Construction Diary -- Part 2, Spinning the Nipkow Disc

Original build of a televisor by a complete novice.

Moderators: Dave Moll, Andrew Davie, Steve Anderson

Postby gary » Mon May 21, 2007 2:36 pm

Andrew Davie wrote:Finally, I removed the 'frame' stickytape -- the one hole in the disc that was covered to indicate start of frame to the circuits. And... whoa... now my motor control synchronisation is ROCK SOLID. It was that missing 'pulse' -- deliberately masked out -- that was throwing the circuit for a loop. I always wondered how/why the circuit was expected to behave itself when it wasn't really getting 32 pulses/rotation, yet it was expected to synch to 32 pulses/frame. Or is it?


An analysis of this circuit by Garth Porter in Newsletter Vol 32 No. 2 makes very interesting reading. Especially as he appears to say that the arrangement that now works for you - shouldn't!, that is unless your synch pulse input also does not have the missing pulse, in other words 31/31 and 32/32 work 31/32 and 32/31 do not. Curiouser and curiouser.
gary
 

Postby Andrew Davie » Mon May 21, 2007 5:17 pm

gary wrote:
Andrew Davie wrote:Finally, I removed the 'frame' stickytape -- the one hole in the disc that was covered to indicate start of frame to the circuits. And... whoa... now my motor control synchronisation is ROCK SOLID. It was that missing 'pulse' -- deliberately masked out -- that was throwing the circuit for a loop. I always wondered how/why the circuit was expected to behave itself when it wasn't really getting 32 pulses/rotation, yet it was expected to synch to 32 pulses/frame. Or is it?


An analysis of this circuit by Garth Porter in Newsletter Vol 32 No. 2 makes very interesting reading. Especially as he appears to say that the arrangement that now works for you - shouldn't!, that is unless your synch pulse input also does not have the missing pulse, in other words 31/31 and 32/32 work 31/32 and 32/31 do not. Curiouser and curiouser.


Just to be really clear on this, when I say "lock" I mean that the frame is consistent and not moving up or down. It is not, however, positioned correctly with scanline 1 at right, scanline 32 at left. It is arbitrary. But even when I was getting "lock" with the single hole masked out... this was the same. I never achieved a good consistent lock with scanline 1 "about" where I'd expect it to be.
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

Postby gary » Mon May 21, 2007 6:23 pm

Andrew Davie wrote:Just to be really clear on this, when I say "lock" I mean that the frame is consistent and not moving up or down. It is not, however, positioned correctly with scanline 1 at right, scanline 32 at left. It is arbitrary. But even when I was getting "lock" with the single hole masked out... this was the same. I never achieved a good consistent lock with scanline 1 "about" where I'd expect it to be.


No, that was understood, this circuit is not designed to frame lock (this is still somewhat of a holy grail). The article was referring to a steady line lock which is what you now have. Due to the fact that there must be a sync pulse for every disk pulse, having one at 32 and one at 31 will cause the speed to change every time a pulse coincides with a missing pulse (if both have a missing pulse then there is no speed change).

However, if there are missing pulses in both signal and disk, Garth appears to have shown that there must be frame lock before line lock can occur so that the missing pulses coincide, and this apparently can take quite some time to occur. Also, because of the long time constant of the motor it seems that when this does occur the frame lock is one line out.

Since most NBTV 32 line video has the missing sync pulse it is hard to see why your system is line locking when you unblock the 32nd disk hole, you would expect that the coincidence of the missing sync pulse and the non-missing disk sync pulse would cause the motor speed to change. Very odd, something seems to be wrong somewhere.
gary
 

Postby Andrew Davie » Mon May 21, 2007 10:11 pm

gary wrote:No, that was understood, this circuit is not designed to frame lock (this is still somewhat of a holy grail).


I spent an hour or so tonight experimenting and analysing the behaviour of the circuit, with various placement of the masked hole, and adjustments to the speed control. I did not play with the dampening setting -- I should probably do that next.

What I did find, much to my surprise, is that there are settings where I get line lock AND frame lock, with the one proviso. That is, occasionally when attempting to lock the circuit will go haywire and the Nipkow disk will accelerate and never come back down to speed. If I manually slow the disk and let it try again (especially if I do this when it's near frame lock) then it tends to lock line AND frame OK.

