An Inteview with Carl: The inventor of HDZero, the technology behind Sharkbyte

hi everyone um i'm here today with carl from divermath we're going to have a bit of a conversation around the hg0 technology and hopefully uh ask and answer some of those questions that maybe you you had about the technology but also maybe some questions that you didn't know you had and hopefully you'll find this interesting so i'm going to launch right into it now thank you so much carl for for joining me for this for this call i'm sure that everybody out there is going to be super interested to hear everything you have to say about hd0 and the technology that you have because it's it's really impressive so um before we jump into the technology itself um can i ask a little bit about your personal background you know where were you born where did you grow up where did you study and how did you come to to to be where you are today all right thank you chris i was a boy in china in 1972 so i'm almost a 49 50 years old so i get my all the education phd from china and in 2001 and in 2001 i after graduation i i went to silicon valley for for a job and stayed there for 20 years yeah so all my career is based on communication and video processing my first job is the uss technology i was doing the modem the modem the modem right the internet with the modem yeah the young guy doesn't know that but the older generations know the modern world what's a modern year after that i was doing the video decoder like the cbps exactly using it's still using in our lpv is a dominant as of today now i was doing the dbpt demodulator which is a digital video terrestrial for european that's cbt after that i was uh right i joined the intersil and you know i started a company in 2005 and the company was acquired by tech wheel attack war is is kind of very interesting the tech world that we i designed the chip is called tw a836 the chip is using inside the goggle for driving the oled display right that's right that that that's very interesting i i don't know that until i get into ap mpv industry uh yeah in 2018 that's about 10 years later after i designed that ship so in 2014 i started a company in united states right and how the branch is in china so in xi'an sending the indian manufacturing and changing but the company tv mass inc is a california-based company in fremont the same city as a tesla resident yeah so the background of myself fantastic oh thank you so much carl it's really interesting it sounds like you've just got an enormous wealth of experience in in this sort of area in the in the field of uh of video and digital communication so that's wonderful to hear about um can you tell me a little bit more about the the background of the company divermath so you said you founded it and it's based in in fremont california um but obviously you yourself and the company have um strong links in in um for manufacturing and things of that nature i guess in china is that right that's right so the company yeah so because i'm chinese and and most of the manufacturer of the pv i think it's the best on china so so tvms will do the same but i was not going to do the like the vtx vrx those kind of thing i was the only we we call is called ic a fabulous ic design company our product is cheap everybody use our chip to build up like the goggle right build up the vtx vrx that's my initial target i steal my target but hd zero is kind of very new nobody was going to risk themselves because the new technology and also because another digital solution is so strong so nobody want to compete with them so i have no choice but to design the vtx and the vis myself so that's i don't have a choice to do that if i chose if i if i do have the choice i will do what i could add that's design sailing a chip right so that's really your your core expertise is in i guess asic design is that right application specific integrated circuit design exactly i always talk to people i'm not more than a qualifier ask designer that's me fantastic and i think i think for for those for those people maybe who aren't aware the the application of digital video in fpv really does require specific silicon like specific integrated circuit design because because our requirements are so are so extreme in terms of like latency and and throughput and that sort of thing so um yeah that skill set's obviously super valuable in this field and um so you say you started off as a fabulous semiconductor company and certainly uh i i also work for a for a fabulous semiconductor company as well um based here in cambridge um so i know all i know all about that um so how did you come to sort of transition from that into products right it's kind of very weird so when i doing the low latency video link it's it's not for apv i don't know hv at all so i was back to 2015 2015. so i i bought a tier is htc vive so at that time i see i'm very curious about what is very very fun but there is a hdmi cable i mean underneath your head that's kind of very strange can i get because i'm engineer like i think can i get rid of that hdmi cable yeah i did a study and just just found out he's so tough because basically the 4k resolution and the 90 frame per second that's a lot of data through the wireless but i i i did the study i think the latency is a very important issue even using the cable right so it has to be even that cable is very expensive but because it's longer the latency from there so when you move your head and that the sensor will send the signal to the to the graphics card and a graphic card will render a radio a video and send to your cargo right this is the kind of loop so if the overall latency is over 30 milliseconds so people will feel dizzy so the move ahead is a picture didn't resound yeah the people were very busy so that's my initial package to do that low latency video link is not for a tv right right so i was doing that but so when i doing that design that i i talked to a lot of my advisor and classmates so because all co-workers all of them are video communication i say i want to do a very low latency 30 minute second everybody saying it's not possible that's why httc even the oculus the facebook they cannot have such a solution the wireless solution so especially the solution is on the six gigahertz so they right now they have for the yeah for the uwp right they can have a very wide frequency range they can do that but on