PLL~ Cascaded divide by N....

[krazatchu]

Hi ppl...

I'm trying to build a sinewave generator and my requirements are...

1) stability, therefore xtal clocked...
2) frequency range, 20 kHz to 30 kHz, ultrasonic...
3) resolution of at least 1 Hz, and some fine tuning...
4) not using pic, since I don't have experience with them...

I searched the web a bit and found this excellent site,
**broken link removed**
Go to (top menu) Projects and then Synth circuits, (left menu) Synth basics and CMOS synth...

I read thru that a few times to get a clear understanding of PLL based synths...

So... It seems I want to create a 1Hz refernce signal for my minumum resolution...
I can add a flip flop to the output of the 4060 and use a 32.768 kHz xtal...
or some other mathimatical rendition of the same...
absolutely no problem there...

The part that I'm having trouble with is the divide-by-N (4059)... It's max is 14,999...
I need to cascade that with another divide by N so I can continue counting up to my desired range (20 kHz to 30 kHz)...

Now... I've searched hi and low and haven't found a suitable replacement or example of a (working) cascade... I tried a 74161 circuit from here,
https://www.wenzel.com/pdffiles1/pdfs/161divdr.pdf
It didn't simulate properly...

So... my simulator doesnt have the 4059.. also doesnt have any pll stuff.. Is there a simulator that has support for ppl??

Anyhoo.. I was hoping someone could drop me an example of a cascaded 4059 or other relevant divider...

Also.. another question, using the 1 Hz reference signal, is that going to increase the lock/hunt time? Should I use a higher frequency/lower resolution and then divide the output down before I convert it to sine?

Was think of using a max292 for the sine conversion, it's a switched cap butterworth filter... but thats for another day...

thanks,
Michael

 

[krazatchu]

Just had another thought...

From a mathimatical prespective... I could add a count of about 20,000 before the 4059... then use the 4059's range...

Problem is... I don't know how to add (mathimatically challenged)... Multiply is easy but that reduces my resolution~~

Thanks

 

[krazatchu]

Having a lot of thoughts today ~~

If I want to add an anologue offset for fine tuning...

Changing the values around the xtal wouldn't give me a +/- 0.5 Hz so...

Could I strap a pot to pin 12 of the 4046? Or would I have to use an op amp to combine the offset after the LP filter and before the VCO?

Thanks,
Michael

 

[Diver300]

I would not use a PLL, certainly not with a reference frequency as low as 1 Hz. It will be very difficult to get stability.

I think that you should use a DDS (direct digital synthesis) IC. They will take in a clock signal, for example at 10 MHz, and a high-resolution tuning word that is typically 32 bits long and the output is a sinewave.

32 bit resolution on 10 MHz gives about 0.0023 Hz resolution. If you make the clock signal a power of 2, for example 16.777216 MHz, then you can easily go up in 1 Hz steps, but still have finer tuning if you want

The biggest problems is getting the tuning word into the DDS chip. DDS chips are available as serial load or parallel load. Serial load needs something like a PIC to produce the serial string. Parallel load needs lots of pins and the DDS ICs that I know of are only available in very fine pitch packages.

I would use https://www.electro-tech-online.com/custompdfs/2007/11/ad9831eb.pdf

which is the AD9831 DDS IC on an evaluation board. That lets you set the frequency directly from a PC, or you can use switches and load the data in parallel. You still have to load the data in two halves, 16 bits at a time, but that is possible from switches.

I have used a PIC to drive a serial DDS IC. That is used to get any frequency from 0 - 3 MHz with a resolution of 0.03 Hz. The frequency is dialed in on thumbwheels, which the PIC reads and send to the DDS IC.

 

[krazatchu]

hey~

Thanks for the info~

Originally, I looked at DDS, but for the exact reason that I have no PIC experience I was avoiding it... My soldering skills are not that great either... and that evaluation board is about $150... I'm keeping that in mind (DDS) as a last resort...

So, back on the subject of PLL...
If a low refernce freq. is going to cause problems... then I should perhaps use 100Hz as reference and then divide the final output by 100 resulting in 1Hz resolution...

