Question about Modulation Difference

[QUOTE=mixedgas;128410]What Tocket could do is write us a frame statistics analyser,
Or get one into ILD SoS
Or laserboy

Preferably have two pieces of code to do this so the biases drop out.

I'm not sure a histogram is the best way of doing this, but we need to know what about 1500 graphics frames picked at random look like in the time domain, and the times and slopes between points.

Otherwise we end up with the brightness version of "I hate PPS because" or "The ILDA test Pattern is worthless because...."


The video guys used to use a series of VITS stuff like this:

http://elm-chan.org/docs/vits/test.html

A Analog of NTC7 and Sin X/X come close to what we need.
I'm betting we need some ringing bursts of varying times and frequencies at various offsets or on ramps.

3DB bandwidth just doesn't cut it for this application.

And the test needs to cost less then 40$ to build with basic lab equipment. Ie eprom, counter, and DAC to make the waveforms. That structure could exist in a FPGA or Microcontroller as well. Since I dont know if you could force Beyond/LD into a true point by point mode, this could cause issues for Pangolin users. Having looked at my RIYA on a scope over the weekend, I'm pretty sure this is a dedicated hardware job up to a point.

The VITS waveforms were dependent on the fact that the phosphor had a constant rise and decay time, so they are used for time domain on the video signal. VITS was used to adjust the transmission and drive hardware in the frequency domain, as the phosphors were standardized. We're a bit different in that we need to look at the final product. We have a long lag time, so I'm just using the video signals as a example to visualize the concepts.

The LDS test pattern DZ uses is a start, DRLAVA has a nice one, too.

But in reality, this is a job for a photodiode, not a eye.

See also : http://www.mediacollege.com/video/te...ultiburst.html

My two main greenies do not make a good example. One is fast and the other lab one takes about 15 points to turn on and off. I don't own any dpss blue. No use testing the PCAOM except for linear response because it has a minimum 100 Khz bandwidth.

Suggestions?

Steve

I think I can supply all necessary software/hardware needed to test this out...

I've got GPIB on my computer, and I've wrote hundreds of GPIB capable applications.

I think all that would be needed a high-speed, high-accuracy, photodiode to measure the modulation response. I have access to the equipment needed at work :)

--DDL

I had a phone call with Mr. Benner of Pangolin. I have sent him a link to this thread and the recent one on modulation.

Rough Summary of His suggestions:


0. A standardized photodiode and supporting circuit. The photodiode is tested for linearity and frequency response by a standard method. This is easier then it might sound.

1. The input range of the laser is 0 to 5 volts, single ended. The laser starts to emit at .2 volts and is at full rated power at 4.8 volts.
(the reason, noise immunity and cable drop)

2. Long ramp linearity, a ramp from 0-5 volts lasting 1 minute, then another lasting 30 seconds. Score the deviation from linearity in percentage.

3. Sine, Square, and Triangle wave, starting at 1 hz to 30 Khz, again score the deviation from desired output.

4. Make the manufacturers improve their product, there is NO reason we should have to do the improving.


-----------------------------------

My suggestions

0. A standard warm up time and a infinite heat sink are specified.

1. Addition of a burst waveform consisting of sharp pulses
2. Addition of a ramp waveform with a pause, then a sharp pulse
3. A "envelope" waveform, ie Sin x over x

-----------------------

Bill disagrees with me on the need for really special waveforms. I see his point.


Scoring is percentage deviation at a set of frequencies F1, F2, F3 etc And per Mr Benner's suggestion 3, a 3 Db down half power) frequency for the sine, square and triangle waves.

I can easily standardize the photodiode circuit, and justify it based on the math for photodiodes published by Dr Phillip C. Hobbs, as well as Jerald Grame's book on photodiodes.

This is easy to measure, the triangle, 50% duty cycle square, and the sine waves are a standard function of a signal generator, and there are numorous means of measuring the results, either oscilloscope, data aqquistion, eyeball, or sound card based. A subtractor circuit using a high performance op-amp or the inherent X1 minus X2 function in many oscilloscopes can also be used.

I'm a big fan of Dr Phil Hobb's work on detectors, having talked with him several times on various projects.

Steve's References:


Basics
http://www.electrooptical.net/www/fr.../frontends.pdf

http://www.amazon.com/Building-Elect...ref=pd_sim_b_2


Advanced class:

http://www.electrooptical.net/www/ca...andhacks6c.pdf

http://www.amazon.com/Photodiode-Amp.../dp/007024247X



Steve

http://search.digikey.com/scripts/Dk...me=751-1001-ND

That's a very good outline. I like the recommendations for the points at which lasing begins and reaches maximum. I think that aspect should go directly into the spec unchanged. Lasing from .2V to 4.8V should be a requirement for a laser to meet spec. It's the allowable deviation from linear response that might be a little harder to quantize across a range of frequencies and basic waveforms. I feel that we're going to have to accept some amount of instability no matter what. It's just a matter of how much is acceptable, and how much improvment can be afforded money-wise. At some point it becomes impractical to improve further because the techniques required to make the improvements will be too expensive. Then we're stuck buying used units off ebay like we do with argons and other high-end systems.