the sub sub six giga they cannot do that so everybody say no but i think probably i need to try very differently so that's why i say the tssc it's kind of the technology i'm going to use because everybody know you you want to communicate in the video right you first do the completion when you do the ofdm stuff this this is reading in the textbook everybody knows what to do so as of today everybody knows what kind of the compression what's the compression ratio you are having so what's the ofdm what what their data rate what's your boundaries people are always asking me this kind of quality that for the hd0 is very different they don't have those kind of parameters so you ask me those parameters because you think i'm doing the compression this kind of thing but i did it if i did that you you cannot have the low fixed latency yeah fantastic well i think that that brings us on uh really nicely to uh just an overview of the the hd0 system so can you talk us through as a sort of summary of exactly how the hd0 system works from sort of the camera through the vtx to the vrx and then finally to the goggles just as a summary and then we'll go into each section in turn oh okay so so the video from the camera is ramiti and to the to my chip and we have to go to the meeting to get the yuv data then we do the video process i will note say compression but it does remove the redundancy of the video get the video data smaller then pass this data to the baseband and the baseband come out similar to the uh that everybody is the same to the to the to the radio frequency modulator and out to the pa yeah it's kind of this flow perfect and then when that signal arrives in the the hd zero goggles in the vrx is it a similar process in reverse uh this is a reverse direction right and yeah we guess that i think you know we do demodulate everything you know then pass the baseband and the baseband we will get something get the data and pass the data to the video processing again now after the video yeah it's just yeah thank you so i think i think that's a fantastic summary um so let's jump into i guess the the camera side of things um and i was doing a bit of research on the on the different cameras and my understanding is that um most analog cmos cameras and the dji cmos cameras all use a rolling shutter and i was interested to see that um hd0 the the cam gs used a global shutter could you could you talk to us a bit about uh the difference between rolling and global shutter whether whether one is better than the other or what your thoughts are on that uh right global shuttle definitely is a more advanced technology and it's it's very powerful when you like it as of today people scanning the qr code right with the camera even you fast moving so you your software will probably have the problem but with a global shutter you definitely will will make your life easier right and for the ipv when i first jumped into a pv i think global shutter is the best then i was at that time i was doing the fire floss i'm using the global shuttle but the global shuttle we will introduce a one-frame latency that's one of the things so so the rolling starter you the light gets in gets one line and the one line out right you get to the other line right maybe a few lines right but it's not a frame right so the global shutter you get a one score frame so so you get out when you store the other frames so basically the one frame latency and the global shadow but because apv is a high motion the close shuffle will help so we have benefit and it is advantage and disadvantage yeah yeah so it's so for for sort of the the lowest latency a a rolling shutter might give you slightly lower latency but the the global shutter is going to give you uh better capture of the motion that's right it's called blurry right so the the global shadow will give you very clear image without a blurry but the rolling started definitely if you see the props you always yeah they get bent they they bend that's right yeah yeah yeah right but the global shadow will remove that yeah perfect and so um is that is that the sort of direction that you're you're thinking that that your technology will go in the future um using more global shutter cameras or do you think you'll support both types or what what do you what do you see in the future i think i would support both types and the price was people are very sensitive to the price global shuttle is more expensive than much more expensive than rolling stock yeah fantastic the the other one is the frame rate so for example dji is 120 frames frame rate per second and that's another higher barrier to produce the qualified high frame rate camera sensors yeah not not so many companies can do that it's kind of expensive and and speaking about frame rate obviously you know there's the cost associated with the high frame rate um do you think that your technology you would apply your technology to different frame rates in the future um or do you think you know 60 frames per second what what do you what do you see in the future okay so when i first designed these i'm thinking about the cost so i think my system cannot use a high end like a high expensive cmos sensor we have to use the commodity right so that's my plan but my system does benefit if we have a higher frame rate because the people are looking for hammers higher frame rate right starting from the tv before the 60 internet now is fixed p right hundred ndp now is 240p so yeah so as you say it's really a cost thing the system can support higher frame rates but we all need to be aware that those sensors are much more expensive and and and that's going to be a cost that that will be more than a 60 frames per second sensor right the cost is higher and the availability is kind of a problem too right right so they're not always they're not always available and they they go out of stock quite quickly and things like that right yeah okay fantastic so um that uh that data that you capture from the the sensor um that is there's some there's some image processing that goes on in the camera i guess um with with a specific image processing chip is that right right and um does do you use a particularly special type of chip or um or do you use similar similar image processing to other fpv cameras in your system oh actually in my system i uh okay i'm i know something about the images signal processing most people i