Connect with us

CVBS Video Switch (non-genlocked)

Discussion in 'Electronic Design' started by Quack, Jun 6, 2005.

Scroll to continue with content
  1. Quack

    Quack Guest

    Hi Everyone,

    I know i have posted questions on this topic before, but i am still
    trying to find a solution for this one.

    I can not use gen-locked video sources, and the sources are composite

    My circuit has 4 composite video inputs, and a single composite video
    output, its a simple switch.

    But i need to be able to switch cleanly between the sources, and
    because they are not within the same timing, i must use some kind of
    field memory etc

    I notice maxim has a lot of 'video buffer' ic's, but i think this is
    regarding the signal levels itself, not like a data buffer.

    What terminology should i be searching for, or better yet, is there a
    fantastic chip out there that could do this for me ? :) Any examples ?

  2. Quack

    Quack Guest

  3. Quack

    Quack Guest

    This guy should do it: But that does not accept component video inputs, so i would first need
    to convert my signals into digital (saa711x? ic), then back to
    composite after :(.
    No easier way ? If not, what is the best way to go about it.

    Surely there must be some examples or good documentation around for
    making a frame synchronizer ?

  4. Without genlock you are basically lost.

    I think that you need a synchroniser/framestore, or a little digital
    video mixer such as

    It will be a lot cheaper than building your own

  5. Ban

    Ban Guest

    If you want to see all signals at the same time split screen, then you have
    to do that. Otherwise you could maybe modify the monitor to mute the picture
    until the sync has locked again, but this is not so easy as well. Ideally
    there would be 4 sync circuits multiplexed, then there would be only one
    frame timeout, well I don't know if the line transformer likes this.
    Maybe somebody knows more of video than I do.
  6. Well, they convert the video image for you from component to digital and
    store a frame. You can read the frame out at a different speed than you
    write it into the chip, allowing you to 'resync' the video source. And yes,
    you'd have to convert composite to component and back as well.

    But as another poster stated, there are relatively cheap consumer
    videomixers out there that can do this for you.

    I dont see exactly how these can be used, unless i convert the
    composite signal to digital and back again. Whats the best way to do
    that ?

  7. Joerg

    Joerg Guest

    Hello Alex,
    You can't store a frame any other way. There used to be CCD buffers for
    one line of video (but not a frame). However, AFAIK even those are

    Just one word of caution: You might find chips that combine a lot of the
    AD/DA and sync separation functions you need but they can disappear
    quickly. Usually that happens the minute such a chip is no longer needed
    for the production of some consumer equipment. Often there is no 2nd
    source. So be careful if this is for a product with a long production life.

    Regards, Joerg
  8. Quack

    Quack Guest

    Many thanks everyone!,

    It seems far more complex than i ever imagined. Unfortunate, i know its
    possible but it would be too big and expensive for what i am trying to
    do. I can not use a consumer grade mixer, this device must be small and
    light (max: 6cm x 6cm x 200grams), which it is currently but it
    switches dirty.

    I had an idea though, how does this sound;

    The output of this switch goes to a digital encoding device anyway, a
    video capture system (digitizer), but this system seems to be extremely
    poor on 'resyncing' after a dirty switch (sometimes, random of course).

    By that i mean sometimes its fine after a switch sometimes, and
    sometimes its screwed up for a few seconds or more due to the random
    location of the 'cut' when switching in the signal.

    So my idea is to use a LM358 to monitor the position of each of the 4
    incoming composite signals (detecting the vertical sync and counting).

    I would then swith video like so;
    I would 'hold the current voltage level' of the signal (on the output)
    when i reach the next vertical sync in the currently selected video.

    I would then internally switch to the desired video input (circuit
    output still in sync-level voltage)

    I would wait for its next vertical sync (the desired input) and then
    let it go out.

    So that the video signal would always be 'complete frames', just that
    when switched, a particular frame has a very loooong SYNC voltage
    level, perhaps this would _force_ the digitizing device to recalibrate
    itself quicker, in that it gets a known stuffed up frame, rather than
    just 'random garbledness'.

    My question here is that perhaps a 'long sync' can be useful, or
    perhaps not, what other methods are commonly used (using the lm358 and
    a video switch IC [eg: max4545]) to force a receiving device to re lock
    on a new signal.

