High definition FPV

For those interested - a hack for live video output (read “live” to mean “low latency” in reality) for the HD HERO already exists – done by my son and a university collegue of his - via a USB.2 extension cable/adaptor plugged into the SD card socket, taken out to a buffer/filter, onto COFDM modulation, then to a low power PA (around 125mW) and then transmission.

Total latency on the Tx side (i.e. to include COFDM modulation) is around 105millisecs. This is for around 15.5Megbytes per/sec over just around 4Mhz bandwidth – not bad for a DIY/home effort, but can be improved, and we are working on that now.

This system is put together at my behest – and is specifically intended for use in a flying FPV model.

The overall system latency is 120millisecs. We are thinking of ditching the HD Hero altogether, despite how good it is (and make no mistake about it, GoPro have implemented H.264 in the HD Hero very well indeed), and replacing it with component level hardware, using the BlackMagic USB H.264 dongle for encoding – which faster for the same bandwidth, and offers a noticeable improvement in picture quality, again for the same bandwidth, over the HD HERO. The camera will be a Sony block H10 or 11.

Regards transmitting an HD video stream, there are a whole bunch of issues that need to be considered and thought through very carefully if you want to setup a decent quality HD video system to stream back to a ground station good quality content on a reliable digital link.

Let me share with folk the issues we have thought through to date, and their relivent points. They are considerations which need to be kept in mind when it comes to selecting hardware to setup a low latency rf transmitted usable HD video stream from a flying model to a ground station. It’s a bit of a mouthful – but they are unavoidable points.

Analogue or Digital
An analogue HD video stream will require at min 9Mhz of bandwidth (as a 720p/25fps picture format), and more typically closer to 12Mhz (as a 1080p 25/30fps picture format). A 9Mhz – 12Mhz bandwidth stream requires and uses considerably more battery power than a does an SD video stream.

Consider also the image sensor required – around 2Megapixels to get an HD picture. Although small CMOS/HD image sensors are now 2 a penny so to speak, the pixel size of most of these sensors is very very small, and one has to ask if as an analogue transmission, there is really a worthwhile improvement in received picture quality using them as the frontend in an analogue HD setup? In my personal opinion, nope – the difference in received & demodulated picture quality (versus SD video) is not worth the effort. Quite frankly, at the hobby/amateur level (read as: cost/budget level), if you can show me a practical HD transmission system/setup in a small/medium flying model, I’ll show you an SD video stream for the same cost with a better picture quality.

That will change over the next 5years or so, but for the moment, the added hardware & cost requirement to setup implement an analogue HD video stream versus an analogue SD video stream (from a flying model) is hard to justify when viewed from the “improvement in video picture quality” point of view.

However, as a digitaly transmitted video stream the answer has to be a BIG yes – incredibly good video quality is realizable today from HD resolution video hardware – both with and without compression been applied. But, a number of considerations need to be carefully thought through if you want to implement digital HD video of an on-screen picture quality that justifies the added cost and effort that will be incurred. In short the issues are as follows:

Image Sensor.
HD video cameras will have an image sensor of around 2Megapixels (for 1080p resolution). As a rule, the larger the individual pixels and the greater the dynamic range of the pixels, the better the quality of the picture/video the image sensor will produce. In choosing a video camera to use on your model, in so far as your budget allows, pay attention to the pixel size and the dynamic range. The greater the dynamic range and the larger the pixel, the better the image quality will be. Dynamic Ranges of 80dB and more from pixel sizes of around 3.5micron x 3micron will give you stunning picture quality - never mind the already good picture quality achieved from image sensors in modern video cameras with dynamic ranges less than this, and pixel sizes half the above size! The problem is: pixel size palys a big role in image sensor/camera cost. There are some 2Megapixel image sensors around (thats what you need fro HD video - 2Meg's) with pixels that measure around 8 or 9microns in size. The picture quality form them is stunning - but, yup, they are in terms of price way beyond all but the most dedicated hobbiest, sadly.

Analogue to Digital Conversion (ADC) rate
Another spec to look for, is the analogue to digital [conversion] rate i.e. how often per second (usually in Khz) changes in each pixels’ output voltage are read, and at what resolution are those readings noted ( usually in bits – e.g. 8bits,10bits,12bits … or whatever). Typical sampling & bit rates for an HD analogue to digital conversion are along the lines described above i.e. 8bits, 12bits, 16bits 24bits …. 32bits …. and so on, and typical sampling rates are 32Khz, 48Khz … and so on. The more often these readings take place & the finer the resolution at which they are recorded, the better the resultant video picture color, contrast and other parameters that make up picture quality, will be.

