Just need to do some more calculations.
The other one options is to use AVR with biggest ram to fit at least one
scanline. So when buffer has at least one line in then seek for index
pulse from laser position and start scanning. After half a rotation the
line is ended and buffer can be updated. If buffer doesn't fill until
next index is reached then just wait to data arrive and try on next
index cycle. But this approach will probably result with speed reduction
(only to get half speed).
And another news. I just wonder how I didn't burn out my laser.
I ordered same one for spare and 120mW to test if I can get more speed.
The laser of choice was:
http://www.ebay.ie/itm/151023934656?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649and there are stated 3V <280mA. But I quickly realised that I can't
modulate the laser as there are some DCDC converter in module. So I
remove them and connect LD to power supply with current limiter. I
slowly increase current and observe laser output and current and
voltage. It's need around 20mA and 4.5V to start working. After that
when I increase current to 50mA the voltage is around 5V and just little
higher with 100mA. So As I don't have laser datasheet (doesn't know what
diode is in) I just recalculate the input of DC/DC (3V @ 280mA) is same
power as (5V @ 168mA) and assume 90% of efficiency of DCDC converter and
I got 151mA. My BJT driver feed 150mA when it's total cold and 140mA
after it's heated up with hairblower.
After I got stronger laser
http://www.ebay.ie/itm/161132483080?ssPageName=STRK:MEWNX:IT&_trksid=p3984.m1439.l2649the data says <450mA and 3 to 5V voltage. I connect the laser to my PSU
set voltage to 5V and limit current. Then I crank up the current and
when I hit around 25mA at 3V the diode start to lase. After that when I
crank up current up to 390mA voltage stay around 3V and just little
higher current there are strange jump of voltage to 5V and decrease of
current to 120mA (like negative resistance). It was so strange to me
that I remove DCDC from that laser too and connect 6x1N4007 diodes and 1
ohm resistor in series (aka laser dummy load) and check the output what
circuit feed to laser. And I got that device try to feed 110mA to the
laser whatever voltage is. But efficiency of DCDC vary from 60% to 95%.
Then I test the circuit from smaller laser and I was surprised. They
output is just 30ma! I checked other one and got same result!!!
So the DCDC has less than 20% efficiency. And I run diode instead on
30mA with 145mA and didn't dye. I just wonder how long will work..
Well long letter to learn that this kind of diode output is close that
current consumed in mA, or at least that powering diode with asme mA as
mW is is safe.
...and 20mW diode give me same power as 100mW diode if feed with 145mA.
On 14. 12. 2014 00:49, Harvey White
madyn@... [Homebrew_PCBs] wrote:
> On Sat, 13 Dec 2014 14:40:06 -0600, you wrote:
>
>> Hi, maybe you should consider double addressable ram, is a bit more expencive, but it can be written and read at the same time, ( has two address counters in it ), I'm not shure but they call it LIPO or something likethat.
>> Anyway it is sure fast enough and wile your avr is pumping bits in it the cnc can read them. It depends now on what you need be the fastest.
>
> Another way is to multiplex the address and data from application to
> application, so that half of the time, the AVR is writing to the RAM,
> and the other half of the time, the CNC can read it.
>
> This is relatively easy to do, but requires the memory to be twice as
> fast as your cycle time, if not a bit faster for safety. I'd use a
> CPLD or FPGA to do the multiplexing, just because it's easier. If you
> don't have the tools to do this, then counters/registers and
> multiplexers work just fine. You'll have to play games with the AVR
> if you're using direct memory access, because the memory access cycle
> is not absolutely in phase with the clock.
>
> Just because you have the clocks for the processor and the clocks to
> the CPLD in phase does not guarantee that the two requests would be
> synchronized.
>
> Your best bet would be to have the whole cycle time so short that the
> processor can write at any time, then expect to get/write data
> correctly.
