Sunday, 28 June 2015

Archiving Hi8 Video Tapes to MP4

So recently i was reading up about the Hi8 Video tape format, this is what i used to use back in the 1990s and 2000s with my Sony TRV-66E camcorder.

I have a significant amount of old tapes that never get played and read that the Hi8 tape format is prone to degradation after about 15 years.

So with that in mind i began to transfer my tapes into H.264 / AAC streams for proper archiving and uploading to YouTube, so these are my experiences.

Tape Degradation
I have viewed some of my tapes and some do show some form of playback artefacts. As you can see in this picture below the bottom half is affected by distortion and horizontal sparkle effects and runs throughout the entire tape. Visually there is nothing visible on the physical tape itself.

Digitizing The Media
This is a lossy step in the process, that means you can affect the quality of the final recording depending on how you do this.

I dont have a PC based video capture device, it's not something i have needed before. I have used them in the past (during the late 1990s and early 2000s) with mixed results, maybe better products are out there now but i generally dont like them!

I do however, have a decent quality hard disk based DVR and DVD writer (a Sony RDR-HX710). This will allow me to record my tapes into an MPEG2 stream either directly onto a DVD+R or to the internal HDD for later transfer to DVD+R.

The DVR can accept analog video in the forms of composite or  S-Video. S-Video is obviously the preferred format here as it separates parts of the video signal to retain more detail.

Because the DVR has only a single layer DVD writer i am limited to standard 4.7Gb DVD+R or DVD+RW so i need to try and fit one whole Hi8 tape onto one DVD. I want to make the 90 minutes fill the disk. If you encode the tape so it only fills 1/3 of the DVD your effectively losing quality because you may find a slightly high recording quality in the DVR settings that could fill the disk with nothing left over.

After some testing i found the most efficient way to record the tapes was to write DVD+Rs directly. This saves me time as there is no HDD -> DVD recording process. I also found using the DVR quality setting on 'HSP' (this is only a Sony naming convention, yours may be different). This encodes at about 7.7 Mbps.

Using the DVR is great, it means i can set play/record and leave it to finish,  it doesn't tie up my PC for hours and i don't have to worry about wonky capture device drivers that will drop frames if you so much as look at the PC the wrong way, or audio that drifts out of sync. It also means i can just archive my tapes onto many DVDs that can be ripped and edited at a later date. In addition to this the output is pure PAL standard 720x576 25FPS Interlaced.

Transferring to PC
This part should have been the simplest, but in reality it was not but in a way it helped the process in a funny way.

After the DVD is written and finalised i can take the DVD and read off the files on my PC. All DVDs are readable this way with the exception of copyright protected disks that have their files encrypted. Because this is an original recording from my DVR there is no encryption so the files can be copied without any issue.

The files that contain the actual content are the VTS_01_[X].VOB located in the VIDEO_TS folder. Copy and rename these on the PC to .MPG and they will play in most media players.

This is where i hit my first problem, for some reason the first VOB file wont play in VLC or load into Movie Studio. It's corrupt or badly formed in some way. It plays fine on the DVR and in Windows Media Player. Thankfully it does load into Handbrake. I have transferred recordings using this method before without issue so i may look into this further to see which bit i'm doing is causing the problem.

I wanted to edit and make the final videos from this original MPEG2 source in Movie Studio but there was a lot of noise in the encoding, small MPEG blocking artefacts and high frequency noise from the original tape and i had this issue with the corrupt file. Movie Studio has little or no options for removing blocking artefacts or noise too so i needed another solution...

After some experimenting i found i could encode these MPEG2 streams into MPEG4 using Handbrake which can read the corrupt file and apply some additional processing to reduce some of the artefacts so in a way it helps even though it adds an additional encoding step and more time to the process. So my handbrake settings were as follows:

Picture Tab: 

Anamorphic: Strict
Cropping: custom, set to zero. Handbrake does a poor job of cropping and i prefer to use the pan/scan options in Movie Studio. So i disable this option, annoyingly it will always revert back on any new video.

Filters Tab:
Deinterlace: Fast, being an original PAL source the picture is interlaced so  this should be removed prior to encoding into the final video.
Deblock & Denoise: Off

Video Tab:
Quality: 12
Constant Framerate: On
Framerate: Same As Source
x264 Preset: Medium
x264 Tune: Grain
H.264 Profile: High

Alternative Method, less filtering, lower compression.
This method i disable the deblock and denoise functions and increase the compression Quality enough so the inherent grain is retained. This actually looks more natural. There is a balance though, you need enough quality to resolve the noise. Not enough and the noise will turn into very obvious blocking itself. In my tests a Quality setting of 12 and disabling the deblock and denoise functions.

Final Differences
I was going to include some half and half images between the MPEG2 and MP4 versions but the differences are so subtle i doubt you would be able to see them. It really is just some high frequency noise.