CODEC
The next process in the chain of events that has to take place in getting video from transmission point to reception point, is, compressing the digital data from the ADC output – using a codec. The codec everyone knows about of course is MPEG 2, but more recently, as processing powers have increased & become faster, another codec is starting to find favour – H.264 (also known as MPEG2 Part 10). Note that H.264 is more a codec standard than a codec. In simple terms, how well the H.264 codec works is very much down to the individual developer implementing the codec in hardware or software formatt – you get H.264 codecs that are well implemented (fast and with good picture quality) and you get H.264 codecs that are badly implemented (slow and with bad picture quality).

H.264 can be implemented in software or in hardware. As a rule H.264 implemented in hardware (i.e. using a dedicated IC/chip/ASIC) is faster than H.264 implemented in software. But this will change over the next few years as software becomes faster and faster (Moores Law). Still, for the time been, hardware is the way to go for speed.

The big attraction of H.264 over MPEG2 is simply this: for a given bandwidth an H.264 encoded video stream will produce a better quality video picture than will an MPEG2 video stream, or another way of putting it: for an eqivilant picture quality an H.264 encoded video stream will require less transmission bandwidth than an MPEG2 encoded video stream.

This reduction in required bandwidth comes at a price, though: added computational requirement – the amount of processing power required to produce an H.264 encoded video stream versus an MPEG2 encoded video stream at the same speed, or equal latency period, is considerably more for H.264 than it is for MPEG2 - and more battery power will be used.

To be specific in terms of bandwidth, an H.264 video stream of x length and y perceived picture quality, will require between 30% - 50% less bandwidth to transmit than will the exact same video content of the same length (x) and same perceived picture quality (y).

And to be specific in terms of power consumption, using the above as an example and assuming equal latency for the 2 video streams, the H.264 encoded video stream will have consumed roughly 33% more processing power and used roughly 25% more battery power (mA/Hour) to achieve.

The primary motive behind converting an analogue video stream into digital video stream for rf transmission is to reduce the required bandwidth, therefore allowing more video to be transmitted (e.g. some digital sat TV streams transmit as much as 10 x’s more TV content for a given bandwidth than could ever be transmitted in the same bandwidth as analogue video).

Altogether there’s a pretty strong argument for using digital transmission for HD video.

Modulation
The next step in the process is to modulate the digital data for rf transmission and there are a 101 different modulation schemes that can be adopted when it comes to transmitting any video, let alone HD video. This is the digital part from the rf transmission point of view.

COFDM is one of the more common modulation schemes, and is used so much because of its unique resistance to delays and fading problems associated with rf transmission – more commonly known as multi-path problems. An excellent introduction to COFDM can be found at:

http://downloads.bbc.co.uk/rd/pubs/reports/1996-08.pdf

…….. so I wont bother with any notes on this part of the overall process

Amp Stage
The last consideration to keep in mind when it comes to transmitting digital content with wide band-widths and lots of data content (and HD video transmissions are about as dense as you get) is the linearity of the amp stage. Linear amplification is critical when it comes to digital data content. Without excellent linearity data quickly gets distorted in both phase and amplitude, rendering it unrecoverable on the receive end. Even the best Class A type rf analogue amps struggle to maintain sufficient linearity for use as digital video amplifiers. Digital rf amps need to be designed from the start with as digital content amplifiers – using an analogue rf amp will usually end in tears.

When it comes to amplifying IP/802.11XX (i.e. WiFi) the requirement for linearity in the amp stage is even greater.

Typical Gain flatness figures required across the full modulated bandwidth of +/- 5Mhz in digitally encoded HD video signal, or the 10Mhz – 20Mhz band width of z 802.11XX signals, at max power output needs to be somewhere around 0.3dB – 0.5dB – never mind the rms/average output required Gain flatness (the all important figure), which will be much lower than that. Fairly easy to implement when your power output is something like 10mW, gets progressively more difficult technically, as power output increases.

In short, if you want to setup a digital HD video stream from your flying model to a ground station, there are a whole bunch of points to consider when it comes to selecting what gear to use – if you want to retain the HD video stream quality that it is possible to get with HD hardware. Get any one component in the chain wrong and you’ll land up with picture quality little better than SD.

Note – HD video is about to burst onto the UAV scene at the military level. USB dongle sized pro picture quality H.264 engines exist (e.g. Black Magic’s $350 H.264 USB engine – small enough, light enough and power conservative enough to use in most flying models). The choice of board camera’s and block camera’s is endless – from Omni Visions 100dB plus dynamic range board camera’s with auto contrast (meaning you can fly into the sun with it and ground based detail in the lower half of the screen will not be blacked out), to Sony’s H10 and H11 block cameras (although these are a bit above the average hobby flyers price range). Adaptive RF make a range of small COFDM type rf amps that suitable for digital video (http://www.adaptiverf.com/Products.php).

