Sine Wave using a DAC/Microcontroller.

[ShadowPho1]

Hello everyone. I am working on a present for a girl (hehe) which will play a wav sound from an SD card into headphones. So, I have an SD card connected via SPI to a microcontroller which then connects via SPI to a DAC.

Right now I connected my DAC and now I am having problems outputting a sine wave. It comes out "fat". I am not sure if it's because I picked bad numbers or what. Please take a look at it and tell me if anything is wrong here. I am basically lost here.


I have a microproccesor MSP430F2012 (http://focus.ti.com/docs/prod/folders/print/msp430f2012.html").

Full code attached to this post. I am posting small snippets here.

//Main Loop

  while(1)
  {
    P1OUT |= 0x01; //Turn LED on
    
    for(i=0;i<511;i++)
         spi_send_byte(sinwave[i]); //transfer sine wave down-up
    
    P1OUT &= ~0x01; //Turn LED off
    
    for(i=512;i>1;i--)
         spi_send_byte(sinwave[i]); //transfer sin wave up-down
  }

//sindata.h

#ifndef SINDATA_H_
#define SINDATA_H_


const unsigned int sinwave[]={0,201,402,603,804,1005,1206,1407,1608,1809,2010,2211,2412,\
  2613,2814,3015,3216,3417,3617,3818,4019,4219,4420,4621,4821,5022,5222,5422,5623\
  ,5823,6023,6224,6424,6624,6824,7024,7224,7423,7623,7823,8022,8222,8421,8621,\
  8820,9019,9218,9417,9616,9815,10014,10212,10411,10609,10808,11006,11204,11402\
  ,11600,11798,11996,12193,12391,12588,12785,12983,13180,13376,13573,13770,\
  13966,14163,14359,14555,14751,14947,15143,15338,15534,15729,15924,16119,16314\
  ,16508,16703,16897,17091,17285,17479,17673,17867,18060,18253,18446,18639,18832\
  ,19024,19216,19409,19600,19792,19984,20175,20366,20557,20748,20939,21129,21320\
  ,21510,21699,21889,22078,22268,22457,22645,22834,23022,23210,23398,23586,23774\
  ,23961,24148,24335,24521,24708,24894,25080,25265,25451,25636,25821,26005,26190\
  ,26374,26558,26742,26925,27108,27291,27474,27656,27838,28020,28202,\
  28383,28564,28745,28926,29106,29286,29466,29645,29824,30003,30182,30360,\
  30538,30716,30893,31071,31248,31424,31600,31776,31952,32127,32303,32477,32652,\
  32826,33000,33173,33347,33520,33692,33865,34037,34208,34380,34551,34721,34892\
  ,35062,35231,35401,35570,35738,35907,36075,36243,36410,36577,36744,36910,37076\
  ,37241,37407,37572,37736,37900,38064,38228,38391,38554,38716,38878,39040,39201\
 ,39362,39523,39683,39843,40002,40161,40320,40478,40636,40794,40951,41108,41264,\
 41420,41576,41731,41886,42040,42194,42348,42501,42654,42806,42958,43110,43261,\
 43412,43562,43713,43862,44011,44160,44308,44456,44604,44751,44898,45044,45190,\
 45335,45480,45625,45769,45912,46056,46199,46341,46483,46624,46765,46906,47046\
 ,47186,47325,47464,47603,47741,47878,48015,48152,48288,48424,48559,48694,48828\
 ,48962,49095,49228,49361,49493,49624,49756,49886,50016,50146,50275,50404,50532\
 ,50660,50787,50914,51041,51166,51292,51417,51541,51665,51789,51911,52034,52156\
 ,52277,52398,52519,52639,52759,52878,52996,53114,53232,53349,53465,53581,53697\
 ,53812,53926,54040,54154,54267,54379,54491,54603,54714,54824,54934,55043,55152\
 ,55260,55368,55476,55582,55689,55794,55900,56004,56108,56212,56315,56418,56520\
 ,56621,56722,56823,56923,57022,57121,57219,57317,57414,57511,57607,57703,57798\
 ,57892,57986,58079,58172,58265,58356,58448,58538,58628,58718,58807,58896,58983\
 ,59071,59158,59244,59330,59415,59499,59583,59667,59750,59832,59914,59995,60075\
,60156,60235,60314,60392,60470,60547,60624,60700,60776,60851,60925,60999,61072\
 ,61145,61217,61288,61359,61429,61499,61568,61637,61705,61772,61839,61906,61971\
 ,62036,62101,62165,62228,62291,62353,62415,62476,62536,62596,62655,62714,62772\
 ,62830,62886,62943,62998,63054,63108,63162,63215,63268,63320,63372,63423,63473\
 ,63523,63572,63621,63668,63716,63763,63809,63854,63899,63944,63987,64031,64073\
 ,64115,64156,64197,64237,64277,64316,64354,64392,64429,64465,64501,64536,64571\
 ,64605,64639,64672,64704,64735,64766,64797,64827,64856,64884,64912,64940,64967\
 ,64993,65018,65043,65067,65091,65114,65137,65159,65180,65200,65220,65240,65259\
 ,65277,65294,65311,65328,65343,65358,65373,65387,65400,65413,65425,65436,65447\
 ,65457,65467,65476,65484,65492,65499,65505,65511,65516,65521,65525,65528,65531,65533,65535,65535};

