PLL as IF filter

I think you haven't specified your scheme thoroughly enough for folk to respond to it.

If really mean "IF filter" then using a PLL won't get there from here.

An IF filter will take signals in an IF that have a wide spectral content and trim the spectral content down to some specified band. It will do so by introducing a fixed amount of attenuation at any given frequency; it will not add any spectral content, or move the spectrum around.

A PLL, on the other hand, generates a tone and locks its frequency and phase to some other tone. If there is no tone to lock to the PLL will still generate a tone, it will just have a more or less random frequency, depending on the PLL design. This does not come close to the requirements of an IF filter.

If you mean "FM demodulator" then you need to get your terminology straight. PLL's can be used for this, although a 74HCT4046 may not be the best choice it could possibly be made to work in a satisfactory manner, depending on your requirements.

Not really looking for a tone here, but more information must be provided.

The modulated signal is transmitted at UHF (400MHz), using ON/OFF (OOK) keying. Since the transmitter is a SAW based colpitt, each unit could come up as far as 600KHz above or below 400MHz.

The baseband modulation type is PWM such that a "1" is 200uS long, a "0" is 400uS long, and the total period is 600uS per bit. 600uS bit lenght is equal to 1666 Hz.

The message is 65 bits long with 50mS of dead time between messages.

As an example,

PLL center would be 9 MHz and lower lock freq would be 8.4 MHz and upper lock freq would be 9.6 MHz.

The Microcontroller would poll the lock pin on the PLL to determine if a signal is to be recovered and possibly use this lock signal to re-create the baseband signal.

The incoming signal at 9 MHz +/- 600KHz would be used as the reference signal.

The output of the locked VCO would also be divided down to a point the microcontroller could determine the frequency of the UHF transmitter. For example a 9 MHz signal divided by 16 would be 562500 Hz where as a

Any thoughts on a good analog PLL that can operate up into this region?

Of course, I know that there are a few wideband 10.7 MHz IF filters out there, but there are crystals on the PCB which have a second harmonic on or near the 10.7MHz.

This is _not_ an application where a PLL is going to shine, certainly not for 200us long pulses. If I had a potential customer approach me with this sort of scheme I'd try to talk them out of it. If they couldn't be talked out of it I'd walk away rather than having a dissatisfied customer be my only working end product.

There are other common IF frequencies -- Digi-Key carries them at

Keep in mind that you have a signal that's only 5-10kHz wide yet because of your oscillator design you need an IF that's 600kHz wide -- this will seriously degrade your noise response.

It depends. If you're using an FM receiver chip that has a limiting amp as its last stage, you can get enough signal to run a 4046 with the XOR phase detector (PD1).

You really can't use the phase-frequency detector (PD2) for this, because it's edge-sensitive and therefore reacts very badly to noise pulses. In either instance, only phase information will be preserved; amplitude changes are either ignored or cause stability and SNR problems.

In rough outline, a PLL will preserve the phase modulation of its input at modulation frequencies well below the loop bandwidth, and ignore phase modulation above there. This is a rough outline because the loop filter usually has a soft rolloff, so that there are significant phase errors below the loop bandwidth and significant response above it.

Wideband IF filters are usually pretty easy to make--it would be the very narrow ones that I'd want to replace with a PLL.

What are you actually trying to do?

Cheers,

Phil Hobbs

If you have to acquire lock from nothing on a single start bit, for example, this problem can be quite hard, depending on your signal-to-noise ratio. I assume your SNR can be quite low, because otherwise you'd just use AM detection in a 1-MHz BW and be done with it.

The easier case is when the TX oscillator is running continuously, transmissions occur all the time, and it's okay to miss a couple at the very beginning while you're acquiring lock. That is a job for a slow PLL, running with a sweeping acquisition aid or a two-speed loop. If you use an analog phase detector with a low offset, the loop won't drift away between tone bursts, so you should be able to acquire lock on the OOK signal.

I'd probably use a VCO feeding a Johnson divide-by-4 counter driving one PD for the PLL (90 degrees phase) and a second one for the synchronous detector/lock indicator function (zero degrees). This is cheap and painless, and puts most of the pickup and other junk at 4f, where it's not such a problem.

You'll need some sort of wideband AGC as well, because you can't just use a limiter for AM signals. Wideband AGC is better than deriving the AGC signal from the 0 degree PD, because it works even when the loop is unlocked, and will prevent the receiver from clipping. Fancy systems sometimes use some combination of the two.

Slow loops following IF filters are sometimes prone to false lock due to IF filter phase shift. This effect is generally weak, but is troublesome for lock acquisition by loop pull-in, which is also weak. A sweep circuit is a good idea.

Cheers,

Phil Hobbs

Yes, I realized that when Phil responded. You could lock onto a strong enough signal with the PLL, then use the generated tone to demodulate the OOK AM signal.

None the less I wonder if, for any system that isn't on continuously, if the noise performance vs. lock time of the PLL would make for a better receiver than just a plain-ol IF filter followed by an amplifier and a diode.

Without knowing _lots_ more about the application it's hard to say.

