ON Semiconductor Corporation (ON) Earnings Call Transcript & Summary

Hello, everyone, and welcome to today's technology webinar, Automotive Image Sensor Fundamentals - Auto Exposure Control. I'm Kyli Miller with onsemi and I'll be your moderator today. In today's webinar, Gulwinder Randhawa [indiscernible] Hyperlux' simple camera operation will simplify exposure control offers stable alpha images without reciting -- without requiring SOC ISP frame-by-frame exposure, game controls. We will explore the recent advancements and trends in auto exposure control for onsemi-image sensors, including the set-it and forget feature of the Hyperlux, which simplifies sensor performance for all lighting scenes with a single setting. At the end of the webinar, we'll be moving to Q&A session to answer any questions you may have. During the presentation, you could take your answers into the chat box below the slide and we'll answer them at the end of the webinar. You can also download a copy of today's presentation and other resources and the related content. This webinar will be recorded and be available on demand. You'll be notified via e-mail once recording is available. Now let's meet today's presenter. Gulwinder Randhawa is a principal field application engineer at onsemi based in the U.K., covering European automotive Tier 1 for support and developing cutting-edge CML imaging systems to facilitate autonomous driving solutions. Gulwinder has over 17 years of experience working on camera systems, including CMOS image sensors, optics, image, video signal processing pipelines, characterization in camera drivers in R&D and application. Now let's get start with the webinar.

