Shop OBEX P1 Docs P2 Docs Learn Events
Max System Clock — Parallax Forums

Max System Clock

Lee MarshallLee Marshall Posts: 106
edited 2007-10-06 01:02 in Propeller 1
Ive been using the prop for a while now with its internal rc oscillator, and im wondering if there is a max internal system clock frequency.
eg, could i use a 60mhz xtal and use the 16xPLL to bring the system clock to 960mhz(something doesnt seem right)?
if there is an official maximum, can the propeller be successfully overclocked beyond it?

▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
Holy parallel processing, Batman!
«1

Comments

  • deSilvadeSilva Posts: 2,967
    edited 2007-10-03 21:59
    There have been extensive discussions about this - some started by Sapieha... Have a look there.
    Or have look into the datasheet: That is what Parallax guarantees...
    - 10 MHz max crystal, when using the PLL
    - 80 Mhz max clock

    Experiments showed that you can extend this, when NOT using PLL16X (i.e. no PLL at all or PLL8X)

    Main constraint seems to be, that PLL ALWAYS multiplies by 16 and divides as with counters and video. There is a maximum frequency for this 16xPLL: It becomes tricky above 160 MHz...

    Many persons run their Prop with 96 MHz! (Whith 12 MHz crystal and PLL8X or 6 MHz and PLL16X)

    I wonder how much you could have done without a precise timing....

    Post Edited (deSilva) : 10/3/2007 10:07:04 PM GMT
  • LeonLeon Posts: 7,620
    edited 2007-10-03 22:01
    The spec. states 80 MHz (5 MHz crystal and x16 PLL) as a maximum. People have got them to run faster by over-clocking, of course.

    Leon

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Amateur radio callsign: G1HSM
    Suzuki SV1000S motorcycle
  • crgwbrcrgwbr Posts: 614
    edited 2007-10-03 22:27
    mine seems to top out at about 100 Mhz. I got that by using a 50mhz resonater from my SX28 and a pll2x setting. Above that freq, things get a bit funky. lol

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "... one of the main causes of the fall of the Roman Empire was that, lacking zero, they had no way to indicate successful termination of their C programs." -

    "If Python is executable pseudocode, then perl is executable line noise."

    "The best accelerator available for a Mac is one that causes it to go at 9.81 m/s2."

    "My software never has bugs. It just develops random features."

    "Windows isn't a virus, viruses do something."

    "Programmers are tools for converting caffeine into code."

    "Enter any 11-digit prime number to continue."
  • deSilvadeSilva Posts: 2,967
    edited 2007-10-03 22:35
    I have reasons to think that your systems is just running at 50 MHz... Have you validated it by measuring a frequency derived from the clock at an output pin?

    You don't need expensive equipment; this SPIN code will do
    DIRA[noparse][[/noparse]0]:=1
    REPEAT
      !OUTA[noparse][[/noparse]0]
      WAITCOUNT(CNT+ 100_000_000)
    



    Don't use CLKFREQ!
  • Paul BakerPaul Baker Posts: 6,351
    edited 2007-10-03 22:49
    I am currently working on max speed vs temperature measurements, they will be incorporated in the final datasheet, it is not going to be a definitive "your Propeller will do this speed" but a "here is the average of what we've tested", you would necessarly need to use a derating factor to guarantee reliability.

    crgwbr, you cant use an input of 50MHz and a PLL of 2, the PLL won't lock onto a frequency that fast, 10MHz is about as far as we recommend pushing it.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Paul Baker
    Propeller Applications Engineer

    Parallax, Inc.

    Post Edited (Paul Baker (Parallax)) : 10/3/2007 10:55:01 PM GMT
  • Fred HawkinsFred Hawkins Posts: 997
    edited 2007-10-04 00:02
    Paul Baker (Parallax) said...
    ·....you cant use an input of 50MHz and a PLL of 2, the PLL won't lock onto a frequency that fast....
    Uh oh.· Anyone want 4 66mhz crystals?·$1 for all of them.

    (Gonna keep hands in pockets at next computer faire)·
  • deSilvadeSilva Posts: 2,967
    edited 2007-10-04 00:21
    Fred, don't panic! Haven't you read my posting (second above)?
    You can of course use them WITHOUT the PLL!
    66 MHz is a nice speed..