The original MPEG2 .VOB files are around 1Gb in size, after conversion to MP4 they do increase to more than 2.5Gb, i found this was inevitable to retain the grain, this is only an intermediate format for editing though. Once the videos have been rejoined, edited, cropped and some basic level and gamma correction they are exported from Sony Movie Studio as a 5mbps MP4.

Thursday, 18 June 2015

Update: 1980s Digital Video Effects Processor

Just some small snippets of extra information about the CEL Electronics P164 & P152B video effects combo:

The P177 connector at the back is for the 'P177 Component Combiner'.

Also additional devices related to this are the 'P158 Eric Editing System Machine Interface' and 'P163 Digital Effects Controller'.

The P164 & P152 form a system called the 'CEL MS850B Effects System', other variants of this are the MS851B with the addition of a BIM (Built In Mixer) and the MS852B which is basically one P152 Touch Screen Controller and two P164 DVEs with the BIM option. The touch screen controller can operate multiple P164s at the same time.

After some lengthy reading about similar DVE systems from Quantel, Sony and Abekas during the mid-80s i believe this is run of the mill system. Maybe offering similar performance to these other high-end and seriously expensive systems but at lower cost. Given it's date (the user guide is written in 1990) it does seem somewhat lacking in capabilities. But this remains somewhat unknown until i can get the system running.

If anyone has any information about any of these devices, please let me know by leaving a message in the comments.

P152 & P164 EPROMs

I have also dumped all of the EPROMs from both the P152 and P164. Interestingly the P164 does have a number of ASCII text strings contained in the CPU Board EPROMs.

ROMCP0 & ROMCP1 are interleaved so the two 32kbyte EPROM dumps need to be merged to form a single 64kbyte binary.

The strings seem to relate to a DSP programming or debug interface as these code snippets show extracted from the CPU board EPROM.

H-help: M=modifyD=dumpF=fill PROGRAM
m=modifyd=dumpf=fill DATA
R=regdumpG=gotoK=code upload
B=set breakpointC=continue
I=int enableX=int disable
>>> TMS320C25 Diagnostic Kernel V9.0 <<<
NO 16k- 32k- RAM system
EPROM ID: Storage Processor DPRAM
** TEST OK **
EPROM boot block not found
Non-volatile RAM is corrupt
Serial Ports
I have connected to all of the serial ports available, all those on the P152B do nothing as i expected. The 'Controller' port on the P164 makes my USB->RS232 adaptor drop its USB connection so may not be RS232 but the 'Panel' port outputs a continuous stream of '[00]' at 9600/8/N/1

Wednesday, 17 June 2015

Teardown: Professional 1980s Digital Video Effects [DVE] Processor

This is a teardown feature of two units that form a professional grade video effects system from the late 1980s.

The two devices in question are the CEL Electronics P152B Universal Touch Screen Controller which is clearly designed to be mounted into a video editing desk/studio and it's processor which is the CEL Electronics P164 Model 38XP Digital Effects Framestore - TBC designed to be rack mounted. Clearly intended to be used in high-end professional installations.

CEL Electronics which was a UK Limited company does not appear to exist any more but the name does seem to be owned by a company called Snell Group (which is part of Quantel), which does manufacture hardware and software solutions for the broadcast & video production industry. Unfortunately there is very little information available on the internet for these two old devices or even the company CEL Electronics. There are mentions of the company on a few wiki pages relating to video processing and devices but as of writing there is only one page that has any information at all: P152B Information & P164 Information.

From what i can understand the P152B is a universal controller, which is why the legend on the panel only refers to basic functions, the system needs to be booted with system disks (3.5" floppy disks) for the specific device it is to control, in this case the P164. Without these disks the system is essentially non-functional. Sadly i wasn't given anything other than the two units when i bought this so i am unable to make any thing work at the moment.

CEL Electronics P164 38XP Digital Effects Framestore - TBC
This rack mountable unit is clearly the main processor for the system, at the back there are 75ohm BNC connections for what appears to be four video inputs and outputs including additional I/O for syncs, chroma keying, component and composite outputs. There are also several multiway 'D' connectors, their use unknown.

At the front there is an access panel with ventilation guard, behind this are the eleven plug in PCBs which appear to be separated into analog (right) and digital sections (left) which plug into a backplane. Behind this is a switch mode power supply and output connector panels.

It is difficult to determine the exact purpose of each PCB and what capabilities the system has but here i will look at each board in turn and see if we can figure some of this out. Essentially it will digitize incoming video, process it and then output the results back out to analog video so the system will essentially flow as:  Inputs > ADC > Processing > DAC > Outputs.

This is not like the famed Quantel Paintbox, but (i believe) rather a hardware device for basic transitions, fades, wipes and maybe chroma keying, pan & zoom etc. I also understand it can be used for format conversion, ie NTSC to PAL etc.

CEL P164 Rear Panel

SPF1 Board.
The screen print indicates this is processing some of the Chroma information. There are eight large custom CEL ASICs named 'CEL KABOC'.

