1. Home
  2. Electronics Design
  3. 4046 PLL model scarcity.

4046 PLL model scarcity.

I Googled to chance my luck finding a model, after not finding one from NXP.

My observations found a few points:

  1. Lots of people asking, not finding.
  2. One poster saying that PLL's were hard to make because of something to do with two time slices or something that I didn't understand at all.
  3. Jim Thompson saying he'd made one. :)

It may be that I shouldn't be spicing this if it is hard to do, because if the HEF4046B's VCO frequency isn't very linear in proportion with input voltage, I'd need to do something beyond the device itself anyway. I understand that the linearity only matters for stability because the output frequency and phase is what matters, but in my case I'll want to use an LM331 as oscilator followed by a flip-flop to get a square wave, or some other means of coercing high linearity out of the PLL voltage output normally fed to pin 9.

If there IS a good spice model of the CMOS HEF4046B, please can someone point me at it, because it will be very useful to me, and it may be that the device is linear enough as it is. I've already put a working pitch tracker together ona pin deck, but a good spice model would be great to help complete a model to help understand what's going on in there.

PS. (Maybe better posted to SED but neater kept here, I think)... If I want to reduce ripple, and create a faster response to an output frequency for any downstream circuit I might try, it seems that divisions in the feedback loop are a neat way to go. Question is: would this cause the lock to a sudden arrival of a new audio frequency to be slower than if I had no divisions in the loop to multiply it for output? Specifically, if it slows more than the response to an un-multiplied frequency would be slowed by a very linear frequency-to-voltage IC like LM331, then multiplying it to speed the LM331 response could be pointless, as circuit complexity rises for no useful result. Clearly this is a good task for Spice to test quickly, IF there was a good 4046 PLL model to do it with. :)

Reply to
Lostgallifreyan
Loading thread data ...

[snip]

In Message-ID:

I said,

"If someone has a real 4046 to measure and can measure all the breakpoint voltages, I can roll a model that will run on most any Spice variant."

No one has come forth with that data.

In Message-ID:

I said,

"The 4046 is actually a CMOS translation and copy, merging my MC4024 VCM and Ron's and my MC4044 PFD, both done originally in TTL (around

1968). The MC4024 was good up to around 30MHz (it was really PECL with a translator to TTL ;-)

Later there were actual PECL releases: MC1658 VCM, and MC12040 PFD.

Then there's also my MC1648 tank-type VCO... no longer made :-( Though I designed an improved replica on a custom ASIC just this last year.

One of my original OpAmps, the MC1530/31, designed in 1963, is still being manufactured (by Lansdale)... made for 47 years now ;-)"

Now...

If you want to model a 4046 I'd recommend making the PFD from 'HC74 plus 'HC00, in the manner as I've posted before.

Make the VCO portion as purely behavioral.

If you have a P.O. I can fit you in to my schedule in about two weeks and model it for you ;-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

What would I need to do, to get that data? I have time, a scope, a Fluke 79 DMM, and soon, a small Chinese signal generator to 5 MHz. And about 50

4046's... But the last time I heard of a break point it was in an FM synthesizer scaling curve.

Well, I'm glad I found the right person to ask, that's for sure. (And a 50 th year for an op-amp will be worth celebrating, I hate early and frequent obsolescense).

I barely know enough spice to drive it, I wouldn't trust myself as a mechanic. I'm not sure what behavioral means here bit of you mean modelled as an ideal, then it won't work for me. I was considering using spice to look at the nonlinearity of the VCO, but for my own purposes, if spicing turns out to be significantly harder than empirical testing, then I'll stay with whatever happens in practise. (I am one of those who Winfield Hill obliges by looking the other way. :)

I have a strange feeling that if I knew what a P.O. was in this case I'd have one, so I probably haven't. Ò^O

Reply to
Lostgallifreyan

Measure frequency versus control voltage. Measure with several different values of external capacitor.