mean ipv they do but it's kind of scientific i'm not experts there and the wrong time even folks they're using the ethic to do this job we call isp right uh right the sony they have their own isp i think dji is very good at the isp2 they they design in earlier time probably they don't they don't have the isp chip but they are good at tuning it right to make it looks very nice but right now they are too big i think they they might have their own isp for their all the cameras but definitely the dji is very very strong to make the picture looks nice definitely that's for sure i think probably i'm not sure maybe yeah he's very strong but as you as you say there are there are several suppliers there are a few different suppliers of these image signal processing chips so it comes down to the camera manufacturers i guess to to find the the best available chip and to tune it correctly is that right yes perfect so um we'd have to i guess look into whether dji do actually have any proprietary um asic for image signal processing i'll uh i'll put something down in the video description if i if i can find any evidence of that later so when uh so now you've got the the processed image that's transported over a mippy link which is a digital video uh link and at this stage the video is uncompressed is that right that's right it's kind of the raw the raw data um so can you tell me a little bit about the the mipi link that you use any any particular uh specifications that you feel you can share about that okay before i jump that i want to make a little correction it's not called raw data when we say the raw data is kind of the cmos and sensor give it to isp this is the control data it's called basically it's got bear from it it's a there's a yeah yeah this is let's call raw data right yes so so the i the the video data after the isp is called iuv data the yuv data is really visible so it's a wirey data over the meeping link yeah the link there is a protocol is a committee csi csi2 right so the midp is consists of two things one is a file the other one is the controller the file basically is tell you is like a truck right but the controller is kind of what kind of the payload on the truck right these are two things so we have an eptx on the camera side and we have a media i ix on the our receiver side yeah so in our video transmitter we have a project chip let's do it that's the media ix yeah right and uh do you use several lanes in your in your mipi link to to get the data rate high enough oh that's right uh meepee is a very very good standard right because we are only at 720p so we actually can use two links all four links or four lanes sorry for like four yeah four lanes or two lengths that's fine it's because our our data rate is very small right compared to 4k yeah sure absolutely yeah yeah and i guess you you so you use uh for for your system you can use two or four lanes is that right or right we can use both right right that's why i said meteor has lots of parameters so i need to you know i'm following or drooling right and what's the color space what what kind of the yuv there are there so that's why we needed a handshake ic for that yeah absolutely well that that's really great because you know with if you support up to four mippy lanes you know that that potentially uh means that that the data rates could could theoretically go higher in the future i guess with with either higher frame rate 120 frames per second or or even higher resolution is that is that true yeah you are very smart and this is a very good apprenticeship the reason we choose falling falling means i mean the data rate is half of two lengths right so the lower the data rate meaning the mip cable can be longer right right so so this is the reason which was the four lanes right uh one of the reasons and the other reason for the later we can support a higher resolution yeah yeah oh fantastic so that's something that i didn't realize that you know you could you could run the mipi lanes at different rates to to achieve a different um to allow you to use longer wires and and achieve better signal to noise ratio i guess on the on the link itself yeah yeah yeah right okay awesome so we've got the yuv data that's been processed in the camera it's over the nippy link and then it gets to uh your vrx so your vtx module and uh and through the toshiba chip there what's the next stage okay so here is kind of very different this is my very different i i will know right i will try to explain as much as i can okay basically the other solution like wi-fi or dji they will compress the video the video the compressed little video will be the video stream is not kind of streams kind of a package right each package has a header has a lens has a payload kind of thing right and those things will pass through the best planets modulator so so here is a problem so the the the each beat is very important for that screen if you mess up one beat maybe your whole taxi is gone for example you have one bit wrong on your lane so the whole package will be wrong right we'll be very wrong so so my approach is very different but differently is i think about the video so the video basically each pixel is easy to understand is rgb right so so rgb aaa is 24 bit so i can tell you even you have only the msb like the r7 g7 b7 so you basically you will get the image so which means you only transmit three bits instead of 21b 24 bit so you are basically you get the basic image and you sort of get most of you get most of the the data um right and do you sorry what i want to say is so if you compress the data each bit is very important so if you do the video if you analyze the video data the video data actually someone important someone is not it's not equally important but i can say that right so i need this i i utilize this feature to design my file so which means the file mini yeah to design the file so for the very important data i will make sure it's i have more protection for that all right all right so for the last for the less important data i probably i would not have any protection for that so if my boundaries is not enough i just get rid of that the less important data right that makes sense right right so this is a very different with the other solution because i think not every bit is equally important and you you were saying uh you're talking about why uv data for people who who maybe aren't