#endif /*SINDATA_H_*/

 

[vk6kro]

Those figures are a reasonable shape for a quarter of a sinewave.

I graphed them with Excel.

But you need to then step down through the same numbers to go from 90 to 180 degrees and then go negative with the same numbers to go from 180 to 270 degrees and then step down through them again as negative numbers to go from 270 to 360 degrees.

Have you done that? Your display seems to indicate that you haven't done the 180 to 360 degree bit.

Probably you don't need anywhere near this precision, though, to get a sinewave.

Your "fat" waveform might be insufficient filtering. Looks like it.

 

[LewisEE]

As vk6kro said, your code is only computing one part of the sine wave. You'll need to fully compute the sine wave (with the trough corresponding to digital 0) and then feed that into the DAC. Then you'll need to remove the DC offset; a capacitor can take care of this. And finally, you'll need a good low pass filter for a variety of reasons: Smooth the waveform, possibly remove fatness, and to present a low output impedance to the headphones (since the cap will be high impedance compared to the headphones).

 

[ShadowPho1]

...
Your "fat" waveform might be insufficient filtering. Looks like it.

In order to graph negative I'd need a negative power supply. Which shouldn't be a problem, but I wanted to get it to be smooth and one line instead of a bunch of up/down waveform.

Why would I need filtering? Shouldn't the DAC be able to ouput a single point at each of the 16 bits?
Even if it's off at a couple then shouldn't it be more of a straight line with bad angles?

As vk6kro said, your code is only computing one part of the sine wave. You'll need to fully compute the sine wave (with the trough corresponding to digital 0) and then feed that into the DAC. Then you'll need to remove the DC offset; a capacitor can take care of this. And finally, you'll need a good low pass filter for a variety of reasons: Smooth the waveform, possibly remove fatness, and to present a low output impedance to the headphones (since the cap will be high impedance compared to the headphones).

But why do I have fatness in the first part? I mean, I went this way to avoid using a low pass filter :(

 

[LewisEE]

Why would I need filtering? Shouldn't the DAC be able to ouput a single point at each of the 16 bits?
Even if it's off at a couple then shouldn't it be more of a straight line with bad angles?

This straight line with bad angles is precisely why we need filtering. From Fourier analysis we know that a "bad angle" is really comprised of infinite frequency components. The DAC's output will have these high frequency components in it and they must be filtered out; if not, it will show up in the headphones as a very noticeable hissing sound.

 

[vk6kro]

The output from that DAC appears to be pulse width modulation. So you need to filter the fast switching component to get the average level.

