Projection isn’t projecting full size

Is anybody else having this issue? The projector displays in the middle, but on the outside there is blank projection. I don’t know the correct terms. But It’s like an outside bezel that doesn’t display anything. I uploaded some pics to show you. This happens when using the projector by itself and also with an hdmi device.

There will always be some light leakage around the image (the light engine containing the DMD unfortunately isn’t coated in Vantablack internally).

The video black-level looks a bit high though in the letterboxing, is this a cropped video file or does it have the black bars encoded in the video?

The black bars are there from the aspect ratio which is fine. But all around the image you can see there’s still excess run off. The image seems to be cropped. But on the initial start up, the white philips pico display image projects the full image. Then it goes to a crop when things are played. Here’s a YouTube 1080 video and you can see how the image sits in the middle as if it’s cropped.

You did not set a digital zoom, corner correction, right?

Interesting! The white image most likely is the splash screen from the DLPC/FPGA, that will show the full extent of the DMD. It’s possible that the SoC has some underscan going on by default (or some digital zoom), you should definitely be able to show an image as large as the white screen. V keystone might mess with image size though (something that the splash screen would ignore completely). Will an HDMI source eliminate the black border?

That’s correct. No digital zoom and I’ve tried turning the auto keystone off as well. It still stays the same.

The image stays the same using an HDMI source and also just using the projector via the apps.

But is it cropped or down scaled? I have seen photos of a test pattern and it did not look like cropped.

1 Like

That’s really odd. It would be great if you can get a test pattern on there to determine whether the image is cropped or scaled, it’s going to be one or the other as we know the white Picopix logo can’t use more mirrors than the 960x540 (I guess the splash happens without the actuator operating, anybody got a high-res picture of it?)

You have a 16x9 projector, but your video content is at the ultrawide aspect ratio of 21x9, so course you will have “bars” at the top and bottom …

This is normal … and exactly the same way it would be displayed on your 16x9 TV … why would you expect it to be different on the PPM?

If you look closely, apart from the format letterboxing there’s another dimly lighted frame like area all around the image. What @Meemario is saying is that the splash screen at boot goes to cover this area as well, which seems strange.

1 Like

In the image provided by @Meemario, there does appear to be a “frame” around the image … as mentioned, perhaps a crop that has embedded black areas?

I’ll have to check with my PPM with multiple files to see if it is the same.

1 Like

That would be most interesting to see the results from. Compare the size of the power-up splash screen with the content (the splash screen is supposedly filling the entire DMD).

A closeup of the actual pixels in the splash screen would also be interesting to see (to determine whether it’s actually 960x540 or 1920x1080)

1 Like

Based on my experience with the Anker Nebula Mars 2, it had the same area around the image and from my experiments it is a sort of margin that they add in for vertical keystone correction. Even the TI documentation for the chips shows that not all mirrors are available actively at the borders. I think it is DMD/DLP chip based as TI advertises this as one of their features. However, what’s truly curious is how the initial screen manages to show a full white image.

My guess is that at startup, the boot image is loaded as the white label splash image from the DLP chip and since the chip is booting up, it has not enabled the vertical keystone part of its code. A quick way to verify this would be to quickly change the angle of the PPM during the time that the Philips logo is being displayed and see if that triggers keystone correction.

1 Like

That’s interesting, and odd!

My experience with the non-pico DMDs is that they only have the specified amount of mirrors, no additional ones. As long as v keystone is disabled the image uses the full DMD. When v keystone is applied the image will shrink as expected, but always touching the left and right edge somewhere. If there’s a margin added it means that the image is already scaled down a bit (or cropped, the jury is still out on that one it seems).

I guess we’ll have to get some hi-res photos and start counting visible pixels, an image of the splash screen would be awesome!

And yeah, the “auto v keystone” isn’t part of the DLP chipset as far as I’ve seen, that’s got to be coming from code running on the Android side, you just feed the DLP chipset a desired keystone amount.

I think there might be two parts to the “border” issue:

According to the DLP230NP datasheet there are 20 mirrors per side (doubled to 40 when shifting) which will always be off. That means that those should be outside the white area of the splash screen, but they will very much still look like “black pixels”. For a pixel shifted image that means that you have the equivalent of 40 pixels black added to all four sides of the 1920x1080 image. That’s over 4% additional mirrors on the X axis and over 7% on the Y axis!

I learned something new today, maybe this Pond Of Micromirror exists on every DMD…


This is very interesting, actually.
If this was to accomodate vertical keystone, it would mean these mirrors are actually active in some way. If they are, either these are additional pixels, or they’re normally using less than 960x540 pixels for the active area.
Even the excerpt from the documentation that @wernerj has posted doesn’t make it any clearer where these mirrors are and what their purpose is. “structurally or electrically prevented from tilting towards the ON state” would suggest they won’t be able to display the white splash, also.

(edit: other DMD datasheets contain the text “Pond of Micromirrors omitted for clarity” next to the equivalent illustration)


Yeah there’s no way the mirrors would be active only for something like a splash screen, it would be hilarious if TI had implemented something like that though.

The POM of the DLP9500 0.95" 1920x1080p DMD only has a POM 10 mirrors wide per side but on the other hand the mirrors are twice the size in that DMD, so it would indicate that they need a 100um border. I’m wondering if they perform some sort of destructive test using those mirrors during production of the DMD and then permanently disable them as they claim that they are “partially functional". Very little information seems to be available (although digging through the TI patents may very well reveal some interesting info).

A typical configuration includes a uniform band of border micromirrors (referred to as a “pond”) around the perimeter of the a central core of active micromirrors in the array. These border micromirrors are not user-addressable but are tilted into the −12 deg. (“OFF”) position once power has been applied to the device. An example known implementation has 10 border mirrors on each side of a 912×1140 active mirror array. The border mirrors serve the purpose of directing stray incident light out of the field of view of the projected light path.


Very interesting! That explains why there would be 20 mirrors on each side on the tiny DMD, maintaining the ~100um border. I guess that’s the accuracy they require from the optical engine, it’s obvious that they want to avoid having to absorb too much light. Clever!

1 Like