Who has used the 4046 to synchronize their nipkow wheels?

[ac7zl]

If you've read some of my other posts, you have no doubt gathered that I am tinkering with NBTV circuitry in the virtual (SPICE) world. I have modeled and developed various circuits for video, gamma, dc restoration and sync extraction. The one thing that has me stumped is synchronization.

I have worked out interesting circuits using D flipflops and XOR gates to detect phase differences, and I've modeled the Nipkow disk and its sync holes with a voltage controlled pulse generator. Standard spice voltage sources are used to generate sync reference pulse trains.

If the ref and wheel pulse trains are devoid of the (missing) frame pulse, the circuitry seems to work fine. In fact, I can go from wheel at dead stop, spin it up, and synchronize perfectly with the reference pulses, and in fairly short order.

The problem arises when pulses are dropped (as per the NBTV standard) to denote start of frame. My synchronizer hunts around and never achieves lock.

Thinking that there might be something inherently wrong with my phase detector, I employed a SPICE model of the 4046. Again, where pulse trains are continuous (i.e., no missing frame pulse) the closed-loop control works like a charm. If I use signal sources that feature missing frame pulses, it doesn't work right.

I wonder if anyone could give me feedback with regard to "standard" NBTV 4046 circuit that is in circulation.

1) Can you verify that it even works?
2) How well does it work? How stable is the picture and how quickly does it reach stability?
3) How do you effect frame lock?
4) Have you had any problems with it or have you noticed any irregularities?

Many puzzled thanks,

Pete
AC7ZL

[Steve Anderson]

...The problem arises when pulses are dropped (as per the NBTV standard) to denote start of frame. My synchronizer hunts around and never achieves lock.

1) Can you verify that it even works?
2) How well does it work? How stable is the picture and how quickly does it reach stability?
3) How do you effect frame lock?
4) Have you had any problems with it or have you noticed any irregularities?

Pete
AC7ZL

The implication to draw from the above is "That missing Pulse" is the culprit. However, one can only but assume that it does work after a fashion as many have built the circuit and used it successfully. (I haven't, built it that is).

The PPL will not like the sudden temporary drop in pulse rate from 400Hz down to 200Hz. To a large degree it will depend on the loop-filter components hanging on the output of the 4046. In addition there are mechanical equivalents to a filter comprising of the mechanical inertia of the system and friction to think of just two. This is of course different for each built example.

The motor/disc/opto-sensor are the replacement for the unused VCO in the chip. More volts=more frequency. It might not be that linear, but here it doesn't have to be. We're only interested in one frequency, plus/minus a bit.

Keep in mind that the original version of the 4046 (HEF4046/CD14046, the 3-15V versions) have a chronic sensitivity to supply volts in the VCO, it's a good thing it's not being used here. The give-away is the inclusion of a Zener diode within the package, pin 15, so you could locally 'regulate' the VCO supply volts. Of course the manufacturers omitted this fact in their datasheets, but can be seen in the plots for the VCO.

The HC(T) versions (3-6V only) are far better in this respect such that the Zener has gone and pin 15 has a more useful function as the output of a third phase comparator. The fact that the supply is usually a well regulated 5V one helps too.

This missing pulse seems to be a perennial problem and should be easily fixed. I'm quite willing to spend some time on this but it will have to wait a few days as my bench is covered in the GPS project I'm currently working on.

A word to the wise...don't rely on simulators too much, they can trap the unwary, including myself on several occasions. They can have problems with 'circular references', similar to Excel when you create a circle between cells. Oscillators and PPLs fall into that catorgory, anything that to it appears to have positive feedback. They are getting better with each release, but still are not perfect.

Steve A.

[Viewmaster]

Interestingly, if you do a search for '4046' on this forum you will find many references to problems with sync using this pll. Some haven't achieved it and with others the picture bounces.

I sometimes wonder if more attention should be given to the sync holes in discs etc. Maybe they should be deburred more etc so that the output is constant?
As a unburred hole passes the opto fork (especially reflective steel/ali) there maybe unwanted reflections from the various shaped burrs which alter slightly the resultant output, thereby 'confusing' the 4046.....just a thought.

Certainly, the 4046 seems to cause more headaches for NBTV builders than most other items in the build.
Albert.

[Steve Anderson]

Interestingly, if you do a search for '4046' on this forum you will find many references to problems with sync using this pll. ...Certainly, the 4046 seems to cause more headaches for NBTV builders than most other items in the build. Albert.

Yes, very true. To many PPLs are somewhat of a black art. The maths behind them is not for the faint-hearted, but properly designed are no less reliable than an op-amp or a flip-flop.

The problems that I see NBTVers have breaks down into two areas. Firstly as Pete mentioned previously, the missing pulse causes some headaches, secondly the mechanical time-constants which vary from one build to another.

And in fact, thirdly, the stability of the source pulses that the poor thing is trying to lock to, poor disc-drilling of the source being an example. Given all those variables it's no wonder the thing has a hard time sorting itself out, it is after all just a couple of flip-flops and some gating logic.

