Monday, February 7, 2011

Multi-camera HD mobile capture system

My last big project here at the UK's University of Surrey was fitting out a bluescreen motion-capture studio with eight Thomson Viper cameras, together with timecode and audio. Dell Poweredge boxes fitted with DVS cards give us about 40TB of storage.
This was back in 2004 and the range of hardware and cabling for HDTV was nothing like as prevalent as it is today.
Time moves on - now there is a requirement for a mobile version of our capture equipment for use on film/SFX sets and small sporting venues. This will consist of two suitably equipped 19" rack flight cases, eight Vipers in single/dual/Filmstream mode. Low-end Canon prosumer camcorders will also be available for less rigorous work. The disk-recorders should be able to store dual-stream recording for at least 80 minutes.
I didn't mention - due to the analytical nature of our work, we cannot use any compression!
So at this stage we've chosen the capture-record machines and I'm now accumulating a small pile of gear such as the SPG, Timecode generator, Rasterizer, DAs, viewing matrix, etc., ready for installation.
At the moment I'm expecting capture-box delivery Nov/Dec time; then I can get on with the project's design and installation.

5th October 2009
Happened upon The Battle of River Plate (1956) on TV this afternoon. First time it's been shown in HD here in the UK. The original VistaVision sparkles through with nice saturated colours from the Technicolor print. Well done Channel Four!

15th October 2009
Just ordered 200 BNC connectors and 100m of cable - oh I'm going to be busy when these capture units finally turn up!
Started wiring diagrams using Word's drawing function, not ideal but it worked for my previous studio project.
I shall use the primitives (U-links, DAs etc.) from these old drawings as well.
Should really use Autocad or similar but don't have time to learn it!
TVLogic monitor arrived but must find a suitable flightcase for it. Pictures look OK for confidence use but I still prefer CRT's.

12th November 2009
Capture boxes delivery now late. Manufacturer can't get the solid-state storage to work reliably on very long clips. They suggest we now utilise conventional disk-drives.

23rd November 2009
Just completed the eighth and final conversion of our Viper cameras to external synchs. for the above project.
The Viper HDSDI back normally only accepts external refs. via the multiway connector.
Using the thick multicore cables in the field would have been expensive and cumbersome, therefore I decided to convert each camera to accept synchs. via a separate BNC connector.
I did not want to drill or modify the case so a small ABS box was bolted
underneath the HDSDI back, just behind the shoulder pad.
Wiring involved virtually dismantling the HDSDI back to access the rear of the multiway socket.
It is physically possible to break into the circuit at a more convenient point, but that would bypass the common-mode/EMC input filters.

24th November 2009
It's annoying when you try to do the best thing technically - and fail!
Last week I needed to temporarily split a short feed of 800mV HDSDI to two separate monitors. I couldn't justify a $600 DA so decided to build a small 3 x 27 ohm star-splitter using copper board and the smallest resistors I could find.
Did it work? - no.
Next I tried using a 'goalpost' type BNC T-piece for the two-way split - this of course results in a double termination and a return loss nightmare.
Again, did it work? - YES!
Final result:- Botch 1 : The Better Way 0

22nd December 2009
Glory be - the four DVS ProntoXways (http://www.dvs.de/products/video-systems/pronto-family.html) arrived last week. They are mounted in two 16U shockproof KALMS flight-cases and weigh approx. 80Kg/177 pounds each* (this is a transportable not portable system!).
I will have to re-design some parts of the project as there is less space than envisaged; also the racks are powered individually by 16A cords. As each rack, when completed, will additionally be interconnected with data/audio/synch./HDSDI cables etc., I didn't want to create a potential earth-loop. As planned, rack #2's mains-power will be sourced from rack #1, it's earth coming from a star-point located within the latter.
Christmas is nearly upon us now so I won't have a chance to progress until early next month, although I am currently now sitting here re-drawing my plans! Season greetings!
* this will of course increase when I've fitted the rest of the equipment.

26th December 2009
Just thinking of other uses for this system:- stereoscopic x 4 channels, digital Cinerama?