Try searching Google images for small digital video modulators & transmitters - type “low power OEM COFDM video transmitter” – quite a selection comes up. The rf stage for digital video remains the most expensive component in any set up, as till now its been the preserve mainly of the pro community, but that is going to change over the next 24 months or so as HD IP based video content gets better and better faster and faster in terms of picture quality. I don’t think it will be long before HD band-width capable small low power digital amps cost little more than analogue SD video amps cost.

Eventually everyone will be wanting to use HD digital video. There is a good argument that says all digital video links will need to be spread spectrum to avoid conflict and interference issues.

IP based HD video systems may ultimately be the way the mass market goes – not withstanding the transmit power limits, and therefore range limitations. In 5 – 10 years time we’ll probably have 4G mobile phone modules that can handle full res HD video on reliable data links.

Implementing digital HD video is great – the practicalities will be a little more involved when it starts to take off, and for the time been it remains comparitively expensive to implement stand alone HD digital video transmission with the same degree of reliability and range as SD analogue video - no way round it.

Anyhow – that’s my little tech primary for those who have HD video transmission in mind. The practicalities involved in implementing a decent working and reliable digital true HD video system are quite a bit more involved than setting up an analog SD video link.

Replies

Patrick Hammer September 25, 2010 at 6:42am

For those interested - a hack for live video output (read “live” to mean “low latency” in reality) for the HD HERO already exists – done by my son and a university collegue of his - via a USB.2 extension cable/adaptor plugged into the SD card socket, taken out to a buffer/filter, onto COFDM modulation, then to a low power PA (around 125mW) and then transmission.

Total latency on the Tx side (i.e. to include COFDM modulation) is around 105millisecs. This is for around 15.5Megbytes per/sec over just around 4Mhz bandwidth – not bad for a DIY/home effort, but can be improved, and we are working on that now.

This system is put together at my behest – and is specifically intended for use in a flying FPV model.

The overall system latency is 120millisecs. We are thinking of ditching the HD Hero altogether, despite how good it is (and make no mistake about it, GoPro have implemented H.264 in the HD Hero very well indeed), and replacing it with component level hardware, using the BlackMagic USB H.264 dongle for encoding – which faster for the same bandwidth, and offers a noticeable improvement in picture quality, again for the same bandwidth, over the HD HERO. The camera will be a Sony block H10 or 11.

Regards transmitting an HD video stream, there are a whole bunch of issues that need to be considered and thought through very carefully if you want to setup a decent quality HD video system to stream back to a ground station good quality content on a reliable digital link.

Let me share with folk the issues we have thought through to date, and their relivent points. They are considerations which need to be kept in mind when it comes to selecting hardware to setup a low latency rf transmitted usable HD video stream from a flying model to a ground station. It’s a bit of a mouthful – but they are unavoidable points.

Analogue or Digital
An analogue HD video stream will require at min 9Mhz of bandwidth (as a 720p/25fps picture format), and more typically closer to 12Mhz (as a 1080p 25/30fps picture format). A 9Mhz – 12Mhz bandwidth stream requires and uses considerably more battery power than a does an SD video stream.

Consider also the image sensor required – around 2Megapixels to get an HD picture. Although small CMOS/HD image sensors are now 2 a penny so to speak, the pixel size of most of these sensors is very very small, and one has to ask if as an analogue transmission, there is really a worthwhile improvement in received picture quality using them as the frontend in an analogue HD setup? In my personal opinion, nope – the difference in received & demodulated picture quality (versus SD video) is not worth the effort. Quite frankly, at the hobby/amateur level (read as: cost/budget level), if you can show me a practical HD transmission system/setup in a small/medium flying model, I’ll show you an SD video stream for the same cost with a better picture quality.

That will change over the next 5years or so, but for the moment, the added hardware & cost requirement to setup implement an analogue HD video stream versus an analogue SD video stream (from a flying model) is hard to justify when viewed from the “improvement in video picture quality” point of view.