SPF2 Board.
Similar board to SPF1 but this time operating on the Luminance, with the same ASICs as the SPF1 board.

FSM Board.
There are two identical boards like this. Clearly FSM is Frame Store Memory. The DRAMs are Goldstar 256k by 4 bit devices with 16 devices in each bank with a total of 4 banks over the two boards. This totals 2 MBytes RAM for the system. The DRAM controllers are custom marked with CEL but also with the National Semiconductor logo so maybe a custom part.

BBO2 Board.
Not much to go on with this board, mostly populated with 74 series logic and custom PAL/GAL devices with another custom CEL ASIC labelled 'SX-8'

IOT Board.
This board is likely some kind of IO control, maybe interpreting the commands & data from the P152B and the various ports on the back.

CPU Board.
No prizes for guessing this one, with the title CPU is clearly some kind of main processing board. There is a TI TMS 320C25FN DSP with external SRAM and EPROMs. Some of the SRAM is unpopulated so can probably be upgraded. A Rockwell R65C52P device is a dual communications interface.

PM1 Board.
Although this appears on the analog side of the system is does seem to be digital in nature, a good dose of 74 series logic and two large IDT 71325A55P 2KByte dual port SRAMs with 3 more of the CEL KABOC ASICs. Maybe this is some kind of buffer RAM?

BIM Board.
Just a blank spacer board. The BIM is Built-In-Mixer and is an option for the CEL P164 not fitted to my unit.

IVI Board.
No idea on this one!

ADC Board.
The Analog To Digital Conversion board, three Exar MP7684 8 bit 20Msps ADCs. This is in the ADC board to convert the analog composite and component video inputs to digital. I would say this would be one ADC for the YUV component video channels. Notice the large number of adjustment pots and varicaps, also the analog delay lines in the green and black rectangles, you can see how  these work in EEVBlog Episode #381.

ENC Board.
The counterpart to the ADC board, with 3 TRW 1016N7C8 Video DACs to provide composite and component YUV video outputs. Again like the ADC board there are many adjustments!!

CEL Electronics P152-B Universal Touch Screen Controller
This is the controller of the system, although it appears to be rack mountable i dont believe thats how it would have been installed. From an ergonomics point of view i think this would be installed face up in a video editing desk to allow easy use of the faders and the joystick. My assumption is you can create an effects program here which is processed by the P164 unit.

The P152 appears to be generic in nature and probably was a controller for several different systems, the legend on the front only gives basic ideas of the functions. On power up it tries to read the floppy disk and reports no disk found so i would expect a program for each of the processing units this can operate can be loaded in as needed.

The so-called touch screen is a simple IR emitter / receiver matrix over the green screen CRT, placing your finger on a 'button' would break two beams indicating your finger location. Primitive but i am sure it works!

On the front panel we have menu buttons VTR, FX, MIX, SEQ, SET UP & U. Visibly the 'FX' button has been used the most. Below that we have Take; Go & Stop then a Spinwheel which is a nicely damped 360 degree rotational encoder. Finally on this side we have 8 function keys relating to menus shown on the CRT.

On the other side we have three adjustment knobs, the floppy disk drive, two transition faders and a joystick. The faders and the joystick are of very high quality, damped but smooth and light to touch with most of the parts made from machined aluminium or steel. The joystick self centres and is 3 axis, X & Y being up, down, left and right with the 3rd axis is provided by the rotation of the black handle which rotates about 20 degrees in either direction and self centres.

Note the 'Maurice II' underneath the monitor, could be a brand for the system, internal CEL Codename for the project and possibly a reference to Maurice Leblanc?

On the back we have an impressive six serial ports, video output (presumably mirroring the internal CRT) and connection for a 2nd floppy disk drive.

A full teardown video is available on my YouTube channel:

Wednesday, 10 June 2015

Radiation Anomaly A49 Warrington

On Monday 7th June i was on my way to collect my latest teardown project from my favourite ebay seller. It's a short trip along the M56 and onto the A49 Tarporley Road outside Warrington.

Of course i had my Ludlum radiation meter with me (doesn't everyone always have theirs with them!?) as i drove i heard my Ludlum suddenly start to click away much more than before then calm down after a few seconds.

On my drive back i had the same thing happen at the same point so turned around and went past several times to verify.

The next day i went back and took video...

The clip below shows me driving past the area in question, turning around and going past in the other direction.

The inset gauge is my Ludlum with my very sensitive scintillation probe, typically i would expect to see between 700 and 2500 CPM background radiation depending on where i am. On this journey it was showing between 1000 and 2000 CPM.

When i pass through this particular area it quickly rises to about 3800 CPM and then dies away after a few seconds. It seems to be higher on the return just as i pass Brookside Fisheries.

Most likely this is just some mineral deposit giving this area slightly higher background. It is completely normal to find patches but does depend where you are. The South West of the UK is well known for this, but less so for the North West particularly in this area around Manchester, Warrington, Crewe & Macclesfield etc.