asking aout the HDU-1

[Trung Dau]

Hi keith.
WHen will ur HDU-1 be released?
Can it convert from HD analog to the HD-SDI? if not , it's not a low cost solution for uncompression at all, because after 2.5 k ( which is the price, u guessed), we would add for the aja converter( around 1k).

[Keith Wakeham]

Hi, and thanks for your interest.

I understand your argument and I've been in contact with a lot of people. And see that this is a big big issue for some people. Although their are a few people who are involved with the project, I'm the one doing most of the engineering and design. As such small hiccups in the design process can really put a huge delay into the works. These small hiccups are more common than I ever expected at any this point.

I'm quite open to explanations of exactly what is going on, but most of it gets down to much to the nitty gritty and its just problems that all have solutions given enough work.

Any price estimate i did give out is just that, an estimate. So sorry about that and sorry that you feel that way about having to use a converter, though AJA makes some great products.

I would like to say in my defense that I'm taking the "Keep it simple stupid" (KISS) route for now.

Simply put HD-SDI requires a chip to deserialize and a chip to serialize (all automatic as well) and the code structure for finding active video is perfectly Perfect to say the least. Whereas component requires 6 chips all manually controlled, and can go + or - 6 clocks either way for certain stuff meaning their can be variance between frames and clock drift and you need a fairly complex design for decoding, locking on, and staying locked on to the signal. HD-SDI decoding only looks for key words. This is all done, I can lock on to anything HD-SDI (but need more test equipment to verify in real world, working on doing that this week). But in all simulations I can go above and beyond the SMPTE 292 spec and even handle variable framerates over HD-SDI.

As for the other side of this, well, the hard drive interface is where i'm running into delays. Originally I was using PIO write modes, 16.6MB/s per drive, now i'm trying to finish up ultraDMA. This means I can use drives set to rates of 33, 66, or 100MB/s and with the right converter use Sata drives. (Sata needs hard to get transcievers so I'm not native Sata - yet)

So what all this means is i'm hoping to be on track again soon but this work is costly and i've already sunk a lot of money into it and have to be more careful where i put money now and can't take chances as much as before.

I have most of the equipment to put together a couple of prototypes now but the small details of these two things are quite costly.

HD-SDI also allows an upgrade via firmware to support embedded TC, audio and meta data, something also not possible with component.

So latter when I get everything under control in the HD-SDI front which provides me with a workflow that doesnt' require software development I'm going to turn my attention to the data workflow and a component device. Since everything should be in approximate sync a component to HD-SDI converter like aja's could be built rather cheaply as an add on, or vic versa.

I do take a little contest with you saying its not a low cost alternative. I'm not going to be a hard drive reseller, so that cost is entirely up to hard drive manufacturers. Even if I was to stick this into the market now if it was ready what are your alternatives that are portable and able to be battery powered. You have HDCAM, HDCAM SR, Thomson Venom, and in the soon future Wafian HR-1 (currently not battery powered but a version soon coming). And of these only two will be approximately camera mountable. The cheapest solution is the Wafian weighing in at 15K. Great product and great for cineform, but your not going to be putting it on back of a XL H1, or the HD200 that was announced at NAB. The thomson Venom is the only one close to portable on camera shooting and at 50K USD last I heard. So anything near uncompressed at this point sub 10K I think is fairly low cost.

Feel free to email me directly regarding anything, quesitons, ideas, investment (I wish). I'm no Jim Jannard so don't expect me to have such a presence as RED. Its a tiny startup by comparison... very very tiny...

Sorry about the long time to reply, busy on stuff and haven't visted this forum in a little while.

[Trung Dau]

thanks for replying. I doesn't want to have any argument. BEcause I'm a newbee so than i ask some dump questions. And so sorry for my bad English, that's why it's not polite for asking u some questions.

Why do u use 10 80gb hard drives, instead 2 of 400gb HD?

And if i use the aja converter component analog to HD-SDI, Can the TC be kept?

[Dan Euritt]

thanks for the update keith!

[Keith Wakeham]

I didn't mean to be argumentative, sorry if i came off that way, I have a tendancy to throw way to much info in peoples faces.

The hard drive number thing.... Man the number of hard drives is what I will refer to as a can of worms.

Shortest answer follows.

The power consumption of a 2.5" laptop hard drive is at most 2.5 Watt/Hr while a 3.5" is 10 watt hours. 4 x 2.5" hard drives will fit in the exact same space as a single 3.5" drive and have about 2 - 3 times the available actual thorughput and still come in less power hungry. 2.5" drives are also much more rugged and able to handle vibrations a lot more.