That means you pay. Though I'm in a better mood now and might do it for free. Last week I "synchronized" my laptop and tower (on a chip design) and over-wrote new stuff with old :-( Fortunately I found back-ups of most of the various pieces, and managed to paste it all back together... whew ;-) ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

I might have been able to help worth more than pay with that one, though not being at the machine can be more than awkward. The one thing I can think of for future use is FAT32, I find it it usually gives me a fighting chance of recovery of lost data even after over-writes of files sometimes. If I can find where they were... With NTFS (ahich I don't use) I have heard of horror stories and very little success. The other thing I do is manually backup in any location that I will recall easily (and never all unless I'm totally sure I want to), never use automatic tools, RAID, or anything else that can duplicate an error faster than I can think. I have made some of my own, but at least never suffered a machine doing it for me. :) A hex/disk editor called HxD has been indispensible at times...

About the 4046, I'll get you some values for that voltage vs frequency some time in the week. Any preferences for voltage change interval and capacitor values? My scope is good for 100MHz so I should be ok for any frequency the PLL can do even if the DMM won't cope.

Reply to
Lostgallifreyan

Watch for non-linearities and provide more points in those regions.

...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Hello. I started looking into this, and there may be no point, with the tools I have for it. I got a datasheet for HEF4046B at

formatting link

Unlike the 1% nonlinearity I saw mentioned for an HC4046, this CMOS one I intend to use seems to be about 0.25 nonlinear if I choose C1 and R1 appropriately (and I can live with that), but the Fluke 79 series II meter I have is at best 0.3% accurate, plus one digit, so I'm not sure that any measure I can make will do you more good than that datasheet which has more detail than I saw in others for variants of 4046. If you think the meter, in a room held at fairly constant temperature, will give relative measures good enough for you, I can go ahead and do it anyway.

Reply to
Lostgallifreyan

The metal gate parts are very nice, except for the dead zone in the phase detector.

The HC parts are far more nonlinear--more like 3:1 in slope, if not worse, and with VDD = 5V, they crap out at 1 to 1.3 V--they just stop oscillating. We had a thread about this back in July--see

formatting link

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net
Reply to
Phil Hobbs

Phil Hobbs wrote in news: snipped-for-privacy@electrooptical.net:

Thanks, that's interesting. Can you tell me more about that resistor that helps the dead zone problem (which I also don't understand yet)? Is it loading the output to ground?

As far as I know, using CMOS will be better not least because better performance seems to come from larger voltage ranges in all cases. Also, as I mentioned in answer to Jim Thompson, if the HEF4046B VCO isn't a metal gate type, but according to the datasheet appears to be more linear than the MC14046, then I have some confusion that the datasheet won't bail me out of because it doesn't say anything about metal gates.

The main problem seems to be that datasheets (and Spice) tend to emphasise defined behaviour. If I'm after something most PLL makers don't consider important, I'm S.O.L... but as there ARE soem more linear than others, and linearity of VCO is important to me, I'd like to know anything I can about fixing any deficiencies in those 4046's that do have high linearity, and also what undocumented details of manufacture might make them so much more linear than others, often touted as 'improved' in any way.

PS. Added S.E.D to the group list because this isn't so much about Spice now..

Reply to
Lostgallifreyan

On 10/27/2011 12:22 PM, Lostgallifreyan wrote: > Phil Hobbs wrote in > news: snipped-for-privacy@electrooptical.net: >

Ground and VDD both work, but ground is usually best, because the VCO control voltage is ground-referred. You just need to pull the quiescent operating point of the phase detector a bit off zero to avoid the flat spot in the V(phi) curve.

The flat spot is caused by the finite slew rate of the output--as the phase error approaches zero, the pulse width narrows until it's less than the rise time. At that point the output becomes a triangle instead of a trapezoid, so its area starts going as phi**2 instead of phi. Since the average output current goes as the pulse area, that makes the phase detector gain go to zero at phi=0, which is A Bad Thing.

AFAIK all 18-volt CMOS logic parts are metal gate. That's one of the usual ways to refer to the general class of CD4XXX, 74CXX, and other old high voltage CMOS parts.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net
Reply to
Phil Hobbs

Phil Hobbs wrote in news:IPmdnZqZtOQ4FTTTnZ2dnUVZ snipped-for-privacy@supernews.com:

Indulge my weakness, please spell it out for me... How much resistor, and from exactly where, to ground? If I know that I have a better chance of understanding the rest than deducing it the other way round. The thing about loops is that all of it might be influenced by any point in it! Causality is a tad hard to follow in loops...

Good enough for me. :) Thanks. That avoids the confusion and leaves HEF4046B in a context that makes more sense to me.

Reply to
Lostgallifreyan

It depends on the circuit constants. You want the total charge flowing through the resistor per cycle to equal that contributed by some safe number T nanoseconds' worth of the phase detector pulling HIGH or LOW. T is something like 10 for HCMOS at 5V, maybe 50-200 for metal gate, depending on the supply voltage. A current source is probably better, but that's getting a bit fancy.

If you have a spectrum analyzer, look at the output with and without the resistor--maybe use a pot for testing purposes. The change in the jitter sidebands is pretty dramatic when you pull just outside the deadband.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net
Reply to
Phil Hobbs

Phil Hobbs wrote in news:rZudndP6woK-CDTTnZ2dnUVZ snipped-for-privacy@supernews.com:

Please can you give a hobbyists guide type starting point? I.E. starting resistor value to try, and what pin number on a 4046, to connect to ground? From my perspective, trying to understand the circuit behaviour without that info, or to depend on using a spectrum analyser I haven't got, is entirely putting cart before horse. :)

Reply to
Lostgallifreyan

Provided you run the VCO at audio frequencies then for the purposes of testing a realtime spectrum analyser is built into Daqarta and ISTR there is a 30 day free trial period to see if you like it.

I have no connection with the author other than as an occasional user.

formatting link

Price is also within reach of electronics hobbyists - it uses the sound card to digitise the line input on a PC.

--
Regards,
Martin Brown
Reply to
Martin Brown

At audio I doubt that you'll know it even exists because you loop filter corner will be quite low compared to the device delay time. ...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

Martin Brown wrote in news:Uxhqq.2591$ snipped-for-privacy@newsfe02.iad:

Thanks for the idea, I'll look into that. I deliberately stay with W98 so I might be out of luck but if it runs it might be a keeper.

Reply to
Lostgallifreyan

woK-CDTTnZ2dnUVZ snipped-for-privacy@supernews.com:

ng

,
e

d?

hat

y

Do you have a digital 'scope? They have a so-so spectrum analyzer. As far as phase locked loops, I don't know the 4046. But there are a few different approaches. The AoE is as good a place to start as any. I'm very interested in the use of the 4046 as VCO. At the moment I'm using one to drive a stepped sine thingie. I've been starring at the spec sheet curves wondering if there's a sweet spot somewhere in R/C space. At the moment I've got R=3D20k and C=3D470pF.

George H.

Reply to
George Herold

Sure, if you post your loop filter component values and operating frequency. If you have a scope, you can look at the PD2 output and adjust the resistor so that the positive pulse is always at least 100 ns wide. That'll give you a bit of frequency-dependent phase error, but that's usually not a big worry in a lowish-performance PLL.

What happens is that the loop oscillates with a small amplitude around the null, sort of like the temperature in your house as the thermostat kicks on and off.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs
Principal Consultant
ElectroOptical Innovations LLC
Optics, Electro-optics, Photonics, Analog Electronics

160 North State Road #203
Briarcliff Manor NY 10510
845-480-2058

hobbs at electrooptical dot net
http://electrooptical.net
Reply to
Phil Hobbs

Jim Thompson wrote in news: snipped-for-privacy@4ax.com:

Now that I DID understand right on first sight. :) I have to say it's not something I noticed in my design, so far as I know. One quirk I saw was a lock indicator based on the outputs of both comparators failing to indicate lock when the audible output proved there was a lock. I don't know what that means... I considered it might mean that PC2 was basing output on a harmonic. (I'm using PC1 in the loop, so I was never sure what PC2 would be doing at the same time until I considered using it for part of the lock indicator).

Reply to
Lostgallifreyan

Awwwwh! You mean you're not audiophool enough to fret over a few hundred pico-seconds of jitter ?:-)

Most lock detectors don't do anything useful but give marketers something to brag about.

...Jim Thompson

--
| James E.Thompson, CTO                            |    mens     |
| Analog Innovations, Inc.                         |     et      |
| Analog/Mixed-Signal ASIC's and Discrete Systems  |    manus    |
| Phoenix, Arizona  85048    Skype: Contacts Only  |             |
| Voice:(480)460-2350  Fax: Available upon request |  Brass Rat  |
| E-mail Icon at http://www.analog-innovations.com |    1962     |
             
I love to cook with wine.     Sometimes I even put it in the food.
Reply to
Jim Thompson

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.