    Perhaps i should just 'break <pause> before make' ... i think i will
    start experimenting this way rather than try to buffer the video and
    remove the bad frame completely, if i can force it to be bad 'in a good
    way', then thats okay too - i can pause the digitzing device (not
    capture data) for upto 200ms, so the goal is to have it resync within

    More info:
    Digitizing device is based on phillips SAA7113 with NT1004 usb
    ( and
    By reading the above SAA7113.pdf, you can ascertain that the 'host
    controller' can monitor this chip and actually know if there is a good
    sync lock or not, this would be fantastic but we have no 'pipe' through
    the NT1004 to do this, so we are not in control of the way the SAA7113
    is configured or used, only that it is used. Also even if we could
    ascertain on the host system that there is no good sync locked
    currently, we would only guarantee we dont get a garbled frame, we
    would still have occasional 'looong pauses' without any data when it
    decided not to sync-lock properly due to the nature of the switched
    frame (random locations).

    So relating these devices to the above description, how can i force the
    SAA7113 to resync nicely within 200ms after a switch, by;
    -forcing all switched frames to be short
    -forcing all switched frames to be long
    -holding at sync level voltage for the time difference between signals
    -holding at other? level voltage for the time difference between

    or something else perhaps ?

    Thanks again, and i hope this makes some kind of sense :)
  9. This is probably the easiest solution. By removing the signal
    completely for a few frames, the decoder will probably lock to the new
    signal a lot faster, when you apply it.
    You can experiment with the length of the "no-signal" duration to find
    the optimum time.

  10. Quack

    Quack Guest

    Hi Trond,

    Would it be best to cut the signal at the vertical sync or any other
    particular place, and re-establish a signal at its vertical or other
    sync, or just randomly cut it and return it at any location ?

  11. Jim Thompson

    Jim Thompson Guest

    That jogged my memory. The trick is to fade to black, switch to the
    new video source which is held to black, pause for sync to occur, then

    ...Jim Thompson
  12. Quack

    Quack Guest

    But thats only for in-sync signals no ?
    What do i do with the broken short (or long) frames ?

  13. Jim Thompson

    Jim Thompson Guest

    I know it works for monitors, due to black-level clamping, but I don't
    know how your digitizing circuitry might respond.

    ...Jim Thompson
    should give you some clues.
    You only need two tbcs if you are crafty with switching, you can
    probably get away with one by using one camera to reference the TBC,
    cut to black for a couple of frames, and change the i/p to the TBC
    while in black.

    I havent fully read the two website quoted, but it should give you an
    idea of the complexity and if it is within your expertise

  15. My guess is that the whole problem is the decoder not realizing that
    the signal has changed for up to several seconds after switching. By
    removing the signal completely for a few frames, the decoder will be
    forced into some sort of "search for sync" mode, and locking will
    probably occur sooner when you apply the new signal. I don't think it
    will matter much at which point in the video frame you do the cutting,
    but you can always do some experiments.

  16. If cutting out the sync works depends on the digitizer. If it does not work,
    you can try this:

    Since at this point, you are extracting the sync signal anyways, you could
    remove the H and V sync from the input and generate new syncs using a uC.
    Then, figure out by how much your timing can be off between frames to not
    get a hickup in the digitizer and adjust your artificial sync until it
    matches the input sync.

    The effect would be that the contents of the video frame would move around
    until the sources are synced again.

    But this is more of a programming job than an electronics job.
  17. Quack

    Quack Guest

    Since at this point, you are extracting the sync signal anyways, you could
    That sounds like an interesting idea.

    I am using a sync seperator chip to read the sync, how should i
    generate it, just go from black (0.3v) to 0v ? or is generating a sync
    more complex than that ?

  18. Joerg

    Joerg Guest

    Hello Alex,
    A good sync separator chip should be able to generate sync pulses that
    are within spec while it isn't receiving any. Kind of a "free running
    mode". It's been a long time but I believe the chip I did that with was
    the SAA1043 or something (could be wrong on that P/N), probably obsolete
    by now. AFAIR I had to design my own PLL for one of the syncs. Anyway,
    Philips used to always have a chip like that. Then you don't need a uC.

    If V-sync was way off from one source to the other it will take quite a
    while to nudge over to the new sync without deviating too much from
    standard. Else the V-sync of the monitor or whatever is connected could
    briefly "hick up".

    Regards, Joerg
  19. Quack

    Quack Guest

    Hi Matthias,

    Can you point me in the right direction for 'removing the H and V sync
    and generating them in a uC' ? :).

    I have tried all else mentioned here, ill give this a try.

Ask a Question
Want to reply to this thread or ask your own question?
You'll need to choose a username for the site, which only take a couple of moments (here). After that, you can post your question and our members will help you out.
Electronics Point Logo
Continue to site
Quote of the day