familiar with yuv the the why is the luminance isn't it of the pixel and then u and v are sort of the color data bits so i guess uh you you would view the y data as being more important let's say than the u or v data because the brightness of the pixel is what our eye is really sensitive to more than just the color right exactly and also the y the the bandwidth for y is much longer because yeah but the uv is kind of only uh only limited in low frequency so that's why it's easy to transmit yeah okay fantastic so it's sort of um in some ways you're you're taking the same approach that um uh you're taking the same approach to the importance of the different parts of the image um as in an analog video where again certain you know for analog video the why the y data is prioritized and and sort of uh sent with the maximum possible uh signal to noise and the the color data is sort of um the first to get lost in a sense so i guess you're you're taking a similar approach but with a digital with a digital signal rather than with uh with the analog video signal so that's really really interesting um so so once you've once you've got that stream of of data can you talk through how you would how you would process a frame of video you know if you if you receive a frame from the from the mippi link um does the frame arrive all at once oh yeah that's that's a very very good question so so we were talking about the video data itself it's kind of the skill image but video actually they have timing you are talking about the timing right yeah this is a very important topic so we are not only printing the video the pixels we need to transfer the video timing right otherwise you otherwise okay you get one you lost one video right you lost one friend so the other side doesn't know so you will have the freeze video frame right over there so that's a problem so but it is very difficult to transfer even a single bit over air if the signal noise ratio is pretty low so you need to have some way to get your video timey to the vix side this is the kind of those kind of this question is i i won't say the most important thing of this system but it's very important i think we we solve it pretty well because our frame rate is kind of even there's a snow noise and yeah but the frame rate is pretty constant it's because of that but thank you for asking this question yeah this is very important it's not only the yubi data and also the timing data to make sure that yeah the the frames are being displayed at the right at the right time on the other end otherwise you i guess you could get you could potentially get strange things happening where over time the the latency of the link is changing at the other end because of missed frames or or or lost data yeah yeah let's say thank you yeah very very interesting so um you you you get your your data from the mippi link and you say does that data across the nippy link arrive sort of line by line as as it arrives or do you get whole frames at a time how how do you transfer the data for for one for one frame right we we just get a few lines then we start crying because we want basically so if we check our vtrx there's no ddr in our vtx which means even we want the video data is huge the yv data is huge so you cannot save in the sap inside the inside of my chip much if i don't have memory inside right so so so in that case meaning we don't have any butter for the friend so we can spare only a few lines so you're you're just processing that video data the moment it arrives and transmitting it it's there's no there's no delay or any waiting because you've because there's nowhere to there's no where to wait if you like there's no there's no space that's right yeah yeah yeah that's the design purpose yeah the low latency you have buffer everywhere meaning it's not kind of low latency yeah if you have a buffer it's not going to be low latency because you've got data waiting around okay yeah that's that's really cool um so yeah it's transmitted um over the over uh as you said an fdm link um now when we spoke before uh you described your earth dm link and it's quite different from um like a typical wi-fi link can you can you talk about uh how it sort of differs from traditional wi-fi um this i think ui expert but yeah this is a very good question so when we talk about the ofdm osdn normally you need to have a c is called cofdm is called kodiro fdm so my my icon is aosdm so i i invented it the data comes in you you code it it's called channel coding yes yes you you get a one bte for example there is a convolutional coding the one be the kind one bit can become two b or seven b become a b it depends on the ecc the capability the correction this kind of thing right so when you say you have a 15 megabits per second a year but output is over 100 right right this is a one bit become two that's that's already 100 right and also there's another like a greater solomon encoding that's that's one eight eight one eight eight bytes it becomes two seven bytes that's another overhead right we add those kind of overhead before we transmit just for ecc just like some error happening we can't correct it right but we but we don't have it we don't have those kind of thing right this kind of thing like uh like the convolutional korean they have the buffer inside we need to get rid of that above uh dji open hd wi-fi is everybody is called c-o-f-g i can only i never combined ceo game i call myself ospn or asda aospm yeah so my not coded i yeah i'm not coded as a traditional way but as i said because my big stream comes in each big screen have different priority so i need to do different mapping for each kind of stream so right yeah so that's that's the quick part of my design so we have many priority data comes in with different priorities so how you map into the different constitution that's that's one of my quick job right yeah so um you actually you you actually treat data that comes in differently depending on whether you view it as high priority or low priority in terms of in terms of how you is that in terms of if how you encode it you know what modulation uh what modulation you use on the baseband or or where which channels you put it on how do you how do you treat the data differently oh yeah so so first uh yeah is a constitution mapping so you know you need you you need to look about that the constitution mapping then for example i'm not