A Sine Wave DDS in PASM - Comments?

DroneDrone Posts: 433
edited 2011-04-13 - 20:17:15 in Propeller 1
Hello Everyone,

Attached is a simple 8-bit sine wave Direct Digital Synthesizer (DDS) in Propeller assembler. The DDS uses a 256 sample look-up table in cog RAM. The accumulator and tuning word are 24-bits in length. Minimum step size is 0.170Hz. With the propeller clock at 80MHz and no post-DAC filtering, the output is usable through around 150-200kHz. I'm sure you can go higher with filtering, but I haven't tried.

The heavily commented source is prefaced with a bunch of introductory documentation. A linear-weighted 8-bit R2R DAC is used, that is well documented as well. In the documentation there are external references to more information on the subject of the DDS and R2R DAC.

This object is actually quite small once you strip all the documentation out; it has to be in order to be fast. I'm no PASM expert by any means, but I would certainly like to know if anyone can make this run faster (thanks to all contributors for the suggestions so-far).

The archive includes a simple Excel 2003 spreadsheet that shows how tuning words are calculated. The look-up table can be easily modified to produce any waveform. The PASM code can be easily modified to produce square waves, triangles, ramps, etc. without a look-up table.

Perhaps I'll embellish this example with other waveforms later on. Independent I/Q output in multiple cogs is also an interesting target but I have to learn how to launch multiple cogs and keep them somewhat phase coherent.

I post this now because the object has been languishing in my to-do bin and lately someone else in the forum needed an ultrasonic sine wave generator. Any comments or suggestions are most welcome. I'll post the final version in ObEx if anyone thinks it is worth it.

Changelog:
* V1.00, 06 April 2010, first release.
* V1.01, 06 April 2010, Incorporated Kurneko's two excellent suggestions as of post time, the current V1.01 sine DDS loop uses 28 clock cycles, down from 40 cycles. This represents a 48% increase in usable output frequency. Thanks Kurneko! I have verified the number of saved clock cycles by frequency measurement. The Excel 2003 spreadsheet has been updated to allow you to do the calculation for determining number of clock cycles used from measured output frequency.
* V1.02, 06 April 2010, Error in the R2R documentation schematic fixed. The resistor to ground is 2R not R. Thanks to Toby Seckshund for pointing this out.

The attached pictures of the output waveform at 1kHz and 100kHz are for V1.00. These waveforms look about the same for V1.01 but are now at 1.5kHz and 150kHz respectively due to the reduction in clock cycles.

Best Regards, David

Post Edited (Drone) : 4/6/2010 1:45:39 PM GMT
1024 x 944 - 113K
1024 x 934 - 112K
«1

Comments

  • TubularTubular Posts: 3,646
    edited 2010-04-06 - 06:39:09
    Hi David,

    Very impressive work, and well documented. Definitely would be a valuable addition to the obex.

    I look forward to trying it very soon (gotta get some precision resistors in!)

    cheers
    tubular
  • kuronekokuroneko Posts: 3,623
    edited 2010-04-06 - 07:29:49
    Drone said...
    I'm no PASM expert by any means, but I would certainly like to know if anyone can make this run faster. Right now the sine DDS loop uses 40 clock cycles.
    For starters, you could drive outa directly instead of going the long way over DACOut, i.e.

    :inline mov outa, 0-0
    


    Next thing, relocate the table to location 0 (this will save you the add Idx, #Sine instruction). All table values are effectively nops so you either accept a 1K cycle start-up delay (cognew requires 8K already) or you jump over the table and patch the first entry. Either way that's down to 32 cycles.

    One more ...

    entry
    
      '256 sampled 8-bit sine table. Replace with any sampled waveform you like
      'as long as the data table is of the same dimension and depth. Only longs are
      'used to make the DDS loop as short as possible in terms of clock cycles.
    
