ON Semiconductor Corporation (ON) Earnings Call Transcript & Summary

Hello everyone, and welcome to today's technology webinar on overcoming challenging lighting conditions with eHDR, brought to you by onsemi. I'm Kyli Miller with onsemi, and I'll be your moderator today. In this webinar, we will discuss onsemi's AR0822, an innovative image sensor that produces extraordinarily high-quality 4K video at 60 frames per second. At the end of this webinar, we'll be holding a Q&A session to answer any questions you may have. [Operator Instructions] This webinar will be recorded and posted at onsemi.com. You will be notified via e-mail when the recording is available. Now let's meet today's presenter. Steve Harris, works out onsemi as Director of Marketing for the Image Sensors Group. Let's get started with our webinar.

All right. Thank you, Kyli. I appreciate the introduction. Hey, everybody. So just kind of a quick background on me, Steve Harris here and Director of Marketing for the Industrial and Commercial Sensing Division. I've been in this role actually now for close to a year. Prior to that, actually, I was working on image sensors for automotive for about 8 years. And then even before that, I was still working on since it was for industrial and commercial. So kind of coming back full circle here. We have a pretty good range of ideas of the different applications within image-sensing between industrial and automotive. And in all honesty, these are probably 2 of the areas where HDR is very critical. So I'll start it off by just kind of giving a little bit of a teaser here. So I don't know how many people are familiar with ARRI. But ARRI is a leading cinematography camera in Hollywood. And this is public information, but the ARRI cameras used onsemi for their image sensor. And I want to say, I believe, don't quote me specifically, but maybe the past 8 academy awards have been awarded to ARRI. Certainly, you can see here on the right, the best picture went to everything everywhere all at once. That was still with an ARRI camera. And even these already cameras have allowed us to even achieve it in. So they actually were gracious enough to help us receive an Emmy for technology. Now the reason that I mentioned this, and this is kind of in small print here is that really the details behind the award of the Emmy is talking about one of the techniques actually that we use to achieve HDR. So specifically, the cinematography was actually credited for HDR performance. So I wanted to mention that. And if you have a look at the long list of movies that are out there, impressive newbies that are out there. And again, I think this is actually kind of a small list compared to like the total ARRI that uses onsemi. But you watch any one of these bills basically in that HTR and low-light performance is what you're seeing coming from our camera from our sensor, of course, with the ARRI specialized processing.  So I just kind of wanted to put that upfront because this right here is what we call the sidecar. So the reason it's called the side of the chart is this looks a little bit like a xylophone. But basically, in the cinematography and digital camera world, the way that you measure dynamic range is called Xbox . And how many different Xbox you have really determines the dynamic range of the sensor. This particular sensor that move being used had about 17 F-stops. And that's about 100 DB of dynamic range. So this is just kind of translating this language of that stop over into dynamic range, and I believe a stop was about 6 DB. The way this chart works is that there's a light that shines behind it, and then I believe they have kind of a shutter and the filter to the block out light. And so you can see it doesn't actually quite accurately show how intense it is, but you do get intense light here and then it goes all the way down even darker. And I think there are charts to even go out to 26, right? This is a really good, useful tool for measuring dynamic range. Okay. So just before I get started talking about what our eHDR is embedded HDR. Just a really kind of quick primer here on what dynamic range is. So the textbook definition of dynamic range is really the full well capacity of the pixel over the read noise. So that's how we measure it. There's an MDA 1288 standard. I believe that uses a similar calculation, but they do the full level over SNRI. So there are different ways of calculating this, but I think if you look from sensor to sensor across even different companies, generally, I think the way the dynamic range is measured is the bowel capacity over the regions. So that's kind of the textbook definition. And the need for HDR, the biggest need for HDR is really with scenes and I'll show some pictures here a little bit later, it seems that we have right lights and dark areas within the same scene. And this chart here on the bottom kind of represents what the lux level or the light level is in certain lighting conditions anywhere from just deep Starlight to intense sun. So generally speaking, when you're an outdoor, they're going to run into kind of conditions where you've got some really dark light potentially you need that and you've got especially Bright light. So these markets like security cameras, automotive cameras, industrial cameras, et cetera, that are outdoors, really need this technology.  But I'll even add to it, and I'm going to do something here for effect. But right now, I'm sitting next to a window. And I just opened the shutter to the window. If I were to move my head over a little bit, you can see the saturation happening in my head on my head, right? And the reason that that's happening is because this particular PC web camera does not have the dynamic range that's capable of producing a good image. So hopefully, especially as video collaboration grows, we start to see maybe even HDR going into our PC webcams. But that right there was just even a classics, I'm not outdoor of just being indoor and showing you kind of what the effect is of having dark and bright in the same scene, and you just lose a lot of detail. That was kind of illustrated an example, live there for you, but Here's kind of an example for you that is taking in an image, right? So somebody is taking a picture outside from their window. Now this seem right here, you get a lot of good details in the dark