Composite video issue

[mayor]

Hi,

I am bringing back my childhood ColecoVision to life. It's pretty crusty. After replacing the 7.15909 MHz crystal (unknown load capacitance!), a 7404 and a broken ceramic disc cap, I managed to get composite video out (hijacked from the RF modulator input). Only thing is, there's a lot of horizontal bleed. Not too sure what causes that, exactly. Amplitude, timing? The composite signal is about 700mV p/p, with DC of around 5V.

Any help appreciated! Pic attached.

[Renate]

You're stealing the video from a high impedance source.
It's not designed to drive a 75 ohm input.
Also, whatever you are plugging it into may not like 5V right in its ear.
(It also probably doesn't need the 4.5 MHz sound carrier that you are probably feeding it.)

You need a buffer.

[mayor]

I am using a transistor to drive the input, so I don't think current is the issue. The DC level is as intended by the CV designers. I should have added this info in my original post.

[PKTKS]

I would also guess you have an unmatched impedance causing
ghosts on the video signal.

at least distortion is low enough to still have SYNC fine

Paul

[vk6zgo]

I am using a transistor to drive the input, so I don't think current is the issue. The DC level is as intended by the CV designers. I should have added this info in my original post.

The actual video waveform including syncs should be 1 volt peak to peak for a normal standard video signal.
This implies that your composite video level is low.
Your device may, however use some other levels, as it seems to already use a non-standard DC offset.

Have you looked carefully at your signal to check that it is not "smeary" going into the monitor?
A screenshot of what your Oscilloscope sees may be helpful for us to help debug this.

[schmitt trigger]

Scope waveforms of your composite baseband baseband signal would go a long way towards troubleshooting your problems.
It could be that the black setup is too low, or that the video transitions are causing ringing into the blacker-than-black areas (the realm of sync pulses).

Or perhaps the DC level is too high. Or the sync pulses too short.

Also, your statement that "I am using a transistor to drive the input, so I don't think current is the issue" is pretty vague, don't you agree? Where is the schematic?

[Renate]

I am using a transistor to drive the input...

So, I'm betting that you are using an NPN as an emitter follower and that your emitter resistor is too high a value.
That means that the transistor can easily (and quickly) pull to white while decay to black is slow.
So just replace the resistor with half the value and notice the improvement.
(Not that that is the complete solution.)

[mayor]

Thanks all!

PKTKS: not too sure how I'd go about matching it, TBH.


vk6zgo: Yeah, I've probed around the video chips, and the signals are lower than 1V p/p, and don't appear very clean at the source, so yeah, it could be considered "smeary". Not too sure what to look for, or what the cause might be. I'll try to find a usb key for some captures.


schmitt trigger: Agreed. Essentially :


12V -> 1K POT -> Collector
Composite video -> Base
GND -> 1K -> Emitter

-- will provide some traces for what I find on the board...


Renate: will try lower resistor value on emitter. Thanks for the suggestion.

[mayor]

Halving the emitter resistor hasn't produced a significant improvement.

That being said, I feel a little shame at not having paid more attention to the output of the graphics chip -- or, maybe more importantly, one of its inputs. Here are some traces.

1. Main clock (looks fairly clean)
2. Video chip clock (derived from main clock)
3. B-Y signal (ouch!)

So I guess the next step would be to figure out why the clock for that chip is so bad, huh?

[vk6zgo]

Halving the emitter resistor hasn't produced a significant improvement.

That being said, I feel a little shame at not having paid more attention to the output of the graphics chip -- or, maybe more importantly, one of its inputs. Here are some traces.

1. Main clock (looks fairly clean)
2. Video chip clock (derived from main clock)
3. B-Y signal (ouch!)

So I guess the next step would be to figure out why the clock for that chip is so bad, huh?

I googled for Colecovision, & found this:-

https://en.m.wikipedia.org/wiki/ColecoVision

which further lead to this:-

http://www.cs.columbia.edu/~sedwards/papers/TMS9918.pdf

It appears that the TMS 9918A chip doesn't present Y, R-Y, & B-Y to the external circuit, but produces composite NTSC video.

The TMS 9928A doesn't do composite, but supplies 525 line format, Y, R-Y, & B-Y to the external circuitry .
This chip requires a separate encoder to produce NTSC or PAL composite video (there was a 525 line PAL system in one country).

The TMS 9929A doesn't do composite, but supplies 625 line format, Y, R-Y, & B-Y to the external circuitry.
This chip requires a separate encoder to produce SECAM or PAL composite video.

