In my last blog I wrote about the power measurements on the ShowWX and a number of people asking “where did the all the ShowWX’s power go?”
Microvision wants you to think that they just point the lasers directly at their mirror and with next to no power loss the laser light is steered onto the screen, but the truth is anything but this. It turns out that there is considerable electronics consuming power to control the mirror and lasers and are significant light loosing optics required to make it work. In this blog, I we will take a peek behind the Microvision curtain.
First, Microvision has not published specifications on the power consumption of their mirror or other chips in their system. Second, I did opened a ShowWX to take a peek inside (see above), but I didn’t rip it apart to measure the current for each of the components. It was clear with 5W of power consumption and the very poor image (more on that in the next installment) that it wasn’t going to be a serious competitor to Syndiant, so there was no point in our spending the time and effort to do a detailed power evaluation. Even with these caveats, it is possible to get a reasonable understanding of the power consumption issues associated with laser beam scanning with the available information.
So per the above my number are not going to be “perfect” but I do believe them to be reasonable estimates. Microvision could clear this all up by publishing their actual numbers instead of their usual hand waving like “making a 40% improvement” without saying what part of the total power was improved by 40% and what was the starting point. I would welcome Microvision’s corrections with their actual numbers. Personally, I think it is the case that Microvision feels “it is better to remain silent and be thought a fool, than to open your mouth and remove all doubt.”
Above is figure 28 from patent application 20110234919 by Microvision. This block diagram outlines the major electronic components that would be required in a ShowWX (or ShowWX plus). As seen from the picture of the opened projector and in Fig. 28 below, there are a lot of components each of which is consuming power. Fig. 28 also gives some idea as to the complexity in driving a LBS. And the picture of the inside of the ShowWX demonstrates that all this takes up a lot of space (note there is a two PC board “sandwich” with all the circuitry inside the ShowWX).
Some people have made the point that the ShowWX is a standalone projector and that the power would go down a lot if it was embedded. In reality, there is not much from Fig. 28 that would go away. The “media module” in the ShowWX is only an analog RGB to digital converter and for WVGA resolution this should consume about 0.2 Watts. The battery and some of the power management might be reduced which might save another 0.2W to 0.5W. There might be a few other things but most of the rest of Fig. 28 would have to be there for an embedded LBS projector. So maybe out of the 5.5 Watts the ShowWX consumes, at the very most 1W might not be needed with embedding. That still leaves around 4.5W if this was to be embedded which is way too high for any realistic volume cell phone application.
Let’s start with the beam scanning mirror in the lower right of Fig. 28. To make the mirror scan the laser beam even roughly correctly requires actively driving mirror. The shorter the throw angle or the higher the resolution, the more the power goes up. Based off a published paper by Microvision from a few years back and datasheets from other makers of 1-D beam scanning mirrors, the power consumption of the Microvision mirror is about 0.3W to 0.5W (not exactly nothing). There are “free oscillating” mirrors that consume much less power but these don’t produce a good scan for making a projector.
Next in Fig. 28 there is the DSP and MEMs ASIC. The problem is that the mirror naturally wants to oscillate in a squiggly sinusoidal Lissajous pattern (see for example Microvision patent application 20090213040 ) which isn’t very good for generating video image. To somewhat straighten out the Lissajous pattern (it still is not nice straight lines — more on that next time) takes power going to the mirror and power to constantly be calculating and correcting the scanning process. The correction of the scanning process with the DSP and/or ASIC plus losses in the drive circuitry is probably taking about 0.5W. So with the mirror itself and the drive circuitry and control of the mirror alone there is about 0.75W 1W being used.
The next big block in Fig. 28 is the “Video Control Module” (VCM) which is where most of the power is going. Note that the HSYNC, VSYNC, and STATUS signals go from the MEMs Control Module to the VCM. The reason is that the incoming image has to be stored and processed and distorted to match the scan process of the MEMs mirror. Even with the active drive, the MEMs mirror does not move the laser beam in nice straight lines at a uniform speed. In fact it moves in curves at a non-uniform speed and the VCM’s job is to re-shape/transform the image so that after it goes through the MEMs scanning it looks similar to the original image.
The VCM takes the digitized RGB data and stores it in the SDRAM. I then processes/scales/transforms the image base on the distortion of the MEMs scanning process. It then feeds the reprocessed image to the laser drivers. All the read and writing to the DRAM and the processing by the ASIC/FPGA takes power, probably on the order of another 0.5W.
Next comes the power taken in driving the lasers. To make a LBS system work, the laser beam has to be modulated (intensity changed) at high speeds which consumes significant power. The lasers drivers are either analog (which consumes power) or have to be very high speed switching digital (which also consumes power) to give the various intensity levels for an image. Even when displaying “black” this circuitry has to be “idling” to be ready to turn on in a few nanoseconds and is consuming power. Likely 30% to 50% of the power going to the laser is being consumed in the laser drivers.
[Update 2011-12-22: The optics below show a polarization based combiner from the Microvision application. My understanding is that Microvision currently is using dichroic mirrors instead of beam splitters for parts 610 and 612 and a total internal reflectance prism in place of beam splitter 614 to first reflect the light into the mirror and then let it pass out of the projector. The light throughput for the dichroic mirror based combiner optics including the MEMs mirror is suppose to be about 60% which is the same number I used in the laser power calculations]
Now let’s get to the optics. I have show below Figures 20 which is a simplified diagram and Figure 6 showing an optical module from the patent application. To hear Microvision talk about it, you would think that only DLP and LCOS require optics and have losses from the optics.
The lasers 204, 206, and 208 have to have their beam shaped by lenses. 602, 604 and 606 each of which is probably losing about 1% of the light. Then note that in order to combine them in a single beam, they have to go through mirror 608 and beam splitters 610, and 612. There is roughly a 5% light loss in the mirror 5% to 10% in going through each beam splitter (note some lasers go through more than one beam splitter in the combiner). Then you have the beam splitter 614 with another 5% to 10% loss, the quarter wave plate 2002 and another 2 to 4% loss that directs the laser light the MEMs mirror 616 which has about a 15% reflectivity loss, then back through the beam splitter with another 5% to 10% loss. Taking all the optical losses together and only about 50% to 65% of the light from the lasers is going to make it out.
Finally, we have the losses from converting electrical energy to light energy in the lasers. The frequency doubled lasers were reportedly getting about 6% WPE. There is a lot of complicated math involving the wavelengths of the light the efficiency of the lasers for which I will use a spreadsheet to calculate the result assuming the lasers used by Microvision and about a 60% optical throughput from the optics. The lasers themselves are taking on the order of 0.7 Watts. remember that this number has to multiplied by about 1.3 to 1.5 to include the drivers for the lasers. So the lasers and drivers alone are consuming about 0.9W to 1W.
Add it all up, subtract off the little bit from the batter circuit and the video-in chip and there is about 4 to 5 Watts being used by the LBS including its electronics. Microvision can hand wave about saving 40% here and 20% there, but the problem is they have to save about 80% everywhere to get their power down to their “goal” of 1W.
Below is a figure taken from “Scanned Laser Pico projectors: Seeing the Big Picture (with a Small Device)” that shows a more simplified diagram than the one in Fig. 28 above.
Below is a top view of the inside of the ShowWX.
A typical Analog Devices analog RGB to Digital RGB converter AD9883A and it will take about 0.2W for WVGA resolution.
For anyone interested, I have added a picture showing the hottest point on the case. It was roughly just above the large ASIC/FPGA in the picture above: