using multiple laser diodes to scan multiple lines with one mirror movement

[Stoney3K]

No, more like 188.4 Kpps, or 188400 pps. Remember that point speed is not the same thing as bandwidth. (Unless you want your picture to be just 1 point wide...)

I can see your mistake, dear Padawan...

You're still considering video as a digital signal, with discrete points (pixels) where the X scanner should stop. Where actually, it is not, and video fields (or scanlines in the progressive domain) are a time-continuous, analog signal.

Bluntly put: A television doesn't know squat about pixels. It only counts lines, and the video signal on the lines can be continuous (so in theory, you have infinite resolution on the horizontal). That's also the reason why video formats are always specified in their line count, e.g. 576i (PAL), 480p (ATSC) or 1080i (HD).

Now, the reason we're thinking in 'points per second' here is because you probably modulate the laser intensity along with the appropriate scan point. The only reason to do this is because it is easier to implement in software, but nobody states that the intensity signals and the X/Y signals SHOULD be in sync.

Thinking purely in the analog domain here (so no PPS involved), you can make this work. Galvos have mass too, and if they're built right, they should move from one point to another in a linear motion (albeit a really fast one). Pulling someone called Nyquist into the equation, he states that the minimum (digital) bandwidth to reproduce a waveform of a specific frequency is twice that frequency, so the horizontal sync should be possible with 31250 points per second of scan rate. The actual video (RGB intensity) can be modulated in a completely independent way.

That's why I suggested using a TV sync generator. Not only does it generate the right waveforms for video scanning (sawtooth and vertical retrace), it also de-jumbles the video signal and turns it into nice RGB. Plus, using interlaced video might help in dropping down the bandwidth a little.

[allthatwhichis]

I can see your mistake, dear Padawan...

[mixedgas]

Hum, Strong not, is the respect for the force of inertia in these PLers.

You need a special low inertial galvo for video

A resonant scanner at 1/4th or 1/2th the video rate: ie the 4khz or 8 khz resonant scanner.

GSI makes them, and since they are sinusoidal, you have a laser diode projecting a grating of spots on the back side of the mirror to generate the pixel clock.

http://www.camtech.com/products/Reso...resonants.html , you want a IDS, not a CRS, usually. Although there are CRS, and a few were on ebay last year. many people bought them to get the g120s in the auction and probably tossed the CRS when it didnt scan like a normal galvo.

Here is one at one 10th the factory price, but the factory price includes the pixel clock board. This just has the pixel clock sensor, which is a ir laser diode. I have a identical one in a desk drawer back in Akron, if any one is interested. Just drive with a sine wave. You are playing the frequency lottery as you dont know if its a 4K, a 8K, or a faster special. They have a very high "Q", so the excitation has to be right on the money, frequency wise, or they just set there.

Steve

http://cgi.ebay.com/GENERAL-SCANNING...item56360c13da

[Psi]

Thanks guys, this is some really interesting info.

About sync probs at high speed...
Couldn't you have one motor doing both spinning mirrors with some gearing.
That way they would always be in sync with each other.
Any change in the speed would just change the overall framerate.

[buffo]

You're still considering video as a digital signal, with discrete points (pixels) where the X scanner should stop.

No, I'm not. I know scanners are analog devices.

One more time:

I'm considering that in order to move the beam across a horizontal path of roughly 3 degrees at a rate of 15,700 cycles per second, that will equate to a point speed of roughly 188,000 points per second. This is the theoretical worst case - just barely enough deflection to make TV images possible - and it still falls far outside the capabilities of modern scanners.

I know that a CRT doesn't care about pixels. I also know that scanners don't care about points. But the fact remains that the ILDA test pattern is *THE* standard that is used to rate the speed of the scanners under a given set of conditions. (Namely, displaying the aforementioned test pattern at 8 degrees.) Yes, it's lousy, but it's all we've got.

And since the most demanding portion of that pattern is the central circle (where the points all lie outside the square, and the resulting image is severely distorted, since all 12 points are rounded off to perfect curves that lie inside the square), the scanners are operating at their absolute maximum speed when drawing that circle. (They are said to be ballistic.) Thus, this is one (albeit cumbersome) way for us to figure out what the maximum bandwidth of the scanner is.

Now, if we know that it takes 12 points to make that circle, then it's clear that at 30Kpps the circle will be drawn 2500 times. Thus, you have a small signal bandwidth of 2500Hz. Get it?

And since the circle is about 3 degrees wide, that's your definition of "small step". Scale everything up to video frequencies, and you'll see that 15.7Khz = 188Kpps.

If you want to define an even smaller step, you could, but really, 3 degrees is pretty darn limiting already for a television, don't you think?

OK - let's assume that the absolute maximum speed of the scanner is 10Khz over a .1 degree step. Not suggesting that this is reasonable - but for the purposes of this discussion, let's go with it. Fine. That might mean you could reach video speeds with just 47Kpps scanners.

But your display angle would be .1 degrees wide! Thus my wise crack about only being 1 pixel wide... (Do the math... How long of a throw would you need at .1 degrees to be able to generate an image that was the size of a standard 40" TV?) It was a comment on the absurdity of using 30Kpps as the bandwidth limit for the scanners, not some confusion about points being equal to pixels.

but nobody states that the intensity signals and the X/Y signals SHOULD be in sync.

Actually, they have to be very much in sync, or at least an even multiple of one another. Otherwise you'll end up with nothing but noise where you want your image to be. True, they are still all analog signals, but the timing between them is a very important part of *any* laser show controller.