For programming convenience, you could calculate all values of your sinewave at 1 degree steps from 0 to 360 degrees and multiply by 65535 to get suitable values for your DAC. This will involve some repetition of numbers but make the programming a lot easier.
This would involve fewer numbers than you have now and 1 degree steps should still give an excellent sinewave.

How is a sinewave going to entertain your girl friend?

 

[ShadowPho1]

Wow! Thank you everyone!
I figured it out. Apparently in my SPI initialization I needed to set SPIPHC which makes the SPICLK rise at the aqusition level (half phase shift behind the SPIDO) instead of the same time as SPIDO. This means that sometimes I got a random bit shift by one. Now it's all fixed!

I uploaded how it looks now. I got to make it work faster though.. got to work out the interrupt version of it.


Note: She is just a friend-girl ;)

 

[vk6kro]

Not sure what the problem was, but that graph is not a sinewave.
See attachment.

To make it go faster, use fewer points and have the whole sinewave already in the program as data.
I tried 5 degree steps and it still looks like a good sinewave.
That only takes 72 data values.

 

[LewisEE]

As noted above, you do not yet have a sine wave; what you have is essentially a half rectified sine wave. You'll need to generate the "negative" part of the sine wave. And as I said earlier, this will create a sine wave with a DC component which you'll need to remove otherwise the headphones won't play anything. This "bumped up" sine wave is what you will have to deal with because your DAC only outputs positive voltages.

 

[ShadowPho1]

As noted above, you do not yet have a sine wave; what you have is essentially a half rectified sine wave. You'll need to generate the "negative" part of the sine wave. And as I said earlier, this will create a sine wave with a DC component which you'll need to remove otherwise the headphones won't play anything. This "bumped up" sine wave is what you will have to deal with because your DAC only outputs positive voltages.

I thought that the speaker only requires the signal to be changing in order to hear a sound. Does it truly require negative voltages to be there as well?


I actually think I might have two or three modules which can do negative voltages. Do i need them?

 

[vk6kro]

The waveform you have will sound much rougher than a sinewave. A sinewave is a pure sounding note, but a bit boring and irritating after a while.

You don't need negative voltages. Just take your output through a suitable capacitor.

You could try the following values for your data. They give a complete sinewave with a peak amplitude of 25000. Should fit into your D to A OK. They graph OK in Excel.

30000
32179
34341
36470
38551
40565
42500
44339
46070
47678
49151
50479
51651
52658
53492
54148
54620
54905
55000
54905
54620
54148
53492
52658
51651
50479
49151
47678
46070
44339
42500
40565
38551
36470
34341
32179
30000
27821
25659
23530
21449
19435
17500
15661
13930
12322
10849
9521
8349
7342
6508
5852
5380
5095
5000
5095
5380
5852
6508
7342
8349
9521
10849
12322
13930
15661
17500
19435
21449
23530
25659
27821
30000

 

[ShadowPho1]

Thanks!

Here's your data. It looks nice, sounds pretty nice. I used a really small capacitor (1u I think) and it looks great!
Can you please tell me what affects the size of capacitor has?

 

[vk6kro]

OK.
That is showing about 40 steps for the sinewave instead of 72 so the steps are coming from the screen resolution. Anyway, glad it worked.

The reactance of the capacitor should be negligible compared with the impedance after the capacitor.
So if the frequency was 2000 Hz and the following impedance was 100 K ohms, the reactance of the capacitor would need to be less than 10 % of 100 K ohms.
Reactance = 1/(2 pi F C) C= 1 /( 6.28 * 2000 * 10000) or about 0.008 uF, but more C is OK.

If you can add a 1 uS delay before reading each data value, the next frequency after 2145 Hz would be 1858 Hz because this would add a 72 uS increase to the total period. Further delays would drop the frequency more. It may be possible to play tunes by changing this delay.

 

[ShadowPho1]

OK.
That is showing about 40 steps for the sinewave instead of 72 so the steps are coming from the screen resolution. Anyway, glad it worked.

The reactance of the capacitor should be negligible compared with the impedance after the capacitor.
So if the frequency was 2000 Hz and the following impedance was 100 K ohms, the reactance of the capacitor would need to be less than 10 % of 100 K ohms.
Reactance = 1/(2 pi F C) C= 1 /( 6.28 * 2000 * 10000) or about 0.008 uF, but more C is OK.

