High Speed Transmission Between two Props

steprog

Wow, things have changed here since the last I looked.  I got a project where I am transmitting about 1.4Mb/s of data over perhaps 1000 meters on fiber.  (Earthquake measuring equipment)  since I have used props for everything over the years I thought this would be another good app.
Anyway, I saw some discussions on this topic mainly some work by Bueu that looked good.  However I was hoping to find some actual code to use.  For doing this.  Any thoughts where?
BTW, it is a basic system demonstration that gathers 24bit data from 3 adc's packs this data and transmits to the surface.  Once on the surface this data is unpacked and output to 3 - 24bit dacs.  Nothing fancy here really, just a demonstration for now.
Thanks,Greg

MJB

http://forums.parallax.com/discussion/134641/demo-high-speed-multi-prop-to-prop-communication

maybe code is on the OBEX  or drop Beau a PM

DavidZemon

Talking Spin? Or open to C/C++ options?

steprog

Spin is what I do at the moment.  I haven't seen any of this on the OBEX library, but I will check again.Thanks,Greg

Publison

Greg, I'm checking all my drives. I know I downloaded when it was first published, but searches of now 4 TB of drives shown nothing. I may be looking for the wrong Spin file. If any one knows Beau's original Spin file name, please post.. Thanks!
Jim

steprog

Thanks Publison

Leon

You need PASM for high-speed.

Publison

You also may want to look at this thread:
http://forums.parallax.com/discussion/116493/pasm-coding-challenge-4-wire-prop-to-prop-communication

steprog

Yes, it always boils down to PASM for any real processing.  I did look at the 4 wire some, but I just have a single fiber coming from down hole picking up signals for 3 geophones.  So a single - wire or fiber solution is needed.Thanks,Greg

Cluso99

Is it one way transmission?
If one way then a modified fullduplexserial for receive only should work for 1.4Mb/s. I thought this had been done. The mod is to remove the task switching code.

jmg

Yes, it always boils down to PASM for any real processing.  I did look at the 4 wire some, but I just have a single fiber coming from down hole picking up signals for 3 geophones.  So a single - wire or fiber solution is needed.


It's unclear, do you have control of both ends of this, or is the stream coming at you already in some format ?( Usually there are limits on what goes down the hole.)

If you only want to read a stream, then simplex UART is the easiest to code.Where does the data stream end up ?

steprog

yes, it is one way, and the card downhole will have a prop. too.  The data format is up to  myself.  The data consists of 3 24 bit words for each of the 24 bit adc's.  I am assuming conservatively to use 32 bits for over head with a data sampling rate of 14.4 Khz .  
so figure 3 x 32 x 14.4ks/sec  just over 1.4 Ms/s
Greg

jmg

I'd suggest targeting 2Mbd, which gives a 14.4k paced burst of 12 bytes, then
(2M/14.4k)-3*32*10/8 = 18.88888 Stop bits
A FT232H/FT2232H would then be able to generate or receive that stream for easy testing.

steprog

Thanks jmg,not a bad ideal, but the hardware is built and now just trying to bit bang the data into a single pin.  I actually thought this would be no big deal, but I decided to check to see if someone already did this before I started coding up something.  I did see that Beau, had something.  I read the description but didn't see any code.
I am hoping to deliver the data as 1 3x32bit packets, unpack it into 3 memory locations for other cogs to access and send to the dacs.

Peter Jakacki

I readily achieve 2M baud speeds in Tachyon and your application seems to be half-duplex mainly so that is even easier. The PASM for such a driver is only a few lines long. I have also used single-mode fiber and run it with a 10Mbps stream encoded in 20Mbps to zero out any DC components. 

steprog

Sounds Great Peter, is there any cold you can share?Greg

Ariba

Here is the Link to Beau's thread:
Andy

steprog

yes been here but no code..or I don't know what I am doing..

MJB

Sounds Great Peter, is there any cold you can share?Greg

Hi Greg
Peters TACHYON 2.4 code shows this
sysfreq        = 80_000_000
baud        = 230400            ' <-- user change - tested from 300 baud to 3M baud

so taking for example the standard EXPLORER image he has on his DropBox
(see his signature  and debugger)
and changing BAUDRATE
to what you need should be all what is required.

then the command

2000000 CONBAUD

at the commandline and then a reboot should be all you need
to then try it via FTDI USB-serial adapter. (aka PropStick)

MJB

Sounds Great Peter, is there any cold you can share?Greg

Hi Greg
Peters TACHYON 2.4 code shows this
sysfreq        = 80_000_000
baud        = 230400            ' <-- user change - tested from 300 baud to 3M baud

so taking for example the standard EXPLORER image he has on his DropBox
(see his signature  and debugger)
and changing BAUDRATE
to what you need should be all what is required.

then the command

2000000 CONBAUD

at the commandline and then a reboot should be all you need
to then try it via FTDI USB-serial adapter. (aka PropStick)



but --- thus is not SPIN it is FORTH !!!
but you might like it ;-)

steprog

I guess I am spoiled, but the ideal I have to learn forth to finally get a high speed prop to prop interface seems excessive, (although it is in my bucket list someday)  I also looked at the dropbox but unfortunately it is just the binary code.  I was hoping that maybe I could use whatever asm files to roll it again in using a spin header.. no such luck.
hmm...

Peter Jakacki

Pressed for time but the "4th" link in my  sig points to the all the files and folders, you will find the kernel files in the root directory.

Alexander (Sandy) Hapgood

Anyway, I saw some discussions on this topic mainly some work by Bueu that looked good.  However I was hoping to find some actual code to use.  For doing this.  Any thoughts where?
I think this is the file you're looking for.  It has an MIT license so should be no problem using it.

Sandy

LoopyByteloose

Full Duplex Serial may be too slow, but the code may be a useful comparison to Beau's.

Try looking at the FullDuplexSerial.spin object in OBEX for code.  (Search "Full Duplex Spin")

While it is a Spin object, it really has PASM providing the serial speed.  It is a bit difficult to remain in your comfort zone of programing entirely in Spin.


And if that doesn't help you, you can always contact Beau directly.  That was suggested above.

steprog

Thanks all, I got carried away with a 2 year old grandson.. You given me some good stuff here thank you.  I will dig through this code and ideas and see what works.. Thanks again.Greg