regions. And it's probably because this is a pretty long exposure for the sensor. But what you're missing here completely is any details in the sky. Now conversely, now maybe we have a shorter exposure. So a shorter exposure capturing less light, okay, now we're able to make out the details here in the Sky region, but now all of the details here in the dark region have completely disappeared. And so what you'd ideally like is something like this, this is an HDR image on the bottom right. One of the keys to, especially when you're outdoors, that you can always look for that generally tends to be a good kind of guiding principle and dynamic range of Scott. A lot of sensors have a really difficult time achieving this kind of blue. And you'll see even more crisp lose guys later, but that's always a good indicator kind of to me of sensors that have pretty decent dynamic range, especially when there's sunlight of course and here's another example. I think a lot of people have seen the classic funnel. So this is just more of a kind of creative way of showing it.  But controls are a really good illustration again of teams that have dark and bright, extreme dark and extreme right within the same seat. And again, the way the sensor handles that could be that you would have a short exposure where you're getting some of these details in the darker regions in the case of day for the tunnel. But now you're missing the areas in the outside of completely saturated, I can't see any detail. Again, conversely here, now we have a shorter closure. We see all that detail out beside the tunnel, but we don't see anything inside the phone. So obviously, you can think of automotive in this case, where you're going through a tunnel, a car in front of you has brakes and they're just outside of the tunnel and you have some autonomous driving program. can't see the car, and of course, that could be potentially catastrophic, right? So dynamic range really is very crucial in that type of application. But another application, I could speak to here is maybe a person walks into a store. There's a lot of light behind a window or whatever Glassdoor that they walked through. And all of a sudden, now they're walking into more of a dark environment. You're losing detail in that bright line. You may not identify the person who actually came in and maybe did something. So it's really key again. HDR, I think, is really key for a lot of applications. I mentioned automotive, I mentioned security camera, any type of robotics that are outdoors, even indoors where you don't have well-defined lighting, meaning that you may have windows or really dark areas and windows. So this is where dynamic range is really good. There I would say the only application potentially that isn't really as focused as much on dynamic range would be fixed lighting.  So there are a lot of applications out there that are more based on active illumination from infrared LEDs or VCSELs. And in those cases, right, you know the lighting, it's both fine. Maybe in that case, in that circumstance, you don't really need dynamic range. But again, I would argue even just as I mentioned here, just kind of consumer camera this PC web camera, right? I'm going to be doing a lot of video conferencing and I don't have the ability to turn the shutter off on the window, this is going to be my image during the entire presentation. So you can even make the argument even on some of these consumer applications that are endorsed that is needed all right. So kind of transitioning here. I want to talk a little bit about the approach is to actually achieve the high dynamic range.  So generally speaking, there are kind of 3 axis' for different ways that you can solve the dynamic range issue. The first one would be kind of in the spatial bumming referring to space, size, exact type of photo diode, et cetera. And I'll show some images later on what I mean by that. But basically, there's architectures that we have, what we call split pixel, a large photodiode and a smaller photodiode basically all in one. And again, I'll show you back that kind of image to you here soon. There's the ability to do it in time domain. So this is where maybe you have multiple different exposures over the course of time. Again, I'll show that one. And I won't talk about it too much in this presentation. But there is an approach as well that basically, you just you increase the full well capacity of the pixel. So maybe you just have a really, really big pixel. Maybe you have a small pixie but you have a lot of capacity due to it. So it's able to store more charter. And so again, I won't talk about it much here, but what I will say is that I think this is actually a really unique approach to achieving HDR. And it's something that we've used in our products. One of our product families we call [ hyodusa ], which you can google on the web, and we'll take you through that site onsemi.com. It uses this kind of approach. This approach is really good because now you have a single exposure or maybe dual exposure, where you're achieving a really, really high dynamic range. And I think this approach is also potentially really good for LED flicker that you may see again in applications like automotive or break like, tail life, traffic exercise, again, are using LEDs. So this is a really unique approach, a good approach. And certainly, a lot of our automotive sensors use this and potentially a lot of our industrial and commercial sensors of the future may start moving towards this approach as well. But absent that, if you look at the industrial and commercial kind of industry right now, there are really 2 popular approaches that are mostly used. The third approach I mentioned right now is kind of emerging in automotive. We'll start seeing it elsewhere, but the need in automotive was critical for LED so the first approach is the one that we're going to focus on, and it correlates very well to the HDR, embedded HDR technology that we have. So I'll focus a little bit more on this multi-exposure approach. But basically the multi-exposure approach is you have just a single photodiode and you take multiple different exposures sequentially. And in that, you're basically taking effectively different images at high light at low light and maybe at some mid-light level. And then what you do is you merge all those images, you sandwich