If you can add a 1 uS delay before reading each data value, the next frequency after 2145 Hz would be 1858 Hz because this would add a 72 uS increase to the total period. Further delays would drop the frequency more. It may be possible to play tunes by changing this delay.

I am learning about reactances and impedance next semester ^^.
I do remember something simple about it.

I have a bunch of lm741cn. (http://www.national.com/ds/LM/LM741.pdf" ) Their listed input resistance is typical 2 megaohms.

For sound I'd need 22 kHz, and if the typical 2megaOhms is actually 2 megaOhms.. 1% of 2 megaOhms = 20k ohms.
1/(2 pi f C).
so, 20000 = 1/ (2 * 3.14159 * 22 000 * C).
2.76460154 × 10^9 = 1/C
C = 361715779 F or smaller... wow.

As for tunes, I want to connect it to real headphones and play wav music.

Now I am not sure if it's possible without some kind of an amplifier. I have a lot of MOSFETs, but I don't think I can really use them.

Edit: OMG I found a perfect thing! http://kitsrus.com/projects/tda7052.pdf"
How do I know if my headphones are 8 ohm?

Edit2:
100k impedence in this amplifying thing. 1% is 1k ohms.
1000=1/ (2 * 3.14159 * 22 000 * C).
C=7.2 nanoFarads or smaller... :(

 

[vk6kro]

I make that first capacitance 361 pF (pico-Farads ie 10 to the minus 12 Farads.) Pretty small. Anything bigger than that is OK. Just use 0.1 uF.

You can just measure the resistance of your headphones with a multimeter. 8 ohm speakers and headphones just measure 8 ohms.

 

[ShadowPho1]

I make that first capacitance 361 pF (pico-Farads ie 10 to the minus 12 Farads.) Pretty small. Anything bigger than that is OK. Just use 0.1 uF.

You can just measure the resistance of your headphones with a multimeter. 8 ohm speakers and headphones just measure 8 ohms.

Awesome! Thank you! I got a 0.1F, it works heh.

I tested one of my speakers, and it draws 0.1 amp (8ohms), how much will that last a battery?

 

[Phrak]

I'm not sure you're on the right page. Impressing a girl with an audio file is about as effective as piling your rice into a volcano during a dinner date and making explosion noises to show how clever you are. Unless, of course, the audio message indicates some promise on the short-list of what all women want.

 

[ShadowPho1]

I'm not sure you're on the right page. Impressing a girl with an audio file is about as effective as piling your rice into a volcano during a dinner date and making explosion noises to show how clever you are. Unless, of course, the audio message indicates some promise on the short-list of what all women want.

*boom boom!* *vroooom, boom!*

There's no promise in the audio file... I will just be playing "Girl" by Beatles - my personal favorite.

I guess I am not really doing this for her per say. It's a way of learning something new, saving $30 and seeing if the girl is the right one ;)


Here's a question though... Does anyone know where I can get a real SD SPI specifications? So far I only found a 1.0 revision, and a bunch of hobbyist sites which seem to differ for every card.

Specifically when I give the SD card a multi-byte command, do I pull the CS low for the entire command or can I only do it for each byte?

 

[Phrak]

*boom boom!* *vroooom, boom!*

There's no promise in the audio file... I will just be playing "Girl" by Beatles - my personal favorite.

I guess I am not really doing this for her per say. It's a way of learning something new, saving $30 and seeing if the girl is the right one ;)

Sorry. That was really a rough comment on my part. It's an unusual girl that has some genuine interest in technology. Good luck to you.

 

[vk6kro]

I'm not sure you're on the right page. Impressing a girl with an audio file is about as effective as piling your rice into a volcano during a dinner date and making explosion noises to show how clever you are. Unless, of course, the audio message indicates some promise on the short-list of what all women want.

Still laughing at that one, Phrak. I wondered the same thing.