Story Time: (skip if you want)
If you knew home many times i smacked my laptop into the corner of a wall while it was on walking down a hall you'd know what i mean, plus I had an unfortunate bicycle accident with a big SUV, suffice to say my laptop landed 20 feet away from me, i landed 15 feet from the truck, so that laptop landed 35 feet from when it was hit - it is still using the original drive while I've dropped 3.5" drives a foot and end up with 10MB of bad sectors.

The number of hard drives: two things are determining this and i'm trying to fix it so that with a safe raid setup it will end up with 6 drives.

I need 150MB/s to pull of full uncompressed. Hard drives had several modes to transfer data. PIO, multiword DMA, and ultraDMA. Simplest being PIO which i got working early and when from their. UDMA is much more complex with specific timings, responses, and crc calculations. So i went with PIO, so to get 150MB/s I need 9-10 drives. However this caused us a nightmare with the pcb. Multiple layers with complex routing to achine the near 300 pin i/o count we needed. So I designed a pcb and tried to get it made, after weeks of being screwed around with by one of the few canadian companies decided to minimize time and cost by jumping to program UDMA and drop the pin count for a simpler board.

So this means actual transfer rate of the hard drive is the important thing now. With UDMA and using one of 2 available 2.5" 7200RPM 100GB drives we can cut it down to 5 drives with one for parity, so 6 in total. This is what were working for now. Its not the interface limiting anymore but drive transfer speed.

So were back to the pcb but we got that somewhat under control, then after that and making sure it all works together theirs all sorts of things.

As far as i know you cannot keep TC on the HD-SDI line using a component to HD-SDI converter because component doesn't carry time code and it would need external timecode inserted into the hd-sdi stream to deal with it.

If I go the way of a component to hd-sdi converter, which assuming i can ever finish looks like the best way, I'm going to support embedded TC and audio so the HDU-1 can store this stuff.

[Trung Dau]

thanks for ur information KEith.
It's very kind of you, KEith

[Keith Wakeham]

No problem.

I figure it might be best to let people know exactly whats going on, I believe many may have lost faith in me, but eventually I'll get it done.

[Dan Euritt]

What is it? NAND is a type of flash memory that will be used in solid-state hard drives.

Why is it worth watching? Imagine a solid-state drive that consumes a fraction of the power and weight of a conventional laptop hard disk drive, and won't break when you drop it.

When is it coming? The end of 2006.

Tell me more. In the middle of March at the CeBIT trade show, Samsung displayed a hyper-thin Q30 notebook computer. In and of itself, the 2.5-pound, 0.7-inch thin portable was remarkable, but the true stunner rested inside the machine. Samsung engineers had modded the Q30 with a 32GB solid-state disk (SSD).

http://techweb.com/wire/showArticle....9500002&pgno=5

ram drives in a 2.5" ide form factor have been around for awhile, but they were always too expensive... it sounds like the prices have been dropping!

going with the faster udma standard is going to be a good idea in the long run!

[Keith Wakeham]

Flash hard drives have actually been used in the military for years. Their is a SCSI one that is something like 128GB and wicked fast... problem is cost. The drive was not available to the public but it still cost in excess of 15K USD.

I could jump straight to the flash medium if i really wanted to learn the protocol which isn't to complicated (a joke compared to UDMA from a quick skim of the docs) but its still expensive and memory boards are normally a minium of 6 layers. To much for me to handle. Along with the other fact that to handle 4 minutes of 1080p30 or 60i I need 36 gigabytes.... which translates into 72 4gbit chips. The largest chips I've seen are 8Gbit, or 1Gbyte, which means i'd need 36 of them. Thats $4700 CAD for 4 minutes in flash.

Thomson already got that covered with the Venom and to make it worthwhile it will be expensive to get the boards made and chips installed except in quantities. Quantities which I honestly don't have the money for. If Thomson did use 4Gbit flash, it would mean they have over 300 chips in the Venom.

I'm sure their are better prices but I only know a few suppliers and I can't go direct to manufacturers. A 2Gbyte SD card is only 80 CAD, so thats not to bad. If I knew SD commands and they didn't charge liscencing this would be a way to go.

Curse you Dan, you got me thinking... Wonder if Pansonic will let me liscence P2 Tech?

Maybe I'll use what I learnt to have a hard drive offloading system like HVX.

