FSK Radio design

Can you actually get a PLL to track that fast ?

Yeah, good point, perhaps not. In that case I would have to mix it down anyway.

No matter what you do, if you try to do everything at 1GHz you're going to be struggling against the 1000:1 ratio of your carrier frequency to your data rate. Unless your frequency shift is huge (and what _is_ your frequency shift, by the way?) you're going to be comparing relatively teeny deviations against a relatively large carrier.

I'm not sure why you're willing to embrace a PLL but reject anything with an LO -- a PLL needs an oscillator just like a downconverter would, but that oscillator may very well need to be even better performing (because you're making it part of your demodulator) than an oscillator in a downconverter.

A discriminator is going to have a similar problem to a PLL, for the same reasons (you want a Q of 1000???).

There ought to be cheaper demodulators out there -- if not, then perhaps you want to consider a superhet.

That is only 0.1 % deviation, if the decoding is done directly at RF.

If PLL decoding is used, how would initial locking be achieved. Making the VCO start reliably within 0.1 % might be a problem over the temperature range and between different units (assuming no manual adjustment possible).

In that case downmixing would absolutely be needed, due to the 0.1 % deviation at 1 GHz. In audio FM demodulation, about 1 % relative deviation is used, e.g. +/-75 kHz at 10,7 MHz IF for broadcast FM and

Tim, You bring up very good points. The shift is +- 300KHz, three orders of magnitude smaller than the carrier. So is it unreasonable to mix with say ~990 MHz to get a 10 MHz IF? By unreasonable I mean does designing a LO of that frequency present more untenable design challenges?

At 10 MHz the shift becomes +- 3KHz. That seems small, I am not sure how it affects the prospects of using a PLL.

With my limited experience I am definitively inclined towards cookbook solution e.g. mixers with built in oscillators etc.

Thank you again for sharing your expertise.

Rich

Indeed! Everyone overlooks the old simple ways that always work and have better SNR than microprocessor solutions.

I made an active-filter equivalent to demodulate modem signals back in the '70's (Patent 4,472,816) and blew everyone's minds at GenRad... did an acoustic modem loop from Phoenix to Louisville, KY, and back, error-free ;-) ...Jim Thompson

Yes, if the VCO works that fast. Most varicap-based VCOs essentially lowpass filter the control voltage.

Mixing down is appealing, though. If you can mix down really low, so the FSK is, like, 1MHz:2MHZ or 4:5MHz, some good ratio, you can do a simple tachometer type FM demodulator.

If you don't need good noise and image rejection, a simple mixer will do.

Maybe a ceramic resonator and a little mixer thing would demodulate FSK at 1 GHz. The Qs are about right. Cheap.

John

One way to get around the stability issue would be to use a translation loop.

Let the VCO operate at say 50 MHz and mix it up with a fixed stable frequency source at 950 MHz (or 1050 MHz) and run it to the PLL phase detector. It is easier to mix two constant amplitude signals (VCO and XO) than trying to down mix part of the RF spectrum with signals at very different levels.

If the VCO up mixing is done with an image rejection mixer (I/Q), the fixed source at 990 MHz and VCO would be at 10 MHz, but of course, this limits the tunable range, if several RF channels must be handled.

It shouldn't. There's plenty of synthesizer parts in that frequency range.

Your big problem will be that you're IF to RF ratio is 100:1. That means that if you want decent image rejection you'll need one hell of a filter at 1GHz -- this is why downconversion in budget radios is often done directly down to baseband.

The shift is not changed by mixing. You're not dividing the carrier down -- you're translating it, more or less whole, to a lower frequency.

I would expect that there are inexpensive down-converters for 1GHz -- but a quick look only turned up things in the $20 class. I'm not sure if they're really not out there on the market, if they're all buried in ASICs, or if my quick search just didn't find the right part.

Tim, Thanks again. I am not an RF engineer (Big surprise eh?) but I am willing to fake it long enough to see if I can pull this off. If I can get the radio to work then I can move to the embedded stuff, more familiar territory.

My ignorance is showing:

I only understand mixing in the AM sense where the frequency is changed to the IF but the envelope, the AM, is preserved. With FM the frequency carries the information, how does the shift remain the same when the mixing occurs? I guess I don't understand the difference between translating and changing.

The RF front end is band-pass filtered by SAW filter with a bandwidth of ~40 MHz. It goes antenna-saw-LNA-saw with matching. Perhaps I would need a IF filter after the mixer....

I have found the RFMD mixer RF2052 with integrated PLL and VCO, it is not quite as terrible at ~$10.

Rich

Bite the bullet and do it right. Get the proper IF frequency and create an I/Q data stream. I doubt that you are going to get a PLL to work at these data rates.

BTW, are you going to get a TSO on this UAT receiver or is it a non- TSO'd device. If it is going to be TSO'd then cheap is not as important as correct.

Coaxial ceramic resonator. No problem.

John

gh

ur

You are a genius

Bwahahahahahaha! Good retort ;-) ...Jim Thompson

You don't know about CCRs? They have Qs above 1000, TCs in the 10 PPM range, accurate CFs, and are cheap.

You could make a nice 1 GHz FM discriminator with a coaxial resonator and a simple diode mixer. A cheap trimpot-type variable cap could trim the cf.

I use them in 500 MHz triggered oscillators, in digital delay generators. They work great.

Google for more.

John

ll

hough

an

ide

ng

your

ld,

He is trying to discriminate the difference between 978MHz + 300 KHz and 978 MHz - 300 KHz and he has 1 usec to get it right. Are you suggesting that this ceramic resonator can do that?

I do not use these parts directly but I use TOKO preselect filters made from these . Four of them stacked together makes a filter with a BW of about 30 MHz

Yup. Make a 978 MHz single-pole bandpass filter with a Q of around

They are remarkable parts. Not a lot of people know about them. Not to be confused with low frequency piezo ceramic resonators.

The common version is a ceramic tube with a hole down the middle, silvered everywhere but one end. It becomes a shorted coaxial cable, with a very high dielectric constant. If you TDR one, it looks just like a shorted coax, typically in the 10 ohm range, but with a lot of round-trip delay compared to a real coax of equal length. Our 500 MHz parts are around an inch long.

Their parallel resonant frequencies go from around 500 MHz up into the low GHz.

John

am

There's a great phasor diagram, that shows AM, FM and I think PM (phase modulation). I know it from Pippard's first volume of "The Physics of Vibration"

I had to sign in to get that so I don't know if you can see it or not. You may have to back up and read some to understand it.

George H.