What would be far more effective is if you could show you can put up a shelf on a wall, stop a faucett dripping, test torch batteries, fix a flat tyre or cook a chicken (preferably Coq au Vin).
Then show you can remember her birthday and her parents' first names, but still address them as Mr Smith and Mrs Smith (or whatever hot-chick's surname is).
About then, you start to become genuinely useful. Just don't talk about D to A converters.

 

[ShadowPho1]

Sorry. That was really a rough comment on my part. It's an unusual girl that has some genuine interest in technology. Good luck to you.

No offense taken :). It wasn't really any rough, sorry if I seemed offended.
Thanks for the luck wishes.

I'm not sure you're on the right page. Impressing a girl with an audio file is about as effective as piling your rice into a volcano during a dinner date and making explosion noises to show how clever you are. Unless, of course, the audio message indicates some promise on the short-list of what all women want.

Still laughing at that one, Phrak. I wondered the same thing.

What would be far more effective is if you could show you can put up a shelf on a wall, stop a faucett dripping, test torch batteries, fix a flat tyre or cook a chicken (preferably Coq au Vin).
Then show you can remember her birthday and her parents' first names, but still address them as Mr Smith and Mrs Smith (or whatever hot-chick's surname is).
About then, you start to become genuinely useful. Just don't talk about D to A converters.

I can put a shelf on a wall, stop a faucett dripping, flat tyre, chicken (even spanish rice!), paint, add insulation, walls, etc.

We moved here from russia, and I used to do lots of home/car work with my dad. He taught me quite a lot of cars/house/physics.

What do you mean by torch batteries? Car batteries?

 

[vk6kro]

Sorry...flashlight batteries. We call that a torch.

 

[ShadowPho1]

Sorry...flashlight batteries. We call that a torch.

Is there anyone who is incapable of replacing flashlight batteries?

 

[vk6kro]

What would be far more effective is if you could show you can put up a shelf on a wall, stop a faucett dripping, test torch batteries, fix a flat tyre or cook a chicken (preferably Coq au Vin).

I wouldn't expect that anyone reading this Forum would have a problem testing D cell batteries, but we were talking about what would impress a young lady.

 

[Phrak]

I found this on another forum that might be helpful.

http://www.edaboard.com/ftopic218555.html"

Whos microcontroller are you attempting this with?

 

[LewisEE]

What would be far more effective is if you could show you can put up a shelf on a wall, stop a faucett dripping, test torch batteries, fix a flat tyre or cook a chicken (preferably Coq au Vin).

I wouldn't expect that anyone reading this Forum would have a problem testing D cell batteries, but we were talking about what would impress a young lady.

What if the young lady is an aspiring physicist?

 

[vk6kro]

Then you are very lucky because such girls are usually very beautiful, make great companions and produce cute, intelligent children.

(My wife has a Science degree and I'm sure she would agree with the above..) :)

 

[ShadowPho1]

What would be far more effective is if you could show you can put up a shelf on a wall, stop a faucett dripping, test torch batteries, fix a flat tyre or cook a chicken (preferably Coq au Vin).

I wouldn't expect that anyone reading this Forum would have a problem testing D cell batteries, but we were talking about what would impress a young lady.

Ah... gotcha...

Hmm, she actually WAS pretty impressed with how me/parents painted the room.
I love how this thread also generates girl advice ^^,

I found this on another forum that might be helpful.

http://www.edaboard.com/ftopic218555.html

Whos microcontroller are you attempting this with?

MSP430F2012.
It has an onboard ADC which I am not using, it has SPI which I am using heavily and it has no DAC.
It's easy to program though, I have a bunch of them and I have a programmer.

 

[ShadowPho1]

Thank you everyone for your help with this! The DAC worked flawlessly, the power amp works perfectly, the 0.25 watt speaker works perfectly, the microproccesor works perfectly.

But I couldn't read anything from the SD card. Nothing at all. I thought I could do it, but I ran out of the time at the last moment. I even resoldered all the SD connections...

Ah well, so I gave her foundation series by Isaac Asimov.