Pretty interlace vs. ugly interlace

[mazinger-z]

Hello!
I tried to display an interlaced 640x480 screen with my PC. At first, I used a Microvitec M1438 monitor (multisync, Amiga branded) and it looked really nice. Then I built a converter from Hsync/Vsync to Composite Sync and tested the same screenmode with a TV set and a Commodore 1084 monitor. This time the picture looked ugly: some lines were thicker and some were hardly visible. Text was difficult to read.
What can this depend on, the converter or the TV set / monitor?  

[Jack Burton]

The sync shouldn't change the image quality.  

A little bit of looking around on google reveals the Microvitec M1438 has a dot pitch of .28mm and the 1084 has a pitch of .42mm.

http://usuarios.iponet.es/dardo/comu.txt (sorry, in spanish)

http://www.amiga-hardware.com/showhardware.cgi?HARDID=849

The dot pitch of a monitor affects its ability to display overall detail in an image.  

http://en.wikipedia.org/wiki/Dot_pitch

For a monitor of that vintage the Microvitec monitor has a very fine dot pitch and it should be able to display text much more clearly than the 1084 or a tv from the same period.  

However, that's not to say that you don't want to use the Commodore monitor.  In fact, it has a far more authentic image than the Microvitec because it has such a coarse dot pitch.  The appearance of scanlines during gameplay will be far reduced, and it will have an overall smoother, more arcade-like image.  You will not experience a loss of detail since even a .42mm dot pitch is sufficient to display the 240 lines or less of detail present in most arcade games.  

[MonMotha]

What kind of "converter" did you use?  Some of the more popular (simple) methods don't generate a proper composite sync signal, so the monitor will think you're actually running progressive at half the vertical resolution (which will indeed look ugly).

Also, some monitors never designed for interlaced scan can't handle it properly and look bad.  That's just the way they are.  Some arcade monitors have this problem.  The old Commodore monitors might since those Commodore computers generally only ran progressive.

[mazinger-z]

I'm using this one by viletim:



The "enable" circuit is not described on the 'net, as I designed it myself. It's connected to Hsync.

[MonMotha]

Well, that's a reasonable way to do it.  If you really are feeding something with a SCART input (or anything that expects CVBS rather than TTL sync), you may need to play with the value of R4, but 1k is reasonable to start.  You might bump R1/R2 up to 47k or increase C1/C2 to 4.7uF, but they're probably fine as is.

It's entirely possible the monitor with "ugly" interlace just isn't designed for it.  I've found a few (mostly cheap) arcade monitors that don't handle it very well.

[mazinger-z]

I observed some more and I don't think the phenomenon is due to the monitor and TV set I'm using. With the Commodore 1084 monitor, I tried to turn the vertical position and vertical height knobs, to see if the lines affected would change; everything remained the same. If I move a window up or down a little bit, the appearance of the text on the title bar changes, and at some positions, it misses a line of pixels.
Furthermore, I have always used these devices with DVD players, computers (Amiga) and TV decoders which give out interlaced screens, without any problems.
Unfortunately I didn't manage to directly use the composite sync signal that my Radeon 9250 card can provide. This could explain many things.

[Jack Burton]

You're definitely having the phenomenon that mon mothma described then.  Single pixel thick lines are disappearing because the overall resolution is being halved. 

[mazinger-z]

Then again... it could indeed be a matter of dot pitch... I took some pics, here they are.

Here's how a titlebar changes when I move the window vertically:



This is what the monitor is fed with:


My understanding is that too few rasterlines are available to display such detailed graphics, so something is lost. When a window is moved, what is lost and what is retained changes, and so does the overall appearance of the graphics.
At least it's a possible explaination.

@Jack: thanks for your answer. I think I know what you're referring to and I disagree: the screen is clearly interlaced as it flickers heavily. Here's a picture:

If you compare the text "Arcade covers" with "CD Audio 64" you'll see another example of what I'm talking about.

[Jack Burton]

I think you're right.  

Flicker = interlace

Can't really get around that.  That's assuming that this is -bad- flickering.  Not just 60hz flickering that's normal for a pc monitor.  

The first couple of photos look about as good as I'd expect a monitor with a .42mm pitch to look.  

