common heatsink to minimise temperature variations.

It is possible to use light from one the high power, metal cased IR LEDs, but the photo output of a LED shows an exponential decrease as the temperature increases. This makes getting high stability and low noise a little bit difficult. Laser diodes tend to be very noisy. (You can buy laser diodes which have an internal photo diode to 'stabilise' the output.) An easier approach is to use a tungsten filament bulb in a feedback bridge circuit, which stabilises the hot filament resistance. You reduce the voltage on the bulb to < 0.8 x that rated. This gives an essentially infinite filament life. The dimmer filament doesn't effect the sensitivity as much as you would expect, since the sensitivity of Si photo cells increases in the near Infra Red. GaAs photocells may also be used with superbright visible orange LEDs. 'Ordinary' LEDs tend to be quite to very noisy. the superbright ones tend to be quieter.

You need the dimensions of the photocells to be large compared with the wavelength of light, to minimise interference fringe effects. The minimum conduction noise in a photodiode is proportional to the square root of the photocurrent, so increasing the photocurrent will give a lower overall noise. With a significant amount of heat being shone on the photocells + optical shutter, the detector needs to be near the top of the seismometer case, maybe in a semi isolated light box, to minimise heating effects and air currents. The bulb is best mounted outside the main seismometer case, to give adequate cooling by direct contact with the housing. You can make quite good windows using Microscope slides / cover slips.

There was a note on PSN some years ago which said that you couldn't use optical detectors, but this was a misunderstanding. If you try to use interference fringe methods, your resolution will be limited, commonly to a fraction of a micron.

Chris Chapman