17th February 2010
Happy New Year! (Indulge me - my tardiness is due to a busy last few weeks!)
Xways are now being tested for reliable storage over several hours and, most importantly for us, no dropped frames. Each machine has 2 independent channels with it's own Windows GUI running the DVS ClipStore software. There are 4 machines so therefore 8 desktops controlling each channel respectively. To keep the amount of hardware down there are 2 KVM switches feeding 2 Lindy monitor/KB/mouse terminals.
Setting up each machine takes a while due to having to keep switching desktops. A source of potential confusion is that you have to set each channel to it's own storage disk eg., Channel 1 to V: and Channel 2 to W:
It is possible to erroneously set both channels to the same disk - this causes frames to drop eventually as the drives get more full.
Obviously I am keen to get some kind of script running to preconfigure each machine and minimise operator error!

23rd February 2010
Just had a rummage around the disk-drives under the front covers. After this amount of time it might sound like a lack of curiosity on my part but in my defence I also have a sick DMX lighting controller, a burned-up DA, and a multiline to-do list sitting on my desk!
Each machine has 2 x 2.8TB RAID 5 storage which is enough for over two hours of 4:4:4
per channel.
The individual drives are Seagate Cheetah 15K.7 (http://www.seagate.com/www/en-us/products/servers/cheetah/cheetah_15k.7/) mounted in removable harnesses for easy hot-swapping.

23rd March 2010
Just watched Niagara (1953) on TV. This film seems to have noticable colour-fringing in some scenes. Was this print struck from shrunken matrices? I wouldn't have thought the Technicolor guys would have let it pass at source so I wonder what has happened since?
Haven't seen the DVD version yet - any comments?

2nd April 2010
For the purposes of sane documentation I needed to uniquely indentify each of the two equipment racks. To keep the operating ergonomics and inter-unit cabling tidy it was decided that when set up on location the cases would always appear as left and right respectively (does this make sense?).
It was difficult to keep refering to "the one on the left with the SPG" or, "the right-hand one with slightly less stuff inside".
It had to be something easily memorable, for example - Laurel & Hardy, Hewy and Dewy, Ant & Dec*, etc,.
I'm afraid it came out really dull:- left side = Port = Red, right side = Starboard = Green.
Therefore they are now individually known as either the Red or Green rack!

* for non-UK readers they are popular TV presenters here.
7th April 2010
It's been suggested here at the research-centre that we should investigate multi-camera setups without the benefit of synchronised shutters ie., no genlock. This would be advantageous for witness camera applications in green-screen studios - one less cable to rig/trip over.
A couple of years age I built a device which consisted of a 'rolling LED' strip which was synchronised with SD synchs. When individual frames were captured and subsequently analysed the illuminated LED(s) would show at what point the camera shutter was open (and for how long).
I now need to redesign and improve this box for HDTV experiments but this is going to have to follow this current project, so we're talking >June perhaps.....

23rd April 2010
Increasingly irritated with the 3rd-party supplied flight-case racking system. Instead of conventional equally-spaced cage-nut holes it has a conduit slot running from top to bottom. This means having to source and fit (hard-to-find) fiddly spring nuts at the desired locations.
Unfortunately "at the desired locations" can mean anywhere in the slot, so you not only have to space them apart the correct distance vertically you also have to ensure that the fitted unit is horizontally level. I've never before had to use a spirit-level to fit 19" rack equipment!
An additional nuisance is that the rack's depth is quite deep which, while not so important for the short stuff, try fitting a Tektronix TG700 SPG. This unit has a sliding mount which fits not only at the front, but also has to be fixed at the rear of the rack as well. This is accomplished in normal situations by using two supplied slotted metal strips which can be adjusted to the desired depth.
Yes, you may have guessed it, S**'s Law - the Tek's metalwork was 1" too short.
Luckily there is a vertical conduit facing inwards at the back of the flight-case (I'm not sure why but mustn't look a gift-horse etc....). If I bend out the Tek's metal strip 90 degrees outwards then it might just meet. Well after a session involving a vice and a large pair of pliers I managed to straighten out the whole thing.
Offering up the TG700's sliding mechanism showed that the aformentioned Law was still in force, now the slot was 1/10th" short and was also slightly too small for the M5 screw needed to fit it to conduit's spring-nut.
This outcome causeth me to rend my garments and utter bad words. It being a Friday afternoon I decided that someday I will laugh about this and should leave it until next week.
Stay tuned.