    Sine          long $80,$83,$86,$89,$8C,$8F,$92,$95,$98,$9C,$9F,$A2,$A5,$A8,$AB,$AE
    
                  ' remainder of table
    
                  mov      dira,    PinMsk         'LSB to PA0, LSB to PA7
    
    :loop         add      Acc,     M_             'M $1FF max if literal with #$xxx
                  ror      Acc,     #16            'position index
                  movs     :inline, Acc            'store 9 bit index
                  andn     :inline, #%1_00000000   'clamp to 8 bit
                  rol      Acc,     #16            'restore accumulator
    :inline       mov      outa,    0-0            'Output 8-bits of long table lookup to DAC
                  jmp      #:loop                  'Loop again
    
    PinMsk        long     $FF                     '8-bit DAC, LSB on PA0, MSB on PA7
    M_            long     0                       'Tuning word var init
    Acc           long     0                       'Accumulator var init
    

    Post Edited (kuroneko) : 4/6/2010 8:09:40 AM GMT
  • DroneDrone Posts: 433
    edited 2010-04-06 - 08:08:55
    Kuroneko (Black Cat?, he-he you must be in Japan - I'm GMT+7 in Jakarta), your suggestion definitely goes to the top of the change list. Thanks!

    Tubular, IMO you don't need precision resistors just to experiment with the DAC. Usually generic 10% resistors work OK, especially if they come from the same batch. Attached are pics of a little R2R DAC board I hastily threw together. This board with 10% resistors produced the waveforms attached to the top post. The 0.1" female header is used to plug the DAC into a proto-board via jumper wires. I would recommend higher precision resistors in a production build though. I reference a nice little SIP Burr Brown R2R DAC in the documentation, half a buck if memory serves.

    Regards, David
    800 x 530 - 61K
    800 x 604 - 62K
  • TonyWaiteTonyWaite Posts: 219
    edited 2010-04-06 - 08:13:57
    Hi David,

    What a wonderfully useful application!

    Thank you for contributing this.

    On the Propeller, simply to visualise software seems enough to make it appear by magic ...

    Regards,

    T o n y
  • K2K2 Posts: 625
    edited 2010-04-06 - 08:17:47
    Fantastic! Where do I send the royalty check?
  • DroneDrone Posts: 433
    edited 2010-04-06 - 08:24:13
    Kuroneko, I see your second suggestion - thanks gain!

    By the way this DDS thing can be used to verify exactly how many clock cycles the PASM loop uses. For a given tuning word measure the output frequency with a counter (or use the propeller itself as a counter!) Knowing the tuning word value and measured output frequency, the rest of the variables are known allowing us to calculate the number of clocks used via a re-arranged version of the tuning equation. I guess you could insert non-DDS PASM code into the loop to measure the difference in clocks used thereby providing exactly how many clocks are used by the inserted code. Hmmm....
  • caskazcaskaz Posts: 939
    edited 2010-04-06 - 08:42:44
    Hi Drone.
    It's great!!
    I also think this is valuable addition to the obex.

    I tried to generate higher frequency sin-wave ,too.
    But I cannot.

    Post Edited (caskaz) : 4/6/2010 8:51:57 AM GMT
  • DroneDrone Posts: 433
    edited 2010-04-06 - 09:55:07
    Hi Kuroneko,

    V1.01 (see file attached to top-post) includes the two suggestions in your 4/6/2010 8:09:40 AM GMT post. Four clock cycles saved by direct port output. Eight clock cycles saved by relocating the look-up table to zero. Total savings, 12 clock cycles - down to 28 from 40. This represents a 50% increase in output frequency for a given tuning word. Quite an improvement. Thank you again!

    Best Regards, David
  • TonyWaiteTonyWaite Posts: 219
    edited 2010-04-06 - 10:03:02
    Hi David,

    Thank you for V1.01.

    The documentation and spreadsheet are tutorials in themselves!

    T o n y
  • Toby SeckshundToby Seckshund Posts: 2,025
    edited 2010-04-06 - 11:50:38
    Doesn't the bottom resistor, to GND, have to be 2R also ?

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    Style and grace : Nil point
  • kuronekokuroneko Posts: 3,623
    edited 2010-04-06 - 12:13:17
    Drone said...
    Hi Kuroneko,

    V1.01 (see file attached to top-post) includes the two suggestions in your 4/6/2010 8:09:40 AM GMT post. Four clock cycles saved by direct port output. Eight clock cycles saved by relocating the look-up table to zero. Total savings, 12 clock cycles - down to 28 from 40. This represents a 50% increase in output frequency for a given tuning word. Quite an improvement. Thank you again!
    I have one more (got it just when I went out the door for some real life stuff [noparse]:)[/noparse] If you don't need the top 8 bit then you could use M_ := M << 8 and change the PASM loop to:

    :loop         add      Acc,     M_             'M $1FF max if literal with #$xxx
                  ror      Acc,     #24            'position index
                  movs     :inline, Acc            'store 9 bit index (top bit is 0)
                  rol      Acc,     #24            'restore accumulator
    :inline       mov      outa,    0-0            'Output 8-bits of long table lookup to DAC
                  jmp      #:loop                  'Loop again
    
  • DroneDrone Posts: 433
    edited 2010-04-06 - 12:40:09
    1. Tony, you are right. The documentation has an error. The DAC resistor to ground should be 2R not R. The physical DAC I used is correct. Thanks for pointing this out.

