Shop OBEX P1 Docs P2 Docs Learn Events
Digital Audio Master Word Clock — Parallax Forums

Digital Audio Master Word Clock

TheVigTheVig Posts: 41
edited 2013-09-17 09:10 in Propeller 1
In your humble opinions, would it be possible to build a word clock using propeller?

These things from the way I understand it are just a 5V square wave transmitted on a 75ohm cable with BNC connectors that are selectable 44,100Hz, 48,000Hz, and 96,000Hz.

So why do pro models cost so much? Why couldn't I build one? I guess they are pretty accurate..but how accurate can a propeller be? I mean, like using a circuit to filter out noise...and stepping up the voltage to 5V from the propeller 3.3v....

Thanks!
Jim

Comments

  • WBA ConsultingWBA Consulting Posts: 2,934
    edited 2013-09-11 13:57
    The prop can easily generate a clock signal as specified. The catch is the accuracy/consistency and that is where the high prices begin showing up. With a good PCB design, proper crystal selection, and proper output shaping/level circuitry, you can probably come up with a solution but it would be hard to tell how well it would stack up to the pricey models you mention.
  • Duane DegnDuane Degn Posts: 10,588
    edited 2013-09-11 20:31
    I thought this article about word clocks informative.

    I don't think you're going to be able to replace one of these with a Prop.

    I'm also curious why those things are so expensive.
  • jmgjmg Posts: 15,173
    edited 2013-09-11 21:34
    TheVig wrote: »
    In your humble opinions, would it be possible to build a word clock using propeller?

    Sure, but you need more than only a Prop.

    If you went to the cutting edge, you could generate anything you wanted out of something like this :

    http://www.silabs.com/products/clocksoscillators/clocks/pages/synce.aspx

    The Prob would do the control, and could also verify clock precisions against standards like GPS.

    Cheaper, but lower spec, is a device like Si5351A-B-GT.

    Then there is the Si504 - tiny, and MEMS based, and you simply feed it a IEEE 4 byte real number, with an optional 16b integer as fractional trim from there.
  • TheVigTheVig Posts: 41
    edited 2013-09-12 11:34
    I'm just making an attempt to state the obvious and make sure I have the simple concept:
    An ideal clock would be a perfect 5V square wave at the same frequency of the audio sampling rate.
    1. The rise and fall from 0V to 5V is instantaneous
    2. There is no deviation in the frequency.

    "Jitter" is any deviation from the the ideal which can be caused by variations in timing, interference, or a slow charge/discharge of the circuit.

    So in theory, if I plugged the BNC 75ohm cable from a perfect word clock into an oscilloscope, it would look like this:
    48000.png


    So seems like specifications for comparing commercial products are up in the air, http://www.jitter.de/pdfextern/towards.pdf
    Skimming the specifications does give me some idea of what I'd be looking for in the ideal circuit.

    -Jim
    48000.png
    324 x 243 - 856B
  • RobotWorkshopRobotWorkshop Posts: 2,307
    edited 2013-09-12 11:43
    Whenever I have heard people ask about word clocks they have always been ones like these:

    http://www.archieli.com/design/clock-by-dougswordclock/

    http://dougswordclock.com/

    It would be easy to use the Propeller to create one of these.
  • TheVigTheVig Posts: 41
    edited 2013-09-12 11:43
    1. How much would it be for something like the Si5328 ?

    2. Where could you buy just one?

    3. How would it be implemented with a Propeller?

    4. What would be a good, better, best circiut for creating a perfect square wave?

    5. How would could the device have multiple outputs to feed several devices the same square wave?


    -Jim
  • TheVigTheVig Posts: 41
    edited 2013-09-12 12:10
    Yes, I am not a fan of word clocks that use words to tell time because any time I try to do a simple search...BAM!
  • jmgjmg Posts: 15,173
    edited 2013-09-12 12:41
    TheVig wrote: »
    1. How much would it be for something like the Si5328 ?

    2. Where could you buy just one?

    3. How would it be implemented with a Propeller?

    4. What would be a good, better, best circiut for creating a perfect square wave?

    5. How would could the device have multiple outputs to feed several devices the same square wave?

    -Jim

    Digikey show SI5328C-C-GM (36QFN) in stock for for 1 : $12.83, or you can buy a Eval Board for $250

    They also show SI5351A-B-GT (MSOP10) for 1 : $1.34

    The Si504 are not showing in stock yet.

