Freq generation granularity for min phase noise

Toby Seckshund

I had been playing with some, now redundant, analogue TV tuner heads. The control of their frequency is by a long I2C sequence and can give 31.25, 50 or 62.5KHz steps. I did get it to work just before the old TV system got shut down last August (but with an AVR, sorry), Now I thought that a 50-850MHz Rx using these heads would keep me off the streets for a bit. Then I thought .... Then I thought ...Mission creep has got to the stage that I am thinking of nore or less just bolting a Prop onto the head, using the "Bolt an aerial to your Prop" I & Q stuff and keeping the intermedery stages down to a minimum. The heads were used to shoving out 33-39MHz so I think that 35MHz for the first IF and then dropping it down again for demod, or even lower in another stage to the usual 455KHz region.

So the question is, am I right in my assumption that the phase noise would be reduced if the first local osc is on a 5MHz boundry (usual 5MHz xtal for the ppl x16), which would be 35MHz in this senario. I recall that it is dependant on there being all ones or no ones in the settings, somewhere.

I know that there will be all sorts of problems generating the end local osc which would have to mimic the infill for the 50KHz stepping but then the tuner is a PLL off of a 4MHz xtal so it would never be the quietest of recevers, I just dreamt of getting a general coverage Rx for the VHF bits around here ie Wx sats, local airport and (non dx) 2 Metres.

Phil Pilgrim (PhiPi)

Toby,

How about this: use a 6.25 MHz crystal to overclock the Prop. With that you can generate a clean 25 MHz LO which, when mixed with 37.5 MHz from the tuner, will give you a 12.5 MHz IF. At 12.5 MHz, you will be able to generate very clean I and Q sampling frequencies internally, with which to do your demod.

-Phil

Toby Seckshund

That sounds excellent. If only I had some 6.25 rocks!

I will have to get some at some time but I do have a 6.55MHz one from something, and a few 6MHz ones and given the wideband output of the tuner head there should give some flexibility. I saw a Russian artical that slung the output of a tuner straight into a Tayloe (I think) detector and then went the usual PC sound card way for the demods.

I still havent got it into my thick skull how to do narrow and narrowish FM demod yet so hopefully it will be another self education (more keeping me off the streets, always a bonus).

Phil Pilgrim (PhiPi)

Toby Seckshund wrote:

more keeping me off the streets, always a bonus

Are you a danger to the citizenry?

The catch with implementing an I/Q demod with the Prop is that the frequency cannot be more than clkfreq / 4. Otherwise it's impossible to produce the necessary quadrature relationship. And, if the IF is a significant fraction of clkfreq, it needs to be a power-of-two submultiple of clkfreq to avoid as much phase noise as possible.

-Phil

Toby Seckshund

"Are you a danger to the citizenry?"

I don't think so. As none of the injuntions have landed ... :- )

If there has to be exact multiples, and divisions then I may as well use xtal osc. I would like the Prop or a-n-other proc to do the housekeeping like freq selection and read outs, vid or pref VGA readouts would be a bonus in addition to a 20 x 2 LCD hence the Prop.

When I tried to do the Rx stuff on a 40 DIP it didn't like it too much so this time I will try a QFPP

Toby Seckshund

Phil,

One thing I have just thought of, is there a particular pin or cog asignment for the inputs and/or outputs that yields better results.

I know that for audio there seems to be those cogs etc that give better noise etc.

Phil Pilgrim (PhiPi)

I used P3 and P4, but only because it was easy to solder a resistor across those two corner pins on the Propeller Demo Board.

-Phil

Toby Seckshund

I summizes so much. I have never used a Quad Prop as all of my recycling is based around DIPs, in bases usually.

Toby Seckshund

I am wondering if this bit of twisted logic would work.

If I run the Prop from a 5MHz VFO ( X 16) and have a xtal osc as a reference then by forcing an error and measuring it the with some maths I could get a 40MHz from a counter plus, or minus 30KHz, without changing the counters away from the "magic power of two". This would get me the infill between the 50KHz steps of the TV tuner. by picking 37MHz as my first IF then this will give me a 3MHz second IF and that should be ok for your I & Q software.

Phil Pilgrim (PhiPi)

One catch: if the Prop is clocked from a VFO, you will have to keep changing your I and Q frequencies, tracking the VFO, in order to keep them at 3 MHz.

-Phil

Toby Seckshund

That was why I was hoping that the small deviation (60KHz), relative to the final IF (3MHz), it wouldn't matter too much. That 3MHz could be made bigger, if it helped.

Edit Ah I see what you mean. All of this hoop leaping would be got by some "off magic power of two" stuff somewhwere else

jmg

Toby Seckshund wrote: »

I am wondering if this bit of twisted logic would work.

If I run the Prop from a 5MHz VFO ( X 16) and have a xtal osc as a reference then by forcing an error and measuring it the with some maths I could get a 40MHz from a counter plus, or minus 30KHz, without changing the counters away from the "magic power of two". This would get me the infill between the 50KHz steps of the TV tuner. by picking 37MHz as my first IF then this will give me a 3MHz second IF and that should be ok for your I & Q software.

I wondered about a Ceramic resonator, and this suggests it could work well here

http://www.qrp4u.de/docs/en/Oscillators/index.htm
Mentions 35Khz at 7Mhz so 50KHz at 40MHz about one forth of that range.

Toby Seckshund

I have not seen the actual figures for the jitter in different counter modes. I see that the OBEX object to demo two freqs from one cogs counters chooses to use NCO if the freq required is <500KHz and PLL for those above, so would the choise of a final mixer freq be a good thing??

Or should I just accept that there is going to be some and just live with it ! After all the Rx experiment is going to start off with a domestic TV tuner unit and they where never designed for the laboratory standards. Any Rx would be better than what I have now ... ie nothing.

Elektor magazine has recently ran a triple artical on "SDR" using a 20MHz AVR, I must sit down an digest that and then, perhaps, the rudiments of DSP should be a tiniest bit clearer to me.

(Moral of the story :- Kids please, please work harder whilst in your education. It gets harder when you are old and going dense and going long sighted .....)

Toby Seckshund

I have knocked up a 28MHz "known" so that I can try to slew the Prop's clock rate around the nominal 5MHz and so get the jitter free x8 of 40MHz (ish) and the I & Q of 5MHz (ish). this means that the use of a high sided LO of the 40MHz against the 35MHz first IF will give the complication of 8/7 steps because the resultant IF shifts too. Thats where the 28MHz bit comes in (I hope) as with all the scribbled sums I think that 800Hz offset on the 28MHz will equate to the wanted 1KHz at 35MHz.

So the prop will end up using a 5MHz VFO as its main clock and that will be bent around so that the +/- 25KHZ at 35MHz will be locked to the constant 28MHz.

Any thoughts?? Or should I change the flux that I am using ??