1.2GHz CMOS synth

Hi,
I am glad it helped you.

VBR from Ivan

Thanks Ivan, downloaded the program, its just what I needed.

John_1981 wrote:For example, 3541 is a difficult one, the best approximation I could find was N=360, R=1220. This comes out with an error of 16.4Hz. Probably good enough to receive SSB though.

My program found N=306 and R=1037 with the same deviation.

73 Ivan

Hi John,
I have administrator rights, but I cannot see your email address. Perhaps it is because you are a guest, not a registered user.

Currently I do not have a website to put the file there for download, but I placed the file to a public fileserver. Click the URI: [url= https://www.uschovna.cz/zasilka/ROC6I2326NH3VZPT-MWG/] https://www.uschovna.cz/zasilka/ROC6I2326NH3VZPT-MWG/ [/url]
Then click the yellow button "STAHNOUT ZASILKU" (yes, the website is in Czech) and the EXE should be downloaded to your device. When I posted it, it was tested by an antivirus of course. The file will be erased after 90 days by the provider. The file can be run without installation.

73 from Ivan

"Hi John,
no genius was required. I wrote a small utility for Windows. You fill in the reference and the wanted frequency, click RUN and the R and N values are found. I tested it on the examples you state. How can I send it to you? it is an EXE, so it may be blocked by all firewalls and antiviruses.   It might be useful for anyone else.."


Ivan, 


Yes this sounds like something that would be useful. With certain frequencies it is just a matter of finding the highest common factors of the two numbers, but where the wanted frequency turns out to be a prime number the only option is to find the best approximation, trading off having as high as possible a reference frequency whilst minimising the error. For example, 3541 is a difficult one, the best approximation I could find was N=360, R=1220. This comes out with an error of 16.4Hz. Probably good enough to receive SSB though. A quick search has led me to Euclids' algorithm as a method to find the HCF of two numbers but this won't help when i'm stuck with prime numbers and integer R/N values. 


I do like the 12MHz crystal as the basis for the reference frequency as I can readily generate 5, 6.25,7.5, 8, 10, 12, 12.5, 15, 20, 25 kHz etc. 


As it stands I'm not sure how to go about sharing files easily. I do use dropbox but I would have to share usernames etc. and leave it open to spammers. I'm happy for anyone with admin rights on the forum to share my email address with you if it is visible to them. I also have some useful stuff I could share, I wrote an excel spreadsheet for calculating the 28/32 bit tuning word for the AD9833 and AD9850 DDS modules and some PIC programs for controlling the same using a rotary encoder. 


Speaking of the AD9833, I have thought of using this with a PLL to extend its frequency range and clean up the output. As it stands the output gets messy above about 7MHz. The AD9850 modules are power hungry and have got expensive, but the AD9833 units are still fairly cheap and fit on a postage stamp.

John wrote:If there are any mathematical geniuses out there i'm looking for a way to automate finding which R and N values get me closest to each 1kHz step within a given band. At the moment I'm using Excel but I have to manually go through a list of values and select the closest one. I'm looking for the best way to approximate, for example,  3728=(12,000/R)*N where N and R must be integer values and I want R to be within the range 1000-1600. 

Hi John,
no genius was required. I wrote a small utility for Windows. You fill in the reference and the wanted frequency, click RUN and the R and N values are found. I tested it on the examples you state. How can I send it to you? it is an EXE, so it may be blocked by all firewalls and antiviruses.   It might be useful for anyone else...

VBR from Ivan