I am trying to modify a laser-scanner from a laser printer into a blue/ violet photoplotter. I need some kind of driver for my blue/violet laser. I've seen the schematics on Sam's laser faq, and looked at various ic's from maxim, analog devices, and ic-haus. They all seem to have trouble with optical-power-correction. Most of the integrated circuits force the average power to be constant, which might work for data streams, but for images, it is not good. I need this to work at frequencies from DC to 100Mhz. The circuits on sams don't seem to have this problem, but the OPC causes spikes in intensity when going from low to high level because the photodiode current is, correctly, way to low.
I've considered using the equivalent of two diode drivers where my modulator can switch between them, this way, there will be 2 capacitors for storing the photodiode level. I think this would work, but there are lots of parts. And i don't think my breadboard will work at the 100Mhz i need.
I am a little confused as to the bandwidth of OPC circuits. Does the photodiode output really vary at khz or mhz? Is it possible to just calibrate the machine every time it is run? Is there any circuits or chips that do something like what i need?
Most systems i've seen use some kind of feed back to insure the output is where you want it. You'll see diodes with a third leg which is a PD for monitoring the Laser output. This is used to balance the driving circuit of the Laser.
Don't know if that is what you have.
--
"I\'m never wrong, once i thought i was, but was mistaken"
Real Programmers Do things like this.
http://webpages.charter.net/jamie_5
Hi, I'm one of the guys that helps write sam's faq. I do a lot of photonics for a university. You need to answer a heck of a lot of questions for us to help you.
First off , why are you trying for 100 mhz for a photoplotter, to the best of my knowledge the fastest blue laser photoprinter in the world has a 20 mhz wide ecl input and gets there with a AOM, not direct modulation, is printing on the order of 20-200 pages per second and is roomsized. What the heck kind of emulsion do you have that is sensive enough for a blue diode and that high a speed? Seriously, 100 Mhz square wave drive?
Second of all, the new blue diodes are not characterized for that kind of speed. You'd need to characterize them as a load. You will need some form of process feedback to compensate for temperature varaiations etc.
Third of all laser diodes are "soft" they do not handle spiking etc very well, which is why the optical feedback is used, 100 mhz is RF, and thus you need impedance matching to the device, a straight current source chip is not going to cut it. To achive 100 mhz, first you need a pile of diodes to blow up, and a expensive network analyser to shape the matching networks etc. You need to learn to use microstrip and other RF techniques for the cabling. Thats why you wont see that in the FAQ, its almost never asked for. You'll need a DC turnon voltage and a "DC block / Bias Tee" to inject the RF current. Such things are available from companies like New Focus or Thorlabs, but are expensive. the blue diodes are very fragile, they dont stand up to well this kind of abuse. Its NOT a driver chip ,in a production quanity device, you modulate the RF and then inject RF over the cosntant current DC bias. You probably need a fast adjustable RF limiter too, A fast diode driver from one of the companies that specializes in pulsing diode device is what you need to start the charterization with.
My polygonal mirror rotates scan the beam at about 2000 passes/second. Say i am trying to write 100,000 pixels per scan, this comes out to
100Mhz for the worst case of alternating black and white pixels. The laser scanner will be moved over the plot at a rate of about 5 seconds/ inch. So not quite 200 prints per second, but much better resolution. Just about all emulsion is sensitive to 405nm. For one thing, gelatin/ chromium emulsion certainly are. It is difficult to estimate the effect of reprocity, but as a rough calculation, the film will be exposed to about 5 minutes of 5mw radiation / sq meter. Which is more than enough for any film i know of.
Also, square wave is not so important. From what i gather the laser should be modulated between threshold and some high value. The scan speed/filtering need to be adjusted to make this work for the film.
There are blu-ray burners, and 1x blu-ray is ~30MB a second, so say
240Mhz, probably higher to deal with reed-soloman, 8/10b encoding and such.
I think the problem is not so much the RF as getting it to work from RF all the way down to DC.
I've come up with a possibly stupid idea for how to do this.
