Defeating the mantis part 3

oscillator frequency.

are ground, +5V, and video out.

So now you are trimming it out by hand?

If you need to get the two 17MHz clocks to lock WRT each other, try a weak connection between them. Even crystal oscillators can do injection lock provided their frequencies aren't grossly different in the first place.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

Mhz

oscillator frequency.

are ground, +5V, and video out.

Hi, Joerg. I'm very curious about this comment in regards to microcontrollers. I'd like to experiment with that later this year.

Before answering, I already know it's very easy to just use a separate oscillator and drive two cpus through the inputs to their inverters from that... that's not why I ask. What I'm curious about is when I have two duplicate boards from the same design and where, for some experiment in mind, I'd like to closely sync up the oscillators directly like this to remove beat frequencies between their behaviors for a one-off testing case. (If it were a design matter, I'd probably take a different tack.)

I don't know about these particular oscillators Jan is talking about, but I'm imagining the usual case I experience with micros, which is a class-A inverter. Some of these have outputs that are (over)driven pretty hard to cope with higher MHz crystals from a wide variety of manufacturers. Are you suggesting by "weak connection" a high resistance (or low valued cap?) between the two input sides of the inverters?

(I probably won't get to test this, in the coming month. But perhaps a month later. But I'd like to make sure I follow you while the idea is fresh in mind.)

Thanks, Jon

On a sunny day (Wed, 23 Sep 2009 11:47:03 -0700) it happened Joerg wrote in :

are ground, +5V, and video out.

The first test was to see if I could influence the crystal frequency. Syncing was already discussed in detail here considering V and H sync.

As the cameras are powered on at the same time I see V is very close, within a line. So it would seem a simple H compare would do, but I need to have the odd frame from one, and then the even frame from the other, so I need a V compare, and to be precise, one that knows about odd and even frames. I have all that stuff already in a separate box. Will have to see if a varicap or some transistor Cbe works best as tuning element. I can only slow down, so I must modify the right (fastest) camera.

right.

Mhz

oscillator frequency.

are ground, +5V, and video out.

No, that wouldn't give you any lock range to write home about. I'd declare one the master and then feed from its output side to the input gate terminal of the other, with an RC series combination (several k and very few pF).

This isn't about crystals but may help:

With crystal oscillators this is iffy and requires experiments, won't work with cheap and inaccurate crystals.

In Jan's case he could also just rip one crystal and directly feed in the clock from the other camera. However, injection looking offers some advantages:

a. You can use low amplitudes for the link which can seriously reduce EMC issues if any length of cable is needed to bridge a distance.

b. Chances that serious modulation gets is is reduced.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

right.

Mhz

oscillator frequency.

glue...

are ground, +5V, and video out.

I was first imagining the output to input path as obvious, but then figured I might as well just remove the other crystal then and drive two inputs instead of just one. Why not?

Thanks for the thoughts. I'll keep these for later reading.

Jon

right.

4.43 Mhz

oscillator frequency.

glue...

pins are ground, +5V, and video out.

Yes, but: A typical (weak) oscillator stage may not be able to drive the input on another board, especially if there is cable with its capacitive load in between. You may need to use a buffer to do that.

[...]

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

are ground, +5V, and video out.

line.

from one,

precise, one that

element.

Varicaps work quite well on crystal oscillators. That is how the Icom IC202 ham radio transceiver was tuned and they got around 1400ppm of pull range of this IIRC. It remained very stable, I used one for many years while living in NL.

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.

oscillator frequency.

ground, +5V, and video out.

cameras,

rivatv driver).

did I?

Wake me up when you get to Part 12.

John

right.

4.43 Mhz

crystal oscillator frequency.

glue...

pins are ground, +5V, and video out.

place.

I've done it without the buffers, but it's been a while so I can't remember the part (and it was on a single board.) I do imagine that most of the micros use class-A stages that are enough overpowered (wasted energy, "oh, well") that they can drive two inputs. I think they do that because they have no idea what crystal will be used and they expect that customers will probably use really terrible ones and they don't want to waste time in after-sale support calls. So they just crank up the juice and avoid the calls and then only rarely are asked to explain why their part is a little more energy hungry than someone else's (the price of which they more gladly pay than the flip side.)

But I take your point about cabling across boards, especially.

Oh, well. Thanks. I'm curious about the possibility and I'm going to play with it when I get a moment and have the boards handy to try it out and set up some software to observe (attempt to drive outputs synchronously and look on the scope for drift, I suppose, or wire that up to something that varies on phase [xor gate might be enough, with the output to the scope and triggering on some change.])

I don't need it now and may not for years. There was a time some years back and I used a different solution (a wire-or grant and address line for synching up.) So this might be handy to check out and put in the tool bag for some rainy day.

Jon

On a sunny day (Wed, 23 Sep 2009 14:21:25 -0700) it happened Joerg wrote in :

That will perhaps not work. Some theory: Both modules use an ASIC, is it guaranteed that, even when running from the same supply, after a power up (is there a reset circuit, RC perhaps?) all dividers and logic are in the same state?

There are 3 timings that need to be correct, if you want to flip between camera output in a V interval and maintain proper color (phase).

1) V sync 2) H sync 3) 4.43 MHz color subcarrier phase.

I am sort of hoping the digitiser chip will be fast enough to lock to a different color phase during the rest of the vertical interval after a switch to the other camera. That way not needing the 4.43 MHz phase compare, but maybe it is not. The color burst will continue during the V interval, so there is a chance.

And I want to change as little as possible in those camera modules, but maybe I will scope around some, connect some things together, and see if it will oscillate and start up in phase like you say.

But it would also be nice to add the possibility of external genlock, locking to an external source, as I have more PAL sources.

I did not have a very low cap varicap in the box, only 250 pF ones. So I used a transistor circuit, based on the capacitance change between collector and emittor of a NPN when the colector emittor voltage changes,

22 pf from the camara module to the collector, 3k9 collector to +5V, 68 k between collector and base (bias). Now feeding some current via a high resistor into the base (say control voltage), creates a neat linear shift in frequency around the other cameras fc. Used a BFR91, you would have liked that one in your ham days, Cce 0.3 pF at 10 V, Ft 5 GHz

same supply,

logic are in the same state?

Most likely it would not be guaranteed.

camera output in a V interval

You could potentially build a "clock skipper gate" and skip clocks on one camera until it's all in sync.

different color phase during the rest

I will scope around

in phase like you say.

to an external source,

Now you want to turn a 2-bedroom condo into a Ritz Carlton :-)

--
Regards, Joerg

http://www.analogconsultants.com/

"gmail" domain blocked because of excessive spam.
Use another domain or send PM.