However, as a digitaly transmitted video stream the answer has to be a BIG yes – incredibly good video quality is realizable today from HD resolution video hardware – both with and without compression been applied. But, a number of considerations need to be carefully thought through if you want to implement digital HD video of an on-screen picture quality that justifies the added cost and effort that will be incurred. In short the issues are as follows:

Image Sensor.
HD video cameras will have an image sensor of around 2Megapixels (for 1080p resolution). As a rule, the larger the individual pixels and the greater the dynamic range of the pixels, the better the quality of the picture/video the image sensor will produce. In choosing a video camera to use on your model, in so far as your budget allows, pay attention to the pixel size and the dynamic range. The greater the dynamic range and the larger the pixel, the better the image quality will be. Dynamic Ranges of 80dB and more from pixel sizes of around 3.5micron x 3micron will give you stunning picture quality - never mind the already good picture quality achieved from image sensors in modern video cameras with dynamic ranges less than this, and pixel sizes half the above size! The problem is: pixel size palys a big role in image sensor/camera cost. There are some 2Megapixel image sensors around (thats what you need fro HD video - 2Meg's) with pixels that measure around 8 or 9microns in size. The picture quality form them is stunning - but, yup, they are in terms of price way beyond all but the most dedicated hobbiest, sadly.

Analogue to Digital Conversion (ADC) rate
Another spec to look for, is the analogue to digital [conversion] rate i.e. how often per second (usually in Khz) changes in each pixels’ output voltage are read, and at what resolution are those readings noted ( usually in bits – e.g. 8bits,10bits,12bits … or whatever). Typical sampling & bit rates for an HD analogue to digital conversion are along the lines described above i.e. 8bits, 12bits, 16bits 24bits …. 32bits …. and so on, and typical sampling rates are 32Khz, 48Khz … and so on. The more often these readings take place & the finer the resolution at which they are recorded, the better the resultant video picture color, contrast and other parameters that make up picture quality, will be.

CODEC
The next process in the chain of events that has to take place in getting video from transmission point to reception point, is, compressing the digital data from the ADC output – using a codec. The codec everyone knows about of course is MPEG 2, but more recently, as processing powers have increased & become faster, another codec is starting to find favour – H.264 (also known as MPEG2 Part 10). Note that H.264 is more a codec standard than a codec. In simple terms, how well the H.264 codec works is very much down to the individual developer implementing the codec in hardware or software formatt – you get H.264 codecs that are well implemented (fast and with good picture quality) and you get H.264 codecs that are badly implemented (slow and with bad picture quality).

H.264 can be implemented in software or in hardware. As a rule H.264 implemented in hardware (i.e. using a dedicated IC/chip/ASIC) is faster than H.264 implemented in software. But this will change over the next few years as software becomes faster and faster (Moores Law). Still, for the time been, hardware is the way to go for speed.

The big attraction of H.264 over MPEG2 is simply this: for a given bandwidth an H.264 encoded video stream will produce a better quality video picture than will an MPEG2 video stream, or another way of putting it: for an eqivilant picture quality an H.264 encoded video stream will require less transmission bandwidth than an MPEG2 encoded video stream.

This reduction in required bandwidth comes at a price, though: added computational requirement – the amount of processing power required to produce an H.264 encoded video stream versus an MPEG2 encoded video stream at the same speed, or equal latency period, is considerably more for H.264 than it is for MPEG2 - and more battery power will be used.

To be specific in terms of bandwidth, an H.264 video stream of x length and y perceived picture quality, will require between 30% - 50% less bandwidth to transmit than will the exact same video content of the same length (x) and same perceived picture quality (y).

And to be specific in terms of power consumption, using the above as an example and assuming equal latency for the 2 video streams, the H.264 encoded video stream will have consumed roughly 33% more processing power and used roughly 25% more battery power (mA/Hour) to achieve.

The primary motive behind converting an analogue video stream into digital video stream for rf transmission is to reduce the required bandwidth, therefore allowing more video to be transmitted (e.g. some digital sat TV streams transmit as much as 10 x’s more TV content for a given bandwidth than could ever be transmitted in the same bandwidth as analogue video).

Altogether there’s a pretty strong argument for using digital transmission for HD video.

Modulation
The next step in the process is to modulate the digital data for rf transmission and there are a 101 different modulation schemes that can be adopted when it comes to transmitting any video, let alone HD video. This is the digital part from the rf transmission point of view.

COFDM is one of the more common modulation schemes, and is used so much because of its unique resistance to delays and fading problems associated with rf transmission – more commonly known as multi-path problems. An excellent introduction to COFDM can be found at:

http://downloads.bbc.co.uk/rd/pubs/reports/1996-08.pdf

…….. so I wont bother with any notes on this part of the overall process

Amp Stage
The last consideration to keep in mind when it comes to transmitting digital content with wide band-widths and lots of data content (and HD video transmissions are about as dense as you get) is the linearity of the amp stage. Linear amplification is critical when it comes to digital data content. Without excellent linearity data quickly gets distorted in both phase and amplitude, rendering it unrecoverable on the receive end. Even the best Class A type rf analogue amps struggle to maintain sufficient linearity for use as digital video amplifiers. Digital rf amps need to be designed from the start with as digital content amplifiers – using an analogue rf amp will usually end in tears.