>
> Strangely enough, writing with using a port (Programmed I/O) is easier
> to synchronize (you can loop and wait on a status) than direct memory
> write/read synchronized to the processor clock. If your FPGA/CPLD
> has dual port memory (Xilinx 3AN series do...) then this can help if
> your assumed buffer size is small enough.
>
> Harvey
>
>
>>
>>
>> cb
>>
>>
>> On 12/13/2014 11:06:25 AM, Slavko Kocjancic eslavko@... [Homebrew_PCBs] <homebrew_pcbs@yahoogroups.com> wrote:
>> On 13. 12. 2014 11:39, 'Tony Smith' ajsmith1968@...
>> [Homebrew_PCBs] wrote:
>>> At 1,000 DPI, the laser isn't going to be anywhere near it's supposed to be
>>> anyway. Close enough thought.
>>>
>>> The beam of the laser will be probably bigger than a dot a 1,000 DPI (which
>>> is good as the overlap fills in the gaps) depending on a bunch of things.
>>>
>>> Dunno how LinuxCNC does rastering, but often the laser is left on between
>>> dots so that evens things out. Sometimes you want dot - dot - dot lasering,
>>> but not in this case.
>>>
>>> Tony
>>>
>>
>>
>> The laser dot is just little under 1000dpi in size. But I have limited
>> on off timings so the width I can controll is about 700DPI. But this is
>> true only for rotation axis. I advance board for only 0.01mm each
>> revolution so there are plenty of overlap. And I do leave laser on if
>> there are more same dots in a row.
>>
>>
>> My preprocesor program just crunch the coordinates and make file stream
>> of '1' and '0' for laser for each 15us. The linuxcnc with component
>> 'streamer' then stream this each 15us to the laser. And here is
>> bottleneck as I can't go faster (but I want) and I need to make some
>> special hardware to stream much faster than 15us/pulse.
>>
>>
>> I think to use some AVR with plenty of ram and to upload entire stream
>> on it, and after that just stream bits. At 1 bit per 1us the AVR will
>> work hard. But seems that static ram is just too small to fit entire
>> image, I will probably use dinamyc ram (4Mbits ∗4). As I calculated I
>> don't need to worry about refresh as cycling as streaming is fast
>> enought to keep memory refreshed. But need to cary refresh when data is
>> uploading and waiting start of operation where is enought time to proper
>> manage refresh cycle.
>>
>>
>>
>>
>>
>>
>> ------------------------------------
>>
>>
>> ------------------------------------
>>
>>
>> Be sure to visit the group home and check for new Links, Files, and Photos:
>> http://groups.yahoo.com/group/Homebrew_PCBs
>> ------------------------------------
>>
>>
>> Yahoo Groups Links
>>
>>
>> To visit your group on the web, go to:
>> http://groups.yahoo.com/group/Homebrew_PCBs/
>>
>>
>> Your email settings:
>> Individual Email | Traditional
>>
>>
>> To change settings online go to:
>> http://groups.yahoo.com/group/Homebrew_PCBs/join
>> (Yahoo! ID required)
>>
>>
>> To change settings via email:
>> Homebrew_PCBs-digest@yahoogroups.com
>> Homebrew_PCBs-fullfeatured@yahoogroups.com
>>
>>
>> To unsubscribe from this group, send an email to:
>> Homebrew_PCBs-unsubscribe@yahoogroups.com
>>
>>
>> Your use of Yahoo Groups is subject to:
>> https://info.yahoo.com/legal/us/yahoo/utos/terms/
>>
>>
>>
>>
>>
>>
>> ---
>> This email has been checked for viruses by Avast antivirus software.
>> http://www.avast.com
>
>
>
> ------------------------------------
> Posted by: Harvey White <madyn@...>
> ------------------------------------
>
> Be sure to visit the group home and check for new Links, Files, and Photos:
> http://groups.yahoo.com/group/Homebrew_PCBs
> ------------------------------------
>
> Yahoo Groups Links
>
>
>
>