Linearizing VCO transfer function

IMHO you don't want to design a PLL that's that twitchy. Break it into two loops with separate VCOs if you can.

Failing that, you can design for worst-case phase margin. You may have to deal with increased lock times, but it should definitely be possible to make the loop stable everywhere.

ADISimPLL is a great tool for this stuff, if you haven't tried it yet.

-- jm

------------------------------------------------------

------------------------------------------------------

Joerg is just addressing the TC of the break-point system. I was commenting that the tuning curve may have a TC as well (varicap, for instance).

...Jim Thompson

It's not "twitchy", it's just the variation in loop gain that results from a wide tuning range using a non-linear element, such as with a varicap.

True, Just design for critical or slightly over-damped, at the worst-case... than all other cases stay over-damped (and slow closing).

...Jim Thompson

You can sometimes juggle other capacitors in the resonator to linearize the VCO -- caps in parallel with the diode tend to reduce its effect at the high frequency end, caps in series tend to reduce its effect at the low frequency end. Of course, all this effect reduction is exactly what you don't want for a really wide range.

You may also want to consider designing your loop with enough gain margin so that it's stable throughout the range. This will impact your phase noise, settling and tracking performance, so you may not be able to get there from here, but it's worth a look.

Don't worry about it. The loop response varies as the square root of the vco gain, so it is not really critical. You can optimize the response at one point and let it vary above and below this point. As long as you are not trying to operate at very high reverse bias on the varicap, you usually can get quite acceptable response over the entire range.

If your tuning ratio is too high and you are using a regular abrupt junction varicap, try going to a hyperabrupt. If you are already using one, figure out how to reduce the required tuning range.

Mike Monett

many new PLL chips have a programmable gain for the phase detector as well as for the dividers. You can you this feature to change the gain of the loop as required when you change frequenices.

Mark

Hello Jim,

Run another diode in a servo loop, a diode that is located close to the others. Best to use doubles with a common cathode like the BAV70 where the servo diode would be in the same SOT23 package. Has worked well for me.

Regards, Joerg

Hi, I am designing a PLL and have found that the gain of the VCO is very non-linear, and causes all my carefull calculations to fall over and the loop to become unstable at the lower end. I read in Wez Hawards "Radio Frequency Design" that quite often designer uses diode resistor networks to piecewise linearize the gain of a VCO. Does anyone have experience of this method or have any circuits I may look at for clues.

Best regards, John.

yes a simple way to give some level of improvement is to have a voltage divider with the botom end conected to a reference voltage through a diode. this reduces the maximum gain when the control voltage reaches the ref voltage + diode drop and the diode conducts. the reference voltage can be acheived with voltage divider too and the efective resistance of this can function as the botom resitor in the first divider.

of course if your maximum gain is at the low voltage end you can turn the diode around. you would need to carefuly analyze your vco charecterisitc to get it right. this efectivly counteracts a single kink in the characteristic, if this isnt enough you can have two such dividers but this is best done using an op amp as well, but might be more trouble than its worth as temp efects of the diodes means you wil stil have to give a reasonable margin for loop stability.

Colin.

Hello John,

I can't send you one of the actual designs I did because it is confidential information. But there is pretty good educational material on the web about this topic, for example this one from Ireland:

As to Jim's point, of course this does not take care of other non-linearities such as that of the varicap etc. They can be compensated for as well but this gets a bit more intricate. One could, for example, run a separate oscillator with another varicap of same type and measure its frequency with a uC, possibly divided down. But it may be tough to find a dual varicap in the same package. Also, the loop could become kind of slow.

Regards, Joerg

Hi Joerg, This sound simple and efficient, can you send me an example schematic, so I can get my head round it.

Thanks, John.

Hello John,

Come to think of it, if you just want to linearize the transfer function somewhat to make you loop calculation easier, or your loop faster, you probably don't need to tempcomp a piecewise linear control even with diodes. At least not as long as you have a healthy voltage swing in the diode section so your control doesn't linger in any transfer area for too much of the total range.

Regards, Joerg

Back to back varicaps in a culprits oscillator configuration wide band and very nearly linear. One of my first patented design at Motorola many if times ago... public domain now! enjoy!

Shame on you, top poster! Anyway, Marc, very nice, how does that work, I thought the back-back connection merely lowered distortion? Also, can you tell us the patent number?