I think a lot of people might be upset if i changed designs now, but so curious

[Wayne Morellini]

Hi Keith, this raises a number of things. As I mentioned before there is the possibility of getting a component to HDSDI bridge reference design, and either including it as standard or as option plug in board or plugged into the HDMI posts as an optional extra. This would save you the time of designing such a thing into the unit, or delay it until you replace this one of the market. As you make them (and sell them on market separate as AJA does) it can be sourced a lot cheaper then $1K. Even tracking down an original Asian manufacture of a reference design converter box, they might supply at a very cheap cost. The manufactures of such bridges will know who manufactures such boxes.

There is now a move onto HDMI, that also can carry surround audio tracks, one could be made to carry time code, if time code not natively supported. So maybe Component to HDMI converter to HDMI to HDSDI bridge ;). I think that HDMI might be much more p[rices then component and therefore much less complex to design for you.

I also think that if you look at professional video component you might find that it much more matches HDSDI, no information, but just a feeling. The cameras output some odd resolutions, but maybe with precise frame and line resolution timings, you see what I mean. Is this what they do? Even if you can't lock onto the pixel timing easily, it might be easier to just save them at the official HDSDI resolution (1920, 1280 etc) as Canon does.

There is now a move onto 12bit, and eventually 16 bit outputs, I think that the newest HDMI (which forms part of the new Universal Display interface standard to replace VGA, supports greater bit depth. I wish the industry would look at making HDMI support camera control+other signalling, USB/Firewire, so slow (we would then replace all regular standard wired interfaces on video and computer with HDMI, 10-100GBE Ethernet, and one slow desktop interface version for mice keyboard and etc).

There was originally planned a consumer plug converter like dongle to go between SATA and IDE, that might be an effect bridge though you might not get full performance conversion.

I think Hdforindies mentioned a new HDSDI recorder for much less then $5K. I think real competition is Min computer + drives + HDSDI card + power supply at the moment, which can be made small but still uncompetitive if cost is near enough.

So, you are saying the price includes drives, that is different. But because of import duties in some countries, it might be worth offering it without drives as well, and the user can source the drive model from local discount supplier.

Forget flash for the moment, Pana planing to go to 50Mb/s H264, that explains their P2 strategy all along. I suspect they will go to 25Mb/s H264 for consumer camera and maybe even true 1920, big improvement for movement artifacts, but I expect maybe not such a big improvement over all from DVCPROHD, if it is all straight frames with no inter-frame compression.

Cost of flash is way over, but look at the semi-regular estimates of flash price on time line (as Pana most likely did) you can see what the price will be by the time you can get it to market, and for the first couple of years. Flash has a real limit they expect to reach by the end of the decade, so expect the cost might drop two to upto eight times (guess here, I don't remember the stats). They re replacing flash with cheaper alternatives, there are a number, a number of reports have been in the press over the last 10 years, Intel has one. So, 466.560GB/hr 25fps 1920 10bit 4:2:2 is $37324.8CA retail, less discount for single piece, by 210, even with upto 10 times reduction instead, it is $3732.48CA an hour.

In Indie, you can do short shots, but people might want to do longer shots with it, in that case you could have two banks that allow one to be swapped out and backed up at a time. What about Ram?

If it was 50Mb/s drives I think it would be cheaper, though bigger.


Thanks

Wayne.

[Keith Wakeham]

Assuming that analog out is developed from a perfect digital signal then the timing should be good. The pain is making sure of it, but without the making sure a simple converter can be made. The Analog component still has to semi sync up with SMPTE 274/296 so a simple converter can be made on the cheap. A CLC030 and a CPLD with some simple code along with a few ADC's @ 74.25 and 37.125 Mhz respectively. Something marketable in the 400 dollar range might be expected.

As for HDMI its more of a consumer interface, and being based on the DVI spec I think its still 8bit per channel. Again I don't have all the specs and I haven't considered it because on a shoot these cables will not take a beating and be secured.

Keep in mind 10 bit linear can store 12bit linear by converting to 10bit log, bit of loss but no biggie and I always expect that 2 HDU-1's can be linked for dual 4:2:2 or 12 bit 4:4:4 or 10bit 4:4:4.

Mini computers may eventually give me a run for the money but even then it will come to price war and reliability and power consumption. My electronics consume in total under 5 watts of power. You won't find a laptop that can do that.