You assume that I'm thinking about the scanners in terms of points. I'm not. I know they are analog devices. But since they don't say "Scanner A supports a maximum lock-to-lock bandwidth of such-and-such frequency" in their promotional literature, all we have to go by is the lame points-per-second rating. I agree, it's a lousy way to do it, but it's all we have. (I didn't come up with it - don't blame me!)

Now, Steve's suggestion to use a resonant scanner is a good one, as it solves the spec problem for us. A resonant scanner *does* give you the maximum bandwidth information for a given deflection angle. But even an 8Khz unit is still 2X too slow for standard NTSC video. (To say nothing of HD resolutions.)

Does anyone make a 16Khz resonant scanner? That would be ideal. Anyone? Anyone? Buehler?

Yeah, thought so... That's why they use the spinning polygonal mirror...

Adam

[mixedgas]

[QUOTE=buffo;120730]No, I'm not. I know scanners are analog devices.

One more time:


OK, its nitpick with Buffo day :-) Then attack myself a bit.


Delete BLAH BLAH BLAH

I know that a CRT doesn't care about pixels. I also know that scanners don't care about points. But the fact remains that the ILDA test pattern is *THE* standard that is used to rate the speed of the scanners under a given set of conditions. (Namely, displaying the aforementioned test pattern at 8 degrees.) Yes, it's lousy, but it's all we've got.

Its better then what we had before, which was Nothing.....



Now, Steve's suggestion to use a resonant scanner is a good one, as it solves the spec problem for us. A resonant scanner *does* give you the maximum bandwidth information for a given deflection angle. But even an 8Khz unit is still 2X too slow for standard NTSC video. (To say nothing of HD resolutions.)


Now attacking myself:

I should have clarified.

The 8 Khz scanner is tunable to 7,785 hz. Since its a sinusoid instead of a ramp, you scan in both directions. Drum roll please.... instant 15,570 hz.

the 4 khz model is 3,892.5 hz. Two beams and two directions or half NTSC video, your choice.


Does anyone make a 16Khz resonant scanner? That would be ideal. Anyone? Anyone? Buehler?

Yes, but they are much more expensive then the "8 Khz" IDS.

.

Yeah, thought so... That's why they use the spinning polygonal mirror...

Or a wide window AO deflector and some fancy beam compression optics. . I own one of the AO deflectors, Dr LAVA has seen it.

And one hell of a decoding arrangement and buffer memory, Doug Dulmadge on Alt,Lasers built several of them, using AO< Galvo, and Polygons. He can tell you its a witch with the polygons unless you have massive amounts of memory.

There is/was a commercial one from Korea in production with a Purelight Star in it. The staff were/are ILDA members. Samsung I think.

Steve

[Stoney3K]

I get the math oops I made now. I was thinking of the scan rate (points per second) as being the maximum end-to-end bandwith for a given scan angle. So I assumed that on fx. 20kpps at 8 degrees, the scanner could make that 8 degree travel 20.000 times per second (when projecting a point on either side), where in reality, it can not.

Specifying the lock-to-lock bandwidth, as you say, is a much better spec, or even still, the bandwidth-scan angle product, like they do with op-amps. Dropping the scan angle (gain) will increase the practical bandwith, and vice versa.

When you're using resonant scanners, you still have the problem that the waveform you're using is a sine, which will probably distort your video in either direction as you need a sawtooth (or a triangle with some smart deinterlacing circuits)

To go with your claim on the X/Y and intensity signal sync: My point was that the intensity signals don't need to have discrete timing which is synchronised with the scanner's position. In theory, you can output the intensity signals a lot faster if you want to (and if your lasers don't start complaining), as the ILDA spec says nothing about DAC sampling rates needing to be equal. In reality, it's a lot more practical to do so and run everything off a single clock, but modulating the laser in between steps should be possible. (That is something different from being synchronised to the start of the frame, which, obviously, should be pretty useful)

[buffo]

I get the math oops I made now. I was thinking of the scan rate (points per second) as being the maximum end-to-end bandwith for a given scan angle.

Not to worry. It's a common mistake, and you're not the first (nor will you be the last) to make it. I actually used to think the same thing myself.

We actually had a *very* long discussion about this a couple years ago, and the consensus was that the scanner companies don't provide enough useful information with their spec sheets. At the time it was hoped that things would change. But, alas, they *still* don't provide enough information on their spec sheets!

Specifying the lock-to-lock bandwidth, as you say, is a much better spec, or even still, the bandwidth-scan angle product, like they do with op-amps. Dropping the scan angle (gain) will increase the practical bandwith, and vice versa.

Yep - spot on. But it seems that no one wants to publish bandwidth-per-scan-angle information. I have no idea why, however.

If you're interested in this subject, Bill Benner has an *excellent* post from 2007 that explains it in far better detail that I ever could. When I first read this post it was one of those "Eureka!" moments when I finally understood what the ILDA test pattern was doing.

When you're using resonant scanners, you still have the problem that the waveform you're using is a sine, which will probably distort your video in either direction as you need a sawtooth (or a triangle with some smart deinterlacing circuits)

Good point! The sinusoidal movement of a resonant scanner would stretch the image in the middle (or squash it on the sides), but you could compensate for that by varying the timing of your color modulation. (Ugh... That would be tricky to adjust though!)

Anyway, I don't think I'll be building a raster-scanning projector any time soon. (At least, not one that can run at NTSC resolution.) But it *can* be done, if you throw enough money (and talent) at the problem... And some of the videos I've seen suggest that the end result is pretty impressive.

Adam

[Psi]

And some of the videos I've seen suggest that the end result is pretty impressive.

got any url's ?

[buffo]

No, sorry. Hayden had a couple videos on his laptop and we watched them at SELEM 2007, but I didn't get a copy of them and I don't think they're on youtube...

Adam