Why does this animated LED POS display look like its leaning?

[BigT]

So I take it my explanation of calling it motion blur wasn’t sufficient? Perhaps this will help:

You don’t see what you think you see. It’s just like a video. It’s a bunch of still images that create the illusion of motion. The lights are just turning on and off sequentially, but, to your brain, the light is moving between the two physical lights. So of course you brain is extrapolating a bunch of other positions for the light.

Now, ordinarily, this would only create the illusion of motion. But the problem is that the top lights and the bottom lights aren’t “moving” at the same time. So your brain is seeing the bottom move before the top. The only condition where that is true is if the text is leaning.

As for your other questions: it only happens in certain types of displays–ones that do not move the text all at once. And it always leans the same direction because the bottom is always drawn before the top, no matter which direction the text is going.

That’s also why it doesn’t happen if the text is moving vertically. What you would see there is a slightly shorter graphic moving up, and a slightly longer graphic moving down.

Oh, and I apologize for previously saying that the top was moving first. I’m just more used to that form of tearing.

[Mangetout]

I was looking at the departures board on the station platform this morning, and it doesn’t noticeably exhibit the slant effect, despite having horizontally text that scrolls in both axes and also obviously being a multiplexed LED matrix display.

I think this must be because the multiplex refresh frequency is probably a very large multiple of the rate of scroll progression - so anything that’s happening to multiple LEDs in sequence is perceptible as intantaneous.

In fact, I’m sure the text slant effect could be made to happen even on a non-multiplexed display (for example, a display based on shift registers, where LEDs may be illuminated 100% of the time, or not) - as long as there is a small interval between the shifting of each horizontal row on the display, the brain will likely still interpret this as a slanted line moving.

[Musicat]

@beowulff

Take a look at this PDF: http://www.zilog.com/docs/z8/appnotes/an_msgdisp.pdf
Specifically, note the passage on page 6

Scanning top down while scrolling left produces the illusion of a character slanted to the left. And vice versa…Produce slightly warped characters by scanning rows 6 0 5 1 4 2 3.

Thanks for the reference to that document (did you write it?). It seems to explain the technical reason behind the expected illusion.

It is interesting that the display designers considered saving 1K of RAM to be so important that they didn’t want to fill a buffer, then read it out, but calculate the display data on the fly instead.

But I am unable to explain why that kind of display procedure produces the optical illusion I observe. I cannot see why the sequence of lights being turned on has anything to do with the apparent location between the LEDs, and selective location besides.

I think I may have to video the sign at various shutter speeds and play it back at various frame rates to see if a certain speed or shutter rate is critical to the illusion. It also seems possible that the illusion will not happen on a video at all.

I don’t know of any other in-store display near me like this one, but I have been observing outdoor multi-segment displays recently. Some are incandescent bulbs, but most are LEDs (I haven’t gotten close enough to determine). Some don’t use the scroll-left display, but those that do, do NOT exhibit the leaning illusion. Their characters appear perfectly vertical.

I can only assume that the other signs use a different display method.

[beowulff]

@Musicat

Thanks for the reference to that document (did you write it?). It seems to explain the technical reason behind the expected illusion.

It is interesting that the display designers considered saving 1K of RAM to be so important that they didn’t want to fill a buffer, then read it out, but calculate the display data on the fly instead.

But I am unable to explain why that kind of display procedure produces the optical illusion I observe. I cannot see why the sequence of lights being turned on has anything to do with the apparent location between the LEDs, and selective location besides.

I think I may have to video the sign at various shutter speeds and play it back at various frame rates to see if a certain speed or shutter rate is critical to the illusion. It also seems possible that the illusion will not happen on a video at all.

I don’t know of any other in-store display near me like this one, but I have been observing outdoor multi-segment displays recently. Some are incandescent bulbs, but most are LEDs (I haven’t gotten close enough to determine). Some don’t use the scroll-left display, but those that do, do NOT exhibit the leaning illusion. Their characters appear perfectly vertical.

I can only assume that the other signs use a different display method.