them together and you produce a nice high dynamic range image. Now the other approach that I was talking about is what we call split diodes, pixel approach. And this is actually a really good x-ray down to the detail of what this actually looks like. Effectively, within the same VCSEL what you have is a micro wins that's covering a part of the photodiode, and this would be the larger photodiode and then you have a smaller photodiodes. So in this case, right, you have the ability between this to expose for longer without saturating, therefore, creating getting the darker parts. And here, you're driving the lighter parts and you're combining them together. Now this is a patent actually that we had beating over the way back to 2009 when we were actually Micron. So I actually come from the Aptina micron kind of legacy, which is now, of course, broader onsemi. So this is a technology that we've had for quite some time in the patent on for quite some time. Now I'll talk about it here. There are some reasons why there's drawbacks. Now the drawback on the multi-exposure approach, really, the biggest major drawback on the multi-exposure approach is, again, you're taking images 3 points in time. And so if you have extremely fast moving objects, you may get motion artifacts as a result of that because, again, you're taking at different points in time. There are ways for us to mitigate that, and that's why I put it in yellow. And I would argue, too, that in order to see those artifacts, you really have to have a fast-moving object. And again, I'll use automotive classic camp. Our automotive sensors have been using this multi-exposure approach, geez, for 10 years maybe. And recently, because of flick, we may have moved to a different approach. But even within automotive, the motion artifacts were manageable. It's really based on angular velocity, of course, even if you've got a car in front of you changing lanes, may not be fast enough even to produce motion artifacts. Now certainly, if you had a car coming orthogonally to you, that would be an issue. But that might not be the case for why you're actually -- you may have a camera on the side of the car that's actually looking for something more agent. So again, I think the motion artifacts are something that can be dealt with, but really need to be used just for very fast, fast moving objects. And then going over to the split diode, you can just see right here that you're kind of breaking up, you're not using light efficiently, right? You have gaps in between the large and the smaller photodiode and their micro lenses, right? So it's just there's some little light loss in quantum efficiency because you're not making the best efficient use of space here. And then you may have color artifacts as well because obviously, there's going to be quantum efficiency differences here between the larger photodiode and smaller photodiode. And then in multicamera systems, especially ones were unique stitch images, the consistency with the smaller photodiodes and the calibration may be different from sensor to sensor. And so it could make it really difficult for 3D mapping type applications or applications with multicameras where you need to stitch. So high level, that's a part of the reason we really didn't choose this approach to our high dynamic range imaging. And here's just another area right here with the splits kind of architecture. Is that you do get a lot of crosstalk as well. So you have that smaller photodiode. In this case, we're showing the red here, but it's sitting very, very close to the blue and the green pixels, right? And so now you're actually getting crosstalk between the pixels. And the other element here is that unlike the prior approach that I showed here, where you have exposures, and you can change the ratios between these exposure, 16x, 32x,very configurable. Here, you have a really a fixed ratio between the large photodiode and the smaller photodiode. And of course, you can do things with game to kind of maybe modify that ratio a little bit, but you just don't have the flexibility that you do in the multi-day exposure approach. And here's actually a video that actually one of our competitors put out there. It's actually again, I mentioned the classic tunnel gauge here, right. Even within their own actual video that they put out there, they actually illustrated one of the issues again with this crosstalk. And you can see that kind of here in this transition region, especially from the light area to the dark area. So again, this is kind of the reason we, as onsemi, have not gone down that route. So what it looks like actually, and I'm showing here at dual exposure, meaning you have basically 2 images that you're doing here sequentially. The way it really works, right, is that in a rolling shutter sensor here, you have a sequential readout. So you have a longer T1 and that longer T1 is, of course, trying to capture the dark areas. And then you have another sequentially with it, AT2 or TIME-2 that's a smaller piece of time smaller exposure or shorter exposure, meaning that you're able to capture the highlights. And so we have sensors, potentially that can go all the way out to 2 exposures. So the more exposure you have, obviously, more dynamic range you achieved, but also the more complex it is to kind of put those images together. So this is just a pretty easy example of a 2 exposure HDR approach. And the way it actually looks as we're actually exposing and reading out pixels through to the sensor. And here's just kind of a classic to exposure to HDR, where I was mentioning here again, you have that first long exposure and you have a second short exposure. And then all of a sudden, we combine into an image now where you're able to get details on that dark area and also details in the white area. So you can see it's taking this nice blue that I was talking about here from the shorter exposure, and that's where it is right here. And then you can see in the darker regions that you're taking from the other longer exposure, where you don't have the detail here, right? So it's just a very simplistic way of looking at how multi-exposure HDR works. And by the way, to for the most part, most of your phones probably use a very similar approach to what I'm mentioning here. Now what about motion, right? I mentioned motion. Motion depending on application, may not even be an issue. I would say security surveillance, if it's in a convenience store, your doorbell camera, people don't move fast enough generally to really create artifacts that are going to create issues. Again, I would argue in automotive, there are certain circumstances, but not all the time. But here's a case where we have a spinning fan. We put it against an orange background so we could kind of see the contrast here. And what you're seeing on the left is a result of motion artifact between the multi exposures. Now we have a technology that we perfected and worked over, I believe, 10 years on, which we're now calling intelligent linearization. And linearization is basically the ability for us to stick together these 3 images and get back to an image that has really, really high dynamic range, much like what you see here, right? So in this case, the linearization is the combination again of these.  