Open Source/Freeware Budget USB DAC

[mixedgas]

Why do we need a frame buffer? I think it is practical that if enough buffer space were given we could transfer the frames in smaller chunks like the sound cards do. The frame buffer just adds expense and added complication.

Yes I know Atmel does not make PICs... my comparision was just in capabilties, speed, and programmability.

even the USB sound card chip has a buffer.

Keep in mind, people who buy low end scanners often need to tune their scanners out of the box, thus, you need to do the ILDA test pattern at 12 bits and a smooth 30K.

Also you have no control over the windows timing, its not 100% synchronous, even though the USb spec says its pretty good.

Steve

[keeperx]

thats another great point.. if your frame is not updated in time, and its time to scan another frame, you can just replay the frame thats in the buffer..

[cfavreau]

I think that would be the only point. The sound card based DACs do have the problem of if things get too busy they repeat their buffer... albeit not a large buffer. A frame buffer less unit could and should do the same.

The USB sound cards don't have a full frame buffer though. Which is my point. In a cheaper DIY situation I think the frame buffer adds expense and complication I think. It can be done to work as well as the sound cards I think. You do have some control over windows timing. It isn't as bad as everyone makes it out to be.

There is no operating except the real time operating systems that you have 100% control over the timing which I guess is the point.

So I guess 12 bits is the minimum for XY and 8 bits for RGB? 8 bits are no good?

I am not sure I get the comment about the 12 bit and smooth 30k ... not everyone has 30kpps scanners... and you can use 8 bits quite effectively to tune your scanners... Are you suggesting a built in test pattern? Anyone can hook it up to a computer to play any test pattern they want... Smooth... yes it needs to be smooth and consistant, whatever speed it is set for.

One more question I guess:

Do people use a constant point output speed or do they vary it from frame to frame? 12kpps or 30kpps fixed output speed is easy to do. Varying the output speed is not as easy to do.

What about 10 bit for XY?

Alot of the frame buffer you can handle in software on the PC.

[heroic]

I am not sure I get the comment about the 12 bit and smooth 30k ... not everyone has 30kpps scanners...

Given that 30k scanners are so cheap now it seems pretty pointless to cripple the design from the beginning.

Here's my SPI DAC board. I used the single-DAC chips so that you could populate as many or as few channels as you need. This will connect to an Arduino or any other microcontroller that has SPI.

The extra space on the board is for the analogue correction amps which I intend to do using one of those digital potentiometer chips.

[mixedgas]

Chris wrote:

So I guess 12 bits is the minimum for XY and 8 bits for RGB? 8 bits are no good?

12 bits is OK, but 12 bit chips are usually more expensive then 16 bit these days.
Please do not even think of 8 bit XY, it is a waste of time, Norms gets shot down frequently around here for low resolution. At least 8 bits for rgb, NO TTL either.

I am not sure I get the comment about the 12 bit and smooth 30k ... not everyone has 30kpps scanners... and you can use 8 bits quite effectively to tune your scanners... Are you suggesting a built in test pattern?

No, I'm suggesting doing it right.
Look the ILDA test pattern is the standard, it is not scanned exactly right by sound card dacs, its a bit undersampled. There is a purpose for the standard, and many of us are trying to make things better by encouraging the use of the standard.

We have had many people with low end systems who have had to send galvos out for tuning because of poor QC, one axis off, flakey mirror sizes etc. And no way to tune on their own. The ILDA circle in the square does not work properly when truncated to 8. Most scanners today are 30K and getting cheaper., the difference between the so called 20K and basic 30 K is less then 75$, and often less then that. ILDA formats are +32768 to -32769, truncating to 255 x 255 or 1024 by 1024 looses points and is a TOY.

Also many of us are tired of people going cheap, the laser industry is flooded with nonstandard junk, and we get a lot of first time posters saying I bought a CHIN NE MODEL 666, with 8 bit dmx controled graphics, and its pretty useless, how do I upgrade.
We also hear a lot from people who are tired of seeing stuff on youtube that is aweful from the artistic point of view.

So why take a step backwards to crude, when it is no more work to do it right, There is a lot of good artware out there now that demands 24-30K, heck must stuff now flickers unless you run 30K. Images push 700 to 1100 points per frame.

And if a buffer is 10$ worth of parts, its worth it.

Chris, it seems you want to take everyone back to 8 bit 20K open loop. That doesn't cut it these days. You should come to SELEM and see what even a low end system does these days.

Not everybody is a 100% cheapskate, and your not going to get enough user volume if you just target the ultra low end.

And yes, we now push for constant point velocity at 30K for graphics and 22-24 K for beams, to fully take advantage of newer, low cost, scanner technologies and optimization software. We have packages now that render out of 3D studio max, and I'm sure 90% of the stuff on the PL graphics FTP site is 30K PPS, 16 bit oriented material

Steve

[cfavreau]

Well I am not here to take everyone back to the stone age. I am more of trying to fish out what every one thinks is a good bottom line to take. It is good to know why something is and not just that it is. So please don't take my posts wrong.

If 12 bits for XY plus 8 bits for RGB is what it is ... good. Now that we got that out of the way. 30 kpps flat... so you either go 30 kpps or 15k pps (or some multiple there of) if you really want to. I do not want to stray from a standard... one thing the hobby community can use are standards... so we aren't wasting our energy all over the place.

I have a set of 30 kpps scanners myself and a set of 12 kpps scanners.

I am trying to go with the title of the thread here: "Open Source/Freeware Budget USB DAC".

So with that in mind I think we can keep the cost of parts and labor if someone were to sell a kit or something ... to below $100. Depending on the rest of the parts maybe below $75 or lower. I think that would be good value. Maybe even $60 if you leave off some parts like an external power supply.

How bout any other outputs like a shutter?

[drlava]

A full frame buffer is not absolutely necessary, but it is nice under high system load, the buffer underrun effects would not be as visible.

In my development of ILD SOS optimization software, one thing I have found is that, contrary to what you might expect, having more control points at a rate higher than your scanners improves their maximum performance. So if we are going to try to construct this from the ground up, I suggest a max point rate of at least 60k. Also, as Mixedgas said, 8 bit x and y is wasting your time, we should really shoot for 16 bit. For modulation, 8 bit is O.K.

But, also as has been stated, with RIYA dacs available at an inexpensive rate, and given that the sound card DAC fills the low end, consider carefully your goals and market size before jumping in.

[mixedgas]

How bout any other outputs like a shutter?[/QUOTE]

BADLY NEEDED, as are 4 to 8 beamtable outputs.

Oh, and these days its RGBV, the 405 diodes have invaded, and we also use dichros to blend in 633 with 650-660 nm diodes.


also galvos ship these days wired for full diff two wire inputs, so add one more dual opamp. Its Ok to leave off the PSUs, or to find a regulator that can take 30V inputs from the galvo power supplies. You've got 5V from the USB, and a AC wallwart and a couple diodes can provide the rest, or use 3 2n2222s as a oscillator/driver and make your v- with a voltage multiplier. Lexel gets -20 and -15 for their laser power supplies that way. Also, the atmels can take some overclocking. GCC is probably the micro language of choice, but assembler is going to give you more speed. Another trick would be to vary Vref on the dac chips, to change image size.
Steve

[cfavreau]

One more thing. If you get too close to the $200 mark what is the point. I would say the majority of us want to play with our lasers and not have to build everything. Building does give satisfaction but there is a line. Please correct me if I am wrong.

[drlava]

absolutely! it's under $130 or bust