If I turn down the speed control then I tend to get the frame 'pulling' effect that I mentioned before. Basically, not a stable frame, and quite annoying. So I up the speed control until I get a stable frame -- and if I leave things, all is OK. However, when restarting the track and it attempts to lock again, then it tends to get to the "too fast, always too fast" stage. Again, I slow it down, let it lock -- and I have perfect frame AND line lock. Mind you, this only works with one particular position of the masked hole on the Nipkow disk. I can't explain why -- but in any case, it's probably because I hadn't had the exact same settings on things for other holes.

So really the problem I have now is that occasionally the 'kick' that the motor is given by the synch circuit is so great that the motor kicks past the next synch hole, and then it thinks it's too slow, goes faster, etc., etc. That's what I surmise is happening. So my guess is that my gearing ratio on the motor is too high -- there should be a smaller 'hub' on the motor, and a larger one on the Nipkow disk. The other guess I have is that the Nipkow disk is too responsive, and too easy to accelerate. Were it a more massive disk -- and I've been thinking surely one wants a massive disk in the first place because that would tend to maintain a constant speed with greater ease -- then it would also be less likely to be 'kicked' by the motor.

I think the above could explain why I was getting good frame lock with no masked hole -- the motor is kicking the disk so much that it gets kicked back into lock straight away, if that makes sense.

So the next thing I think I should try is another motor, or a smaller wheel/hub combination. I think it's very very close to working properly now -- and as I said, I'm particularly pleased with the fact that it currently frame AND line locks when it does lock at all.

I should note that during my experimentation with this I have seen all sorts of interesting behaviour -- there are definite 'modes' that the circuit/system gets into, for example rapid on/off pulsing of the motor (which is audible and partially visible, as I placed a mark on the motor's wheel) -- producing in this case an excellent lock, but very annoying sound. Also, there's a 'half frame' lock, where the picture is vertically displaced downwards, again with the noisy on/off pulsing -- one would slow the disk down and give it another chance, and it would then tend to lock OK (line only).

But as I said, now I have an almost working system. If I can solve the problem of it kicking too fast and going out of control, then what I have will be sheer perfection ;)
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

Postby Andrew Davie » Tue May 22, 2007 12:41 am

This is very interesting... without changing the settings of the motor speed controller, the system works nearly perfectly for a video signal 25% over-speed. That is, it exhibits locked frame AND line synchronisation from scratch, every time I tried it (which was about half a dozen -- it's late :) and the only bad thing was a very minor "pulling" as noted at the normal frame rate when the motor speed control was set too low. So I know there's some relationship of the timing of the signal and the motor/disc rotation that CAN cause a near-perfect system.

I must say, although this part of the system has caused me the most grief, it's been very interesting. I'm starting to think about writing down a sort of finite-state-system to describe what I'd expect to happen in terms of an incoming IR signal compared to a 'state' driven by a set video signal, and see how that compares to the PLL chip's operation.

I might note, one aspect of the current circuit that I don't understand is why, when there is no video signal, and so no pulses, it insists on slowly rotating the disk anyway.
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

Postby Klaas Robers » Tue May 22, 2007 9:40 pm

Andrew, I think that the problem is caused by the fact that the motor can be speeded up by the servo control, but cannot be slowed down. Slowing down is done by letting the disc run out, which goes much slower than speeding up.

I noticed the same with the cruise control of a previous car. When switching it on in a low gearing the system started oscillating. When the car went too slow, the motor gave power and the car accelerated to above the wanted speed. Then the throttel was closed and the car slowed down much slower.

In your monitor when the speed is already too fast, the system accellerates and decellerates both 50% of the time. But as accelleration is much more effective, this action wins always.

Yes I know, this is not giving you a solution.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Andrew Davie » Tue May 22, 2007 10:37 pm

Klaas Robers wrote:Andrew, I think that the problem is caused by the fact that the motor can be speeded up by the servo control, but cannot be slowed down. Slowing down is done by letting the disc run out, which goes much slower than speeding up.


I did some rough timings tonight -- it takes 50+ seconds for my disk to stop, once I remove the input signal. I wonder how this compares to others?
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

Postby AncientBrit » Tue May 22, 2007 10:40 pm

A more sophisticated motor control unit would provide power to speed the disc up, and power in the opposite sense to effect dynamic braking.

In a simple form imagine the motor being fed from a centre tap on the 12v supply, ie +6v.

The free end of the motor may go to 0v, speeding the disc up or it may be returned to +12v to provide braking in the opposite sense.

The control could be a form of pulse width modulation, where the mark-space ratio gives a proportional control.