Two rather interesting things arise with your 'scope screenshot.

(1)You seem to be looking at the video signal at field rate or thereabouts, so it puzzles me that the display looks as if it is suffering from aliasing, but your sample rate seems quite adequate to correctly show the signal.

(2)Strangely, you have a box at the top of your pic which reads 15625Hz.*
For NTSC it should be 15750Hz or thereabouts.

Can you show us a pix at line rate?

Edit---*Sorry, I read that wrong--- that said, the frequency it does show makes no sense at all.
On the other hand, I don't have a DS1054Z, so maybe that information is nothing to do with the actual frequency components of the signal.

[mayor]

Hi,

I think the frequency doesn't make sense in part because I am unable to trigger on that signal (ie, it's not very constant, even with a static image). I've found the following: https://rigol.desk.com/customer/en/portal/articles/2311798-ds1000z-counter-technical-details, which is interesting. I'll experiment a little later with the vertical scale, since it impacts the measurement.

That being said, the B-Y signal looks pretty bad. All the drops to 0 between syncs is unexpected to me, but then, so are many other things :-) I'm not familiar enough with electronics to know for sure, but doesn't the input clock look bad to you, which might explain the bad output?

[nali]

There's a bunch of info including reverse-engineered schematics here, they show that the RF modulator is a LM1889. According to the app note  dataheet it's possible to hack a composite video from pin 13 via an emitter follower and decoupling cap.

[PKTKS]

Thanks all!

PKTKS: not too sure how I'd go about matching it, TBH.
(..)

That depends on the circuit you have chosen to couple your
composite output.

Essentially  you "REFLECT" the output back to the base of the
transistor calculated to be 75 Ohms.

75 Ohms is expected in any input as say if you have 10.000 Ohms
would be too high causing harmonics to bounce in group delay of
high components.. thus explaining those late line phantoms...

Top list first thing to check by design.

Sure there are other methods to measure with a good scope the
reflected signal.. but far too details to  put here..

Paul

[mayor]

There's a bunch of info including reverse-engineered schematics here, they show that the RF modulator is a LM1889. According to the app note  dataheet it's possible to hack a composite video from pin 13 via an emitter follower and decoupling cap.

Hi, yeah, that's what I essentially did. But the signal isn't very clean at the source, if I'm interpreting it correctly.

[mayor]

There's a bunch of info including reverse-engineered schematics here, they show that the RF modulator is a LM1889. According to the app note  dataheet it's possible to hack a composite video from pin 13 via an emitter follower and decoupling cap.

Hi, yeah, that's what I essentially did. But the signal isn't very clean at the source, if I'm interpreting it correctly.

Ah but I will review: they use a series resistor (75 ohm) and a DC blocking cap. OK, that'll be an experiment for tonight!

[PKTKS]

(.)
Ah but I will review: they use a series resistor (75 ohm) and a DC blocking cap. OK, that'll be an experiment for tonight!

That only fits if your reflected impedance (the HFE reflection)
if low enough compared with the series resistor e.g. it does
not add enough  not causing harmonics delay.

BTW the transistor you choose is also sensitive as low BW
gates tend to introduce group delays as well (harmonics delay)

Paul

[Renate]

Reflections are a considerations, but we're nowhere near that level yet.
If somebody were to recommend a unity gain IC buffer with near zero output impedance then talk of a 75 ohm series build-out resistor would make sense.

The current buffer doesn't need any collector resistor and the emitter resistor is still too big.
You're also shoving 5V down its throat.

While the attached schematic is not the last word in design, it should get you decent video.
(If you take any photos of the video signal, please show 10uS/div sweep.)

[PKTKS]

The  color sub  carrier contains harmonics
more or less necessary for proper color definition

In 3.5MHz at least 8MHz full band pass window
In 4.4MHz at least 10MHz full band pass window

Good color definition requires a careful chosen
transistor so we do not cut off higher order harmonics
which will eventually lead to poor colors.

so assume a proper  safety margin of at least 10MHz band
without reflections whatsoever. (or phantom colors appears)

PAL and NTSC demodulation resides in the higher order
harmonics of the sub carrier.

Paul

[Renate]

The  color sub  carrier contains harmonics more or less necessary for proper color definition

Erm, not really.
The color subcarrier certainly has sidebands.
The second harmonic of NTSC color subcarrier would be 7.16 MHz
NTSC ran on a 6 MHz bandwidth channel for decades with a 3.58 color (and that was already offset 1.25 MHz).