    Uploaded V1.02 to the top-post; fixes this documentation error.

    2. Kuroneko, thanks for your latest tip. I will study it. Not sure if I can get to it tonight my local time.

    Regards, David
  • kuronekokuroneko Posts: 3,623
    edited 2010-04-06 - 14:03:58
    Drone said...
    2. Kuroneko, thanks for your latest tip. I will study it. Not sure if I can get to it tonight my local time.
    PASM wouldn't be complete with at least one counter based version [noparse];)[/noparse] This one - if workable - gives you 20 cycles per loop.

                    mov     dira, PinMsk            ' LSB to PA0, LSB to PA7
    
                    movi    ctra, #%0_11111_000     ' LOGIC always
                    mov     phsa, M_                ' TW
                    shl     phsa, #3                ' TW*8
                    add     phsa, M_                ' TW*9
                    shl     phsa, #1                ' TW*18
                    mov     frqa, M_                ' TW == M*256/20
    
                    mov     addr, phsa              ' -12  %%
                    nop                             '  -8
                    nop                             '  -4
    
    :loop           shr     addr, #24               '  +0
                    movs    :inline, addr           ' -16
                    mov     addr, phsa              ' -12
    :inline         mov     outa, 0-0               '  -8
                    jmp     #:loop                  '  -4
    
    PinMsk          long    $FF                     ' 8-bit DAC, LSB on PA0, MSB on PA7
    M_              long    0                       ' tuning word (initialised to M*256/20)
    addr            long    0
    
    


    Requirements are:
    • ctra is running in LOGIC always mode
    • frqa is loaded with tuning word x 256 / 20
    • phsa is preloaded so that the first mov (%%) would read frqa x 20 (== tuning word x 256)
    256 is the result from shifting the 24 bit accumulator left 8 bits (makes addressing easier), the 1/20 is based on the cycle count between reads from phsa (you only want a single tuning word accumulation between subsequent reads).

    Post Edited (kuroneko) : 4/7/2010 5:55:18 AM GMT
  • lonesocklonesock Posts: 909
    edited 2010-04-06 - 16:27:08
    As requested (in another thread) here is a version using the video hardware. Basic concept: use the VGA mode of the video hardware to shift out 4 bytes at a time.

    Limitations:
    * hard to set up a frequency - because the video HW uses the PLL mode, you can only adjust within one octave before changing PLL divider, etc.
    * you can _NOT_ skip samples, so the waveform table length determines the maximum output frequency.
    * you can only use max 8 bit output
    * you must use one of the pin blocks (0..7, 8..15, 16..23, 24..31)

    Benefits:
    * the video HW PLL does your timing for you
    * you can output 4 bytes per waitvid...in the current scheme there can be 3 instructions between waitvids, but you could unroll the loop, doubling performance

    So the maximum frequency of the sine wave for this code is ~ 73kHz, but if you make the data table smaller, or unroll the loop you could seriously increase the maximum frequency.

    Jonathan

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    lonesock
    Piranha are people too.
    Free time status: see my avatar [8^)
    F32 - fast & concise floating point: OBEX, Thread
    Unrelated to the prop: KISSlicer
  • schillschill Posts: 741
    edited 2010-04-06 - 19:08:32
    I'm looking forward to playing with this. I was just about to figure out something similar myself - although I don't need the frequencies to be quite as high.
  • Bill HenningBill Henning Posts: 6,445
    edited 2010-04-06 - 19:09:28
    Nice work David and Jonathan!

    when I was playing around with something very similar it seemed to me that the maximum practical limit appeared to be 40msps out per second; I based this on WAITVID taking 5+ cycles.

    Using 64 WAITVID's in a row, with a 64 long packed table of values (for a 256 sample waveform) it might be possible to do 64msps out, if we can trust WAITVID to take a constant 5 cycles; but to avoid jitter it might be best to use a 40MHz frequency with 255 samples, using the time of the 256th sample for a JMP to the beginning of the WAITVID sequence.

