Shop OBEX P1 Docs P2 Docs Learn Events
Programmable elliptic filter — Parallax Forums

Programmable elliptic filter

Paul BakerPaul Baker Posts: 6,351
edited 2005-06-03 16:16 in General Discussion
Im curious how fast the the SX can be run using a peltier junction, heat sink and fan. To do this I'll need a programmable clock source, after searching high and low for a programmable oscillator with a·wide frequency range and high top frequency, I think that Analog Devices AD9581·is the best choice (5V, 0-180Mhz). The output is a DAC sinewave·and the package also contains an analog comparator to create a square wave (exactly what I want). The specifications state a low pass filter is placed between the output of the DAC and the comparator to filter out aliased versions of the signal thereby reducing jitter in the square wave. Thier example shows a 7th order elliptic filter whose corner frequency is 70MHz, I will need the filter's corner frequency to be adjustable (at least in the 100-150MHz range). While I studied the basics of analog filter design in college I have no experience and am having a difficult time finding a filter fitting these criteria that doesn't employ an expensive DSP solution to the problem. Can anyone guide me in the right direction, or think of a means of accomplishing the reduction of aliased frequencies without using traditional filters?

I hope this isn't a "cricket chirping" question.

As an aside, is there any problem driving the oscillator pin of the SX running at 5V with a 3.3V signal?

Post Edited (Paul Baker) : 6/2/2005 6:43:55 PM GMT

Comments

  • SteveWSteveW Posts: 246
    edited 2005-06-03 09:44
    Why wouldn't you just use an SK-key? The programmable clock generator works pretty nicely, it's off the shelf, and, it doubles as a rather nice debugger.

    While what you're proposing sounds, err, interesting, I think you're missing the point in a few ways.
    1) A sine wave clock to the SX is likely to be fine, if it meets the voltage thresholds and slew rate erequirements
    2) How much jitter is there on the square wave output? Is it actually enough to be fretting about, in this application?
    3) Since you don't care much about phase or amplitude performance in the filter (since its output is going straight into a comparator), it can be very, very crude. A simple non-tuned LC (or RC) will do - all you're trying to do is remove the quantisation noise. If you're using this to test the topspeed of an SX, I'd imagine that you'll care most about frequencies between 90 and 130MHz. Build a crude RC filter with a 3db point at, say, 150MHz, and I suspect it'll do all you need. (DSP-based filters aren't the answer in any way, shape or form, for this. The cost will be ferocious, and an ADC -> DSP -> DAC capable of filtering 150MHz, would _still) need a recnstruction filter on its output, so you'll have achieved nothing (except funding TI for a year...)

    However, teh SX-key gets my vote [noparse]:)[/noparse]

    Steve


    Steve
  • BeanBean Posts: 8,129
    edited 2005-06-03 11:03
    The SX-key is limited to 110MHz.
    Bean.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "SX-Video·Module" Now available from Parallax for only $28.95 http://www.parallax.com/detail.asp?product_id=30012


    Product web site: www.sxvm.com

    "It's not getting what you want, it's wanting what you've got."
    ·
  • SteveWSteveW Posts: 246
    edited 2005-06-03 12:27
    Ah, I thought I remembered a bit higher than that - can't get to mine at the moment, but I've not had much joy above 100MHz with heatsunk (not actively cooled) SX28s anyway. Still well worth a go, for our original poster, before embarking on his massive clock-building project...

    (Also, the original poster might want to go for something cheap & cheerful (and designd for the job) - a video clock synthesiser.
    ICST ( http://www.icst.com/ )make some very usable candidates. Serially controlled (I2C or some other, according to the device you choose), low jitter, and easy to drive (or, at least, as easy as anything clocking at hundreds of MHz is...)
    ICS1523 would seem to be a reasonable candidate. My old favourite, that I used to use for faking up clocks, seems to be obsolete [noparse]:([/noparse]

    Steve
  • Paul BakerPaul Baker Posts: 6,351
    edited 2005-06-03 15:11
    Thanks Steve the ICS1523 looks promising (I was hoping for upto 200Mhz, but 150 is acceptable), I noticed it is a 3.3V chip, which brings up my original aside question regarding 5V SX being clocked with 3.3V signal.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
  • NewzedNewzed Posts: 2,503
    edited 2005-06-03 15:13
    If I am not mistaken, I think the logical 1 level on the SX/B is about 1.7 volts.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Sid Weaver
    Need a bezel for your LCD?

    Newzed@aol.com
    ·
  • SteveWSteveW Posts: 246
    edited 2005-06-03 15:25
    Yeah, as long as you make it past the thresolds (and you will, with an LVTTL output), it'll work nicely. Going from 5V to 3v3 is always more of an issue.
    The ICS8402 is good out to 350MHz...
    (A little PCB with a SX, LCD and PLL chip, with a few buttons / serial port to set the frequency would be a very handy project, if anyone's bored...)

    Steve
  • Paul BakerPaul Baker Posts: 6,351
    edited 2005-06-03 16:16
    Im familiar with the switching levels for the I/O pins, but I wasn't sure if the OSC pins are the same. After perusing ICS's products, I think the ICS525-02 is a better fit, It permits 5V operation and can operate upto 250Mhz, it requires 19 pins to operate the inputs, but this is a testbed so using half of the SX52's inputs isn't a serious issue, plus I can throw some '595s in if I want those pins back.

    Thanks to everyone for thier input.
Sign In or Register to comment.