30th April 2010
Recomposed this week. TG700's metalwork placed in vice, end holes cut through into U-shape and then offered up back to the rackmounts - perfect fit - job done.
The disk-recorders are now being used for a couple of weeks (on a limited basis) for an upcoming European Union project. Rather than lose any assembly time I decide to
separately build the wiring loom for the termination panels at the rear of each rack.
Each unit will have two rows of 20 x BNC which provide interconnection with inputs from the cameras (links A and B), outputs for TLS reference synch., and LTC.
Eight of the lower BNCs will mate with a main interconnecting 'umbilical-cord' which carries HDSDI signals and control data signals.
The rear termination cable for each camera BNC is 1694 type. After a long cable run the signal will be of low amplitude so any further losses must be minimal. The other cables used within the racks are much less critical and are of a smaller, slightly more lossy type.

5th July 2010
Well my last 'more soon' turned into 'much much later' - it all turned very busy, very rapidly. A European Union sponsored research project turned up at the last moment, they needed my mobile HD capture system and all 8 Viper cameras in a special inline configuration.
The camera views needed would give 4 x stereo and 8 x 1 for later use with a pseudo-holographic monitor.
The only silver-lining was the fact that they would be using the existing studio's infrastructure.
As usual, the fact that the units were not yet complete didn't come into the equation, so there was a lot of last-minute 'temporary' cabling.
They wouldn't have the viewing matrix, rasterizer or patch panels - the DAs for the synch. and LTC were wired and cable-tied to the inside of the racks.
Unusually for an EU project the requirement was for 720/50p. The existing Tektronix TG700 doesn't do this standard so an AJA GEN10 spg was pressed into service.
To cut a long story short the capture sessions were very successful and I think everyone was pleased with the results.
Just as that storm had passed it was decided then that London-based Framestore would like to use the mobile units for witness camera tests! As this would be in the field we would need 24 x 30m cables (Return HDSDI, TLS, LTC for 8 cameras). There was no time for making neat multiway cables so I ordered a drum of BD SD19 coax, a lower-loss version of 1694.
Of course the devil is in the detail, I also needed BNC connecters plus strain relief boots of various colours. No time for neatness, each cameras cable was assembled from 3 coax cables wrapped at 1m intervals with cable-ties. Not tidy, and a pain to spool up, but time was against us.
As the units would have to endure a journey in a truck I had to anchor down my 'temporary' system with many more cable-ties and velcro-wraps!
I spent 2 whole days measuring cable and making over 60 BNC terminations!
To make matters worse I would be on vacation the week the units were to be used, therefore I had to train up a non-video engineering person as to how it all went together. All the cables were copiously labelled.
As with the last project I was told that it all went well with no serious problems - in fact user's feedback is vital in getting the finished units just right.
Now things have quietened down I can hopefully finish this project soon(?)

6th July 2010
'Proper' cable for the cameras now on order for future jobs - Belden 7710A (3 x 1694A inside a 20mm PVC sheath). This will give us adequate performance over 30m with predicted loss of approx. 9dB @ 1.5Ghz. One of my future projects will be to try and measure the loss over the various BNC connections, I know it's small but I just like to know these things.
There is a possibility of dirt ingress and the PVC splitting at the 'Y' point where the 3 individual cables emerge from the sheath. A neat (but expensive) solution is Tyco Electronic's 462A023-25-0 heatshrink boot/transition.
I've tried an intial test and, once you've got the hang of heating it into shape with the heatgun, does do a good job of sealing the cable.

7th July 2010
Not many cartoons about HDTV but take a look at http://www.xkcd.com/732/
I shall say nothing. . . . . . . .

8th July 2010
I'm no football (soccer) fan but whilst watching last night's semi-final of the World Cup 2010 it occurred to me how far we've come in television engineering. Live high-definition TV pictures from South Africa where you could almost pick out the facial features of the crowd. I remember back in the early 1970s watching analogue pictures from South America of the World Cup through a fog of spatial and motion artifacts - cutting-edge 525 to 625 line standards-converters built out of discreet TTL logic!
I believe the BBC had a analogue converter built out of ultrasonic delay-lines, but that's just ancient history!
Looking back at literature of the time it's interesting to note how much successful R&D into spatial and temporal components of the video signal was being done. This was at a time when microprocessors were way too slow for video - all signal processing was achieved using 5 volt TTL adders/multipliers plus some early RAM chips.