doing anything about the correction i do for the priority one especially the timing one i need to do the protection i need to do the channel code it probably is not a one piece become too big probably the one bit become more much more than two so because as i said the timing is very important i can lose the whole frame data but i cannot lose the timing yeah so that's a very important one so yeah right it's kind of yeah so that's how you do it you you sort of as you say you convolve the the important data to protect it and make it easier to decode at the other end accurately and then you you um yeah the less important data um you you don't do so much of that sort of error correction um convolution to sort of help that makes that makes a lot of sense i think i think that will yeah um that will be really clear to people um so can you sort of say what what is the rough kind of data rate that you that you send over the the ofdm link in a normal in a normal situation uh so first of all the data rate it has to be variable because it depends on the air condition uh snr yeah it has to be different otherwise the the fixed data rate is kind of suicide to me yeah so that that's that's one of my design spec so the average i think i i don't remember it it's close to 100 it's 97 or something yeah yeah wow so that's but sometimes these over 100 that i reached averages would be 97 yeah right so i think that's that will that will sort of uh surprise and i think impress people that actually although you've only got a you're using a narrower band you know 27 megahertz versus i think the 40 or something that dji use you know your the way you do your modulation uh and the way you you prioritize the different data in the link allows you actually to move more data through that through that channel than um any of the other systems i think uh certainly certainly more than uh dji are able to anyway um so i think that's um it's quite impressive in terms of in terms of the amount of data right here is kind of little so we are talking about actually the different thing so when dji says 50 megabits per second is quite useful used for bit rate ah these are useful beautiful right but i'm talking about the 97 actually the five bit rate right that's the actual link that's the that's the the link the link rate basically right so when dji is called 50 megabits per second i think it is on the air the link is over 100 right right right but because it has all the correction for that it has all the data have the error correction right i mean the redundant speech but for my stuff it's not some of them have lost some of the project so right i have 100 but my useful bits stephanie's exchanging right right yeah so talking about the difference yes now that that makes a lot of sense so yeah you've got about 100 megabits per second of the link rate and um how much of that roughly would you think is is is going to be uh you know useful data and how much would be that error correction right it really depends on the snr in the low snr when i only principal in the uh only transferring the the timing let's say uh the worst scenario i use in one frame time frame one what should i say one frame during the time of a one frame let's say one over 60 seconds i printed only one beat right you think how much beat i uh how much protection i have put it on i put it a lot right so during like uh it should be eight something eight minutes ago i only placed only one beat just for the timing right yeah right or not yeah but if this is the signal is very good so i transfer all the details and the other case so my 97 maybe we have a 80 megabits per second it's all useful right it really depends on the current the the channel condition if the channel condition is bad maybe all the yeah the useful data rate will be very small but if the sni is good the useful data rate is close to the five data rate right that makes a lot of sense so yes your your link rate will always be the same but your as you say the useful the useful data varies to to accommodate the different channel condition actually yeah yeah thank you yeah perfect um so you've transmitted it over the air and it uh it arrives at the goggles and as you say you some of that data is going to arrive um undamaged uncorrupted and some of that data is going to have been subject to some interference how how do you manage um data that might have been damaged during transmission when it arrives at the goggles yeah first of all we have a kind of the indicator it's called flag right so hold this bit call this block i would say block for this block uh call the data the status over there so if you found this data is it's cut is not the white or black meaning so this is a blog data maybe only only some of them are getting damaged right or we will identify that so if we damage it is totally damaged we will considering using this level code to regenerate it right right but most of the case we we get only half wrong half right so we need to make that decision to do that right so you've got you've got a sort of uh error correction code when you're in your asic that that can take a block that's only been slightly damaged and and sort of repair it in in some way right right because we have some kind of error correction right it's even a minor damage i can fix it and and how about for a block that's been significantly damaged where it's mostly mostly missing that's right i need to use this neighborhood pixels because i transferring the raw data so if the image is gone i can look at its neighborhood right it's neighborhood or blue i would think probably i'm blue too right right yes so you're using you're using uh effectively a sort of intra-frame prediction technique to look around the pixel and and figure out what it should have been that's right that's the right common knowledge i've been i've been practicing my terminology um yeah great so uh i think that that sort of explains things and then when you're when you're displaying that data into the goggles do you display it all at once in the way that dji does or do you display line by line as the data comes through the system right when i design this system later when i design this system the latency is one thing so when i get the one only one line of data no a small piece of the data i will start transmitting and here once i get the small piece of data to display i would throw it out right once