The PNP transistor is a MPSH81, the 3 npn's are MPS5179D75Z
The idea is that there are a 5 DAC external to this that is able to set both the bias and modulation current. My plan to do the output correction is to calibrate the system during the end of a scan. After the scan is done there is like 10us of dead zone. During which i can set the laser to both high and low state, digitize the output of the photodiode and update the DAC. I assume the temperature of the laser diode can't change much in the 500us of single scan.
Thanks. You seem to know what you are talking about.
In , snipped-for-privacy@gmail.com wrote: (And I edit for space)
Believe me, many, probably most photographic emulsions do very well at
405 nm. I don't consider that the higgest obstacle.
You could only need to pass a sine wave of 100 MHz with frequency response at least somewhat flat from DC to somewhat over 100 MHz, and not too much ringing at any frequencies at or near where gain starts varying heavily inversely with frequency.
I see the big challenge making a broadband RF circuit that works from DC to a little past 100 MHz. This touches heavily upon such things as stray inductance, stray capacitance, and within that distributed inductance, distributed capacitance, stripline PCB principles, knowledge of a stripline transmission line, and principles of transmission lines such as characteristic impedance and need for matching load impedance to characteristic impedance to flatten frequency response, and need to match source impedance to characteristic impedance to minimize resending any reflections from the load if the load mismatches the transmission line.
This gets a bit easier if you understand characteristic impedance, impedance matching to characteristic impedance, use of the Smith chart to determine theoretically effects (and degree thereof) of for mismatches, what the characteristic impedance of a coax cable and a twinlead cable are as a function of dimensions and dielectric constant of insulating material involved, "velocity factor" and relationship to dielectric constant of insulating material exposed to electric field, and principles of characteristic impedance and velocity factor of a stripline transmission line.
Heck - video amplifiers, which only need to work from DC to about 4 or 5 MHz or so, get to be a slight bit of a challenge! If you want practice, try making a video amplifier set and a pad-down to gain and pad back down a signal set for an SVGA monitor. (No, I have not done that!) If you can do this without horizontal blurring or ghosting, then you are probably most of the way there to amplifying or buffering a signal of DC to 100 MHz. I do suggest this as practice!
One more note - if a transistor has Ft of 300 MHz, then it's "beta" or Hfe at 100 MHz is close to 3! If you use a transistor good for microwave frequencies, then your circuit needs to be well-behaved at frequencies up to a goodly fraction of the transistor's Ft. Watch for parasitic oscillations! Know common-collector and common-base as well as common-emitter transistor amplifier circuit theory! One classic oscillator is the Colpitts, which uses a transistor in common-base mode! Know the treatments for parasitic oscillations, and I suspect they are best-published for different frequencies and different applicable impedances than those you will run into! Know how to translate component values to different frequencies and different "working impedances", and hope that you are treating the same oscillation mode that the treatment is for, hope for lack of 2nd and 3rd or whatever order complications in your application, etc!
Also, you might be exploring into territory that has not all of its charts well-published, which means you could be a prime victim for Murphy's law finding a way to effectively supersede Ohm's law!
Oh, one more thing - if you have a bipolar transistor in a Class A amplifier (better know what that is), if the transistor is a fast one then design for lowish supply voltages preferably 30 volts or less, with the transistor having average collector-emitter voltage while conducting around or under 20 volts. For that matter, in Classes B and AB (better know what those are), if the transistor is conducting close to half the time its average C-E voltage could easily need to be under 25 volts or so if the transistor is a fast one. One key here: "SOA" or "Safe Operating Area", and that is a graph of allowable combinations of C-E voltage and collector current. It is common for bipolar transistors to have allowable power dissipation decreased when C-E voltage is more than somewhere around 30 volts, and this gets worse for faster ones. The principle here is "forward bias second breakdown", where a hotspot develops in a localized region within the transistor "chip" / die.
I have made my living doing stuff other than this for the past 23 years, and I have this little bit of knowledge of this that could be "dangerous" here... I suspect a possible hazard is need to kick in a few dozen hours of extra work time on weekly salary or per-diem time or unbillable hourly as "continuing education", and that gets to be hell when deadlines make you spend an extra 15 or 50 hours fixing something in a matter of days to a week!
I did some work ten years or so ago on improving printer and scanner technology by combining holographic (hologon) scanners with analogue current-modulated diode lasers. Since the whole laser goes on at once (except for a few possible mode hop funnies during turn-on), you can go quite a bit faster than with an AOM.