Like I said, not a drive reseller but the drives need to be fast enough to save the data, so only a handful exsist that can do that. I have to pay tariffs on anything that passes over the US/Canada border. It sucks but its not my fault, so eventually when I am further I will do my best to help these people out by making sure that they have access to drives that I've tested to be fast enough.

Unfortunately I can't dwell on what could be, only what I have access to. From where I'm sitting thats hard drives and possibly flash mem. Although an SD card based recorder could be ultra tiny, 4 mins isn't much. But then again their is a wide range of uses. Some people migh want standard Film-style shooting with short takes, some want to be able to record straight for an hour.

Not a big fan of compression in general, and more inclinded to intraframe over interframe - no delay for GOP and not mem intensive.

Ram in the ultimate in speed and not safe. Loss power its all gone. Then again its reliable and makes it very much like the digital equivalent of handling film. Theirs still a job for a guy to install film / transfer data off ram.

[Wayne Morellini]

Originally Posted by Keith Wakeham

Assuming that analog out is developed from a perfect digital signal then the timing should be good. The pain is making sure of it, but without the making sure a simple converter can be made. The Analog component still has to semi sync up with SMPTE 274/296 so a simple converter can be made on the cheap. A CLC030 and a CPLD with some simple code along with a few ADC's @ 74.25 and 37.125 Mhz respectively. Something marketable in the 400 dollar range might be expected.

A good bridge design would have buffering to handle all timing problems, and from a low cost Asian manufacturer could be a lot cheaper (but as Pro Video is dominated by the first world consideration, there might not be out there at a cheap price).

HDMI has been upgraded, as part of the Universal Display interface to replace VGA, I am pretty certain it has increased bit depth and data rate, for pro quality in the computer field. I think I might have sent you an email on it some time ago with links to the upgrade of the standard. But as UDI, they have scrapped all but video channel, but still this is what cameras will be moving to in the consumer/prosumer ranges.

If they go to HDMI, component becomes irrelevant 90% of the time, and signal timing becomes assured, making it easier for you to design for the component ;) This also will mostly, become the only option that people with Prosumer and consumer cameras to sue your recorder directly from camera.

My electronics consume in total under 5 watts of power. You won't find a laptop that can do that.

Ahm, yes you can, Intel and VIA are moving towards smaller power consumption and even though you can use the screen as a monitor, it can be turned off to save more power, but this power. Already I think that the Ultra mobile Pc's are marching to under the 5W territory, but they are underpowered, non pun intended ;). But all told it is totally impractical to do 4:4:4 1080 on a laptop, even 4:2:2 8 bit is a bit of a long stretch on a powerful laptop. Bayer si the only really credible option for laptops.

I have to pay tariffs on anything that passes over the US/Canada border. It sucks but its not my fault, so eventually when I am further I will do my best to help these people out by making sure that they have access to drives that I've tested to be fast enough.

Yes, that is what i mean by shipping without and sourcing nationally.

Unfortunately I can't dwell on what could be, only what I have access to. From where I'm sitting thats hard drives and possibly flash mem. Although an SD card based recorder could be ultra tiny, 4 mins isn't much. But then again their is a wide range of uses. Some people might want standard Film-style shooting with short takes, some want to be able to record straight for an hour.

Yes, an event filmer could want to go many hours, not that many will use the extra quality.

Ram in the ultimate in speed and not safe. Loss power its all gone. Then again its reliable and makes it very much like the digital equivalent of handling film. Theirs still a job for a guy to install film / transfer data off ram.

Low powered laptop ram as well. Battery backed Ram, and constant transfer to drives, you only loss a bit if something goes wrong (hopefully, but a fault then it's possible for the Ram to glitch ten to a hundred times a day).

Anyway, back to everyday life.

[Glenn Chan]

Just some ideas, I don't know if they'll help:
1- You know about R&D tax credits already right?
http://www.bdo.ca/en/library/publica...axbenefits.cfm

If you want non-repayable loans, there's some government programs for that (i.e. for young people starting up a business). I'm pretty sure they're non-repayable.

2- You might want to try picking the brain of Jeff Kreines, who worked on the Kinetta camera. It records uncompressed onto a RAID of hard drives.
He posted a little bit about that on the CML list if you want to do a search (i.e. search for 'sorbothane').

Good luck with your product by the way... hopefully you'll solve all the problems you might run into (i.e. hard drive firmware being different, UL/CE certification, avoiding dropped frames, marketing, etc.).

[Wayne Morellini]

An excellent idea for you Keith.