Now we have a process in our linearization where we do motion compensation and we're able to really mitigate the effects of the emotion. So even when we do have motion like this ban is something that we can actually correct for. And again, going back to it, right? So now you kind of have the best of both worlds. You've got a lot -- you have multi exposure, so you're getting a total flexibility in the amount of ratio that you can choose between the exposures, you're achieving up to really high dynamic ratings. You don't have to worry about the crosstalk. And oh, by the way, that one thing you did have to worry about, which is motion, we've got a methodology for us to correct it. So again, leading to why this is such a good approach.  Now again, I really think a really unique approach that we do in automotive is this high capacity for well type approach. And that really, again, is something that may be a multi-exposure could not solve. And that is LED flicker where you have LEDs that are on for a period of time under a short period of time. And if you don't if it misses in one exposure or misses it may be in multi-exposure you got 3 and 2 exposures. If the capture it in one, it's going to create some difficulty. So LED flicker I think, is probably the only seam drawback that you may receive from this approach. And even with LED flicker LED or sorry, intelligent linearization has some ways of mitigating that as well. So even in that case, right? -- there's ways we can get around it. Just another example right here, we just have something sitting on a vibration play. And you can even see on this vibration plate, is obviously moving pretty quickly here, and you can see those artifacts as a result of motion when we don't have the motion compensation or intelligent linearization turned on. And here again, sitting on this vibration plate here. The device is moving, but you're not seeing those artifacts that you were seeing. Again, just speaking of the back that you can compensate for the motion. And then finally, one of the things that we do, again, as a part of our intelligent linearization as we smooth out regions of transition. So as I was mentioning, you have multiple different exposures. And it produces the curve it kind of looks like a shark tooth or a triangle wave of kind of waveform. when you switch from more exposure to the next. And so you do get S&R drops when you switch when you take one exposure to the next exposure, right? And sometimes we'll at cause some issues in the transitional regions between those exposures. Again, the intelligent linearization that we actually build on embed onto the sensor is able to compensate for that. Okay. So I mentioned again, the kind of benefits of this multi-exposure approach, especially against some competing type architectures. And I mentioned some of the things that we perfected in terms of dealing with the significant drawback, which is the motion. And again, motion could happen may not even be a case in a particular application. So there's also another reason that we want to embed the sole sensor. The intelligent linearization, we know our sensor. We know how to do the linearization of the HDR. All of that is great. But what I'll say is that when you start to have higher resolutions and rather than doing the HDR and the sensor, we decide, you know what, we're going to output those 3 exposures to a processor. And then the processor can do the linearization. Well, if you think about it, you're basically in that case tripling your frame rate you're tripling the speed of which that information needs to come across in the 3 exposure case or in the 2 exposure case doubling your or exposure case quadrupling it. And now as you start to get the really high resolutions, if we want to produce 120 DB HDR at 4K 30 frames a second, we're going to have to go up to 12 gigabits per second per camera. And there's not a link in them. I mean there's very few things, I'll say, in the world that can really add a long distance handle that. So where cameras are remote to the processor, certainly, extremely difficult if the HDR is not done on the sensor. Automotive is a great case for that. But if you start looking at some of these smart warehouses and you start looking at some of these smart buildings, even smart homes, you have many cameras that are located either in close proximity to each other or maybe even far away from each other and maybe instead of having a processor sitting next to a sensor on every sin camera, you may want to do that sensorized. And again, if you don't do this HDR linearization on the sensor, it's virtually almost impossible for it to happen on higher resolutions over the cabling that we have. Now you could also mention here, well, hey, why don't you use compression? Well, in a lot of applications, again, compression, especially the decompression part of the date to produce artifacts. And you don't know necessarily where those artifacts are coming from. So if you're in some sort of application that has ERP and machine learning, et cetera, or anything that's basically what we would call computer vision, not pretty human eye. You don't want those artifacts, especially the artifacts that you don't know and are maybe not deterministic. So you really do need the raw bandwidth. And then I'll just even add to that, right? You may have the case where a processor just does not have the horsepower. Maybe it's even sitting next to the sensor, right?  