Regards,

GL
AncientBrit
Green padded cells are quite homely.
 
Posts: 858
Joined: Mon Mar 26, 2007 10:15 pm
Location: Billericay, UK

Postby Klaas Robers » Wed May 23, 2007 5:45 am

I think there is a different way, i.e. limiting the current that the motor can draw:
- Suppose the current that the motor draws at correct speed is 300mA
- Then the minimal current is 0 mA
- Now make the maximum current 600mA.

In this case the acceleration is equal to the decelleration.

Wouldn't this solve the problem? Or isn't it enough?
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Andrew Davie » Wed May 23, 2007 11:29 pm

Today my new lot of 100 ultrabright LEDs arrived. I've been playing with packing them into efficient patterns onto a perfboard. It's an interesting problem, actually -- achieving the highest density whilst still using the square-grid hole placements. Of course the best possible would be an insect-eye-like hexagonal packing and this is almost achievable, given a bit of alternation of the orientation of LED pins. Almost, but it is a bit squashed and so the LEDs don't sit evenly.

What I'm going for is a square grid with all pins facing the same way, with enough room in the 'middle' of each square for a LED with pins at right angles to the others. I'm placing these in eight rows of 5, giving me 40 LEDs in total. It should be wickedly bright.

Each row will require 25mA, and the total matrix will therefore use 200mA. Ignoring, for now, the diode's max 100mA capability -- I would assume that this means my array will be half-brightness. With 200mA, and the array strings using 5 x 3.5V = 17.5V, then the resistor for the whole array will be (21V (supply) - 17.5V (matrix)) / 0.2A = 17.5 ohms.

Now assume that I put some very small balancing resistors in series with each of those 8 strings of LEDs (say, 5 ohms or so)... I assume that the resistance calculation for each is not the parallel calculation (which is the resistance for the whole matrix) but instead just the balancing resistor -- because each string is 'resisted' by only the resistor in series with it. Then I would assume that this would leave 17.5 - 5 = 12.5 ohms resistance required.

But I have the resistance of the gamma circuit -- which is using the original resistor values. And here things get interesting. From the handbook for the LED driver board, I see that the resistors are calculated at 3R, 1.6R and 1.3R for the gamma circuit. The resistors are also given as 150, 82 and 56 ohms. So R in this case is about 50.

Using the parallel formula for resistance, R = 1/ ((1/150) + (1/82) + (1/56)), so the total resistance of the gamma circuit would be 27 ohms(approximately). Now this is confusing for me, because I would have expected this to come to something like 50 given the resistor values chosen and the formula given. I'm missing something.

But, let's plough on. The calculated value for resistance of the gamma circuit is 27 ohms. Given that I'm using 5 ohms or thereabouts for balancing resistors, and I calculated I'd only need 17.5 ohms for the total resistance... I'm not going to see as much current through the LEDs as I would like.

(21-17.5)/? = 27 ohms
? = 3.5/27
= 0.13A (roughly)

This gives me just 0.13/8mA per string... 16mA or thereabouts.
That's not good enough -- I want BRIGHT!!!

OK, so I'm going to have to reduce the current required. I'll reduce it to 7 strings of 5. Now the total current is 7 x 25 = 175mA. Calculating the resistance required to safely drive the matrix at 21V -- (21-17.5)/0.175 = R --> 20 ohms. Take out the 27 ohms from the gamma circuit, we're still short. So let's look at 6 strings of 5... that comes out at 23mA. Close enough (10%).

Perhaps, though, I could increase the number of LEDs per string. Given 3.0-3.8V rating (I assume 3.5) and I had 6 per string the max would be 6 x 3.5 = 21V which is the same as my supply. I'm not comfortable with this, but it does 'sort of' fit. Since there's no voltage left over, no resistors would be required, so the gamma circuit would just be draining the current/brightness achievable.

So strings of 6 are out. Looks to me that 6 strings of 5 are perhaps the best I'm going to get, and I'll just need a very very small balancing resistor across each string.

6 strings of 5. Each LED at 3.5V requiring 25mA. Each string of 5 has 5 x 3.5V = 17.5V. That leaves 21V-17.5V = 3.5V to calculate the resistance required. Each string uses 25mA, so 6 require 150mA. So the resistance is 3.5V/0.150A = 23 ohms. The gamma circuit already calculated at 27 ohms (or 50, which is it?) will adequately limit the current already -- so we put (say) a 1 ohm resistor in series with each string.