Yes, he's still going to have his PLL running off of the same noise and interfering signals that he'd have to demodulate. For a PLL to give him any advantage it'd have to be locked, and with any kind of a low signal/noise ration it'll take ages to lock.

So unless the transmitter's on all the time, he basically needs something that emulates a superregen with a wide-open IF, and don't expect long range performance.

schreef in bericht news: snipped-for-privacy@g14g2000cwa.googlegroups.com...

Paul,

The 74HCT4046 requires logic input signals. So you can't use analog signals unless you make them "digital" by a comparator for instance.

petrus bitbyter

Hello Tim,

It depends on how often Paul sends the signal. The challenge will be the first "catch", identifying that the picked up signal is the right one and not somebody else's. After that, with a regular data rate, he could have a PLL lock onto it. But I would use a hybrid approach. Heterodyne down to another lower IF where cheap filters with reasonable performance can be had. The oscillator can be PLL, controlled by a nifty uC or something to handle the detect and locking.

Regards, Joerg

Hello Tim,

The plain old filter wouldn't work in Paul's case because the oscillator in his transmitter seems to be as runny as maple syrup.

Regards, Joerg

Hello Tim,

Maybe it isn't so bad in Paul's case since he said he has a SAW oscillator with a large tolerance and some drift. If his circuit could clearly discern the signature of his signal from anything else that's within his deviation range he could lock a PLL onto his signal. Basically just an AFC. Pretty much like finding a good station on an old FM radio. Hand turns the dial, brain says "that's my kind of tune" and the AFC then keeps it there. In Paul's case the hold time of the AFC needs to be long enough and the LO stability good enough to bridge the pauses.

Regards, Joerg

Hi there all,

Thanks for taking the time to add the two cents worth. I see that Maxim makes a 16 bit ADC that could run up to 150MHz. I wonder if a direct conversion for the 10.7MHz could be DSPee such that digital filtering could apply

Back to more information,

On the baseband signal, the message has a preamble with is a square wave signal of a series of 8 "O". This would be eight symbols were the on-time is equal to the off time, and that time would be total of 600uS where 300uS of time is high and the remaining 300uS is low.

The preamble is basically a signal to help get the ACG level on receiver adjusted before the data appears.

OK we have the preamble,followed by 2mS of no transmission and then the start of first bit and ends at the final bit some 64 bits later.

The signal will appear again in 50mS if the user continues to press and hold a control button. . Is there a mixer circuit out there that would take the radio's 10.7MHz IF down to 455KHz?

Given that my transmission frequency could be 400MHz plus/minus 600KHz, a translation down from 10.7MHz to 455KHz would cause a transmitter at the minus side to be lost?

For the plus side, the 455KHz added to 600KHz would be 1055KHz, so what I would require a "bandpass from 455KHz to 1055KHz or more and also require that any transmitters on the low side of 400 MHz be retuned to a frequency on the plus side of 400MHz.

I could translate the 10.7MHz IF "range" down to 455KHz by mixing the IF signal with 11.155000 MHz crystal (NE603?). Where the desired range of 455KHz to 1055KHz could be easily done with a op-amp filter.

At these lower frequencies a easier PLL implimentation.

What you think group?

Just some thoughts.

thanks for the interesting discussion.

Not at the phase comparator input. The input is self-biasing and I've succesfully fed signals as small as a few tens of mV via a coupling capactor. I think the spec depends on the brand but it's way less than

Very little about the 74HCT4046 is digital. IIRC, only a single input, "Inhibit", requires real logic levels.

Tim.

Hello Paul,

A direct conversion makes you susceptible to the other sideband. It's tough to create a steep enough pre-filter at 400MHz. Why not use TV IF filters, IIRC in the 30-45MHz range? Since these TV filters are about

Then downconvert to 10.7MHz, then to 455kHz and use a uC or DSP or whatever you have on board to "fish" for the correct signal. If you set up a good tracking filter at the TV IF filter range you can single convert but another conversion really doesn't cost much, it's easier.

This preamble needs to be unique enough and long enough to use for fishing. Besides your signal there can be lots of others in that

That sound feasible. 50msec isn't likely long enough for your LO to drift away too much. But it may take some time for your circuitry to find that brief 2msec burst and lock onto its frequency. Not quite the needle in the haystack but close.

I've always done them discrete but then again I am a cheapskate, especially if it's design for clients where cost needs to be low. You could also use a 1496 mixer chip or any other, maybe the NE series. Haven't used Philips lately so I am not so familiar with NE anymore and what's available or what it costs.

You really need to scan (fish) for your signal, using some intelligence on the receiver. A uC, a DSP or a PC if one happens to be conneceted anyway.

I am afraid an ordinary PLL won't catch your signal. You need a scheme that finds your preamble and locks onto that, not just onto any carrier that happens to be there.

Regards, Joerg

Hello Paul,

Just an add-on: If this application isn't critical in terms of cost and power consumption you could look into mixers from Mini Circuits. They are a few Dollars a pop. With a little care they can be driven by logic chips and all the filters in between the conversion stages can be conveniently calculated at 50ohms.

Regards, Joerg