24th September 2010
Time scoots by and I'm falling behind, not my fault, these units are just too darned popular!
I've fitted both the viewing matrix, rasterizer and made the 10-way interconnection cable but haven't been able to get near the machines!
Note to up and coming project engineers - never let the users near your installation until you've finally finished it!
In the meantime I've been assembling the eight 3 x multiways with the boots, BNCs etc., mentioned above.
Highlight was melting the Tyco fittings mentioned above into shape in a room that's already 27C (80F). These heatshrink boots take a fair amount of heat so a 2kW heatgun takes at least two minutes to become pliable. When shrunk the fittings are still very hot so they need cooling under a fan before wound onto the cable drum.

At least now I could use my latest toy, the Harris HDStar to test the cables.

This is a convenient test generator/monitor for SD and HD. The in and out BNCs are at the top of the unit so making cable checks really quick and easy.
I love this little box although it's quite heavy on the Li-Ion battery; also the recharge rate seems spectacularly slow.

23rd December 2010
Christmas break at last so now I have a chance to update the blog. Someone uses the cable reels mentioned above in the studio and decides that it's a nuisance having to unravel the whole 30 metres in order to get to the BNCs at the far end. I pondered earlier on whether to terminate the ends with just BNC tails or to fit bulkhead sockets into the cable reel's centre plate. Looks I'll have to do this after all - that will mean making 8 x 2 metre multiway cables and fitting BNCs, boots and transition joints. A job for the new year - deep joy!!

29th December 2010
Channel 4 here in the UK is, at this moment of writing, showing David Lean's This Happy Breed (1944) - what a stunning print and a fine example of 3-strip Technicolor from the British plant.
Talking of Technicolor I recently caught the end of Anchors Aweigh (1945) and noticed some horrible RGB misregistration in some shots. It looked camera-specific as it would alternate with 'good' registration between shots within the same scene.

14th January 2011
Oh dear, a cloud has appeared in the paradise of ProntoXway ownership. Number one machine had some flakey bootups and has now died completely.

17th January 2011
New system drive received from DVS so machine is removed from it's flight rack.
Wow - it's !*@@*!! heavy but once put on my (straining) worktop I'll open her up and take a look around inside the machine.

18th January 2011
Of course the ProntoXway is typically German top-dollar engineering - can't fault it (except the dud hard-drive of course!). From the top I can see a large motherboard kept cool by a small hurricane from the fans mounted across the case. System-drive C: is mounted in a bracket at the front, held in a by a screw and mounting lugs so it's a breeze to change.
OK, disk now replaced so just have to re-image with the GHOST disk supplied by DVS.

4th February 2011
NOTE: Old/seasoned IT hands may wish to skip this entry.......
Oh brother, what a saga that GHOST reimage turned into. Loaded the disk and the usual Linux screed raced down the screen. When prompted which device to install Windows OS onto you have (in Linuxspeak) a choice of sda, sdb, or sdc. The prompt suggested ">sda?" which, in a hurry I did - big mistake! As I found out later the Xway's devices sda, sdb, and sdc relate to RAID array #1, RAID array #2, and the 4GB system drive (C:) respectively - therefore sdc should have been selected.
To cut a long saga short I now had Windows installed on a RAID drive which would normally be used for media files only. Windows will not allow you to destroy itself through disk manager so somehow I had to get 'outside' the system to flatten the RAID drive. Thankfully I found this wonderful program SystemRescueCd on the net which creates a bootable Linux CD with a miriad of disk maintenance tools, some of which work with NTFS.
Using this wonderful software I was able to wipe the RAID drive and load the image onto the correct drive - sdc (aka C:).
Moral of the story is clear - before pressing the Enter key read the options given to you in the small print!

5th March 2011
Readers of this blog may be interested in an article that I have written for the Spring 2011 issue of the UK's Guild of Television Cameramen's magazine 'Zerb'. This article concentrates on a multiview research project held last Summer 2010 at the University of Surrey, and hints at the (still) distant prospect of holographic TV. The aforementioned mobile capture system and Viper cameras were used for this session.
Copies for non GTC members can be ordered from zerb.production.07@gtc.org.uk