If I build it and find it's not bright enough, I can disable one of the strings, giving me 5 strings of 5 -- so the current total would be 125mA, so 3.5/.125 = R --> 28 -- JUST bigger than the gamma R, so my 1 ohm resistors would be perfect. Maybe I should just do a 5x5 and be done with it.

How does that all sound? I'm sure glad I didn't go ahead and build my 7 x 7 matrix without thinking about this first!
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

Postby Klaas Robers » Thu May 24, 2007 7:45 am

Andrew, for the balacing resistors I advice 1 volt across the resistor at in your case 20 mA per string, i.e. 1 / 0,020 = 50 ohm, so 47 ohm or 56 ohm in series with each string.

The gamma circuit calculates like this:
The voltage swing on the input of the transistor = 2 volt from black to white. Then if you want to have 200 mA peak at white, the reference resistor R should be 2/0,2 = 10 ohm.

Now you can calculate the real resistors. Use diodes 1N4001 and not the 1N4148 at these high currents. The BD139 is strong enough.
User avatar
Klaas Robers
"Gomez!", "Oh Morticia."
 
Posts: 1656
Joined: Wed Jan 24, 2007 8:42 pm
Location: Valkenswaard, the Netherlands

Postby Andrew Davie » Thu May 24, 2007 11:33 am

Klaas Robers wrote:The gamma circuit calculates like this:
The voltage swing on the input of the transistor = 2 volt from black to white. Then if you want to have 200 mA peak at white, the reference resistor R should be 2/0,2 = 10 ohm.

Now you can calculate the real resistors. Use diodes 1N4001 and not the 1N4148 at these high currents. The BD139 is strong enough.



I don't understand any more how R is calculated. Assume that I use the example above, where R is calculated at 2/0.2 = 10 ohms. Using the formula 3R, 1.6R and 1.3R this gives me resistors of 30 ohms, 16 ohms and 13 ohms.

Now let's assume that I wanted to find R given the above resistors. I'd use the 1/R = 1/R1 + 1/R2 + 1/R3, so R = 1 / ( 1/R1 + 1/R2 + 1/R3 ) = 1/ (1/30 + 1/16 + 1/13) = 5.78 ohms.

Now this isn't anywhere near the 10 ohms that was used as the basis for calculating the resistors in the first place. As a rough approximation, it's about 50% of what is needed.

Assume that instead of 3.0R, 1.6R and 1.3R we used 4.5R, 2.4R and 2R then we'd have 72 ohms, 39 ohms and 32 ohms. Using the parallel resistance formula from these we get R = 1/((1/72)+(1/39)+(1/32)) = 9.8 ohms -- much better agreement.

So my question is -- what am I not understanding about R, and in particular the multiplication values used to calculate the resistors? I understand that the different values are giving us a non-linear greyscale that more closely matches the eye's responsiveness. What I don't understand is why the values chosen do not give a reasonable R when back-calculated from the resistor values. To me it looks like the recommended values are 50% too small.

Could someone please enlighten me?
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

Postby gary » Thu May 24, 2007 12:25 pm

Andrew, the key to producing the gamma correction effect is the diodes not the resistors. You have to take into account the voltage drop across the diodes (0.7 volts for silicon) to see what effect each individual resistor is having on the current flow:

The maximum current flow (and hence LED brightness) occurs, by Ohms law, when the video signal at the base of the transistor is at a maximum (in this case 2.7V) there is a 0.7V drop between the base of the transistor and the emitter (you can imagine that there is a diode between base and emitter inside the transistor) so this gives 2 volts at the emitter pin.

Now, at this input voltage there is 2 volts across the 30 ohm resistor, and, taking into account the voltage drops across the diodes, 2 - .7 = 1.3 volts across the 16 ohm resistor, and 2 - 0.7 - 0.7 = .6 volts across the 13 ohm resistor.

So the total current running through the gamma circuit at maximum input voltage is:

I = 2/30 + 1.3/16 + .6/13 = 194 mA

Which is close enough to the 200mA you require.

Now, a diode won't begin conducting until it's forward voltage (0.7V) is exceeded, so as the voltage rises from 0 to 2 volts you can see how each diode is switched on, one at a time (one at .7V and the other at 1.4V), and thus bringing in the other resistors, giving the required non-linear current/voltage curve.