again i'm using the same chip on the transmitter and receiver side which means i don't have the memory on this side either right so i don't have buffer right so the data is going straight straight out of your chip um into the hdmi link or or or directly you know into the hdmi link from the from the vrx hdmi link is a final stage right so for me the output is called bt 1120 this is kind of standard the bt 1120 can't go to hdmi chip right there the chip i mean converts bt 1122 hdmi signal oh come convert to sdi or whatever thing yeah the chip you just see of is come out is bt 1120 right and then that hdmi chip um does that have any any uh date any data storage in it any buffer or does it convert and and transmit down the hdmi cable sort of instantly i i'm not an expert there but i don't think they have any buffer maybe line buffers but i don't think that they have buffers because the hdmi thing is kind of like the meep thing yeah right so yeah it would be so we you don't think there's there wouldn't be like a whole frame buffer to just store everything it's likely to just send it to the screen as you send it the data right that's on the transmitter side on the receiver side may be a little bit different so this is a hdmi test meter right there is a hdmi receiver right on the hdmi receiver it needs a little bit of time to train its circuit to trace the incoming video incoming clock or data those kind of things then convert into the bt 1120 to display right right so yeah um it's going to be up to the the way that the goggle is implemented in a way into how it's going to treat that data that it receives whether it's going to display it immediately or whether it's going to um buffer a frame and then and then display a frame so that's something that that uh yeah i guess is is important uh that one i am not very sure about that actually each display is very different like olay and our osd oh no sorry lcd is a very similar some of the ltd they have a very springy chip it's it's it's kind of the kind of a muscle chief is like a square right small square it's kind of a yeah that the spaceship is very different and some of these flagships do have the memory inside yeah because they even use a driver but once again i'm not expected there so what i can do is i get the data i will transmit immediately i get it and once i receive it i will transmit uh right right so any any latency that might you know any latency that might be added by the the goggles as a function of how it handles incoming hdmi data is sort of is separate to your system and and that's going to be down to the uh the way the goggle is made right right and even the cargo maker probably that doesn't know no nobody knows the inside only the like when we're talking about the display let's say oled right the oled so they have something you know we pass the data to that meaning we only pass the data to the interface right or you can display the real image on the panel it's there's a chip to do that how that chip works we don't know right yeah yeah i think this is i think this is going to be interesting because in the next few days and weeks i'm going to be doing some high-speed video of of the screen inside the goggles and i've already done some testing and uh i've looked at an analog goggle um the onway commander and i've looked at the dji goggle and with the analog goggle you can see it drawing the image line by line yeah yeah at the same at the same rate that the analog scanning is is happening so it takes a whole um 17 milliseconds to draw a whole field as it scans line by line but the dji system the the screen just instantly updates the whole screen in less than a millisecond so it's sort of it's not line by line it well it probably is line by line but it's it's buffering a whole frame and then displaying a whole frame right no yeah yeah that's my expectation right yeah so your expectation is that is that most most hdmi goggles will work similarly that they'll they'll buffer a whole frame and then when they've got that whole frame they'll display it all at once no i i'm thinking the tgi right the dji they they do is not dji i mean the other solutions it there was like they have the data they do the compression right decompression they have the whole frame at once that's why they display it right right right but for us like analog we only get a few lines that's what we can do right so actually that could be that could be a benefit because um you know if you've got that few lines and you display those few lines if the object that you're looking for happens to be in those lines then you see it sooner yeah but when you test it probably the the food who died so it's kind of the the curve maybe it takes time to to to rise up right because the tgi is a pump up immediately so he immediately indicated that i mean when i read the report from ryan and the chief about their testing measure i think they do a very good job that that's that that's my expectation so something that something that i'm working on and i'm going to be doing is a similar type of test but rather than a photo diode i'm going to be using a high-speed video camera so you'll be able to see the screen and you'll be able to see how the screen is updating and it's it it's similar to the photo diode but it's like having you know lots of photodiodes all over the screen i guess right yeah there is a photo died theoretically there's no problem but the thing is how to select set the threshold is kind of the problem right as i said tgi they from zero to 100 instantly right yeah for the analog for the for we we we gradually so how to set the threshold is kind of very tricky if you say to the same threshold so definitely it could be different yeah yeah right right yeah i think i hope that when i hope that when i do the high-speed video it will be clearer to see what's going on because you'll actually be able to see the screen as if you were looking at it and actually see how it's updating in real time as the image comes in so that hopefully should clear a bit of that up first of all i think the latency test actually is very difficult it's very difficult it needs equipment and needs technology i don't need the knowledge yeah well um well what i'll do after this cool car is i'll send you over the videos that that i've prepared so far on analog and dji with my high speed camera and we'll see what you think of them and maybe