Moreover, if you really want things to look good you should just buffer the Y, R-Y, B-Y out of the Coleco and avoid the whole encode/decode process.

[PKTKS]

Erm, not really.
The color subcarrier certainly has sidebands.
The second harmonic of NTSC color subcarrier would be 7.16 MHz
(..)

Yep, I am sorry about the crude terms..  but the real thing
is only put in proper terms with a not trivial math.

The first correction is that what I meant by "harmonics"
actually are the vestigial bands of the HORIZONTAL component
which actually fall shortly on the primer sub carrier (by figure)

The vestigial bands are different for PAL and NTSC thus
explaining the different precise XTALs used.

Any further understanding unfortunately goes through the
proper Fourier expansion which including the periodic square
wave adds a plethora of things not  visible at first glance

Attached figure is the result of Laplace and Fourier plotted
with proper math done and proper vestigial band (for NTSC-M only).

That I meant for "harmonics" ... (actually vestigial bands)
Just in case you would like to see formal equation I can add too

Paul

[vk6zgo]

https://en.m.wikipedia.org/wiki/PAL

The PAL wiki above is a good resource, but I will attempt to provide my own description.
As I am more familiar with PAL(B), I will use it as an example:-

The RF channel allocated for transmission in this system is 7MHz wide.
This must allow for "guard bands" at the upper & lower extremities of the channel width.

The video signal,which is limited to 5MHz bandwidth by specification in PAL (B), modulates the Vision Carrier (Vc) using Vestigial Sideband Modulation(VSB)'with an Upper Sideband extending to 5MHz above Vc.

There is a "Guard band", then the first sound subcarrier at 5.5MHz above Vc, then another sound carrier a5.75MHz above Vc.(for the stereo system employed in Australia).

As the term "Vestigial Sideband" implies, the lower Vision sideband is partially supressed, only extending out to 1.25MHz below Vc.

All the above goes to point out that it is a band limited system.

Video signals in PAL(B) are rolled off quite rapidly after 5MHz.

The band occupied by the video signal is often drawn, as if it continuously occupies the whole spectrum, but this is not so.
The (black& white) Luma video information appears at multiples of line frequency, like a lot of "carriers", with field rate sidebands, & areas of spectrum between these "carriers" carrying little or no energy.
Providently, this allows a method of transmitting colour information.

The human eye has relatively poor resolution for colour, compared to black & white, so it is possible to reduce the required colour signal (chroma) bandwidth to around 1.3 MHz.

Instead of having to transmit  R, G, & B signals, colour difference signals R-Y & B-Y are used.
These are Quadrature amplitude modulated onto a 4.43361875MHz subcarrier.
The reason for the strange subcarrier value is that the video sidebands from this modulation fall in the "empty" spaces between the luma "carriers" ----so called frequency interlace.
The colour signal is placed at the high end of the video passband because the average signal power at that end is quite low.

Very early Colour receivers used Low pass filters to separate the Chroma & Luma components of the received signal, but this caused a loss of Luma resolution.
A later iteration was to use "notch"filters, but the definitive method employed "comb" filters, which caused much less loss of Luma resolution.

The PAL wiki page shows a diagram of the frequency components of a transmitted TV RF signal, followed by a spectral display of the same signal


OP--- your pix of the B-Y signal at field rate looks a lot more like a composite signal.
The colour difference signal will not normally have syncs attached.

In any case, you need to drop vour volt@/div control so the video signal takes up most of the screen height.
The "scrunched up" version you provided makes it very difficult to get a good idea of the signal's appearance.

Ohh, & a line rate screenshot would be nice, as I suggested earlier

[schmitt trigger]

Indeed!
As a fact: a line rate screen shot is the only way to determine whether the active video setup is correct.

[mayor]

Here's some more info. With the same circuit I described above, I have managed to get a good (albeit a little dark) image. What I had to do was trim the pot down initially so the TV picked up the signal, and once it did, I could increase the resistance until I got a decent pic. I measured the signal with this adjustment, and it looks like the TV "blows out" above 6V. I'll give Renate's circuit a try after this to lower the DC component. Hopefully, this will produce the desired result, while keeping signal amplitude high enough.

Here are some captures of the original signal, and the output with my current circuit. The image signal is somewhat "skewed", which seems to happen with lower resistance on the emitter (500 ohm here).

[mayor]

Indeed!
As a fact: a line rate screen shot is the only way to determine whether the active video setup is correct.

Sorry about the noob question, but not sure what the "line rate" is!

[mayor]

Input and output with 1K resistor from emitter to ground.