    40MHz/256 samples = 156,200Hz

    40MHz/200 samples = 200,000Hz

    and so on...

    I used 1% resistors, and I hand-sorted them to match within .1% (by measuring and binning a large enough number of resistors to get enough for an R2R ladder matched within 0.1%)
    lonesock said...
    As requested (in another thread) here is a version using the video hardware. Basic concept: use the VGA mode of the video hardware to shift out 4 bytes at a time.

    Limitations:
    * hard to set up a frequency - because the video HW uses the PLL mode, you can only adjust within one octave before changing PLL divider, etc.
    * you can _NOT_ skip samples, so the waveform table length determines the maximum output frequency.
    * you can only use max 8 bit output
    * you must use one of the pin blocks (0..7, 8..15, 16..23, 24..31)

    Benefits:
    * the video HW PLL does your timing for you
    * you can output 4 bytes per waitvid...in the current scheme there can be 3 instructions between waitvids, but you could unroll the loop, doubling performance

    So the maximum frequency of the sine wave for this code is ~ 73kHz, but if you make the data table smaller, or unroll the loop you could seriously increase the maximum frequency.

    Jonathan
    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    www.mikronauts.com E-mail: mikronauts _at_ gmail _dot_ com 5.0" VGA LCD in stock!
    Morpheus dual Prop SBC w/ 512KB kit $119.95, Mem+2MB memory/IO kit $89.95, both kits $189.95 SerPlug $9.95
    Propteus and Proteus for Propeller prototyping 6.250MHz custom Crystals run Propellers at 100MHz
    Las - Large model assembler Largos - upcoming nano operating system
    www.mikronauts.com / E-mail: mikronauts _at_ gmail _dot_ com / @Mikronauts on Twitter
    RoboPi: The most advanced Robot controller for the Raspberry Pi (Propeller based)
  • GreenTHHNGreenTHHN Posts: 11
    edited 2010-04-15 - 02:50:13
    A couple bug (minor) reports and a question.

    In section 2 of the program documentation a website is listed as www.bournes.com. I believe this should be www.bourns.com as www.bournes.com is an an auto service site.

    In the program comments in the DAT section: "mov dira, PinMsk 'LSB to PA0, LSB to PA7" I believe it should be 'LSB to PA0, MSB to PA7".

    Now my question is, can it be easily changed to use pins PA16 (LSB) to PA23(MSB) instead of PA0 - PA7?
    Can I just change the line "PinMsk long $FF" to "PinMsk long $00FF" ? Would the "Mov OUTA 0-0" line need to be changed?
  • kuronekokuroneko Posts: 3,623
    edited 2010-04-15 - 02:58:07
    GreenTHHN said...
    Now my question is, can it be easily changed to use pins PA16 (LSB) to PA23(MSB) instead of PA0 - PA7?
    Can I just change the line "PinMsk long $FF" to "PinMsk long $00FF" ? Would the "Mov OUTA 0-0" line need to be changed?
    Sort of, the pin group is defined by the table values. In order to use P16..23 you'd have to shift the whole table accordingly (the values that is). You also need to implement the jump-over-table-replace-first-entry as the table values then no longer appear as nops. The pin mask will have to be set to $00FF0000 ($FF == $00FF). Everything else stays the same.
  • GreenTHHNGreenTHHN Posts: 11
    edited 2010-04-15 - 03:55:19
    kuroneko said...
    You also need to implement the jump-over-table-replace-first-entry as the table values then no longer appear as nops. The pin mask will have to be set to $00FF0000 ($FF == $00FF). Everything else stays the same.


    Well I'm still new to propeller programming. That is a little over my head. I think I will re-arrange my circuit to use PA0 - PA7 instead of monkeying around with the code.
    I haven't actually built the circuit yet but am just drawing the schematic.
  • kuronekokuroneko Posts: 3,623
    edited 2010-04-15 - 04:04:23
    GreenTHHN said...
    Well I'm still new to propeller programming. That is a little over my head. I think I will re-arrange my circuit to use PA0 - PA7 instead of monkeying around with the code.
    If P16..23 is more convenient for you I can produce a suitable driver. No problem at all.
  • CoronaKidCoronaKid Posts: 25
    edited 2010-04-15 - 05:06:41
    Hi Gang,
    We have been discussing Drone's (David's) code in this topic heading for the past few weeks.
    http://forums.parallax.com/showthread.php?p=895566

    Kodos to David and other contributors like·kuroneko who took the code from 40 to 28 and now only 20 clock cycles.· I'm greener than GreenTHHN as far as spin code is concerned, but am very impressed with what some of you can do.