When it comes to amplifying IP/802.11XX (i.e. WiFi) the requirement for linearity in the amp stage is even greater.

Typical Gain flatness figures required across the full modulated bandwidth of +/- 5Mhz in digitally encoded HD video signal, or the 10Mhz – 20Mhz band width of z 802.11XX signals, at max power output needs to be somewhere around 0.3dB – 0.5dB – never mind the rms/average output required Gain flatness (the all important figure), which will be much lower than that. Fairly easy to implement when your power output is something like 10mW, gets progressively more difficult technically, as power output increases.

In short, if you want to setup a digital HD video stream from your flying model to a ground station, there are a whole bunch of points to consider when it comes to selecting what gear to use – if you want to retain the HD video stream quality that it is possible to get with HD hardware. Get any one component in the chain wrong and you’ll land up with picture quality little better than SD.

Note – HD video is about to burst onto the UAV scene at the military level. USB dongle sized pro picture quality H.264 engines exist (e.g. Black Magic’s $350 H.264 USB engine – small enough, light enough and power conservative enough to use in most flying models). The choice of board camera’s and block camera’s is endless – from Omni Visions 100dB plus dynamic range board camera’s with auto contrast (meaning you can fly into the sun with it and ground based detail in the lower half of the screen will not be blacked out), to Sony’s H10 and H11 block cameras (although these are a bit above the average hobby flyers price range). Adaptive RF make a range of small COFDM type rf amps that suitable for digital video (http://www.adaptiverf.com/Products.php).

Try searching Google images for small digital video modulators & transmitters - type “low power OEM COFDM video transmitter” – quite a selection comes up. The rf stage for digital video remains the most expensive component in any set up, as till now its been the preserve mainly of the pro community, but that is going to change over the next 24 months or so as HD IP based video content gets better and better faster and faster in terms of picture quality. I don’t think it will be long before HD band-width capable small low power digital amps cost little more than analogue SD video amps cost.

Eventually everyone will be wanting to use HD digital video. There is a good argument that says all digital video links will need to be spread spectrum to avoid conflict and interference issues.

IP based HD video systems may ultimately be the way the mass market goes – not withstanding the transmit power limits, and therefore range limitations. In 5 – 10 years time we’ll probably have 4G mobile phone modules that can handle full res HD video on reliable data links.

Implementing digital HD video is great – the practicalities will be a little more involved when it starts to take off, and for the time been it remains comparitively expensive to implement stand alone HD digital video transmission with the same degree of reliability and range as SD analogue video - no way round it.

Anyhow – that’s my little tech primary for those who have HD video transmission in mind. The practicalities involved in implementing a decent working and reliable digital true HD video system are quite a bit more involved than setting up an analog SD video link.
Roman Krashanitsa September 17, 2010 at 3:28pm

Using an on-board encoder will introduce a delay. A low-power light board will produce a bigger delay.
I do not have numbers right now, but delay could be very significant ( ~1sec )
Jack Crossfire August 26, 2010 at 1:14pm

Got to hack it yourself. It's too expensive to start a business selling closed circuit HDTV so most everyone makes their own using 802.11 modules.
AKRCGUY August 26, 2010 at 11:44am

GoPro HD camera's support an output that requires a hack to input to Video Tx's but it's doable.
JOHN PIZZICARA August 7, 2010 at 8:28pm

Got Money?!? Check out Nucomm's CamPac2.
Anner J. Bonilla August 7, 2010 at 7:24pm

I am interested in this for a UAV. I want to build i thought of many options.
Xbee
2.4ghz
UHF
WIFI

So far only feasilble ones appear to be the Xbee pro which claims 144kbits. But I do not funds to buy the equipment and test it right now. Anybody that have tried it or thought about it?
Paul Briesch August 7, 2010 at 3:21pm

Roberto,

it is possible to transmit higher definiton video with amateur radio DVB-S or DVB-T tranceivers. However, this is like running your own TV station and usually to big for UAVs we make.

Recently I was looking into a high resolution web cam with wireless lan and H264 codec. According to the specs the camera should deliver 25 fps. However, I could not get a fluent stream.

Regards,

Paul

of 2

This reply was deleted.

Discussions

High definition FPV

Replies

Activity