    And BTW: I pay 0.25 Euro per HC-49 crystal in Germany - so: No, thanks smile.gif

    Post Edited (deSilva) : 10/4/2007 12:26:23 AM GMT
  • Fred HawkinsFred Hawkins Posts: 997
    edited 2007-10-04 01:15
    yeah, well these have four pins so they are worth more.
  • deSilvadeSilva Posts: 2,967
    edited 2007-10-04 01:20
    So they might be true oscillators?
    Or - in fact - they have to be smile.gif
    This indeed is 1.50 per piece!
    However they are generally rated for 5V supply. Do they work with 3V3?
  • Fred HawkinsFred Hawkins Posts: 997
    edited 2007-10-04 01:39
    The closest thing to a clue that I think they are meant for PCMCIA cards. They are marked:
    CXO-389C
    66.0000MHZ
    3C2
    KSS
  • deSilvadeSilva Posts: 2,967
    edited 2007-10-04 02:14
    That's fine. CXO means "Clock Crystal Oscillator" - I presume you know the pinout and how to use them?
  • Lee MarshallLee Marshall Posts: 106
    edited 2007-10-04 02:20
    it runs with a 14mhz xtal(from a dead zx81) and 8x pll(112mhz) shocked.gif it doesnt at PLL16x
    thats 224 MIPS with all cogs runnin...

    curious here, what sort of damage can you do if you run it too fast? or is it just a matter of reliability?

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Holy parallel processing, Batman!

    Post Edited (Mr Crowley) : 10/4/2007 2:29:53 AM GMT
  • Mike GreenMike Green Posts: 23,101
    edited 2007-10-04 02:29
    I believe it just stops working.
  • deSilvadeSilva Posts: 2,967
    edited 2007-10-04 02:44
    Mr Crowley said...
    it runs with a 14mhz xtal(from a dead zx81) and 8x pll(112mhz) shocked.gif
    This is very common...
    said...
    it doesnt at PLL16x
    Of course not! Why do you mention that?
    said...
    curious here, what sort of damage can you do if you run it too fast?
    You will heat it up, may be in places where it does not like it too much. See the dissipation curve in the datasheet. The diagram stops at 100 MHz where the Prop draws 100mA.
    The dissipation increases quadratically with the frequency! Don't let you fool by the straight lines - this is a double-logarithmic diagram smile.gif
  • Fred HawkinsFred Hawkins Posts: 997
    edited 2007-10-04 03:07
    deSilva said...
    That's fine. CXO means "Clock Crystal Oscillator" - I presume you know the pinout and how to use them?
    No, all I have seen is text descriptions so far.
  • deSilvadeSilva Posts: 2,967
    edited 2007-10-04 03:20
    sco020pecl.jpg
    They are all alike. The pins count as if they were 14 pins DILs.
    Add 3V3 and GND and output to Propeller's ClockIn port.
    Many SHOULD work @3V3.
  • Fred HawkinsFred Hawkins Posts: 997
    edited 2007-10-04 06:46
    Thanks deSilva. I think you saved me 2 spin commands in every program.

    Post Edited (Fred Hawkins) : 10/4/2007 5:56:32 PM GMT
  • crgwbrcrgwbr Posts: 614
    edited 2007-10-04 15:29
    Paul Baker (Parallax) said...
    you cant use an input of 50MHz and a PLL of 2, the PLL won't lock onto a frequency that fast, 10MHz is about as far as we recommend pushing it.

    Thats odd, I've used that setup to do serial communication with a PC. I haven't had any problems so far.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    "... one of the main causes of the fall of the Roman Empire was that, lacking zero, they had no way to indicate successful termination of their C programs." -

    "If Python is executable pseudocode, then perl is executable line noise."

    "The best accelerator available for a Mac is one that causes it to go at 9.81 m/s2."

    "My software never has bugs. It just develops random features."

    "Windows isn't a virus, viruses do something."

    "Programmers are tools for converting caffeine into code."

    "Enter any 11-digit prime number to continue."
  • Mike GreenMike Green Posts: 23,101
    edited 2007-10-04 15:44
    Maybe the PLL is locking onto a submultiple of the clock. Perhaps it's failing enough so it's producing a usable clock anyway.