No, I didn’t write it.

I have been thinking about how to explain this better, and this is what I’ve come up with:
For a Multiplexed display (one where only one row is illuminated at a time), how can a scrolling vertical line NOT appear slanted? After all, the lit LEDs in each row are separated in time by some small amount, right? Your eye/brain is “fusing” the flashing LEDs into a solid line, and tracking with them as they scroll, so the LEDs on the top and bottom of the line are going to appear displaced, compared to the center of the line (since they are illuminated earlier and later than the center LED).
If you can get video with a very high frame rate camera, you will see how the multiplexing works, but I’m not sure it will make this visual artifact any clearer.

[Musicat]

Let’s try it this way. Assume I have masked off all LEDs except two adjacent (side-by-side) ones. Now the scrolling-left text display starts. What will the eye see? Will it look like there is a bright LED inbetween the two actual ones? Will it look like the first (right) one is lit, then the second (left)? Will it look like both are on at the same time, and it looks like a wide LED instead of a pinpoint?

[beowulff]

@Musicat

Let’s try it this way. Assume I have masked off all LEDs except two adjacent (side-by-side) ones. Now the scrolling-left text display starts. What will the eye see? Will it look like there is a bright LED inbetween the two actual ones? Will it look like the first (right) one is lit, then the second (left)? Will it look like both are on at the same time, and it looks like a wide LED instead of a pinpoint?

I don’t know.
That’s too narrow to get the illusion of scrolling. I think you would see a line move from the left column to the right. But, a lot depends on how often the line is displayed. It’s an interesting experiment,

ETA:
1K is a lot of memory for an embedded controller - My original sign design used 256 bytes!

[Musicat]

@beowulff

I don’t know.
That’s too narrow to get the illusion of scrolling. I think you would see a line move from the left column to the right. But, a lot depends on how often the line is displayed. It’s an interesting experiment,

My next question was going to be: What if you expanded the mask to uncover two horizontal lines (2x2, 4 LEDs) – would the line appear to lean? And if not 2 lines, what about 3 or 4? What about 3 across, 2 down?

ETA:
1K is a lot of memory for an embedded controller - My original sign design used 256 bytes!

I can sympathize. I used to work interfacing designs not unlike the sign one you referenced, and my boss said if program didn’t fit in 4K, it wasn’t worth writing.

[Mangetout]

@beowulff

For a Multiplexed display (one where only one row is illuminated at a time), how can a scrolling vertical line NOT appear slanted?

I can think of two ways:
[ul]
[li]If the multiplexing frequency is a vast multiple of the scroll progression rate, the slant might not be perceptible.[/li](I’m pretty sure I’ve seen some displays where this is true)
[li]For a display that is only capable of displaying text at one rate of scroll, the LEDs could be physically arranged with a real slant in the opposite direction, which should cancel out the perceived slant caused by the multiplexing.[/li][/ul]

[scratchmarc]

Sorry to dredge up this old thread, but was wondering why segmented digital displays are usually italicized (leaning)?

This thread is not about those kind of displays, but was all I could find that was similar, to see if anyone had an idea.

Thanks!

[beowulff]

I think it’s just for appearance. I’ve seen 7-segment displays that are vertical, but they just don’t look as nice as the ones with some tilt to them.

[Half_Man_Half_Wit]

@Mangetout

Our brains perceive the transition of LED P turning off and LED Q turning on as movement between P and Q - if this happens at the same time as LEDs V and W (stacked above), we see a short vertical line moving. If P and Q transition ever so slightly earlier than V and W, the only way our brains can make sense of this, and still see it as a connected line, is if the line is tilted slightly.

Just for completness’ sake, this is known as the phi phenomenon: if two adjacent lights are lit in short succession, we perceive the light moving from one to the other, due to our visual circuitry filling in the gaps (there is no discontinuous movement in nature). It’s perhaps more clear in this discussion of the color-changing version of the effect. You can play around with an interactive demonstration of the phenomenon here (although that never worked too well for me).