I posted a uBeam promotional video earlier today, and a comment had a link to a presentation given by the current uBeam CEO at the OurCrowd summit a few months ago (OurCrowd were one of the larger funders for uBeam).
Most of it is unremarkable, with the usual marketing claims, but three slides have some interesting info - this is a post that gets reasonably technical, just warning, so a summary of my opinion of this video is:
- The uBeam proprietary transducers seem a variation on existing commercial devices
- Measured surface data shows the don't perform evenly
- They can't focus sound where they want to (2 meters target, 1.2m actual)
- The transducers may be highly directional, limiting steering in the array
- They will never deliver more than 1W to a phone, even under ideal conditions (10 hours to charge)
- The transducers are 5 to 10mm thick, likely doubling the thickness of a phone
- Estimated phone case cost - $150 minimum
These transducers seem to be the proprietary designs that are essentially (IMO) a variation on the standard Murata devices used in car parking sensors, with what looks to be a bar rather than a disk of piezoelectric material and then a cone shaped 'speaker' on the front. I expect the bar to be either prestressed or a bimorph layered component. The transducer diameter is large, appearing to be larger than a wavelength of sound in the ~40 kHz range, which would mean in phased array operation steering would be hard as you'd send energy in unwanted directions. The other image shows what I expect is a "heat map" of the vibrational characteristics of a single transducer in operation - where you are looking at the round face, and red indicates a lot of motion, green indicates little. If this is a magnitude map, then it indicates this transducer is not operating particularly well, as you'd want red evenly around the surface to get maximum power out for the area available. Now they may have chosen a bad result just to "throw us off" but I don't see what that would buy them, and this tells me that the transducers are, unsurprisingly, inefficient.
The next interesting images appear at around 3:22 and seem to show data on performance (not clear if simulated or measured). Axes on the left image are hard to read but seem to show a linear reduction of power with distance, though some of the markings on the y-axis indicate it may be logarithmic. The very interesting one is on the right and says "Focused Power @ 2 Meters", which is really weird because if you look at where the peak amplitude is it's around 1 to 1.25 meters, so not particularly good targeting there if you are looking to deliver power safely. What surprises me though is that there are no grating lobes, as mentioned above the spacing means in phased array application I'd expect some beams at an angle, but it's really clean.
Two possible reasons for this - one is that the transducers are highly directional and really only send energy forward, so there's just no power at angles to form strong grating lobes - the other is that they aren't steering or focusing at all, just driving all over the elements together and letting a 'natural focus' form, same as if you had a single large vibrating plate and not many individual elements. Were would that focus? Well, from the Y-axis it looks to be a 20cm wide panel, and an approximate distance for the near-field/far-field transition peak is (width^2/(4*wavelength)) then for 40 kHz (8.6mm wavelength in air) that gives a "focal peak" at 1.16 meters - amazingly right on top of where that peak shows up in the plot.
I don't think they are focusing at all, something essential if you want to carefully target devices for efficient and safe power delivery. If they want to prove me wrong there, the same plot as the one on the right, but steered at 45 degrees up or down, and placing the focus at various depths along the Z-axis, would make it clear. Not sure if they aren't focusing because they can't, or because showing that data would be ugly, but I'd like to know.
Last slide of interest starting at 3:59 or so, has power delivered versus distance. Now I'm skeptical of these numbers, but let's take the uBeam ones at face value (not the Next Gen ones of course, since they fired the entire team that would be working on them). What this claims is that a phone sized object can get around 1 Watt of usable power, and given that at 145 dB a phone gets around 3W of acoustic power incident, then that implies a ~30% reception efficiency, which is impressive if they can show that (to be clear, I don't think it's that high in reality). So impressive that I'm shocked there isn't a demo that makes that clear and proves the doubters like me wrong. Hmmmm. If you notice, that implies the next gen devices would have been 2x as efficient, or near 60%, given a 145 dB limit, which is an incredible feat and humanity is worse off for us not receiving this technology. Given 'circles in squares' leads to an effective usable area of around 79% max, that's getting to the limit of what's achievable (yes, I know hex layouts are more efficient, except in finite sized areas where the perimeter is a substantial fraction of the area)
More than that though, the numbers are just weird. The limit on usable power received is "Incident Acoustic Power Density x Area x Efficiency" and given the latter two are fixed, only the Incident Acoustic Power Density is the limit. That is set by the regulatory limit at 145 dB (really it's 115 dB, 1000x lower, but we're being nice) so 300 W/m^2, and the area of the transmitter. Let's say a 30 x 30 cm transmitter is used, that's around 30W acoustic transmitted and with really good focusing you can target most of that to the receiver. That means as you move further out, with a large transmitter you can maintain power at the receiver by using more transmitter area and so power received should remain constant for a period (when limited by 145dB ceiling), then dip when you run out of transmitter area to compensate for the increasing in-air losses. Indeed because of the natural focusing effect you want to be increasing the array used as you move the receiver further out. Of course, efficiency continuously decreases, and transmitter cost goes up, but that's 'hidden' in this plot. There's no reason for a plot like this not to assume a huge transmitter, so maybe they are admitting the can't steer/focus arbitrarily, or perhaps they are simplifying for the lay audience, or maybe it's just kinda nonsense.
The power received here is for a phone sized object, around 0.01 m^2, but an individual transducer is around 100th of that. That implies at point blank range the max power per transducer is around 10 mW, and barely 3 mW at 2 m, under ideal conditions. Note for IoT charging, you don't get much space.
Nothing on cost here - a Murata transducer in bulk is a bit under $1, so even assuming they do an awesome job and produce at 50c (doubtful, most likely they'll be more expensive), then a phone case has around 100 of these, so $50 in COGS before anything else like electronics, putting the case at $150 retail, minimum. Ouch.
More than that though, the numbers are just weird. The limit on usable power received is "Incident Acoustic Power Density x Area x Efficiency" and given the latter two are fixed, only the Incident Acoustic Power Density is the limit. That is set by the regulatory limit at 145 dB (really it's 115 dB, 1000x lower, but we're being nice) so 300 W/m^2, and the area of the transmitter. Let's say a 30 x 30 cm transmitter is used, that's around 30W acoustic transmitted and with really good focusing you can target most of that to the receiver. That means as you move further out, with a large transmitter you can maintain power at the receiver by using more transmitter area and so power received should remain constant for a period (when limited by 145dB ceiling), then dip when you run out of transmitter area to compensate for the increasing in-air losses. Indeed because of the natural focusing effect you want to be increasing the array used as you move the receiver further out. Of course, efficiency continuously decreases, and transmitter cost goes up, but that's 'hidden' in this plot. There's no reason for a plot like this not to assume a huge transmitter, so maybe they are admitting the can't steer/focus arbitrarily, or perhaps they are simplifying for the lay audience, or maybe it's just kinda nonsense.
The power received here is for a phone sized object, around 0.01 m^2, but an individual transducer is around 100th of that. That implies at point blank range the max power per transducer is around 10 mW, and barely 3 mW at 2 m, under ideal conditions. Note for IoT charging, you don't get much space.
Nothing on cost here - a Murata transducer in bulk is a bit under $1, so even assuming they do an awesome job and produce at 50c (doubtful, most likely they'll be more expensive), then a phone case has around 100 of these, so $50 in COGS before anything else like electronics, putting the case at $150 retail, minimum. Ouch.
So, nothing surprising here, but a few extra bits of information that we didn't have before that mostly confirm our existing expectations.
Very nice write up. I think they keep missing the key details from their presentations as they know it just doesn't work. Surprising how you can pull the wool over investors eyes for so long! It's quite impressive actually.
ReplyDelete