It looks like the text and folder icon for "Arcade covers" is kind of compressed vertically.  That's a geometry distortion and is also pretty normal for a monitor that age. 

[mazinger-z]

Just tried a different sync converter and, guess what, I had a different picture.

Old converter (74HC86):


New converter (transistors) (please don't mind the different colors as I was playing with the contrast knob):


The photos were taken without moving anything, it's exactly the same screen, just the converter was swapped.

[MonMotha]

The various sync combiner circuits posted on the internet work in various different ways.  An exclusive-OR (74xx86) is generally logically the most correct, but not all monitors are designed for TTL level inputs.  Your circuit above would output TTL levels but with a 1k source impedance which is quite high.  An HC series will also have very sharp edges that can cause problems; you might have better luck with an LS type.  What circuit did you end up using?

[mazinger-z]

I had an 'LS86 right on my desk, so I tried it, but got the exact same results as with the 'HC86.
I took the new circuit from here: Coelho's Site - VGATV Hardware Page [MC1377P]. I only used the part of the circuit that combines the sync signals, but it's probably out of specifications, so I'm not sure I want to use it in my board. The really strange thing is that it only works with supply voltage disconnected. 
Do you think my problem is most probably due to impedance mismatch or the 74xx86's signal shape itself?
Thanks!

[MonMotha]

Doubtful it's the signal "shape" (meaning edge quality), especially since the LS and HC behaved the same.  That was a long shot (but easy to test), anyway.

There are two major differences between those two circuits.

One is that it's logically different.  Rather than XORing the syncs, it creates a multi-level OR of sorts then runs that through an inverter stage.  The behavior will be quite different depending on the polarity of the input syncs (and if they're matched or not).  In general, if the polarities are matched, you'll get "little" h-sync pulses during a frame, and "big" h-sync pulses during v-sync.  This isn't really composite sync, but people do it all the time and, due to the way the sync separators on most analog monitors works, it often works.  It can confuse some digital monitors, though. The non-linearity of Q1 will play with this some, so without simulating it I'm not entirely sure exactly what will come out the end.

The other big difference is the signal levels and source impedance.  Your old converter put out 5Vpp from 1kohms.  This one puts out ~2.5V from 70ohms.  If you're still curious as to what's going on, this is easy enough to test.  Replace the output resistor on your old circuit with a T-pad where every component has a value of 47ohms.

How are you hooking sync up to the commodore monitor?  Does it have a dedicated sync input, or are you coupling the sync onto the video signal somehow eg. SoG?

Also, if you ever actually want to do an RGB to Composite/S-Video conversion, look up the AD725.  It's much easier to use than that MC1377.

EDIT: Wait a sec, you say you're running it with the supply disconnected?  What parts do and don't you have hooked up?

[mazinger-z]

How are you hooking sync up to the commodore monitor?  Does it have a dedicated sync input, or are you coupling the sync onto the video signal somehow eg. SoG?

It has a dedicated composite sync input. It's equipped with a Scart socket.

Also, if you ever actually want to do an RGB to Composite/S-Video conversion, look up the AD725.  It's much easier to use than that MC1377.

No doubt about it! The Sony CXA1645 is also nice, but almost impossible to find in DIP.

EDIT: Wait a sec, you say you're running it with the supply disconnected?  What parts do and don't you have hooked up?

Here they are. I just wrote "+5V" where the output of the 7805 was. It won't work with that voltage connected, though.

[MonMotha]

Ok, that's a bit weird of a circuit to run without that supply present, but I think I know what it's doing (not going to explain it since it's a bit complicated and I'm not entirely sure without simulating it).  I think that you've effectively got the classic "hook 'em together" "hacky but it often works" style combiner circuit.

I suspect your issues have to do with SCART sockets not expecting TTL level sync inputs.  That "sync" input is actually a CVBS input.  It expects a composite video signal (the video on which is ignored unless you actually do the fast switching thing that the RGB inputs on SCART were originally intended for) at typical 1Vpp levels, of which the sync signal is only about 0.3V worth.  Good chance you're just feeding it "too big" of a signal with all your other circuits.

But hey, you've got something that works now, so use that

[mazinger-z]

Thank you very much for your patience and time!