It's a credit to it that it has been as successful as it has been, teething problems and all.

Steve A.

More to come on this.

[Panrock]

As is well known, the in-depth electronics understanding of all this is well above my head. However the circuit works for me, and works well.

Moreover I am using not one missing pulse but a bunch of four. This helps to ensure my heavy wheels synchronise not only in line lock, but automatically take up frame phase too.

A couple of my old friends the 555 clean up and shape the triggering pulses before they are presented to the 4046. There is a more elegant way to do this I know, but it's legacy circuitry on my board so I might as well use it.

Klaas has suggested a method to remove hunting by adjusting the proportion of the two resistors off Pin 13. I haven't found this necessary in my application, though I do find a power resistor in series with the motor - to 'slug' the action - beneficial.

Steve O

[Steve Anderson]

Moreover I am using not one missing pulse but a bunch of four. This helps to ensure my heavy wheels synchronise not only in line lock, but automatically take up frame phase too. Steve O

Conceptually that's going in the direction of my 'one reference per frame' idea I did for the 48-line display. And indeed a proposal/idea raised by Pete a few days ago.

With the mechanical inertia of a disc/drum rotating at 750 RPM are line syncs required? Here I have to take the advice of the mechanical guys.

What is the variation of a typical disc/drum speed per rotation without speed control, just set for a nominal 750 RPM? For the brushless DC motor I got around ±20µs in 10ms, around 0.2%, at least as I recall.

That 0.2% is short-term, long term it never dropped even part of a line or frame. In 24hrs the input counter showed over a million rotations, the output counter was exactly the same.

Steve A.

[Panrock]

With the mechanical inertia of a disc/drum rotating at 750 RPM are line syncs required? Here I have to take the advice of the mechanical guys.

Well I'm using the same circuit, with only one pulse per frame, on one of my stereo shutter discs. The locking action is fine, if a bit violent. Again, I'm using 555s here for pulse shaping.

The other (camera) shutter uses two pulses per rotation (slots in the rim of the disc) to divide the camera frame rate by two, for left-right, left-right...etc. operation.

However the stereo frame shutters don't require the same precision as a line lock. I would think hunting - or rather 'slop' - would likely be a problem in that application.

Steve O

[Steve Anderson]

Gents,

Here's an interim report on the good old (and it is old) 4046.

As suspected it doesn't like the missing sync pulse at all, although not 100% upset in the sense that it maintains lock, it's not very hygenic.

I imagine that the inertia of the disc and motor iron out a lot of these irregularities.

Now although the 'regenerated' pulses are not syncronous at first, their overall frequency is still the same at 400Hz, they make up for lost time later in the frame. So the PPL is doing it's job, it's maintaining lock, it's just not phase accurate.

Now gents, I had to go out and actually buy some HEF4046s, In February I threw a few thousand LSTTL and HEF-style CMOS chips away. I couldn't even give them away, no-one wanted them. I hadn't used them in a decade and a half, they were just taking up room. In amongst them was of course the HEF4046! Sod's law.

Steve A.

[ac7zl]

You are correct in your cautions regarding simulators. First, simulation is, at best, an approximation of reality. Then, the simulation is only as good as the models used. Your point about circular referencing is a good one.

As I see it, there are two major failure modes of simulations. Either the modeled circuit "works" in simulation when it doesn't (or physically cannot!) in the real world. Or, the modeled circuit does not function in simulation, but does in the real world.

My experience is that the first is the more common fault. You can easily see an example of this by connecting a 1000 V voltage source to a little 1n4148 diode. The simulation shows 1700+ amps going through the diode! Yeah right! (The simulator does not show the flash of light, the puff of smoke, or the black stain left after the diode vanishes from this dimension.)

When Spice says a circuit won't work, on the other hand, my experience is that reality tends to agree. Granted, I've seen electronic parts used in unorthodox ways to exploit some obscure electrical parameter for the benefit of the circuit. If the component model does not model the obscure feature, then a simulation will likely fail. For the most part, however, I trust simulators when they say "no-go."

Pete
AC7ZL

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

A word to the wise...don't rely on simulators too much, they can trap the unwary, including myself on several occasions. They can have problems with 'circular references', similar to Excel when you create a circle between cells. Oscillators and PPLs fall into that catorgory, anything that to it appears to have positive feedback. They are getting better with each release, but still are not perfect.

[ac7zl]

I think signals from sync holes should be passed through buffers (schmidt trigger) with heavy hysteresis. It might also be of benefit to feed both the wheel feedback and reference sync signals through flip flops. This converts both into square wave streams with a 50% duty cycle. It seems to me that it should be easier, then, to make meaningful phase comparisons.

Pete

Interestingly, if you do a search for '4046' on this forum you will find many references to problems with sync using this pll. Some haven't achieved it and with others the picture bounces.

I sometimes wonder if more attention should be given to the sync holes in discs etc. Maybe they should be deburred more etc so that the output is constant?
As a unburred hole passes the opto fork (especially reflective steel/ali) there maybe unwanted reflections from the various shaped burrs which alter slightly the resultant output, thereby 'confusing' the 4046.....just a thought.