    GreenTHHN - PA0 to PA7 works good on the Demo board.
    Fred
    cool.gif

    ▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔▔
    ________________________________

    Work:www.vhf.ca··Play:www.Costalegre.ca
  • Andrey DemenevAndrey Demenev Posts: 362
    edited 2010-04-15 - 05:15:25
    If there is a spare pin, you can chain 2 counters and save another instruction. If FRQB = 1, one waveform cycle will always utilize all 256 values from table, and maxim frequency is limited by 156250 Hz (at 80 MHz clock). Manipulating both FRQA and FRQB would allow higher frequencies, with signal quality degrading at higher FRQB values

    :sample     MOV     OUTA, 0-0
                MOVS    :sample, PHSB
                ANDN    :sample, #$100
                JMP     :sample
    
    
    Day OFF? You must be pulling my leg! Stop making humor before someone sees you, fool!
  • DroneDrone Posts: 433
    edited 2010-04-15 - 09:26:02
    GreenTHHN, Thanks for the tips on the documentation. I'm currently pulling the documentation out of the source as a separate file. I'll incorporate your changes.

    As for the DAC pin numbers. If I remember correctly, the V1.02 code needed the DAC to be on PA0-PA7 for minimum clock cycles. I'll revisit this in light of new comments here and the upcoming release of V1.03 using Kuroneko's brilliant counter method after a bit of beta testing.

    CoronaKid, Have you had time to beat the beta V1.03 to death yet? If so, send me an Email with results if.

    I've been busy with U.S. tax stuff as of late. Nightmare to say the least. Sorry if I've been late to reply. I'm also not getting reliable email notifications when this thread is updated even though I check every notification and filter all propeller email forum notifications from my spam folder to my in-box. Hmmm....

    Best Regards, David
  • kuronekokuroneko Posts: 3,623
    edited 2010-04-15 - 10:10:09
    Drone said...
    As for the DAC pin numbers. If I remember correctly, the V1.02 code needed the DAC to be on PA0-PA7 for minimum clock cycles.
    Shouldn't be an issue, after all, you only transfer longs to outa. We could be nice and - based on the current table - optionally shift the table values to a different pin group (0..7, 8..15 etc).

    @Andrey: I toyed with the idea of chaining counters but figured that I get too many limitations for free (while preserving current functionality).

    Post Edited (kuroneko) : 4/15/2010 10:16:51 AM GMT
  • tdlivingstdlivings Posts: 437
    edited 2011-04-01 - 11:49:01
    I am trying out Drones code using a Gadget Gangster USB and the ProtoPlus as shown in the pictures.
    The original code does DDS with 28 clocks and works fine but is not adjustable with out recompiling or
    starting and stoping the cog it runs in.
    I wanted to build a sweep generator so I modified the code to use a rdlong and pace itself with a clock
    generated on pin 8 by a counter putting out a 1.25MHZ signal. The pacing was done because the
    rdlong takes a variable number of clocks 7.. 22 if I remember and would cause jitter. As long as the code is
    completed before the next 1.25MHZ edge I get even pacing of the DDS.
    Drone's org code
    VAR
    long cog 'cog id
    PUB start(M) : okay 'M is the tuning word
    stop 'Stop DDS if it is running, frees a cog
    M_ := M 'Assign passed variable (tuning word)
    okay := cog := cognew(@entry,0) 'Start DDS, uses a cog.
    PUB stop
    if cog 'If cog is running
    cogstop(cog~) 'Stop DDS, frees a cog
     
    DAT
    org 0
    entry
    '256 sampled 8-bit sine table. Replace with any sampled waveform you like
    'as long as the data table is of the same dimension and depth. Only longs are
    'used to make the DDS loop as short as possible in terms of clock cycles.
     
