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

Factual Questions
1

I don’t know what to call this point of sale LED display unit, although it is similar to an older, single-color gadget called “silent radio.” It is a matrix of 7-high LEDS by about 100 wide, and each LED/dot can take on one of 3 colors. (Maybe each “dot” is a triple-LED behind a lens; I can’t tell.)

view 1
view 2
view 3

This device is in a convenience store near me, and continuously displays text messages with limited graphics. Nothing surprising so far, but here’s the illusion. When it scrolls to the left, to the human eye it appears that all the characters are distinctly leaning, with the bottom leading the animation and the top following.

Sorry I don’t have any video. I may have to take some if no one has seen this phenomena and/or can explain it. It’s in someone else’s store, and they might think it strange if I want to videorecord their sales counter. And I can see I will have problems with frame sync right off.

My question: since static pictures such as those above always show completely vertical lines even during animation, how does the illusion of the character tilt happen?

Pseudo-animation by blinking non-moving lights I can understand, but this is different from a theater marquee.

My first thought is that the bottom 3 or 4 dots are turned on ahead of the top 3 or 4, which would give a leaning look, albeit crudely staggered at the middle. But this is most certainly not the case, as shown by the static snapshots and by careful closeup examination by eye. The amount of lean appears to be slightly less than one column, as if there were some way to physically tilt each column (there’s not). It appears to lean evenly, that is, as if each LED were physically ~15% farther to the left than the one above it. The illusion stops instantly when the animation stops.

Just guessing from memory, it scrolls left at about 12-15 columns per second. Persistence of vision could have something to do with it.

I’ve never seen it scroll to the right. Some graphics scroll up, but they are very slow compared, and I don’t see anything odd there.

So what causes this illusion? Has anyone seen this device and does anyone know anything about it? Googling “silent radio” turns up many false leads and humorous sites, (“radio was silent until the invention of sound in 1920…”) but nothing I can sink my teeth into.

2

It appears to be tearing. You know, the top of the display moves faster or slower than the bottom part. This happens because it takes a finite amount of time to reset the lights, and apparently this finite time is pretty slow on this device. Every time the text is moved, the top moves before the bottom

As for the illusion–it’s just that your brain is combining the straight up and down text with the one pixel off text, and essentially averaging it out. It’s perceiving it as just a bit of motion blur.

I’m sure an enterprising Doper could demonstrate this with an animation.

3

But it doesn’t look like one pixel off. It looks like 1/10 pixel off for each row (and didn’t you mean the bottom moves before the top?).

It looks like at least 5 LEDs are on BETWEEN two columns, and at different amounts of “between”.

If what you are suggesting is possible, why doesn’t a theater marquee do that? I’ve never observed that on a theater, but I have seen these gadgets for many years and they all exhibit the same behavior. The color can’t be a factor, because the original silent radio was only red LEDs.

And if your theory is right, why does it always lean in the same direction?

I don’t see any delay in the on/off status of any one LED. It seems to be instant if you look at just one.

4

I’ve designed LED moving-message displays, and Big T is correct. Generally, the LEDs are “multiplexed” - there is only one row driver, and each row is connected to it sequentially. In the best case, each row is on for 1/nth of the time, n being the number of rows. The row driver has registers that latch the LED data for each LED in the row, and then are ready to receive the data for the next row, while the first one is being displayed. The time to write the data and update the latches is insignificant compared to the total display cycle time. The reason that incandescent marquees don’t show this effect as much (I’ve seen it) is: 1) The design of the sign, where each lamp may have it’s own driver, and 2) the thermal lag of the filament of the lamp, which tends to cause a distinct blurring of the character displayed (in the direction of motion).

5

So the bottom row is turned on first? I can see how that might cause a blinking/strobe effect (I can cause that with a LED clock by rapidly moving an open hand past it). But why the lean? How can a rapidly turned off/on light look like it’s coming from someplace to the left of where it actually is?

I can see a LED turned on at position 1, then on the next pass, position 2 (left one column), but it looks like some rows are at position 1.1, then 1.2, etc.

That gives me an idea. I’ll try creating a strobe effect with my fingers and see if that changes the illusion next time I’m at the store. Maybe I could mask off all but a single vertical column and see what that looks like. Hope I don’t get the store manager mad at me – I’ll buy a sixpack first.

6

Those stills look a lot like there is tearing to me. The top two rows in the “badger 5 drawing” image (#3) appear to be lagging behind the others. And you were correct, it must update from bottom to top. Some of the “doubling” of the lights in the middle may actually be an attempt to counteract the perceived leaning, although to really know you’d need more examples of the same letters at different times to see what’s being done.

