Sinewaves using interrupts.

[cncmachineguy]

Ok, I'll give part of it a go:

For a 5K signal with 32 samples, thats 32 x 5000 = 160,000/sec times to hit the interupt. With no real numbers to use here, lets say it takes 100 instructions to "hit" the int. So thats 1,600,000 ins/sec or 1.6mips. so to start off, i would think at a min 8 meg clock (2mips)

The rest of the math depends on how the signal is generated. for instance using hpwm or maybe old school R2R D/A.

Yes i get tired of the question

[Mike, K8LH]

Someone told me 16 PWM outputs per cycle could generate a decent filtered sine wave. If PBP can handle a much lower 20-KHz (50-usec) interrupt (with 20-MHz clock), maybe we could get Hank to check out the waveforms on his scope? For 16 outputs per cycle he could use an Fout frequency of 20000/16=1250-Hz. For 20 outputs per cycle he could use an Fout frequency of 20000/20=1000-Hz.

So how would you PBP gurus convert this DDS engine for Hank's 20-MHz 16F690 to PBP (Accum and Phase are 24-bit variables)?

//
//  setup CCP in PWM mode for 20-KHz (50-usecs) with prescale 1,
//  postcale 1, and PR2=250-1 (20-MHz clock)
//
//  Fdds = 1 / 0.00005 = 20000 Hz
//  Fres = 20000 / 2^24 = 0.0011920928955078125
//  Phase (1250-Hz) = INT(1250 / 0.0011920928955078125 + 0.5) = 1048576
//  Phase (1000-Hz) = INT(1000 / 0.0011920928955078125 + 0.5) =  838861
//
void interrupt()               //
{ pir1.TMR2IF = 0;             // clear TMR2 interrupt flag
  Accum = Accum + Phase;       // add phase offset to Accum
  ccpr1l = sine[Accum >> 16];  // setup next PWM duty cycle
}

We'd also need a 256 element sine table with PWM values of 0..250. Here's the output from my handy dandy Sine PWM Table Generator;

Number of steps: 256
   Output range: 250
 Table or Array: a

        125,128,131,134,137,140,143,146
        149,152,155,158,161,164,167,169
        172,175,178,181,183,186,189,191
        194,196,199,201,204,206,208,211
        213,215,217,219,221,223,225,227
        228,230,232,233,235,236,237,239
        240,241,242,243,244,245,246,246
        247,248,248,249,249,249,249,249
        250,249,249,249,249,249,248,248
        247,246,246,245,244,243,242,241
        240,239,237,236,235,233,232,230
        228,227,225,223,221,219,217,215
        213,211,208,206,204,201,199,196
        194,191,189,186,183,181,178,175
        172,169,167,164,161,158,155,152
        149,146,143,140,137,134,131,128
        124,121,118,115,112,109,106,103
        100,097,094,091,088,085,082,080
        077,074,071,068,066,063,060,058
        055,053,050,048,045,043,041,038
        036,034,032,030,028,026,024,022
        021,019,017,016,014,013,012,010
        009,008,007,006,005,004,003,003
        002,001,001,000,000,000,000,000
        000,000,000,000,000,000,001,001
        002,003,003,004,005,006,007,008
        009,010,012,013,014,016,017,019
        021,022,024,026,028,030,032,034
        036,038,041,043,045,048,050,053
        055,058,060,063,066,068,071,074
        077,080,082,085,088,091,094,097
        100,103,106,109,112,115,118,121

[HankMcSpank]

Sure, I'd be more than happy to run any test code posted up here, and respond with related scope screen shots.

But please note, I only have a PIC16F690 & 20Mhz resonator at my disposal

(that said I think I have a 20Mhz Xtal i my drawer somewhere if thats significant!)

We could change it so the answers would be nice and long like:

a pesticide used to kill lice?
DICHLORODIPHENYLTRICHLOROETHANE

When I buy the stuff every other week, I just ask for the shortned name - DDT ...my chemist normally charges me one thousand, one hundred & thirty three pence per bottle

[Jerson]

Hi Hank

If you're planning to use the digitally synthesised sine wave for audio purposes, please re-consider the number of samples. 16 samples are good enough to generate a sine wave for the CRO, but for the ear,.... well, it's a different game altogether.

I had used the DDS technique, albeit with a PSOC, to generate pure tones (sinewaves of fixed frequency) for a clinical audiometer. You will get a clean sine wave on the scope, but, the distortion present in the wave will make it unsuitable for any audio purist. So, I had to switch over to the AD9832 which should be quite similar in function to the 9835(I haven't checked)

The AD9832 is a 1024 point sampled wave DDS IC. The results I obtained from this are just excellent. Before this, I had used a ML2036 which is now obsolete.

You might achieve the tones, but, you may not like the sound you get using a 16 point DDS. Let me not discourage your collaborative effort, but look at it as a gentle helmet I've gifted you to save the remaining hair on your head LoL

Cheers

[HankMcSpank]

The AD9832 is a 1024 point sampled wave DDS IC. The results I obtained from this are just excellent. Before this, I had used a ML2036 which is now obsolete.