Hello. Good morning, good afternoon, everybody. Thanks for joining us. In this webinar, I'm going to talk about autoexposure control and its fundamentals in high dynamic range image sensors. As an agenda, we start with the basics of exposure and then cover different technical parameters, which directly impact exposure time in rolling sensors and image signal processor, in short-form ISP. I will cover how image quality engineers test robustness of the port exposure control and digital camera systems, give you a brief overview of different type of HDR techniques for the image sensors and how different parameters in the HDR sensors can directly impact the control of the exposure. Also, we learn how no exposure control is needed for automotive applications with Set it and Forget it feature from onsemi, Hyperlux family sensor. So what exposure is? So exposure is the amount of light needed to properly expose the image sensor to create induce or video over a period of time. So I'm going to switch this one slide. Actually, there is some technical difficulty there slide, this one before the other one. So these are the 3 photos actually I have captured from on top of the Edinburgh Castle looking towards the city. The first image is overexposed. The second is perfectly exposed. And you can see the third one is overexposed. You can clearly see the difference. I will come back on the histogram later. So for easy understanding of the exposure, let's take the water tap example where knob is shutter, tap is aperture and water is amount of light entered in the camera and time to fill the bucket is shutter speed. So when we turn the tap, aperture on, the time taken for the water to collect shutter speed is shorter. So when tap is turned up more water will flow and time taken for the water to collect shortens or in other words, shutter speed become faster. On the other hand, when the tap is turned down, less water flows in, the time taken for water to collect becomes longer, shutter speed become shorter, so easy and simple. Let's cover this parameter in bit more detail in next slide. Starting with the exposure triangle from water tap diagram, we can conclude that the exposure is determined by 3 main variables: aperture, shutter speed and ISO value. ISO is also called gain at the sensing. Shutter speed measures how long the actual shutter stays open to capture a photograph at the correct exposure with measures typically in millisecond or second. The long shutter produces brighter image, shorter speed results in dark image. The typically long shutter speed grades, motion blur with moving up subject in the same. And at what the aperture, it's actually blade in the lens, which control how much light enters. So it is a physical measure of how open lenses expressed as F number or F stop. It also affects the depth of field. In the diagram here, opening up the aperture F 1.4 will allow more light in, creating a shallow depth. Closing down the depth, this top to F11 will allow less lighting, creating a wider depth field. So that means everything in the frame is in focus. ISO or gain is how sensitive the camera is to the light. Typically, it's value range from a ISO 100 to 6,400 in digital cameras, each incremental ISO number up or down, represent a doubling or halving of the sensor sensitivity to light. For higher ISO, the output image will be brighter at the expense of extra noise in the imaging. It is typically used in low light scenes. Now let's talk about histogram, which is a bar graph. It shows how the tonal range of the image is distributed between pure black and pure white. The best way to see if a picture is correctly exposed is to look at its histogram as someone in these 3 example pictures here. The first one shows the correctly exposed picture. The histogram tail off to nothing at the highlight and telling us that only fewer pixel, white pixels are in the picture. At the shadow and the drop off is quite steep, and there are a few pixels in the photo that are pure black. So basically, we have captured as much as information possible with this thing. The second picture is badly under exposed. The pixel distribution points down the shadow end, and there are virtually no pixel within value greater than [indiscernible]. Finally, in the overexposed version, you can see that we have a large number of pixels with the value of 255 or white. So basically, details in the highlights is lost forever. So let's look at the exposure value, which is simply a way to combine shutter speed and an aperture to a single value. This calculation is done through this formula where N is the F number and T is the shutter speed. For a bright mid-day scene you want to have TV like plus 15 or 16 to avoid too much light with the aperture and shutter speed combination. For the dark scenes, we need much lower value like minus 5 VV in order to avoid under further. For depth of field, it refers to the amount of an image, which is in focus at any given time. So to put it simply, the shallow depth of field is a small portion of a short is in focus, whereas deep depth of field has a large area in focus. Depth of field is affected by lens aperture and focusing distance. In the camera system, the image sensor received instant light photons that is focused through a lens or optics. The camera optics is the eye of the camera and the lens we choose to have a big impact on the outcome of our photos from the camera. In image sensor when light strikes the lens of a camera, it captures that light, convert it into electronic signals and then transmit it to [indiscernible] device processor, which transform the electronic signal into a digital image. Now let's talk about the image signal processor or ISP. It's a dedicated processor that converts a raw image data from sensor into a high quality colored output image through various processes inside it, like land sharing correction, Gamma correction, color correction, noise reduction, auto exposure and auto white balance and few other blocks. For some application like rearview camera system, image sensors do come integrated with ISPs, we are some -- whereas some other application like ADAS, or advanced driver assistance system and machine vision required stand-alone sensors with uncompressed and unprocessed raw digital image. In such system, actually, the external ISP processor is used for viewing part, which is either a discrete ISP or embedded ISP in the SoC system on chip processors. So let's look at the -- how the exposure time has been normally referred. So it's typically referred as a shutter speed or integration time, major in milliseconds. So integration time is the sum of course integration time and fine integration time, typical sensors. The course integration time is counted in a number of holders, whereas fine integration time is the fraction of row defined in units of the 6:00. Exposure control mechanism of CMOS sensor fall into 1 of 2 categories, either global shutter or rolling shutter. The global shutter appraised by exposing old pixel in the F2 area at a single instance in time, whereas rolling shutter exposes different areas of the pixel array at different points in time. So the exposed timing is different line by line with reset and readout happening at shifted times. Looking at the analog gain function. It's designed to multiply pixel value by the analog game than reading the pixel values. So using the game control, we can boost the analog signal from a pixel before conversion into digital value by ADC, which is analog to digital converter. So gain control drawback is that it increases noise. Similarly, digital gain is applied after the ADC conversion, it's basically a multiplication of the digital data, image digital data after ADC. Mentioned over here. Now let's look at the gamma correction and tone mapping, which are indirectly affect explosive performance. So as you -- they know, the light intensity is nonlinear in the display. Gamma correction is the technique used to compensate for the nonlinear display characteristics of a device. It is used to properly display brightness on computer and other display screens. So basically, pixel has linear response, and our eyes have [indiscernible] response. So gamma put nonlinearity back into the image to look right. For tone mapping, most displaced videos have limited time range. For example, standard computer display uses 8-bit [indiscernible] channel to present image data. So this presents a problem for trying to view HDR image content that has 14 or 16 bits per color channel. So the main goal of tone mapping is to transform or compress higher-range input image data down to the range of the intended display medium while preserving contrast and color as you can see from these pictures, here. So after going through basic of exposure technical parameter, let's dig deeper into exposer algorithm, which typically involve 3 processes. So light metering. So this is typically done by either an external exposure detector or the camera sensor itself and for scene analysis using imaging metrics, brightness image techniques [indiscernible] of the scene. The best exposure can be estimated by adjusting the brightness based on the overall lighting value. And for image brightness correction by altering the lighting and shutter time settings, it makes sure that the appropriate quantity of light reaches to the image sensor. So based on the image stats and average luminous is calculated and compared to a reference level. Luminous is then adjusted by images to the exposure value as required. There are 2 approaches commonly used in implementing auto exposure metering in the optic center approach and electronic centering. So in optics centric approach, an independent sensor detects light and the exposure is adjusted by controlling the aperture or shutter time. This method is typically used in film cameras and in some digital camera systems. In electronic centric approach, the image sensor meters, the available light through the lens, the exposure adjusted it electronically by bearing the integration time of the sensor and the gain of the amplifier in the analog domain before the digitization. Okay. So this slide actually covering the auto exposure algorithm mixing matter, typically used in image sensors. So center weighted average metering takes the entire surface of the image into a consideration to adjust exposure. In partial metering, the middle of the image is more significant for exposure, especially when the subject is in the center. For spot metering, as name suggests only evaluate the light around your focus point, nothing else. Metrics metering, which is also known as valuated or smart metering, drives the frame up into separate zone and analyze all the light data available in this zone. By reading this information, it aims to produce a balanced exposure over the whole scene. So what happened to verify the auto exposure, test engineer to auto exposure testing in image quality labs with gain and expose function of light lab. So as lab images are typically captured under different illumination level, it is possible to plot the exposure and gain as a function of Lux as shown in this example. In the middle here, what I'm pointing on my screen, in this example, host switches sensors from 30 frames per second to 15 frames per second to utilize high exposure time of the sensor. In this case, it's around 60 millisecond for low light scene. So this is called night mode in some camera application. This type of test allow checking that the auto exposure algorithms are properly working as expected. This testing can be very complicated, consisting capture of thousands of images with a different combination of exposure, gains and lux levels. So it needs to be done in controlled environment for repeatability if any of the parameter needs to be retained. Normally, further field testing done to validate the auto exposure fully. So in some examples, external light is also used with the exposure function. Typical example is mobile phone camera auto exposure. In this example, gain and exposure as a function of light level, xenon flash is kicked at 300 Lux. So this type of implementation at another variable to verify the auto exposure functionality. So other example may be external infrared light trigger in cabin rather monitoring application where manual or autoexposure control function needs to be tested. So I'm going to cover a bit of about HDR image sensors for an audience who might want to learn more about these type of sensors before delving further into complexity of the auto exposure. So typically, the humanized can easily adapt to a wide range of elimination conditions. The typical bright range visible by our eyes within the fixing is about 10,000:1 or around 80 dB. So most image-sensor used in consumer cameras for short of capturing high dynamic range scenes because the pixel of these sensors have a linear response to light input and have limited pixel well capacity to store electronics, resulting in saturation before bright regions can be captured. So many HDR capture techniques have been developed to use in specialized security, automotive, industrial and military camera applications where the ability to capture high dynamic range scenes is important. For HDR sensor lenses are equally important to get high dynamic range to achieve full potential of the HDR sensors. So autoexposure parameter we discussed so far are equally valid for different type of HDR sensors. So here is a sematic diagram of a standard for transistor CMOS image sensor pixel circuit, which shows the role of conversion gain in the typical pixel design. So the basic 40 pixel architecture consists of a pinned photodiode reset transistor, a transfer gate to move charge from the photodiode to the floating diffusion sensor node [indiscernible] transistor. So the TX transfer gate enable or disable charge flow between the photodiode and the floating diffusion. So capacitance is connected in parallel with the floating diffusion node. So that more capacitors can be added to hold charge. So one transistor I'm pointing it to the diagram here called the dual conversion gain or DCG switch is added here above the capacitance. So when imaging in high light condition or bright light condition, the DCG switch is turn on, connect physical capacitor to the floating diffusion node. In this way, the large capacitance of the floating diffusion node is used to enable low conversion gain mode, which can handle a large amount of segment charge. And then in low light condition, the DCG signal is turned off, disconnecting the cap from the floating diffusion and enable a high conversion gain mode, which can be used as an extra analog gain inside the pixel. So the figure shows curve of signal in units of millivolt at the floating diffusion node versus light exposure in units of lux second. It is the measure of a sensitivity of an image sensor called responsivity. In the DCG sensors or dual conversion gain sensors, each pixel can upright either in high gain mode or low conversion gain mode as an adaptive gain investment. So in summary, high commercial gain mode is beneficial for capturing better detail in darker areas, whereas low conversion gain mode helps to reduce noise in brighter areas. Okay. Let's briefly talk about different type of HDR techniques being used in high dynamic range sensors. First one is multi-exposure approach of traditional HDR. This technique captured multiple exposure of the same scene at different exposure level. Typically, 3 exposures HDR sensor will capture short, medium and long exposure, where the short exposure image reserve details in the high light areas of the scene, while the long exposure image details in the shadow areas. So these multiple exposure have been processed and merged using specialized HDR algorithms to create a final combined HDR image. This technique helps further details in high contrasting and low-light conditions, resulting in image with improved dynamic range. The second one is split pixel. This technology involved dividing each pixel into large and small photodiodes or subpixels. The large photodiode has a higher sensitivity to light and the small photodiode has a lower sensitivity. So the large subpixel is used for long exposure capture and the small one for short exposure to extend the dynamic range. While split pixel technology improves dynamic range, it often comes at the cost of reduced sharpness and color accuracy. This is because each subpixel is smaller than traditional pixel and can exert in less spatial recognition. In overflow multi-exposure also called Super exposure, the HDR sensor extend dynamic range by increasing pixel full value capacity and lowering the noise mode. When photodiode is up, overflow charges saved in large in pixel capacitor, optimized for high dynamic range. This technology is used in onsemi Hyperlux family sensor, which achieved more than 120 dB of dynamic range that is flickered and motion architects free from super exposure and it extends further to 150 dB with this very short second exposure. For split pixel HDR sensor exposure merging, a small photodiode used to capture bright part of the scene and the large photodiode pixel flex more light therefore, exaggerate in brighter condition. For the HDR merging, there are 4, 12 bits readouts from each pixel. So first readout is large photodiode with high conversion gain. Second is the large photodiode with the low conversion gain. Third one will be the small photodiode with high conversion gain and fourth is small photodiode with low conversion gain. For each exposure, 4 readouts and are then combined and generated as a 24-bit values. So each of the -- as you see from this diagram, each of the blending shows maximum SNR level for each exposure. There will be SNR drop in exposure transition region. So this SNR transition limit is defined by the pixel full value capacity, the exposure ratio and the gain of the sensor. So the choice of exposure ratio, it will directly impact the capture above scene dynamic range and the SNR at the exposure transition. For example, higher exposure ratio capture high dynamic range scenes whereas lower exposure ratio has higher SNR at the transition region. There are different type of on-chip motion correction algorithms available on HDR sensors to get rid of promotion artifacts for low and high exposure ratio set ups. Now for super exposure plus T2 mode merging even high commercial gain readout is used for the low light region, followed by the E1 and E2 mid conversion gain and then T2. So this figure shows the relative signal versus elimination behavior of each readout. In E1 HCG or high conversion gain, it's a high analog gain read of photodiode, which is -- has optimal low light performance. In E1 MCG is a read of photodiode, which has short noise limited transition. So no noticeable drop from HCG or high commercial gain. E2 MCG overflow read is optimized for LED flicker range. The [indiscernible] T2 read optimized for full range exposure ratio to manage transition. So this technology extends dynamic range by increasing linear pixel capacity and lowering the noise lower. As I mentioned earlier, can achieve uptake 120 dB flicker free and 150 dB with second exposure. Now some image sensor contained on chip autoexposure control, this diagram shows typical autoexposure control, which is implemented as 3 main blocks, a stats calculation, a target selection and an exposure control system. A stat calculation block takes the user-specified regional fund trust and create a histogram for the host from which all relevant autoexposure stats are generated like AE mean. The exposure target selection block determines how much and in what direction to adjust the exposure relative to the current exposure value as per target mean value specified by the user. And the exposure control system output, the new integration time required. If enabled the analog and digital gains will also be selected. The control system will also monitor the document of the sensor and the user or host interacts with sensor AAC through I2C register interphase. ISVs have similar AA blocks with addition of tone mapping gamma correction and extra digital gain correction. Yes. So, so far, we have learned how camera exposure relies on different parameter of the sensors and factors which make it quite complex to accurately implement in camera system. And also, it requires significant effort in verification for reliable and accurate functionality of the camera exposure. So it is very complex autoexposure controls chain to implement it, host needs to set multiple autoexposure like T1 and T2 or T3 exposure, juggle with exposure ratio as per dynamic range or SNR at exposure transition. Also, the analog and digital gain needs to be adjusted in each frame as per the lighting condition of the scene. The low conversion gain and high conversion gain needs to be taken care as per the bright or dark environment. And most of the HDR sensors have pre and post HDR gains, which also need adjustment as part of auto control loop for white balance or control of brightness. So typically, the host needs to control up to 20 I2C registers for exposure control frame by frame. It's very complex. I hope you agree with that. So this is one of the main reason why many third-party ISP vendors never able to integrate HDR sensor to its best performance. So as a world leader for automotive sensor supplier onsemi has come up with innovative Set it and Forget it approach to simplify autoexposure control of HDR sensors or Hyperlux family. This feature is a result of many years of sensor HDR, AAC or autoexposure control implementation experience with onsemi and third-party ISPs. I have highlighted many of the features of set-it and forget-it approach on this slide. This slide is very busy, but I kept this information here for easy reference for you after the webinar. In summary, host doesn't need to retune frame by frame, the exposure and the gain controls. Autoexposure control doesn't need to worry about frame drops or frame exposed incorrectly. In these 2 pictures, you can see 2 type of frame drops and incorrect exposed frame issues solved with Hyperlux sensor control. This sensor can have same exposure setting for viewing and sensing which make them excellent choice for cameras used for ADAS and viewing application simultaneously. As per my understanding, no other HDR sensor had this functionality in the automotive market yet. So in summary, Hyperlux sensor offer, simple camera operation with stable out due to its Set it and Forget it exposure control. So as I mentioned, Hyperlux image sensor are first in the market with engine sensor that does not require autoexposure loop because we have the highest dynamic range in the automotive industry, which does not saturate in most practical situation. So sensor takes care of analog gain, high commercial gains, low commercial gain and simplify exposure control for the host. So literally 3 I2C are enough to cover a wide range of lighting condition in Hyperlux sensor exposure control. So now I'm going to play a couple of videos captured with -- the first one, this video is captured with typical HDR sensor. Their host control -- the exposure with adjusting exposure time gain and other discussed parameter when entering and exiting the [indiscernible] area. In this video, you can see dark and incorrect exposed frames where it is difficult to see, but it's happening in the scene. So that directly impacts automotive safety, where system can be blind and not able to see internal or outside impact machine vision, ADAS algorithm or even the human driver. So the video over here is captured with Hyperlux Set it and Forget it exposure control, where host has to do minimum in terms of exposure control due to simple camera operation with a stable output images, where you can see no frame is blind or dropped which make perfect choice of automotive ADAS and [indiscernible] application. So this clearly show the main benefits you can get out of the Set it and Forget it approach from the Hyperlux sensor family for the host to do the minimum for autoexposure control. This slide actually shows details of onsemi Hyperlux sensor family. It comes with 3.2 megapixel, 8.3 megapixel and other higher resolution sensors. These sensors contain 2.1 microns super exposure pixel with highest dynamic range of 150 dB and flicker-free operation over 120 dB dynamic range. Highest temperature stability at automotive temperature and it also offer world-class low light performance. This family also have option of cybersecurity available in 8 megapixel sensor called AR0823. You can contact your local sales office for more information on these sensors. So in conclusion, exposure control is clearly a crucial function in digital imaging system. Complex STR exposure control mistakes can limit the potential of any best HDR sensor. So it's very important to have similar approach what we have mentioned here in Hyperlux sensors Set it and Forget it using the state-of-the-art approach for human and machine vision automotive application. Thank you. I think this is the last slide of our presentation. Thanks for listening.

All right. Thank you for the excellent presentation. We have received a number of questions, so we'll just jump right in. If you'd still have to submit any questions, just type them into the chat box. All right. Let's see, first question nice video of Hyperlux Set it and Forget it difference. Are the videos available on onsemi's YouTube channel or are they on their website.

Yes. This video will be part of the presentation, which you will get after this webinar. Yes. So you will get those as part of the presentation or you can contact us separately, if you don't, and we can work on that.

Is there a detailed document describing the embedded stats in the footer of the output data.

Yes. We have actually -- we call it developer's guide. Each onsemi sensor have the full detail of embedded and that sticks on the data from the image sensors. So you can refer those developer's guide. If you don't have, please contact your local sales office. They can help you with that.

All right. What are the Johnson criteria for object detection on the road?

Yes, this actually. This is quite old type of a standard, which is for detection, recognition and identification. It's basically calculated based on how many pixels are necessary in order to make an accurate evaluation of offshore object. So the typically, actually, if I remember correctly, the Johnson criteria for detection is typically for 2 vertical pixels of the target to detect it, for recognition it's around 8 vertical pixel of the target and for the identification, 14 vertical pixels. From my personal experience, I believe this Johnson criteria is outdated. It doesn't provide the right measure of resolution for today's camera system. So we need a better way to fund the information required to calculate the expected camera performance. I think we are also working in P2020 to do such requirements.

Does ISP need HDR processing on video stream coming from Hyperlux sensor or is it streaming HDR process to [indiscernible]?

Yes. Good question, actually. So Hyperlux is HDR sensor, but it can offer you up to 26-bit depth to do the 150 dB, for example. But for the ISP to take advantage of all this high dynamic range, it needs to process that much data. It can handle that bit depth, then it has to possess to further pass into the space. So yes, you do need to have the ISP which has high-dynamic-range processing capability.

Then next question, how is low light performance of Hyperlux compared to the AR233.

Yes. We have done a number of studies on that AR233 is the 3-micron pixel sensor and Hyperlux is 2.1 micron Hyperlux pixel. So this -- when we compare it at the higher temperature, we can clearly see there is big benefit for the Hyperlux sensor families to have the better dark current performance. So it give the no-line performance better than the AR233 and in terms of the ambient temperature, we do see some difference between 233 because of this higher 3 megapixel with more light. We do have some minor difference on ambient light. But on the high temperature where you typically use your autoexposure camera, the Hyperlux sensors stands off low light.

Could you elaborate a little more on F number.

F number is actually, as I mentioned originally, actually, it tells you actually how wide your aperture is. So the bigger the F number, more wider your aperture is that means the more light can go in. So it's a physically bigger diameter. Obviously, when you have a bigger hole in front of lens, you expect more light to flow in. So obviously, that inversely impact the depth field of the sensor, which can give you the shallow depth. So for the higher F number, obviously your diameter is smaller, so that means the less light goes in, then obviously you have better depth of field for that sort of higher F numbers.

All right. What is the relation between image sensors modulation transfer function and aperture exposure tag.

Okay. This is an interesting question. So typically, I think you know that MTF is the ability of the optic system to resolve the contrast at a given resolution or special frequency. So different special frequency correspond to a different level of detail in an image. So what happened actually when you have a larger aperture, which means small F number, generally allow more light, high-frequency stills to be captured, and that contribute to higher MTF values. However, very large aperture maybe -- may also introduce abrasions that can affect MTF negatively. So the relationship between the image sensor MTF or purchase size and exposure time is complex and involves many trade-offs. So the choice of aperture and exposure time should be made considering the fact like depth of field, motion blur and sensor noise to achieve the desired balance between capturing defined details and maintaining the high image quality as you except from any high good MTF sensor. So typically, maybe different scenario may require different settings. So understanding the statistics of the lens and the sensor is crucial for making informed decision for your application.

Next one, does the autoexposure change based on CFA color, is the blue AE, for example, different than green AE.

Autoexposure typically won't have that much impact, mostly it will be the auto-wide balance where you need to have -- when you actually balance the RGB color, for example. But then if the QE of 1 channel is way too higher than the other than, obviously, you need to balance it that way. So that will mostly affect the wide balance. For autoexposure point of view, it won't impact that much. But then they are kind of interrelated to some extent here. So technically, wide balance required to do that sort of correction.

In your explanation of the Pixel schematic on Slide 15, you showed one cap CFD. Is there actually 2 caps, one for charge capture and one for overflow?

The one I was showing on the screen was the typical actually -- it wasn't Hyperlux pixel, so that if you have the overflow capacitor, then that will be there, yes. So the DCG switch is not shown in that particular figure actually. So -- but that's -- I was pointing to that picture that you need to add DCG switch to get that -- to work with the overflow question. So please do contact your -- or my e-mail should be on the screen. You can contact me. I can give more details on that if you need more information.

Can you explain more about tone mapping local versus global?

Okay. That's a good one. So typically, that means you already know that the tone mapping can be global or local depending what sort of application you have. So what happened actually when you have a global tone mapping, one tone curve is computed at and applied the same over the entire image. So it's all global across the whole image. But in local tone mapping, it calculates the tone curve based on the local spatial features or brightness level. So it's more depending on what sort of scene you are looking at with your camera. It's taking the more informed decision based on that scene contents rather than the applying the tone curve to the whole global image. So typically, for onsemi ISPs, we use local adaptive tone mapping, what we call the ALTM, adaptive local tone mapping.

Do you have a database of images for different light and conditions, which can be used to meet qualifications.

Okay. I understand question could actually -- so for qualification, I mean we do characterize our sensor. We have a number of different types of -- different teams actually, which works on the [indiscernible] of our sensors, where they capture a number of light scenarios, different types of environmental. So this sort of images have been captured and kept in our database to compare or to use as a reference for any other testing. So I not fully understand unless you are saying that the database for the third party as well. But internally, we do have these databases or all the sensors we test and then we use as a reference for the next [indiscernible] sensor for example.

Okay. You didn't specify whether it's internal or a third party or so. Next question, is there any test report based on P2020 standard for the [indiscernible].

Yes, we are part of the P2020 group, we have many of our engineers are part of those standards we are writing. For example, the noise measurement, also the flare is done by the engineers actually from onsemi. So we tend to -- even most of these sensor used for developing the P2020s is being qualified or tested in our labs as well in parallel with many of the third-party labs. So yes, we do reduce report as per the P2020 standard? So for the customer facing document if you are looking for the full P2020 report for Hyperlux sensor, currently that is not available purely because the P2020 still have some more work to do which is ongoing right now. And once that standard is published, then we will have more concrete reports available to share with our customer. But we also have EMV1288 reports, [indiscernible] reports for our sensors and [indiscernible].