    Sine long $80,$83,$86,$89,$8C,$8F,$92,$95,$98,$9C,$9F,$A2,$A5,$A8,$AB,$AE
    long $B0,$B3,$B6,$B9,$BC,$BF,$C1,$C4,$C7,$C9,$CC,$CE,$D1,$D3,$D5,$D8
    long $DA,$DC,$DE,$E0,$E2,$E4,$E6,$E8,$EA,$EC,$ED,$EF,$F0,$F2,$F3,$F5
    long $F6,$F7,$F8,$F9,$FA,$FB,$FC,$FC,$FD,$FE,$FE,$FF,$FF,$FF,$FF,$FF
    long $FF,$FF,$FF,$FF,$FF,$FF,$FE,$FE,$FD,$FC,$FC,$FB,$FA,$F9,$F8,$F7
    long $F6,$F5,$F3,$F2,$F0,$EF,$ED,$EC,$EA,$E8,$E6,$E4,$E2,$E0,$DE,$DC
    long $DA,$D8,$D5,$D3,$D1,$CE,$CC,$C9,$C7,$C4,$C1,$BF,$BC,$B9,$B6,$B3
    long $B0,$AE,$AB,$A8,$A5,$A2,$9F,$9C,$98,$95,$92,$8F,$8C,$89,$86,$83
    long $80,$7C,$79,$76,$73,$70,$6D,$6A,$67,$63,$60,$5D,$5A,$57,$54,$51
    long $4F,$4C,$49,$46,$43,$40,$3E,$3B,$38,$36,$33,$31,$2E,$2C,$2A,$27
    long $25,$23,$21,$1F,$1D,$1B,$19,$17,$15,$13,$12,$10,$0F,$0D,$0C,$0A
    long $09,$08,$07,$06,$05,$04,$03,$03,$02,$01,$01,$00,$00,$00,$00,$00
    long $00,$00,$00,$00,$00,$00,$01,$01,$02,$03,$03,$04,$05,$06,$07,$08
    long $09,$0A,$0C,$0D,$0F,$10,$12,$13,$15,$17,$19,$1B,$1D,$1F,$21,$23
    long $25,$27,$2A,$2C,$2E,$31,$33,$36,$38,$3B,$3E,$40,$43,$46,$49,$4C
    long $4F,$51,$54,$57,$5A,$5D,$60,$63,$67,$6A,$6D,$70,$73,$76,$79,$7C
    mov dira, PinMsk 'LSB to PA0, LSB to PA7
    :loop add Acc, M_ 'M $1FF max if literal with #$xxx
    ror Acc, #16 'Postion index
    movs :inline, Acc 'Store 9-bit index
    andn :inline, #%1_00000000 'Clamp to 8-bits
    rol Acc, #16 'Restore accumulator
    :inline mov outa, 0-0 'Get long value from table
    jmp #:loop 'Loop again
    PinMsk long $FF '8-bit DAC, LSB on PA0, MSB on PA7
    M_ long 0 'Tuning word var init
    Acc long 0 'Accumulator var init
    {{
    

    The main spin code just launches the above code into a cog
    ddscog := DDS.start(Tuning_Word) 'Start the DDS, cog no.in ddscog.

    End of review and now my questions.
    The modified code I wrote as a exercise works but I would like to modifiy the Freq Control variable
    without a rdlong. FreqControl or M_. I do not think it is possible but will ask because there are many
    experts on the forumn. The other question is with the data table at org 0 and the label entry at org 0
    why does the cog not try and execute code in the data table but instead it works fine and executed code
    starting at the first ASM instruction. It works but it tells me the prop loads in a cog is loaded with the code
    starting at where you tell it in cognew and then knows where the code begins skipping over the data table
    long def's. I moved the label entry below the table and the code quit working.
    My modified code
    VAR
      long  cog                                      'cog id
      long FreqControl
    PUB start(M) : okay                              'M is the tuning word
      stop
      FreqControl := $41893 ''$E560                            'Stop DDS if it is running, frees a cog
      M_ := M                                         'Assign passed variable (tuning word)
      okay := cog := cognew(@entry,@FreqControl)      'Start DDS, uses a cog.
    PUB stop
      if cog                                         'If cog is running
        cogstop(cog~)                                'Stop DDS, frees a cog
    PUB ChangeFreq(NewFreq)
      FreqControl := NewFreq
     
    DAT
                             org      0
      entry
      '256 sampled 8-bit sine table. Replace with any sampled waveform you like
      'as long as the data table is of the same dimension and depth. Only longs are
      'used to make the DDS loop as short as possible in terms of clock cycles.
     