    The fact that it seems to work doesn't mean that it's working properly or reliably or accurately
  • RaymanRayman Posts: 14,162
    edited 2007-10-04 16:04
    What's the best way to get to 100 MHz reliably?·

    I have some ideas for timing devices where an even 10-ns clock would be convenient...

    Anybody know if VGA, TV, keyboard and mouse drivers will still work "as-is" ?

    Is there a good test program to stress the chip and make sure it's working reliably?·

    Any chance Parallax would test and sell some as being rated for 100 MHz?·

    What is the internal problem at higher speeds anyway?· (I have a feeling it's not overheating.· I'm guessing it's propagation delay issues.)
  • Paul BakerPaul Baker Posts: 6,351
    edited 2007-10-04 18:06
    Ok, common misconception: "a stated crystal frequency is the stated crystal frequency", this is not the case! For crystals under 25MHz, this is the case because they are fundamental frequency crystals, but for crystals between 25 and 75Mhz they are 3rd harmonic crystals and the Propeller's crystal driving circuitry is not a 3rd harmonic driver, so the Propeller sees only the fundamental. For a 66Mhz crystal, this means the Propeller actually sees a 22MHz waveform. Now if you are using an active oscillator, it does the driving and amplification of the clock itself, so no worries.

    To achieve 100Mhz, you'll need an active external oscillator, they only come in surface mount and aren't exactly cheap. The cheapest I can quickly find is the Connor-Winfield CWX823-100.0M for $3.11

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Paul Baker
    Propeller Applications Engineer

    Parallax, Inc.
  • RaymanRayman Posts: 14,162
    edited 2007-10-04 18:33
    Any chance of using a 6.25 MHz crystal at PLLx16?

    (assuming that I can get Vectron or somebody to custom make me one...)
  • Paul BakerPaul Baker Posts: 6,351
    edited 2007-10-04 18:37
    If you can find one, but that is a non standard frequency, it would cost more to custom make than getting what I've already pointed out.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Paul Baker
    Propeller Applications Engineer

    Parallax, Inc.
  • RaymanRayman Posts: 14,162
    edited 2007-10-04 18:51
    Probably, but it'd be a whole lot easier to solder into the proto board!
  • Paul BakerPaul Baker Posts: 6,351
    edited 2007-10-04 19:03
    You can have a breakout board made for it, if your in no hurry BatchPCB will fab it for you for $12.50.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Paul Baker
    Propeller Applications Engineer

    Parallax, Inc.
  • RaymanRayman Posts: 14,162
    edited 2007-10-04 19:24
    You mean a tiny PCB just for the clock chip? That's a good idea! I already emailed Vectron about a custom oscillator. But, I have a feeling they'll want a few hundred $$ at least.
  • hippyhippy Posts: 1,981
    edited 2007-10-04 19:47
    If you are looking for cheap and simple, why not a more common 100MHz oscillator module divided down by 16 using external hardware then use XINPUT+PLL16x ? 100MHz/16, 50MHz/8 and 25Mhz/4 will give you the same 6.25MHz clock.
  • TookingsTookings Posts: 18
    edited 2007-10-04 20:01
    Would these work? Very little specs...

    www.alltronics.com/cgi-bin/item/XTAL_6_25MHZ/search/6%2E25MHZ%2DCRYSTAL

    (Checked out batchPCB...found my way here: www.4pcb.com/index.php?load=content&page_id=128 -- reads to me like a PCB order less than $500 would be free?!)

    -Rick

    Post Edited (Tookings) : 10/4/2007 8:27:34 PM GMT
  • Paul BakerPaul Baker Posts: 6,351
    edited 2007-10-04 20:04
    hippy, if he got a 100Mhz oscillator, there's no need to do division, just set clkmode to xinput and don't use pll.
    tookings, excellent find, only wish there was more info on packaging specs.

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Paul Baker
    Propeller Applications Engineer

    Parallax, Inc.
  • RaymanRayman Posts: 14,162
    edited 2007-10-04 20:39
    OK, so I ordered 15 from alltronics (thanks Tookings). I don't know how legit they are and I'm concerned that they don't provide any additional info or stock status. But, we'll see what happens!
Sign In or Register to comment.