Homebrew PCBs

Volkan Sahin wrote: 
> 
> 
> Current printer speed is 32\u201d/sec in bidirectional mode and 16\u201d/sec in
> unidirectional mode.
> I always use unidirectional mode because of quality of the output it
> gives sharp well defined lines.
> 
> For example it takes 8 minutes to print
> ~8\u201dx1.3\u201d board in unidirectional mode at 720dpi resolution. 
> 


Great! You area already at top speed in the simplest manner, and is
faster than i imagine.

> New generation rotating mirror motors are PLL based
> they have a clock input and PLL locks the motor speed to external
> clock, so they should be accurate. For example, if we're printing at
> 720dpi with 2.5microsecond exposure time, scanning of 8\u201d width
> material requires 14.4ms and if we have a 6 faced mirror motor speed
> will be 694rpm 
> I did some experiments at 200mW and did some
> rough measurements it seems most of the power loss is because of the
> 45 degree mirrors and it is ~20% and total loss is around 30%

That 2.5�S value is experimental? Or how did you calculated it? Which is
after all the size off the printed dot?
I was following Adam Seychell. 2mils dot (~0.005cm) printing a 20cm line
(~8") (like a A4 laser printer) is 0.1cm� area so is 6mJ and with 85mW
laser is ~0.07s which gives the 14Hz (lines per second) and 140rpm with
the 6faces mirror. With the 30% loss slowing to 100rpm is a good number.
And there is the thing you can't use all mirror surface but that needs
higher accuracy and longer distance from mirror to pcb  and not slower
rpm. You have to wast a portion off the surface for start counting only
when the reflected beam touch the light sensor on side off your
lenses/printed material.
But still is almost 10times faster than your actual machine. Making a
20x30cm (~100inch�) printed in 7minutes plus turnover mechanical delays
in each line (ok 10minutes). Which is as good as 30.000+eur film
photoplotters (ok those make 8000dpi not 720) but directly on pcb! 
Your 2.5�S number is about 5 times faster, so please tell me i am wrong
because i want to be :>

But can't the actual machine print in both directions? If the
misalignment is because delay between encoder signal and firing the
laser, which makes printing differ in the reverse directions. You could
try (if you not made it yet) synchronise the laser with the signals
coming from the encoder directly. Between positions it 'loads' (shifts)
the next bit, and is the encoder signal which will fire it. 

BTW i refer various times 'printing traces and not dots' (compare when
to turn on and off - less bandwidth) because is how seemed to work the
crap noisy photoplotter i have used. I think LPKF laser photo engravers
work like that too (just guessing). But if you are loading the all dots
line in to the buffer and shifting them, is after all less computation
work, and you already have it done for the epson print head.


> I think it is a good idea, as far as I
> know Sun Microsystems used similar concept in their laser printers.

Which similar concept? Connect the printer to VGA port? Or use the
framebuffer in some mcu chips to build the raster processor?
I will look better if is still possible the hsync as dot clk (since it
could avoid the fpga with memory and usb connection). Hsync as dot clk
needs a much simpler circuit but requires the all thing completely
synchronous, if pc fails to update one single line all print gets
twisted. The idea behind 1 line per frame made it asynchronous.