Daedal just reminded me that LEDs are pretty much 1 photon per electron, so we use a high power LED to calibrate the linearity, and the leds I've tested hit 200-300 nanosecond rise/fall times, so that handles the calibration.
I was gonna do it with a AO, but the LEDS are 2-4$ each.

Daedal and I like this:

BPV10 Digikey http://media.digikey.com/photos/Vish...otos/BPV10.JPG751-1001-ND

And this:

Thorlabs FDS100 http://www.thorlabs.com/NewGroupPage...=285&pn=FDS100

I've used FDS100 at my old job, its plenty fast enough.

For test purposes, ie a Silicon photodiode with roughly 1 square MM or less of surface area. With 12V of reverse bias, such a diode will easily hit more then 500 khz with just a terminating resistor. With a transimpedance amp, 1-10 Mhz are possible.

Steve



Steve

Nice! 500 kHz is more than sufficient for this.

I don't really care about the ranking of modulation characteristics, but if i may chime in with the methodology...

Si photodiodes are extremely linear by nature (with proper biasing at least), way more so than any led I have seen. I am not sure what led you tested that showed a good 1photon/electron output (if you did let me know I want to buy some!), and the effects due to heating of the die can give for a pretty nasty P-I curve. Of course the photodiodes response relative to wavelength is utter crap, if you have a laser that has a significant amount of wavelength 'chirp' as it heats up it could cause issues (I would expect almost all of the direct diodes lasers to show a nm or two), but it sounds like this is aimed more for dpss lasers, so that shouldn't be much of an issue.

I suppose it really boils down to now good of a number are you looking for, 10% 1% .1% .01%? .1% should be achievable with a decent photodiode and proper baising/amplification (it will take a little thinking to get your amp flat to .005db across dc-100khz).

As to the pattern generation, I would personally recomend using a PIC/AVR combined with a dac and an adc, pick as many bits as you need to get enough resolution (a 12bit should be more than enough for this application, and is resonably priced). Make sure that they both share the same Vref, so you don't have to worry about that drifting around as well. Pics are dirt cheep, as are the programmers, and its easy for a pl member to burn a pile of them with the offical pl test pattern and mail them out to members too lazy to build a programmer. Also, once the code is loaded you can have a bootloader to update programs at a later date. The only issue I see is the limited memory for program storage, would something like 32kbytes be enough for your guys? For the simple repeatings waveforms that would be no problem, but for the sudo-random test it could be a little limiting.

Quote:

---

Originally Posted by krazer

I don't really care about the ranking of modulation characteristics, but if i may chime in with the methodology...

Si photodiodes are extremely linear by nature (with proper biasing at least), way more so than any led I have seen. I am not sure what led you tested that showed a good 1photon/electron output (if you did let me know I want to buy some!), and the effects due to heating of the die can give for a pretty nasty P-I curve. Of course the photodiodes response relative to wavelength is utter crap, if you have a laser that has a significant amount of wavelength 'chirp' as it heats up it could cause issues (I would expect almost all of the direct diodes lasers to show a nm or two), but it sounds like this is aimed more for dpss lasers, so that shouldn't be much of an issue.

I suppose it really boils down to now good of a number are you looking for, 10% 1% .1% .01%? .1% should be achievable with a decent photodiode and proper baising/amplification (it will take a little thinking to get your amp flat to .005db across dc-100khz).

As to the pattern generation, I would personally recomend using a PIC/AVR combined with a dac and an adc, pick as many bits as you need to get enough resolution (a 12bit should be more than enough for this application, and is resonably priced). Make sure that they both share the same Vref, so you don't have to worry about that drifting around as well. Pics are dirt cheep, as are the programmers, and its easy for a pl member to burn a pile of them with the offical pl test pattern and mail them out to members too lazy to build a programmer. Also, once the code is loaded you can have a bootloader to update programs at a later date. The only issue I see is the limited memory for program storage, would something like 32kbytes be enough for your guys? For the simple repeatings waveforms that would be no problem, but for the sudo-random test it could be a little limiting.

---

I'm biased toward AVR/Arduino,I use OLEMAX's AVR boards myself. As simple as this task is, with proper corrections to the counter timer scheme, no reason why a PIC could not be used as well.

I'm not just shooting for PL here, I would like it to be a ILDA standard. Mr Benner agrees this needs to be done at the laser factory as well.

Ok, so 2.6 electrons per photon, but most big leds are linear for our purposes. Remember the people on the manufacturing end probably do NOT have nice test gear. Hence the LED. And from what I've seen, I'd be thrilled at 1% compared to what we have now.

I'd like to scale this to 12-16 bits and use a real dac chip. R2R ladders fall apart at more then 8 bits. We have long known that 8 bits is fine for show performance, but its just a bit slimey for test, a 1 minute or 3 second 8 bit ramp would be a travesty. Of course a good manufacturer is going to be using a professional waveform generator in R&D, but I doubt that the line kids get nice gear.


I've found some AVR code, but it needs recoded to use a dac chip and more then 8 bits off of the accumulator:

http://www.myplace.nu/avr/minidds/index.htm


Steve