Direct Digital Synthesis. (DDS)

AdrianJ · Joined: 16 Jan 2006 Posts: 2483 Location: Queensland

Sorry for that, comes from writing code "on the fly"

the statements
lds zl,{low wWavetable}
lds zh,{high wWavetable}

should be
lds zl,{wWavetable}
lds zh,{wWavetable+1}

to get z pointed at the value loaded into the word wWavetable in ram.

It is really easy to get confused about the various forms of syntax for pointers in rom and in ram, and whether you want the value of a variable, or its address.

davidapex · Joined: 12 Mar 2006 Posts: 42 Location: Auckland

Hello Luciano,
Many thanks for the comprehensive answer to my post. I believe I can follow your code but could never had originated it - still many months away from this level.

That's a busy interrupt routine - more than I would have comtemplated. I look forward to trying it.

BTW - back in the '70s when function generators were high tech there was a great myth about dovetail waveform generators. Imagine a square wave with a rise time that goes back in time (T anticipation) and a fall time that also goes back in time (T hesitation) While I never saw a dovetail waveform generator I know a few who made raster scans on their 'scopes that looked like they had "cracked" it.

Thanks again.

Cheers,
David

davidapex · Joined: 12 Mar 2006 Posts: 42 Location: Auckland

Hello Adrian,
Thanks for the update.
I will go back and revisit this. Luciano has kindly posted all I need to get underway but I also need to understand the process involved rather than just have code handed to me so I will go back and play a bit more to help my understanding.
I agree - even without syntax it's very confusing trying to learn about variable contents and variable addresses. Will keep chipping away.....

Cheers,
David

Luciano · Joined: 29 Nov 2004 Posts: 3149 Location: Italy

Hi,

Just one quick note to avoid confusion:

The code sample I have posted yesterday and the code sample I have posted last
week have the table(s) in ROM only. (The table is not copied from ROM to RAM).

As far I understand, for now, David is working on code with the tables in
ROM only. It makes sense to load the table in RAM, but if we jump back and
forth between the two solutions this thread will be difficult to follow.

Please specify clearly in your post if the table is copied first from
ROM to RAM and then its values read from RAM.

* * *

David,

You are welcome!

I have edited the sample code posted yesterday.
(Eliminated 3 lines of redundant code, "Set Eifr.6" moved down, comments improved).

About the interrupt routine:

Yes, it looks strange but it will work. If you add a 4x4 keypad
and an LCD, the keypad will be read in polling mode from within
the interrupt routine. (The keypad will also trigger the interrupt).

Best regards,

Luciano

davidapex · Joined: 12 Mar 2006 Posts: 42 Location: Auckland

Hello Luciano,
Yes I am still using the wave tables in the code space and not in SRAM. May try that later. Will also try the keypad - I had this on my Pic version and it was really nice to tap in the numbers.
I don't have a need for a signal generator as I have good audio generator but DDS has always fascinated me and it also allows you to write your own waveform (dovetails) as well as having programmable precision.

I think it should be possible to make an FSK sender from the DDS by converting the input data to binary and switching between two set frequencies. As mentioned before, it should also be possible to run two (smaller ) accumulators and generate two waveforms, sort of analog DTMF

Thanks again for your input. I will take little steps so I don't break anything.

Cheers,
David

jenalcom · Posted: Fri Mar 31, 2006 5:34 am Post subject:

Bascom already does that to produce DTMF!

And it will also work with a standard 3.579545MHz crystal so you can run a decoder of the same clock.

Alan

Luciano · Joined: 29 Nov 2004 Posts: 3149 Location: Italy

David was talking about FSK, not DTMF.
(Read "sort of analog DTMF").

Luciano

(Click to enlarge the picture)

jenalcom · Posted: Fri Mar 31, 2006 12:18 pm Post subject:

He actually did mention DTMF Smile

<quote>
As mentioned before, it should also be possible to run two (smaller ) accumulators and generate two waveforms, sort of analog DTMF
</quote>

Using the same (similar) code it should also be able to generate FSK.