You might achieve the tones, but, you may not like the sound you get using a 16 point DDS. Let me not discourage your collaborative effort, but look at it as a gentle helmet I've gifted you to save the remaining hair on your head LoL

Hi Jerson,

Weirdly the AD9835 is available much cheaper on Ebay than the AD9832 - http://cgi.ebay.com.sg/AD9835BRU-Com...item35ac588755 )

Ok, I'll give part of it a go:

For a 5K signal with 32 samples, thats 32 x 5000 = 160,000/sec times to hit the interupt. With no real numbers to use here, lets say it takes 100 instructions to "hit" the int. So thats 1,600,000 ins/sec or 1.6mips. so to start off, i would think at a min 8 meg clock (2mips)

And what about generating such to two decimal places (eg 4999.22Hz) - what impact does that have on the number (100x more?)


Actually my intentions aren't particularly music driven, but while on that vein, here's someone pushing the PIC DDS (ADSR) envelope...



I actually totally agree about the moderation (a big high five from me)...it was just a bit of tongue in cheek - I too normally forget to answer the question (& for a good while, whenever I got presented with the coffee question I always got the answer wrong - I was just putting in the missing letter as the answer .....a case of RTFQ Hank!)

[Mike, K8LH]

I agree that if you're doin' audio you should look at one of the many DDS chips from Analog Devices. They used to be very nice about samples too, which I took advantage of when building a DDS Ham Radio project a few years ago.

Regards, Mike

[cncmachineguy]

I want to get clear about something interupt related here. If I have a pure ASM isr, does that mean I don't have to worry about PBP stuff at all when I enter and exit? So for instance, if I set PIE for timer 1 and GIE, I know my interupt will fire any time. I also know the PIC will tend to W,STATUS, and PCLATH, (more or less depending on device). So I don't need to worry about anything else? And if I address a user defined variable in my ISR, I do that with _VariableName? If its that easy, thats wonderful.

Sorry Hank, not trying to hijack your thread, just seems a good place to ask. I am cooking up something related to sine wave generator for you.

@Dave and all the other kind folks that moderate for us, Thank You. I HAVE noticed this is one of the best spam free forums I have seen. I do know thats takes a LOT of work, and if my having to answer a question everytime I post helps, its the least I can do. Have to admit though, I preview my post most times, then forget to answer again. But I hope sooner or later I will learn. I have telsa spelled tesla almost every time now

[Jerson]

I want to get clear about something interupt related here. If I have a pure ASM isr, does that mean I don't have to worry about PBP stuff at all when I enter and exit? So for instance, if I set PIE for timer 1 and GIE, I know my interupt will fire any time. I also know the PIC will tend to W,STATUS, and PCLATH, (more or less depending on device). So I don't need to worry about anything else? And if I address a user defined variable in my ISR, I do that with _VariableName? If its that easy, thats wonderful.

I can answer that for you. The ISR only needs to save as many registers as you are going to use in the ISR. So, if you plan on using W, Status, FSR, and any other registers that are used in the main program code, you need to save them on entry to the ISR and restore them while getting out. The moment you mix in PBP in an ASM ISR, you need to worry about the variables that PBP starts using.

PBP uses a lot of internal variables called T1, T2, R0, R1, etc
The moment you start using PBP math, logical and, if x and y then type constructs, these variables are used. So, you are better off not using PBP inside interrupts if you mark the ISR as ASM code with DT_ints The best way is to look at the generated LST file for the ISR code; you will learn a lot by just inspecting it.

A simple x = y type assignment does not need any PBP variables and it will most certainly work in ASM.

Regards

[HankMcSpank]

Ok, old thread of mine & hands up...short of dabbling with scalerobotics cool sinewave generation program I never actually progressed much further - but I keep coming back to this DDS issue....I reckon I can do so much cool stuff once I've mastered the art of generating sine waves under my control

Now I'm never the brightest button in the sewing box, but I do have dogged determination (rough translation "a pest")...I'll often kludge my way through without necessary understanding every part of the process - So with this in mind, I hope to start soon again on this pesky DDS melarky.

So some questions.....

all this talk of a 24 bit accumulator - how do I achieve that on a 12LF1822 with PICbasic?!!!

Would up to 4.8Khz be do-able in HPWM for the DAC aspect?

Can I really expect to get granular with setting the frequency without recourse to floating point? (I want to be able to set the frequency via another PIC by fiielding an Eusart interrupt to the DDS PIC)

There are those on the net, that have done this with other processors. The Atmel AT90S2313 seems to figure large due to the lower number of clock cycles it takes per instruction. This chap has posted a great webpage...


http://www.myplace.nu/avr/minidds/index.htm

Including code....

http://www.myplace.nu/avr/minidds/minidds.asm

for those that have a clue (rules me out), how much of an ordeal would it be to move that over to a PIC?

Is this a project where I really ought to start learning assembly?

[mark_s]

Hi Hank,

Note that the avr project needs at least 8 I/O pins to drive the resistor ladder(DAC). So with a PIC12f series you are stuck using PWM, serial DAC or a dedicated DDS chip.

I tried a spi DAC, too slow. Couldn't get over 600hz. It wouldn't be too hard to do the avr dds with a pic. Use a look up table and a loop and write each value to an 8bit port. Not sure how the speed would compare with an AVR?

This is the best, but it is expensive
http://www.sparkfun.com/products/9169

I wrote some code for this in PBP if you are interested.

cheers