Certainly, the 4046 seems to cause more headaches for NBTV builders than most other items in the build.
Albert.

[Steve Anderson]

I think signals from sync holes should be passed through buffers (schmidt trigger) with heavy hysteresis. It might also be of benefit to feed both the wheel feedback and reference sync signals through flip flops. This converts both into square wave streams with a 50% duty cycle. It seems to me that it should be easier, then, to make meaningful phase comparisons.

The idea of passing the signal through some form of 'clean-up' operation is wise. Below (Motor 1.gif) is the 'classic' implimetation of the motor control circuit as I understand it. It's my reference at the moment.

Motor 2.gif is a version that Klaas has developed that rightly adds some hysteresis to the opto-detector output.

PLL 2.gif is how I have done it using a LM311, but basically it's the same idea. It needs a a minimum of 600mV centred around 6V to get any output into the 4046. The edges are sharp (10ns or so) and don't bounce. The rest of the circuit to the right is unchanged.

On the subject of duty-cycle, as we're using Comparator II it doesn't care at all about the wave shape, it's activated only by the rising (positive going) input pulse edges. It couldn't care less about the following waveform until another positive-going edge comes along.

This is not true of PCI which does require a 50:50 duty-cycle input. But we're not using it, so no problem.

I also added a resistor in the emitter of the photo-transistor to stop over-dissipation should the 47k preset be turned to zero.

Steve A.

P.S. PLL 2.gif is very interim, hence Rxxx and so on....

[Steve Anderson]

You are correct in your cautions regarding simulators.... You can easily see an example of this by connecting a 1000 V voltage source to a little 1n4148 diode. The simulation shows 1700+ amps going through the diode! Yeah right! (The simulator does not show the flash of light, the puff of smoke, or the black stain left after the diode vanishes from this dimension.) Pete AC7ZL

I must try that! It could be quite a giggle! 1000V, 1700A, that's only 1.7 Mega Watts! Poor little 1N4148, what a way to go.

A real 'blaze of glory', even if it's for just a few microseconds!

Steve A.

[gary]

You are correct in your cautions regarding simulators.... You can easily see an example of this by connecting a 1000 V voltage source to a little 1n4148 diode. The simulation shows 1700+ amps going through the diode! Yeah right! (The simulator does not show the flash of light, the puff of smoke, or the black stain left after the diode vanishes from this dimension.) Pete AC7ZL

I must try that! It could be quite a giggle! 1000V, 1700A, that's only 1.7 Mega Watts! Poor little 1N4148, what a way to go.

A real 'blaze of glory', even if it's for just a few microseconds!

Steve A.

...and totally in the spirit of Bairds experiment in making diamonds...

(esp. if you embed the diode in a bucket of concrete first!)

[ac7zl]

How do you guys effect frame sync with this circuit? It seems to me that the best this will do is line sync.

Do you have to adjust the wheel manually?

Pete
AC7ZL

I think signals from sync holes should be passed through buffers (schmidt trigger) with heavy hysteresis. It might also be of benefit to feed both the wheel feedback and reference sync signals through flip flops. This converts both into square wave streams with a 50% duty cycle. It seems to me that it should be easier, then, to make meaningful phase comparisons.

The idea of passing the signal through some form of 'clean-up' operation is wise. Below (Motor 1.gif) is the 'classic' implimetation of the motor control circuit as I understand it. It's my reference at the moment.

Motor 2.gif is a version that Klaas has developed that rightly adds some hysteresis to the opto-detector output.

PLL 2.gif is how I have done it using a LM311, but basically it's the same idea. It needs a a minimum of 600mV centred around 6V to get any output into the 4046. The edges are sharp (10ns or so) and don't bounce. The rest of the circuit to the right is unchanged.

On the subject of duty-cycle, as we're using Comparator II it doesn't care at all about the wave shape, it's activated only by the rising (positive going) input pulse edges. It couldn't care less about the following waveform until another positive-going edge comes along.

This is not true of PCI which does require a 50:50 duty-cycle input. But we're not using it, so no problem.

I also added a resistor in the emitter of the photo-transistor to stop over-dissipation should the 47k preset be turned to zero.

Steve A.

P.S. PLL 2.gif is very interim, hence Rxxx and so on....

[Steve Anderson]

How do you guys effect frame sync with this circuit? It seems to me that the best this will do is line sync. Pete AC7ZL

Quite correct, it doesn't provide any means of frame sync. Which is why I have dumped the whole concept of sync pulses and the like in recent times.

To get a 'sort-of' frame reference one needs to retrigger a monostable such that it will give you a pulse each time the sync is missing.

The current 'nominal' standards call for a line sync that is approximately 5% of line-time, say 125µs. In the 48-line system I posted recently I used one line for frame and line sync as well as real black-level clamping, just over 2%. I didn't use the full line-time, but it was there for the taking.

Steve A.