Personally I always use square waves for FSK, especially as most FSK (I use) is sent over severely limited bandwidth links (eg radio with 2.4KHz b/w). I normally use a timer in auto reload mode to generate the square wave. I have thought about using DDS for this but never got that far with the idea.

Of course if you wanted to get really clever you would be able to do Phase Shift Keying as well or perhaps even QAM and other modulation schemes. Unfortunately I think the loop size would probably get too large too quickly. May be worth investigating if you have a high enough cloak frequency.

Alan

davidapex · Joined: 12 Mar 2006 Posts: 42 Location: Auckland

Hi,
I did indeed mention FSK and DTMF, not for any specific task but more to highlight the flexibility of the DDS.
Phase shift keying should be straight forward as would quadrature waveforms (or any phase)
It should be possible to generate 3 phase sinewave using three output ports but a single accumulator and three table reads. These would be the calculated table value, table value plus 60 degree offset, table value plus 120 degree offset. Not sure if it would have any applications but the theory is nice.

Cheers,
David

AdrianJ · Joined: 16 Jan 2006 Posts: 2483 Location: Queensland

Yes, I do more things with this DDS than just a single tone. In fact I have twin outputs from the same wavetable, but I can adjust both independently for both amplitude and relative phase. So I can synthesise a "stereo" effect by both delaying one wave with respect to the other, and reducing its amplitude, to give the effect of turning the head. Only works with phones on, but its fairly effective. The amplitude effect seems to be the most obvious, the phase effect is difficult to pick by itself.

davidapex · Joined: 12 Mar 2006 Posts: 42 Location: Auckland

Hi,
Over the last month I have had stolen just enough time to make a real pigs ear of the code that Luciano posted and have got myself in to trouble.

I have learned how to add the keypad and LCD to this Mega168 DIP but was not impressed with the chip pin layout or the peripheral feature placement.
Why did they put the xtal on one of the full 8 bit ports and the UART on the other full port, leaving the incomplete 7 bit port alone? Huh? Surely you would trash an incomplete port first?

Anyway, I have removed the xtal and are running off the internal osc for the time being but I have made a mess of the interrupts. I currently have portb.7 as an interrupt but would like to use the keypad eventually. I have used a GOTO instead of return for the ISR just to make something happen but I know it's not correct. The present code starts with the preset frequency values then the interrupt clears it and waits for the keypad input. The value input is then loaded in to the phase register and the new frequency is output. Then nothing else happens - no interrupt. BTW the lookup table offset is not indexed to a boundary either but this is minor compared to the other problems.

Can someone show me how to get my interrupt to be more than a one shot?

Best regards,
David

Luciano · Joined: 29 Nov 2004 Posts: 3149 Location: Italy

David,

From the Bascom Help file:

ON INTERRUPT

The first RETURN statement that is encountered that is outside a condition will generate a RETI instruction. You may have only one such RETURN statement in your interrupt routine because the compiler restores the registers and generates a RETI instruction when it encounters a RETURN statement in the ISR. All other RETURN statements are converted to a RET instruction.

From the code you have attached:

davidapex · Joined: 12 Mar 2006 Posts: 42 Location: Auckland

Hello Luciano,
Thanks for that. That return is the return from the regval gosub. I tried restructuring that to in-line code rather than the gosub and although it appears to be working it does not load the correct values in to the phase accumulator anymore.

Will do some more fiddling now you have pointed that out.

Thanks again.

Cheers,
David

Luciano · Joined: 29 Nov 2004 Posts: 3149 Location: Italy

Hi David,

The 6 assembly lines below are the main loop. When the
interrupt occurs, the execution will stop somewhere in
these 6 lines of code. When you return from the interrupt,
the execution will resume where it was left when the interrupt
occurred. Again, you don't know where in these 6 lines of code.

davidapex · Joined: 12 Mar 2006 Posts: 42 Location: Auckland

Hi,
I think I follow that. That also explains why it sort of worked when I used a goto Begin: instead of the Return. I had got confused about where the interrupt was actually happening. I need to have a long hard look at it again - good thing it's the weekend.

Cheers,
David