But maybe we have a 20 megapixel camera. Okay. Now does the first of physical link work? Do we have are we able to output 20 megapixels at, let's say, multiply 30 frames a second, 3 exposure 90 frames per second. It's a lot of data. And maybe the processor doesn't even have the capabilities to do it. So this is really what kind of pushes the HDR linearization onto the sense of this embedded HDR. So multi-camera systems, high-resolution systems where the interfaces really haven't stepped up and processors just purely do not have the resources. And again, what I'm showing here, by the way, was no HDR and HDR-enabled sensors literally over a 5 times improvement. So again, it's just it's a really strong case to be made that the HDR should be done in the sensor. And maybe even you could in theory you move from an architecture with a security can that doesn't have a processor on board because you know you may have it centralized. So that's kind of what it gives you the ability to do here. Okay.  So just kind of shortly here talking about some of the other benefits, I did mention in my verbiage before that it really reduces the need for the processor, the system on chip to combine these multiple exposures. So multi-camera systems are really high-resolution systems. You don't want to use your SoC bandwidth to focus on doing multi-exposure to HDR because you're going to sacrifice potentially that horsepower in other areas that's needed for the actual ISP or the processor, right? So it just really optimizes the system resource by allowing the processor to focus on other tasks instead of linearizing the multi exposure. And I mentioned again back here, but there's no need any more for kind of a sensor SoC co-location for bandwidth purposes because we're now like in this case, right here, 5 to 6 times kind of improvement on the length. So we can use lower speed cabling and lower-speed connectors, et cetera. So cost savings as well by doing that. And that's my third bullet point right here. Cable in these systems can be quite extensive. The shield the twisted pair, if you print the cables, you could break them. So the ability for you to actually use less cable, less expensive cabling is actually a system benefit. Same with connectors as well.  And then the other thing that I mentioned here is the motion compensation. This is our sensor, we're taking these multi exposures. We have the ability through our ISP teams and our design teams to compensate for that motion, and we can do it right at the sensor look. So you really do get some of the best motion compensation you can get because it's done by the actual sensor vendor itself. So a lot of benefits again of this kind of approach of embedded HDR. And then some and honestly, in some cases, almost mandatory. Okay. So not really tying all of this together. What I wanted to do is talk a little bit about our latest 4K sensor. We're calling the AR0822 here that we just introduced in press release at Embedded World in March. So a little bit over, I think, about a month ago now. So the 822 uses this embedded HDR approach that I spent some time talking about here. Going back to the xylophone chart slide that I showed earlier, I was showing that, that ARRI camera achieved a 100 of dynamic ranch.  And here it is winning all of these awards, all of these Academy awards, have gotten us any for high dynamic range. And again, it's roughly at about 100DB. Of course, they're working on improving that. We are, et cetera. But this particular sensor that I'm showing here can achieve up to 120 DB. So all of those beautiful films and pictures that you saw with basically 100DB type sensor, now we've cranked it up to 20, even higher. And I think you could even make a case in automotive, and I actually don't even think you could make a case, I think you need to have maybe even higher than 120 DB is good, but some of our automotive sensors will push even higher up into the 150 DB range. So really, really necessary. Now for, I would say, in scenarios like I'm presenting here, if you get to 120 DB, this is not going to be an issue, what I'm doing right now. It really is mostly going to be in your staring directly into Sunlight. And again, I think you still be pretty good images. But when you come to safety and an automotive, we maybe not acceptable. Now having said that, the sensor is itself capable of doing that off exposure, what we call Wild West, and that's outputting those exposures to a processor. So it is flexible in that manner. But of course, I think I'm here to talk about the embedded HDR. I think that's something that obviously we recommend. And again, we have all these years of experience in how to deal with motion during this linearization process and even potential to kind of mitigate some market back as a result of LED. The sensor itself is a 4K sensor can run a later mode up to 6 a second in HDR mode, the 3 exposures up to 30 frames a second, built on a 2.0 Micron pixel is he's really, really good, low light. The sensor itself is what we call stack sensor. And this is where the array sits on top in a wafer and then you have a bonded wafer beneath it that is dealing with the logic, the ADCs and anything digital that we need to do. And so when you have that, you literally aside from the fact that you have some packaging tolerances here that you see, you're achieving almost the best-in-class sizes that you can get. So from a size perspective, as a one-of 1.8 inch sensor here, they're getting the smallest size that you could possibly get by using this stacked architecture. And the other thing, again, I'll mention here is just power. One of the things we've really been focusing on, especially as we go into battery type applications or there's multiple cameras, and we really want to start saving power. An area that we focus on is power. 460, 400 milliwatt we can go down by 2x, potentially 30 frames a second. And we have another line of sensors even more optimized for power, maybe with some trade-offs and some other areas. We do have a wide portfolio here. Another key kind of interesting technology, I didn't spend much time on it here to talk about because this is mostly an HDR presentation is a function that we call Wake-on motion. But basically, the sensor itself has the ability when it gets certain different light levels from pixel variation will basically be able to take itself from a very low power, low resolution node. And basically send an alert to a processor that has detected motion and turn it off.  