      Sine          long $80,$83,$86,$89,$8C,$8F,$92,$95,$98,$9C,$9F,$A2,$A5,$A8,$AB,$AE
                    long $B0,$B3,$B6,$B9,$BC,$BF,$C1,$C4,$C7,$C9,$CC,$CE,$D1,$D3,$D5,$D8
                    long $DA,$DC,$DE,$E0,$E2,$E4,$E6,$E8,$EA,$EC,$ED,$EF,$F0,$F2,$F3,$F5
                    long $F6,$F7,$F8,$F9,$FA,$FB,$FC,$FC,$FD,$FE,$FE,$FF,$FF,$FF,$FF,$FF
                    long $FF,$FF,$FF,$FF,$FF,$FF,$FE,$FE,$FD,$FC,$FC,$FB,$FA,$F9,$F8,$F7
                    long $F6,$F5,$F3,$F2,$F0,$EF,$ED,$EC,$EA,$E8,$E6,$E4,$E2,$E0,$DE,$DC
                    long $DA,$D8,$D5,$D3,$D1,$CE,$CC,$C9,$C7,$C4,$C1,$BF,$BC,$B9,$B6,$B3
                    long $B0,$AE,$AB,$A8,$A5,$A2,$9F,$9C,$98,$95,$92,$8F,$8C,$89,$86,$83
                    long $80,$7C,$79,$76,$73,$70,$6D,$6A,$67,$63,$60,$5D,$5A,$57,$54,$51
                    long $4F,$4C,$49,$46,$43,$40,$3E,$3B,$38,$36,$33,$31,$2E,$2C,$2A,$27
                    long $25,$23,$21,$1F,$1D,$1B,$19,$17,$15,$13,$12,$10,$0F,$0D,$0C,$0A
                    long $09,$08,$07,$06,$05,$04,$03,$03,$02,$01,$01,$00,$00,$00,$00,$00
                    long $00,$00,$00,$00,$00,$00,$01,$01,$02,$03,$03,$04,$05,$06,$07,$08
                    long $09,$0A,$0C,$0D,$0F,$10,$12,$13,$15,$17,$19,$1B,$1D,$1F,$21,$23
                    long $25,$27,$2A,$2C,$2E,$31,$33,$36,$38,$3B,$3E,$40,$43,$46,$49,$4C
                    long $4F,$51,$54,$57,$5A,$5D,$60,$63,$67,$6A,$6D,$70,$73,$76,$79,$7C
                    mov      dira,    PinMsk         'LSB to PA0, LSB to PA7
                    mov      AddrFreq,par            'Get Addr of Spin Freq Control Value
      :loop         rdlong   M_,AddrFreq             'Get Freq Control Value
                    add      Acc,     M_             'M $1FF max if literal with #$xxx
                    ror      Acc,     #16            'Postion index
                    movs     :inline, Acc            'Store 9-bit index
                    andn     :inline, #%1_00000000   'Clamp to 8-bits
                    rol      Acc,     #16            'Restore accumulator
      :inline       mov      outa,    0-0            'Get long value from table
                    waitpeq  state,Pin8Msk           'Make sure pin 8 is high
                    waitpne  state,Pin8Msk           'leave on pin 8 going low
                    jmp      #:loop                  'Loop again
      PinMsk        long     $FF                     '8-bit DAC, LSB on PA0, MSB on PA7 
      Pin8Msk       long     $100                    'Mask for pin 8 TDL
      state         long     $100
      M_            long     0                       'Tuning word var init
      Acc           long     0                       'Accumulator var init
      AddrFreq      long     0                       'Address of Freq Control in Main Spin Code
    {{
    
    Again I am looking if there is a way to modifiy M_ on the fly from the main spin but do not think it is possible
    but have seen some posts on people updating vars in DAT sections so I ask.
    The maine spin of my quick sweep code
    CON
      '-----[ User definable parameters below ]----------------------------------
      'Uncomment ONE of the lines below to test different frequencies.
     