The site I linked states the max freq at 9VDC of the 4046 is 3.5 MHz.. So using a 100 Hz reference and a divide by N max of about 32,000... I would output 3.2 MHz and then divide by 100...

Now.. the part I'm having trouble with is figuring out how to cascade the divide by N ...
Perhaps I'm using the wrong words for googling... My results have been minimal...
Is cascade the correct term? cascade counter? cascade divider?


Thanks,
Michael

 

[RadioRon]

Hi Michael
I tend to agree with Diver300 that, from a theoretical point of view, going with DDS is likely to work a lot more easily. However, if you absolutely insist on using PLL techniques, then I still have a problem with 100 Hz resolution. I am assuming that you would build a simple integer divide-by-N type of loop, not a fractional-N loop and so your resolution of 100Hz is also your phase detector comparison frequency. A common tradeoff that you will need to face with such a low comparison frequency is that you will have sidebands in your signal at multiples of the comparison frequency and since you haven't mentioned what your goals are for phase noise and spurious content I take it that such sidebands don't matter very much. Suppression of these sidebands often forces designers using the 4046 as a phase detector to push their loop frequency quite a bit lower than you would need for basic stability. For example, in designs using 5 KHz as a comparison frequency, I have seen loop filter cutoff frequencies down as low as 10 Hz! So, I'm wondering about how well your idea for 100Hz will work. If you are forced to push your loop frequency down as low as 10Hz or less, then the lock time may become too long and/or you may not be able to achieve a decent damping factor and therefore suffer poor stability.

Perhaps you should consider synthesizing at a much higher frequency, like 300 MHz using a single chip synthesizer from National and a simple FET VCO, and then dividing down the output by 10,000? But this seems like overkill considering you only want 20KHz to 30KHz and the VCO would have to cover a pretty wide range, from 200 to 300 MHz. Hmmm, maybe too tough to do. Another approach would be to go with a two-loop topology where one loop is responsible to give you a narrow range of frequencies with fine resolution and the other loop looks after the coarse resolution and broader frequency range.

 

[krazatchu]

hmmm...

Guess I wil lhave to bite the bullet and learn how to program PICs...

I was hoping to avoid that until I needed it later... I really have alot of other things to focus on right now ...but I might as well learn it now..

I don't supose anyone knows of a function generator kit thats not too expensive...

As for signal noise... I was planning on using a max292 to convert the square to sine... the purpose of this circuit is too drive an amp then some piezo donuts as part of a transducer...

That two loop idea sounds interesting... I have no idea how to implement that tho... What would I search for if I wanted to learn more about that?

Thanks,
Michael

 

[krazatchu]

Hey~~

I went to the electronics market yesterday to check out PIC writers...

They didn't have much selection at the 300+ shops... The cheapest I found was around 250USD...

They did however seem to have an awful lot of 8051/AVR/Atmel writers... at around 20 USD ...

I guess the PICs are not so popular here in South Korea....

Now... I don't know either way the differences... Is there any specific reason I should go with PIC over AVR/Atmel ?? or vice versa??

Thanks,
Michael

 

[RadioRon]

Of course there are fans of both kinds. I'm only familiar with AVR and find them to be very good. The PIC series is better known. I'm not sure what the technical advantages are of AVR vs PIC, but I think either series is fine.

 

[krazatchu]

Thanks~~

I'll browse around the micro controller section and figure out what to do~~

Michael

 

[Nigel Goodwin]

I'll browse around the micro controller section and figure out what to do~~

As suggested, either PIC or AVR would be fine - as would the 8051, but these are really low spec antiques, and only seem popular out in India and such places.

 

[krazatchu]

Hi ya... back again...

I decided to go with the AVR for DDS... found a couple of sites with somewhat detailed instructions... then I found out the AT90S2313 is obsolete...

I started a new thread over in micro controllers... I'm hopeing someone can give me a hand there...

https://www.electro-tech-online.com/threads/avr-dds-rs232-problem.34479/

Thanks, Michael