So it's a really good power saving, unique kind of feature that we've built in here. The applications that I mentioned here, and these are all applications that we do have customers for and we're seeing I made the case for the home security camera be the case for the professional security camera, certainly, body cameras now with police and other industries are becoming popular. They're going to be outdoor, right? So they need that kind of dynamic range. Machine vision in cases where you have factories that don't have controlled light, maybe you want to take a really piece of expensive machinery and move it into another warehouse. You don't know what the light is going to look like in another warehouse. You want color images and you need the HDR and the sensor kind of gives you the cover to do that.  Dash cams, I mentioned, this is more of a commercial type application. It even though it's automotive, it's sitting on the dash of the car. Now we'll say this a lot of the OEMs are starting to integrate some of these dash cameras. But as camera is still a very popular application. A lot of people have insurance rates, for example, that may decrease if they have a dash camera because it's going to record potentially any accident that may happen, right? So there are a lot of good uses here for these dash cameras, especially insurances and accidents, et cetera, or even just gelling for your own personal purposes. And surprisingly enough, like I said here, I showed it again earlier, and I won't do it again, but the note put camera. Video collaboration is becoming huge. And I can't tell you how many times I've said in a conference room, and I've been here just on my web camera just like I am right now, and I've got a colleague that's sitting across from me, and that colleagues using their web camera. Well, what if we had a bar here that had high enough resolution and you had a wide enough field of view, but that person could actually sit next to me, and we both could be talking on the same stream, right? So that's one for the resolution. But again, the HDR I pointed to. And what I'm presenting, I don't know where I'm presenting what the light conditions are going to look like when I'm on the road. So again, having this ability, even on the PC web cameras is something I think that hopefully, we start seeing. And certainly, within videoconferencing, I've seen a lot of requests now for HDR. And that means more kind of a discrete kind of camera that's just purely focused on doing video conferencing for a room, for example, right? So the applications for this are numerous, as I've mentioned, HDR critical in every single one of the applications. And especially this kind of integrated embedded approach, really, really important when you have higher resolutions because the multi-exposure approach where it's not the linearization is not done a sensor is just really, really difficult to achieve.  And of course, we're always going to keep pushing the boundaries, right? We're going to keep pushing resolutions higher. And the rest of the world doesn't necessarily follow at the pace of the sensor. We may be able to put out a sensor or a year, every 2 years. The processors may not be able to keep up pace with that, right? They have a longer life cycle. Maybe some of the interface techniques have a longer life cycle and maybe you're off generation. And so I think you can always make a case here that sure, at some point, we're going to get there with the cabling. We're going to get there with the processors. But that's going to be a generation or 2 may be removed. If you really want to stay in kind of the tip of the technology curve. So that is what I had here kind of summarizing. So the beginning case here, HCR really needed for any application with mixed lighting and especially for harsh lighting conditions where you've got bright sun land or you've got really dark regions with price on land. There's all techniques. As I mentioned, there's the pixel approach. There's the multi-exposure approach and even some of our automotive sensors, the big full well capacity approach and there's many different ways to achieve it. Multi exposure seems to be the most popular within the industry, but again, what you see from a lot of different sensors in bet even with the multi-exposure approach, they're outputting those multi exposures onto a processor who's doing that combination. And again, with what I talked about with the resolutions and the speeds, et cetera, just may not be achievable, multi-camera system, et cetera. It maybe doesn't have the kind of knowledge that we've acquired for the motion, right, doing the 10 years, 10-plus years of motion compensation may not have that kind of ability or capability on the SoC. And then so that's where this HDR comes in and really provides value to you, the customer or you the partner, whoever by solving these issues with basically doing this HCR linearization on the sensor. And again, I point back to this 822 since a great amazing sensor, lot-like capabilities. But again, just thinking back to some of these cameras that are winning Academy Awards and in these brilliant pictures, the sensors achieving 120 DB. So it really is quite remarkable. And we're going to continue to push the limits even higher, of course, in certain generations moving forward. We made our honestly, as I would be lighting starts to become more popular, maybe we started adopting kind of that charge, the main type HDR that I talked about within our sensor here even for industrial. So right now, we're really kind of at the tip of the technology curve for the industrial for these high resolutions and we're solving issues at the system levels. But of course, there's always room for improvement. And onsemi, we continually push the limits of HDR. That's one of our biggest specialties in terms of sensors is the safer capability. And we're going to keep pushing higher. But I would say, from where we're at right now and the applications that we're serving here. It's had a really, really good level. Okay. So now I'm 45 minutes into the presentation here. Maybe I'll throw it back to you, Kyli, then for any questions that may be coming from the chat here.