When using variable exposure time, how do you prevent flickering due to [indiscernible] of artificial urban light?

It depends on for the -- depending which flicker you're referring here. If you're talking about the [indiscernible] flicker, then obviously, you tend to take multiple of 50-hertz and 60-hertz exposure for your application that can go flicker and for LED flicker actually typically for automotive application, you used 11 millisecond for example as one example. This is one of the other reason actually why the Set it and Forget it approach is quite convenient to work with the LED flicker mitigation, where you don't need to do number of variables exposure to use with it. So you can use just one or 2 different type of exposure, and that will work perfectly with your LED flicker mitigation.

Next question, sensors are targeted for automotive applications. Would these sensors be available for other markets.

I think it's best to -- I don't want to comment on that. But typically, we -- onsemi don't only do automotive, we do have industrial and consumer application sensors as well, which have similar technology being transported to there as well. So if you're just referring to just Hyperlux sensors, it might be able to use for other applications, but please contact your sales office for more information.

Are Hyperlux parameter going to be different for a human vision application versus a computer vision application?

Yes. I think this is one of the -- another key advantage of Hyperlux sensor actually. Typically, if you look at any automotive HDR sensors out there, mostly for [indiscernible] and machines using application, you have a totally different setup for exposure and gains or even other parameters like brightness control. But for Hyperlux, because we have very high capacitance, the [indiscernible] capacity there for 150 dB. So that can work in almost every standard condition we have in automotive. So for that sense, you don't really have a very big difference between the viewing and ADAS settings for the exposures. So this is the reason why it will make a perfect choice for viewing and ADAS applications.

How does gamma correction impact the [indiscernible] exposure response to varying lighting conditions.

So gamma actually, it's like a mostly how your display is being shown to you. For example, there are many different types of display available in the market. So gamma does actually -- it's just a composite for the nonlinear display characteristics. So what happened actually if you look -- if you take your display and look at in the bright light and then you look picture in low light, it will look very different or even in some cases, like for example, the mobile phone display [indiscernible] adaptive to the light environment. So that's where the gamma plays a role for the exposure to control the brightness so that it can be usually displayed properly on the screen. In other words, actually, because as I mentioned earlier, actually as well, the pixel -- the camera pixels has linear response, our eyes are different [indiscernible] to response. So what gamma does actually. It put the non-linearity back into the image. So it looks more pleasing to the eyes on the display.

We have time for couple of more questions. The Set it and Forget it exposure approach proprietary to onsemi's sensors..

Currently, actually, Hyperlux image sensor are first to the market with imaging sensor that does not require auto exposure loop this Set it and Forget it approach. One of the main reason is that because this -- we have the highest redundant range sensor in the automotive industry from Hyperlux image sensor. So personally, I believe the most other sensors will follow the similar exposure control approach, but no other sensor is available in the market yet. So this is one of the reasons why the Hyperlux sensors are unique with this Set it and Forget it approach. But it's a matter of time, I think other will follow the similar approach soon.

Last question. How do you integrate your HDR sensors exposure control loop with the third-party ISPs.

Yes. For the onsemi, we have dedicated [indiscernible] team with many, many years of sensor HDR autoexposure control loop implementation experience with many onsemi and third-party ISPs. So this team actually worked with most of the system-on-chip and ISP providers directly actually to integrate their onsemi image sensor. So what happened actually with their experience, they managed to get the best possible optimization of the high dynamic range and low light performance. So if you want to know more about any of the particular third-party ISP functionality with onsemi sensor, please contact your local onsemi sales office for more information.

All right. Well, thank you very much, Gulwinder. These are all the questions for today. Please take a minute to answer our short survey and let us know how we did. Don't forget to sign up for coming webinar to learn about our AC solutions. And on behalf of onsemi, I would like to thank everyone for attending, and I wish you have a nice rest of your day.

Thank you.