      'Tuning_Word = $24B                             'DDS tuning word, 100Hz
      'Tuning_Word = $16F0                             'DDS tuning word, 1kHz
      Tuning_Word = $E560                           'DDS tuning word, 10kHz
      'Tuning_Word = $8F5C2                           'DDS tuning word, 100kHz
      'Tuning_Word = $599999                           'DDS tuning word , 1MHz
      '-----[ Typically no user definable parameters below ]---------------------
     
      _clkmode = xtal1 + pll16x                      '5MHz * 16PLL = 80MHz system clock.
      _xinfreq = 5_000_000                           '5MHz external crystal.
    VAR
     long FControl
     
    OBJ
      DDS : "DDS_PASM_Sin_SWP2_V1.02"
    PUB Main | ddscog, tmp, stat
      {Set up Counter A to put out the a clock signal at 1.25MHZ}
      ctra[30..26] := %00010
      frqa :=  268435456          '2^28 an even power of 2
      ctra[25..23] := 1            'PLL divide 
      ctra[5..0] := 8              'Pin 8
      dira[8]~~                    'Set it as an output
     
      'Start the DDS. We pass the pre-calculated tuning word shown in CON.
      'Alternatively the user may include code to calculate a tuning word here.
      'The DDS cog number is returned and stored in the variable ddscog.
      'To stop the DDS use "DD.stop".
     
      FControl := Tuning_Word
     
      ddscog := DDS.start(Tuning_Word)             'Start the DDS, cog no.in ddscog.
      waitcnt(2 * clkfreq + cnt)
      tmp := FControl
      repeat
        repeat 20
         FControl := FControl + $16F0
         DDS.ChangeFreq(FControl)
         waitcnt( clkfreq/10 + cnt)
        repeat 20
         FControl := FControl - $16F0
         DDS.ChangeFreq(FControl)
         waitcnt( clkfreq/10 + cnt)
    
    Again credit to Drone and Kuroneko for great work my code is just a mod to try sweeping
    320 x 240 - 78K
    640 x 360 - 239K
  • kuronekokuroneko Posts: 3,623
    edited 2011-04-01 - 18:49:30
    tdlivings wrote: »
    The other question is with the data table at org 0 and the label entry at org 0 why does the cog not try and execute code in the data table but instead it works fine and executed code starting at the first ASM instruction. It works but it tells me the prop loads in a cog is loaded with the code starting at where you tell it in cognew and then knows where the code begins skipping over the data table long def's. I moved the label entry below the table and the code quit working.
    Let's look at the first table entry long $80. This can also be written as
    Sine    long    %000000_0000_[COLOR="red"]0000[/COLOR]_000000000_0[COLOR="blue"]10000000[/COLOR]
    
    Does that ring a bell?
    tdlivings wrote: »
    Again I am looking if there is a way to modifiy M_ on the fly from the main spin but do not think it is possible but have seen some posts on people updating vars in DAT sections so I ask.
    DAT updates from SPIN are only effective before the cog is started. After that it's rdxxxx or grabbing something from pins you probably don't have available. What's the required max frequency?
  • tdlivingstdlivings Posts: 437
    edited 2011-04-01 - 20:11:15
    @ Kuroneko
    Took me awhile and at first I thought you were saying $80 was a jmp instruction but that did not look right.
    The -con- part of the op code of all the table entrys in 0000 the whole table is a bunch of nop's.
    Went back and did a re-read of the original thread to figure it out but this was a much better lesson that
    I will remember. I am sure I read you post back last year but it went past me as to why as I was learning
    DDS concepts at the time.
    You show the $80 in blue %10000000 only to show where it is as the table is only byte size any long with
    only byte values will have the -con- part of the op code 0000 none of the table values byte values matter.
    If you meant to show something else I missed it.

    Thank's
    Tom
  • IanMIanM Posts: 40
    edited 2011-04-01 - 20:27:20
    Hi Guys

    Just noticed this thread. Here is code for a DDS I've been working on. It generates quadrature sine waves at an equivalent clock frequency of 20MHz (so up to about 6MHz). It uses 7 cogs. The frequency can be changed on the fly using a spare pin to communicate the change. If you xor the two outputs (after filtering and squaring) you can double the frequency. I haven't optimised the code so there is probably a way to reduce the number of cogs!

    Hey, could someone please show me how to insert code so it formats nicely?

    Cheers, Ian
  • tdlivingstdlivings Posts: 437
    edited 2011-04-01 - 20:47:08
    IanM
    Sounds interesting.
    Insert your code between"
    and
    
    "
    I had to add the double quotes to prevent it from inserting the word and in a code section

    Tom
  • tdlivingstdlivings Posts: 437
    edited 2011-04-01 - 20:50:44
    woops still did not show.
    I will use curly brackets only to show but you use "[" brackets to do it.
    Put your code between {CODE} and {/CODE}
Sign In or Register to comment.