All right. Well, thank you for the excellent presentation. We have received a number of questions, and we'll jump right in. If you would like to submit any questions, just type them into the chat box. All right. So some of the questions that we received. Here we are. Out of the 3 techniques that you described, which ones do or does AR0822 used to achieve eHDR of 120 DB.

Yes. So of the techniques, it is the 3 exposure technique. So it's the just go back to the kind of the diagram here really quick. It's this approach in the time domain, the multi-exposure approach. And again, the reason that we did it is we just saw too many challenges with the split by pixel producing artifacts and cross talk. So yes, it is this multi-exposure approach. All right.

Next question. Do you have experience with commanding 140 DB into 12 bits and then decompounding into 20 bits. Some ISP are limited to 20 bits. What are the effects of that dynamic range limitation in the ISP...

Yes. So a terrific question again. And I'll explain the companion for people. Obviously, the question is an educated question. I'll speak to the companion for people that maybe have not heard that term order not familiar with it. companioning is basically kind of a loss is lateritic compression. So the person who asked the question mentioned 140 dB of dynamic range. Okay, that's going to require 24 bits to get that up. Now we have the ability to [ unrate ] can pan that down to 12 bits and put it over an interface at a much slower speed. Now what happens on the other end of the spectrum here is that the ISP eats decomp, that's the person mentioned. And they do that kind of through a lookup table. Now if the processing element itself doesn't have above 20 DB. So or sorry, not 20, 20 bits, then you are going to be limited effectively to 120 DB. There may be some things that they may be able to utilize in the 24 bits. But for the most part, you're kind of limited to the 20 bit. But what I've seen, and it's a little bit more on the tip of the technology derive again, SoCs being on a little bit longer life cycle, I have seen that there are certain SoCs that are out there that are now actually expanding to 24-bit pipes. It's a really good question, really educated question.