    The SI5328C & SI5351A both need precision reference sources, so a (VC)TCXO is a good start.

    Digikey have FOX924B, which are TCXO, affordable and easy to apply.

    They also have a series like ASVTX-09, which adds Voltage Trim to the TCXO, so you can lock this to a global standard like GPS.

    Or, you can go on eBay and chase down a Rubidium Oscillator....

    A series of 5V 75 Ohm signals, would need multiple 75 ohm line drivers - this needs to drive 10V into 150 ohms. (75 ohm terminator + 75 ohm cable in series)

    Examples & circuits :
    http://www.analog.com/static/imported-files/seminars_webcasts/36888492358237Section2.pdf

    Addit: Worth testing would be modern MOSFET gate drivers.
    These will do 10V comfortably, and are pushing faster as SMPS move into MHz ranges.
  • Jack BuffingtonJack Buffington Posts: 115
    edited 2013-09-14 20:56
    I don't know much about audio production but I would think that it would take a LOT of jitter to make ANY difference in the perceivable quality of recorded audio. As far as reading what I have heard referred to as 'master clocks', the prop can definitely do it. Over the summer I worked on a project where I needed to read VITC and LTC timecode from an Evertz master clock unit. It has a similar price range. It is possible that in the future, if I need a second master clock, I will just make a small propeller board that generates VITC and LTC simultaneously that remain in sync with each other. If I can read time code, it will be even easier to generate it.
  • TheVigTheVig Posts: 41
    edited 2013-09-15 20:24
    If you read about clocks, they are spoken about in very non-scientific terms. People will just say something sounds better / worse. Sometimes I wonder if jitter can actually cause some distortion that warms up the audio for people where they say it sounds better when it is actually performing worse.

    It all seems very rediculous to me. So being able spec a clock that doesn't have a huge price tag and performs is the name of the game here.

    -Jim
  • jmgjmg Posts: 15,173
    edited 2013-09-16 14:35
    TheVig wrote: »
    If you read about clocks, they are spoken about in very non-scientific terms. People will just say something sounds better / worse. Sometimes I wonder if jitter can actually cause some distortion that warms up the audio for people where they say it sounds better when it is actually performing worse.

    It all seems very rediculous to me.

    Of course, the Audiophile area is full of delusional fluff.
    TheVig wrote: »
    So being able spec a clock that doesn't have a huge price tag and performs is the name of the game here.

    If you do not like the universal clocks I listed above, then there is a LCM for your 44,100Hz, 48,000Hz, and 96,000Hz that is within Prop range.

    Just connect your 70.560MHz oscillator and the Prop SW does the rest.

    Finding a 70.560MHz Osc is more of a challenge :)
    - you could use 4.410MHz and x16 PLL, but now the PLL phase noise adds to the clock, and 4.410MHz Xtals also looks to be rare.

    To avoid the Prop PLL, buying a programmed Osc, like Si590 (590CD-ADG) is probably best.
    This specs 1 ps max jitter, and is still under $10 at Digikey.
  • Brian FairchildBrian Fairchild Posts: 549
    edited 2013-09-17 09:10
    Jack Buffington wrote: »
    ...it would take a LOT of jitter to make ANY difference in the perceivable quality of recorded audio.
    Not true. Whilst people argue about how much jitter causes audible effects, it's effects are real.

    When dealing with digital audio we are dealing with a sampled analogue original signal. The digital chain often has many many steps which act in series on a number of samples from different sources. Delay lines to equalise for delays in one part of the software are common. Jitter means that the samples we are dealing with are no longer related to the other samples.

    Try it sometime. Draw a sine wave and sample it in a number of places equally spaced. Now phase shift those results by 180 degrees and add it in to the original. The result should be perfect cancellation.

    Now randomise your sampling points and do the same sums. There is no longer perfect cancellation.

    In an audio system there can be hundreds of sources each of which might have hundreds of mathematical operations performed on them. Jitter of the master sampling clock will soon have drastic effects on the output.

    Now take a master clock with slow rise and fall times. As different logic gates it is fed to will have different logic switching levels you can't guarantee which gates will detect the change in logic state when. The result is yet another source of jitter.

    So we need a word clock with both low jitter and fast edges.

    If you dig around you can find several papers (the AES is probably a good source) which run through the maths to correlate jitter with distortion of the signal.
Sign In or Register to comment.