7

I’m not sure the doubled dots are meaningful here; they may be just the font design. Notice how many perfectly vertical lines there are in some of the image samples at the same time as some “italic” type fonts exists. If the italic lean were an artifact, you would expect all LEDs in the same row to exhibit the same way (unless the camera shutter was super-fast and it caught a line in partial-update).

I have drawn an example of how it looks to my eye, here. It would be interesting to see if a video motion capture at various shutter speeds would look the way it does to a human eye, but my drawing clearly shows how it appears. I don’t see how the progressive update from bottom to top could cause this.

8

Your drawing is wrong.
What is happening is this:
(assume a single vertical line consisting of 5 LEDS is moving across the display from right to left). Times are 1/5 of the total refresh cycle, probably around 1/50 of a second. Rows are labeled from the bottom, 1-5
At t0, LED 1 is turned on.
At t1, LED 1 is turned off, and LED 2 is turned on.
At t2, LED 2 is turned off, and LED 3 is turned on.
At t3. LED 3 is turned off, and LED 4 is turned on.
At t4, LED 4 is turned off and LED 5 is turned on.
The cycle then repeats, with the next LED in the row. You can see that a) only one LED is on at any time, and b) the LEDs are staggered by 1/5 of a cycle. The stagger can be made quite small if you can refresh the driver fast enough, but it’s still present, unless you use a separate row driver for each row.

At t2 the

9

Which part, Fig 1 or 2? Fig 2 (on right) is exactly the way it appears to my eye.

I’m having a bit of trouble understanding your text chart. Let me redraw and label each LED.

10

Fig 1.

11

OK, I’ve redrawn the pic, labeling the rows & columns of the left group. If you refer to any one LED as R,C, like Row 1, Col 3 (1,3), then we can communicate better.

You might have to refresh the image to get the new one – it’s the same URL.

Tell me which R,C should be red and when, and I’ll draw a series of charts to match.

12

To a very quick eye, that would appear to be a dot moving from bottom to top, in 5 steps, in a vertical line, right?

13

Correct.
Now, do that very, very fast, and then move the line of LEDs over one. Keep doing that, and you get a line moving across the display, tilted slightly.

14

Why? If the display started at the top, would it appear to lean the other way? (I don’t question the moving part; it’s the lean that puzzles me.)

Do you think that would work if I make a video animation out of it, maybe 2 frames for each “on” dot?

15

Yes, it would lean the other way if the multiplexing started at the top.
Sure, try an animation. I suspect it will be easy to see the effect.

16

Here’s a WMV video file.

I only did two columns (it’s a lot of work this way). Each red dot is 2 frames at 1/30 sec per frame. The video editing program I use does not allow fewer than 2 frames for any image.

This animation may be too crude to determine if your theory is correct. LEDs light much faster than 2/30 of a second and even though you specified to turn one LED off before the next one came on, I think they may stay on longer than that in the actual display.

But within the limitations of my animation, I don’t see any leaning to any direction.

I fail to see why this theory should work. Just because a column of LEDs is lit sequentially vertically, why should the illusion be of fractional-column horizontal displacement?

17

I don’t know why Beowulff described it that way, as it doesn’t correspond with the sign. More than one LED is on at a time in the sign. What your animation should do is leave all lights in a row in the state they are in (on or off) and change them only once all other rows have been updated first.

What this results in is that the first picture you showed (with some lights in Col 1 and some in Col 2) is essentially a single frame of an animation. As it progresses, more lights in Col 1 become lit, and fewer in Col 2. It may be difficult to see it using only two columns, though.

18

Think about it: Your eye is tracking the LEDs as they scroll to the left. If we assume that your eye is tracking the center of the line, the bottom LED is slightly ahead of the top LED, since it came on and went off first.

19

It’s easier to understand if you only worry about a single LED in a row being on.
There is indeed more than one LED on at a time in the sign - but only one ROW on at any time.

20

I don’t think video animation is going to be fast enough to create this phenomenon. I estimate that each vertical line is re-formed 12-15 times a second, so that would be only a frame for each, not enough time to show 7 dots lighting individually.

Unless someone can come up with a better illustration, I am going to see if I can videorecord a few minutes of the device. Then I will be able to step single-frame to verify what the columns look like, and show it at normal speed to (hopefully) create the illusion. I may have to play with the shutter speed, though – too fast or too slow, and it won’t work.