All right. Thank you for your response. Next, will this tech be built into an LVDS model?

I'll say the answer is yes. I can't reveal too many details about that. But the answer is yes. We do have devices on our road map that will have that capability.

Wonderful. All right. Is tone mapping and color correction happened on sensor 2.

Yes So we have yes, so we have a process, call LTM, but basically it's adapted local tone mapping, what it stands for. And yes, we do have sensors with that capability on board. It does not necessarily need to be done by processor.

All right. So a bit longer, starts off with hello. First of all, thank you for the excellent presentation. And then the question is about the multi-exposure approach. Can the user select a number of images to take the exposure parameters for each image.

Yes. Exactly. You can. You could select -- this is the great thing about the approach of the multi exposure. You could choose any one of these, right? You can go up to 3 exposures, if you like maybe you don't need through exposure. Maybe you need 100 DB, you'll need 2 exposures. So you through exposure. Maybe you want to alternate. These sensors have what we call context switching, you can switch from one frame that's a linear frame to one frameless an HDR frame. You could then to an ADP frame from 4K and maybe have a native 4K another mode. The other question here is, do you have the ability to change the ratio Absolutely. That's the benefit. The big benefit of this approach is that you do have that ability to change those ratios for your lighting conditions. And again, that's just something that is very fixed by the loss of physics here that you can't do. On the Pixel architecture, so yes, everything is completely adjustable, selectable , very flexible in the sensor configurations here.  Now by the way, I actually started before I finish that question, though, I do want to kind of plan to see here, we do have sensors that you're going to start seeing them even get up into the 140 DB, maybe even 150 higher type range. And once you get to that range, you almost start to cover basically all lighting conditions in one set of settings. So if you really even get up to 140 DB, there may be the case where you don't need to do anything, right? You just need it. We call it set it and forget it. So that is something that I would be aware of in the future is that once that dynamic range on the sense of doing, you may not have to adjust any of the ratios or the auto exposure, which is really interesting.

All right. Wonderful. Next question. How is the AR0822 image sensor better than similar image sensors in the market today?

Yes. So I mentioned, obviously, the 120 DB HDR. You don't see a lot of 4K sensors with that capability, especially not in the industrial and the consumer space. I mentioned that. The 2.0 micron pixel BSI architecture, we've measured against competitors and is best-in-class and little light. It had that wake on motion feature that I mentioned briefly that allows a sensor to go into a low-power mode and basically only wake up when it detects motion. So very critical for doorbell camera type operations. Many of the competitor sensors you see for industrial-owned commercial, other times, they're repurposed for mobile sensors. And mobile doesn't have some of the usage cases that we do here. You're more concerned about clicking on a single photo most likely than running a camera 100% of the time. So we are building in function specific to industrial. So another key differentiator.

Wonderful. All right. Do you plan to have other variants of the sensor with eHDR?

A good question, good question. Yes. Answer is yes. We I can't reveal too many details, but keep your eyes posted 8 is a really popular resolution, but there's still a lot of need for 1080p. And down the road, I think you can see it's NAP sensor from us and maybe even higher than 8 mg without getting into too much on the specifics. One reason I do like that question though is because the way that onsemi has been defining our sensors is using a common architecture. So what I mean by that is that you may have basically the exact same sensor, it just has a different pixel different resolution. So if you're working on a process here, I'll throw out Qualcomm or NCP, and you've written a driver for one of our sensors, let's say, it's the A22 here.  Now all of a sudden, on synergy releases in 6 months, a different resolution. Guess what, you don't need to write that driver again. All of the functions in the most part because this is a family registered almost completely identical. You of course, you're going to have to tune for the different optics you're going to have to tune for the different resolution, but that family-based approach that one takes really allows you to scale your platform using one common architecture. So keep your eyes posted, you'll start to see sensors on our road map here publicly soon. If you are already a customer of ours and you do have an NDA, we'd be happy to show you our road map and show you a little bit more detail than what I gave you there.

All right. Last question. Do you have an application not or white paper related to this webinar?

Yes. Yes. So if you search for 22 in our product family, it will come up. So all of the really detailed details behind this part are going to be in the white paper. I kept it fairly high level here because of the range of audience that we have, judging from the questions we have some smart audience. But I kept at kind of a higher level, and we really do want to get into a deeper technical level, 2 things, right? We do have a white paper. Second thing is just people in contact with our sales. We'd be more than happy to even step through some of the specifics here with you as a customer.

All right. Wonderful. Thank you, again, very much, Steve. These are all the questions that we have for today. So on behalf of onsemi, I would like to thank everyone for attending, and I wish you a nice rest of your day.

All right. Thank you, everybody, for similar all right. Cheers. All right.