Of course there is some resistance in the diodes themselves which will have some effect but hopefully this is negligible. I think Klaas alludes to this elsewhere in this forum, it is one of those areas where a little experimentation may be in order to get it near perfection.

In summary, because of the voltage drops across the diodes, you cannot just use the standard resistors in parallel formula.
gary
 

Postby Andrew Davie » Thu May 24, 2007 10:31 pm

gary wrote:In summary, because of the voltage drops across the diodes, you cannot just use the standard resistors in parallel formula.


Many thanks once again, Gary. I've learned a little more. I understand now exactly how the gamma circuit works, which I didn't before.

Following your post, and advice from Klaas, I felt that I would be OK building my super-large matrix consisting of 8 columns each of 5 LEDs with a 47 ohm resistor for each column. And to test it, I figured it would be safe enough connecting to my current gamma circuit/setup -- that is, with the original resistor values (although I have higher current-rated diodes, I think). My thinking was that it shoud work, but be less bright.

Well, the nice surprise is it worked first off -- all 40 LEDs shining very brightly. Very very brightly! I'm half tempted to leave the gamma circuit alone -- but then again if I can double the brightness -- well, why not. I have the correct resistors, so tomorrow I might just see how much improvement I get with the values recommended/calculated above, for the 200mA supply.

The thing that I've noticed consistently, as I churn through various LED matrix display combinations (my 3rd matrix now) is that the brighter I get the LEDs to light, the better the contrast I get. Put another way, the whites are whiter, and the blacks are still black.

There's still the problem of diffusing the light, of course. I wonder if anyone has tried a thin layer of insulation fibre glass type material?
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

Postby Andrew Davie » Sat May 26, 2007 12:30 am

I decided to install the 'correct' resistors for the gamma circuit, but when I went to do that I noticed that there was so much delicate wiring around the resistors on my single-board implementation that it would take ages to do, and I decided to junk the idea, as things were 'good enough'.

But then I had an idea. What if I placed resistors in parallel with the existing ones? So I calculated what would happen if I put my 33 ohm resistor in parallel with the existing 150 ohm one -- (1/150)+(1/33) = 1/R, R = 1/((1/150)+(1/33)) = 27 ohms, close enough to 30. So I placed the 33 ohm resistor in-circuit manually whilst it was on, and the picture brightened up *considerably*. The grey-scale looked a bit odd now, of course, as the gamma was behaving differently. But still, much brighter, higher contrast picture. So I soldered this one in, piggybacking on top of the original resistor.

Next came the 82 ohm resistor, which I wanted to get to about 16. Again, I had a 13 ohm resistor, so R = 1/ ((1/13)+(1/82)) = 11 ohms. Close enough -- well, not really -- but it did look OK, making just the slightest visible difference to the picture. So bango, soldered that one in too.

Next came the 56 ohm, and there was no noticeable difference when placing a 11 ohm resistor in parallel with it, so I left that alone. Wanting about 13 ohms, I figure R = 1/((1/11)+(1/56)) = 9 ohms, well that's close enough -- but visibly I couldn't see it so I left the original there (for now, anyway).

The difference in the picture is quite noticeable now. There's a definite 'harshness' to the highlights on the balding heads (sorry, guys!!!) on CD#1. But this is corrected by turning down the contrast knob a bit. And here it's actually a huge improvement. Prior to this, I've had to have the contrast turned up to absolute maximum to get anything really good looking. Now there's quite a range over which I can adjust the contrast, and then adjust the brightness, to get various feeling for the visuals -- in short, they now appear to operate similar to a 'real' TV and are both useful.

The picture itself -- well, about twice as bright, I'd say, as my version from earlier today, which in turn was about three times as bright as the first LED matrix that I'd built. So I'd say my picture is roughly 5 times brighter than yesterday -- and in fact it's probably brighter than a typical black and white CRT TV in similar light. Certainly much more recognisable and detailed images -- my wife commented on the teddy bear's textured fur from about 8 feet away, whereas previously it was recognisable as a teddy bear, but not contrasted nearly as much.

I really have to stop playing with the display now, and move on to something more useful -- like that negative-video circuit thing. I seem to be always putting off things to tinker with what I already have :)
User avatar
Andrew Davie
"Gomez!", "Oh Morticia."
 
Posts: 1590
Joined: Wed Jan 24, 2007 4:42 pm
Location: Queensland, Australia

PreviousNext

Return to Andrew Davie's Televisor #1 (2007)

Who is online

Users browsing this forum: No registered users and 1 guest

cron