uh maybe uh you can you can make some comments and i'll put them down in the in the video description what you think right i i'm very interested i i really like a zoo kind of thing i i like challenge yeah absolutely um so we've we've i think we've kind of talked about the the latency side of things and i think that that's going to be uh something that obviously we're going to get more information about um when i do my testing and as other people kind of refine their testing methods and it's great that you're that you're engaging with um with all the people who are doing the testing i think that's something that you know we don't see uh we don't see dgi doing like they're not they're not engaging with the with the community in the way that you are um so i think that's that's really great uh right everybody say that but but i'm small right i have nothing to do dji is big so yeah well you know i think it's i think it's wonderful that you take the time um i just as we sort of uh um come to i guess the the end of our time here i wanted to ask you a few questions about um you know the future the future technology um that you're working on with the system and um how how you see how you see hd0 progressing into the future um what what types of new products you're thinking about um obviously you've got a relationship with fat shark uh uh do you do you expect to to build relationships with other companies as well do you wanna you wanna talk a little bit about what you see for the future for for the technology oh okay uh for the current technology so we are weak in the penetration and the range so we definitely we need to hire watch vtx for that right and also the current camera we because we we don't make a camera right we we ask we put requests to long time or for folks ask them to make camera but it seems like we need to uh co-work with them more because for example if we can reduce our frame rate from 720p 60 to 720p that's our bandwidth will reduce from 27 to 13 which we have a much better range and a much better uh penetration if you think about this kind of range and this range passing through the area this range is much easier right so this is the kind of thing so the camera before a lot of people say oh the ht0 is analog plus it's like a 360 right it's well it does look like but until we have the htc will come out and say oh okay this system has some potential they were right but right but actually they are wrong i then the penny patient and the range we can if we can manipulate the camera setting into 720p 30 if we think some something is wrong so definitely we can improve the range of penetration without changing anything right so so you're you're looking to actually integrate more with the camera and and actually if you're in a if you're if you're going behind a wall or something and you suddenly get very low signal you could actually switch the frame rate and yeah and then and so okay it's a lower frame rate but you're still getting that image through it's you're still seeing you're still able to fly that's right yeah that's one of the that's one of the the other our chip actually supports 1080p 30. so yeah i know i i it's a kind of lower priority because uh the cargo itself no goggles support seven no uh 1080p yes right but this is the kind of thing we can't do do you ever think you're you're i know you've got your own camera now the the hd0 camera uh i think it's made by runcam is that right yes so do you think you'll continue to to do your own your own cameras in the future or are you hoping that the partners are going to take over most of that actually i want to get rid of the camera and the vtx and the vrx everything i want to sell chip right so you that's my ultimate goal i'm an isaac design company yeah like daniel english he has a video he asked me about that i i told him about the rich way i want to be a reach away nobody knows me but i'm doing the i don't know the video transmitter receiver thing right yeah nobody knows which way but actually the 100 market share in yeah but but that's that's could be really great so it wouldn't just be it wouldn't just be hd0 and it wouldn't just be um sharkbite but potentially everyone who currently makes analog vtxs could be making um could be making hd zero vtxes in the future right actually everyone come to me who want to like want to achieve all the mojo i i i think i'm i welcome them anytime contact me if they needed a chip it needed a sample needed a reference design no problem i'll give them fantastic well i think i think then we have to we have to throw the gordon out to uh you know all of those all of those manufacturers who currently make analog vtx we want we want you to go and make uh sharkbite digital vtx's as well you know we want to see what those products look like and um and get some get some more competition in the market and and get some more products in people's hands i think that would be really cool yeah that's the thing right we are the new technology we are a small company and we have a huge competitor so nobody wants to get in there there's nobody i mean making the hdd0 products you know because we are not open because they don't want to come in right we have to prove ourselves okay this hg0 is okay yeah then they were jumping yeah so people misunderstand that that's right well i think it's i think it's wonderful to hear and obviously i think everyone is going to be very pleased to understand that in the future we could see um you know people like i don't know tbs and rush and um akk and and all these other companies that currently make analog vtx's and make really good ones um could also uh turn their hand to making um you know shark bite vtxes and and potentially uh that would be a great partnership you know a great combination yeah i i think i i made a lot of mistakes so for example i i i still think i'm a very good esc designer an algorithm designer too yeah right but i'm not a qualified system designer so we made a lot of mistakes when we make the vtx the first thing like so when people and so that's right kind of we will have the tough situation right now is because of that so for example we when the first the first 200 units so we shipped it has some problems some of them like some guy only tested on the bench is dead so actually yeah we it's it's it's our mistake we don't have much experience on that and we made mistakes actually it's