RE: [avr-chat] will an avr be fast enough...

[Alan Kilian]

> I mean, can you possibly fit 16 channels worth of 10bit ADC data 
> down a 38khz IR pipe???

    You can fit 16 channels of 10-bit data down any sized data-stream 
    you want. You will need to choose the rate at which you sample the
    data to match the maximum transmission-rate of the data stream.

    So, let's be generous and say that a 38 KHz IR link can transmit
    2,000 bits-per-second. It probably can't, but let's go with it 
    for now. (The data sheet for this product shows an 800 bit-per-
    second data stream which is much more reasonable.)
    http://www.hvwtech.com/pages/products_view.asp?ProductID=166

    So, if you have 2,000 bits-per-second, and a data packet of 
    10 bits, you can transmit  2000/10 = 200 samples-per-second.
    (I'm ignoring that you will need some headers and some inter-
    sample start and stop bits, but we are just making an estimate
    right now, so I think it's OK to do that)

    Now, if we look at the original poster's information:

> 16-channel 10-bit A2D ... data out via an infrared link.  

> The max SPI clock of the MAX1031 is 10mhz

> I need a sampling rate of around 20khz/channel.

    We can start at "20 KHz per channel" and multiply by (let's say)
    15 bits-per-channel to be reasonable, so we need to read the A/D
    converter at 20,000 * 15 = 300,000 bits-per-second which is WELL
    under the maximum SPI clock of 10 MHz, so that's just fine.

    Now, what exactly does "20khz/channel" mean? Maybe it means EACH
    of the 16 channels needs to be samples at 20 KHz. That means we need
    to multiply our 300,000 bits-per-second by 16 channels to get
    4,800,000 bits-per-second which is still well under the maximum
    SPI clock for the A/D converter.

    (Note that this is grossly greater than the 2,000 bits-per-second
    that we optimistically estimated was possible with w 38 KHz IR link)

    So, can any AVR set its SPI clock at 4.8 MegaHertz? If I look at the
    data sheet for the Mega48/88/168, I see that the maximum SPI clock
    in master-mode is the crystal divided-by 2, and the maximum crystal
    frequency is 20 MHz, so we can indeed generate a 4.8 MHz SPI clock.

    Last but not least, "data out via an infrared link". We saw that a
    38 KHz IR link is not going to work, so how about IRDA?

    I went to digikey.com and searched for IRDA and found that 
    microchip.com makes IRDA control ICs, so I looked there. I quickly
    found an application note "Selecting an MCP21XX Device for IrDA®
    Applications"

        http://www.microchip.com/stellent/idcplg?
IdcService=SS_GET_PAGE&nodeId=1824&appnote=en012109

    It says:

        The IrDA standard SIR specification has a maximum
        baud rate of 115.2 Kbaud. If both ends of the system
        are controlled by the developer, this specification can
        be ignored. The MCP2120 allows a maximum baud
        rate of 312.5 Kbaud (although some work may need to
        be done with the optical transceiver), while the
        MCP2122 has a maximum of 115.2 Kbaud.

    So it looks like standard IRDA can't even come close to 4.8 megabits
    per second.

    That's it for me.

    Can anyone else come up with an IR transmission scheme that will   
    work?

                                -Alan

-- 
- Alan Kilian <kilian(at)bobodyne.com>