The other useful trick is use the chirp of the laser diode to get a factor of between 10 and 40 improvement in the attainable scan rate with lowish-speed hologons. You use a couple of diffraction gratings in series, coming in near normal and coming out near grazing on both gratings. This turns the long skinny stripe from the DL into a more or less circular beam, and makes it steer in angle with drive current.
So instead of scanning a single spot, you effectively scan a whole stripe of 10 to 40 pixels at once. You adjust the exposure dose by changing the dwell time on each spot.
My preliminary scanner demo worked great (minus the hologon--it was just the stripe part), but my management changed before I got the whole thing working. It was also capable of measuring the range to an uncertainty of about 1 part in 10**5 of the distance to the object, by using the tuning trick with a solid interferometer. The current tuning produced an optical frequency shift (i.e. the chirp) that became a baseband modulation of the interferometer's output.
It was intended to be a pocket-sized scanner and coordinate measuring machine, fast enough to give good results handheld, and could have sold for a few hundred bucks. It never got finished, unfortunately, so I don't know if it would have worked as I expected. The new part, namely the chirp-scanned X-Z measurement, worked great at about 40 Mpels.
So there are several ways of getting beyond 20 Mpel/s.
One, get a copy of Phil Hobb's book on electro-optical system design, its excellent
Two HP's (now Agilent) Appcad freeware has much of what you need for the board layout in terms of transmission line and matching.
formatting link
you'll need to know the E sub r of your board material
Three, minicircuits.com
Four, start looking at patents
such as 6667661
and here:
by Marc Thompson
formatting link
Mr Thompson is missing some spike protection circuitry, and startup circuitry, that would need to be there as well as some impedance matching details
In my experience anything from Shuji Nakamora's family of blue devices (led or laser) looks more like a SCR then a Led when driven, they tend to act like 4 level devices. (PNPN) Rather expensive device to put on a curve tracer, but.... Start with the nichia blue led as a test load, then scale to the lasers, no guarentee they are that similar, but the led is a heck of a lot more rugged.
look here:
A 10Gb/s SiGe compact laser diode driver using push-pull emitter followers and miller compensated output switch Maxim, A. Solid-State Circuits Conference, 2003. ESSCIRC apos;03. Proceedings of the 29th European Volume , Issue , 16-18 Sept. 2003 Page(s): 557 - 560
here:
take a look at Analog Devices AD9661A diode driver. (usually do free samples)
If you have a spare that isn't too precious, that would be wonderful. I have a recently-repaired HP4145B semiconductor parameter analyzer I can try it on. Just so I don't blow it up, can you give me a ballpark idea of how big the negative resistance is, and whether it's fast or slow? I don't want to wind up with the poor thing turning into a transmission-line oscillator, because it might not survive.
(I haven't priced those Nichia things in a couple of years, but last time I looked they were $2k or so.)
snipped-for-privacy@gmail.com snipped-for-privacy@gmail.com posted to sci.electronics.design:
I like the plan some, but for better tracking use an optical splitter during the scan and compare optical output versus intended value continuously. 500 uS is well into thermal response time.
The laser diode has an integrated photodiode, i don't know if this is any better or worse than an optical splitter. Leaving that aside, comparing the output to the intended value continuously is probably impossible. Say there is a random stream of bits of ~50% duty cycle. Then one would need to use an integrator on both the photodiode and the incoming signal. Then we could say the integral over some period was like 5000 units when we should have had 6000. Then the question is, is the bias current of the modulation current to low. This is not too easily solvable. There will be situations or time periods where the duty cycle is significantly more/less than 50% allowing better guesses on what to adjust, but these cannot be guaranteed. Really, i don't think the integrator will work either. It will absorb so much data from transients that the average will be totally screwed.
On another note, i've gotten a hold of some MAX3701 ic's. I've designed a prototype board that is being produced. Whole thing has come to about $300 so far. I'm using a 100mhz LVPECL link from an fpga to the laser module for the data, and the laser diode will be directly soldered to the driver board. I hope to get rid of most of the RF design problems by keeping things smaller than lamba/10.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here.
All logos and trade names are the property of their respective owners.