not our expertise it's not our expertise is the algorithm under the chip we want to do that right but with the chip with a huge competitor those kind of companies they won't jump what can i do yeah what can you do and i think that what you've done is you know you're showing you're showing a proof of concept of proof of principle uh an mvp product as it were to demonstrate that the system really is um you know really is has that potential to be great and that and that really you know we should be seeing we should be seeing people who currently make analog vcx's thinking seriously because in a few years time no one no one will be flying no one is likely to be flying analog anymore um i i'm i'm a firm believer that uh the digital technologies uh dji um hd0 uh i know we've got open hd as well you know these technologies in a few years time are going to be are going to be the the dominant ones right oh yeah so the digital let's say we have eight channels in channels i mean i i don't know why they say to the 25 milliwatts for the reason right because it will not affect each other but for our for our system if you said at the 200 milliwatts you have eight players play together no problem you don't need to say 25 right because our speaker from is like the door right yeah analog is like this yeah it goes out that's right yeah it will occup it will invade the neighbors yeah well the digital is the gate it's it's no problem to do that yeah i think that that's also you know something that's that's really great to hear is that we might we might be racing on 200 milliwatts or or 500 milliwatts or even more um in the future if everyone's running if everyone's running digital systems like like hd zero uh and that means that we could have you know even even more exciting racing through through courses where there's more obstacles more more things that interfere with rf because we we can actually fly at higher powers for racing so um i'm i'm excited to see how that will affect like the race course design and stuff in future when you can have pilots flying through buildings and through trees and not worry about them losing signal oh yeah and also such digital they occupy the less uh frequency bind so right now the r1 to r8 actually is only i'm not sure is only two or three three or four of them are legal legal illegal the other is not it certainly depends on on the country doesn't it depends on where you're from oh yeah yeah yeah that's true yeah yeah so what i want to say that digital they occupy let's say we are only using 27 right but from r1 to r2 is the the spine is 30 something right 36 or something yeah it's wider than 27 yeah he's a wider right either wider than 27. i mean we can squeeze more channels yeah and get more more pilots in the air yeah wow fantastic um well i uh i i come to the end of my questions i think you you know carl you you've been wonderful and thank you so much for for answering um is there anything that you want to ask me or um or ask anyone who who might be watching this and about to type a comment do you want to ask anything of anyone uh okay i want to say something sure so i i know the the ht0 right now it's still unmature immature to to somebody i think that's true but we are improving we are listening the requirement we have some very good purpose to help me understand the requirement uh i'm yeah that's my that's my last scene before my retirement i i need to do it right i yeah i want to do something like i was doing the video surveillance thing so it's i don't know where the ntsc from 2005 to 2014 i'm converting the analog to digital now i'm doing the same thing i want to convert the iron over to digital for the apv domain for that yeah that's my goal this may be my last job to do that yeah fantastic i yeah i want to make money but it doesn't make money i'm okay here i i just i have a dream to do that fantastic well i think that i think that uh your passion for it really comes across um and i think that i think that that will really connect with people um yeah it's about uh it's kind of why we all do it right we all do this because at the end of the day we want to try and make uh the fpv hobby better for pilots and uh and that's the goal that's right i actually i love the atv these those people those pilots i mean not all of them definitely but most of them they are warm-hearted when you ask them they're very glad girls i i won't i will not name anyone right but they are you my heart they they support me so even i made mistakes after mistake yeah but you forgive me and give me chance yeah yeah well i think we're we're all well i certainly am i'm 100 um behind what you're trying to do um and i'm very excited to be making all this content about about your system and uh um you know i think hopefully uh people who watch this will will understand you know what you're trying to do why you're trying to do it and and why um it should make things better for for pilots in the future um and that we can all do what we can um to to support you uh in trying to make that dream a reality which i think is is really great yeah that's a that's that's really cool i i i i want to make the fpv this hobby better that's not for sure right yeah so the hd zero doesn't occur it doesn't exclude anywhere right we want to convert analog to digital digital have many beauties so we need to convert right but for example as of today we i lost a lot of money for to do that but i'm not regret i'm never requested i think the ltv pilots they are i love them really i love them yeah they help him a lot they're helping me a lot yeah yeah oh well that's fantastic i i think uh i think that's all we all we probably need to say today what do you think carl is there anything else you want to you want to say i think that's a lovely note to end on uh thank you chris give me this chance thanks no problem uh it's always lovely to speak with you carl and i wish you all the very best thank you for taking this time and i hope everyone who's watching this video uh also enjoyed it and um and we'll we'll leave any questions they have down in the comments uh i'll certainly try and reply to as many as i can and uh and um yeah we'll we'll talk about this a bit more in the future yeah thank you thanks carl all the best take care bye

Channel: Chris Rosser

Views: 29,050

Keywords:

Id: 5YYr2gUEgLk

Length: 68min 33sec (4113 seconds)

Published: Fri Sep 10 2021