Quantel Ramcorder - Digital Video RAM Store

[Alex Eisenhut]

I like Mike Oldfield but he's been like a cafeteria for decades now: serving reheated leftovers.

But so has Jean Michel Jarre...

[dexters_lab]

I like Mike Oldfield but he's been like a cafeteria for decades now: serving reheated leftovers.

But so has Jean Michel Jarre...

yea, i know what you mean... Guitars is a nice change from his other works

[Willyarma]

Wow, I've only just found out about this! I've been watching a lot of your Quantel videos today and am absolultely blown away with what these devices were capable of considering how old they are! I'm very impressed with the repair jobs too, especially the PAL hacking!
That Mirage unit is something else! It blew my socks off, its incredible. It got me thinking if I could do something Mirage-ish with my last FPGA project which was an SRAM based live HDMI full HD roto-zoomer, I can address any pixel in any order. VIDs of it have been on youtube for about 4 months.

[dexters_lab]

Thanks, threre is more quantel magic on the way to the channel soon, i have been trying to coax a Quantel Hal into full life recently and it makes the Paintbox and Ramcorder look a bit basic really.

[mikeselectricstuff]

Wow, I've only just found out about this! I've been watching a lot of your Quantel videos today and am absolultely blown away with what these devices were capable of considering how old they are! I'm very impressed with the repair jobs too, especially the PAL hacking!
That Mirage unit is something else! It blew my socks off, its incredible. It got me thinking if I could do something Mirage-ish with my last FPGA project which was an SRAM based live HDMI full HD roto-zoomer, I can address any pixel in any order. VIDs of it have been on youtube for about 4 months.

If you can do any pixel in any order, presumably the next step is to have a pixel address lookup table so you could do any arbitary transformation. Size of LUT might be an issue though...

[dexters_lab]

nice videos (thank Mike for finding and posting)

yea that's just the sort of thing quantel were doing with the Mirage

most of the effects though that looked like that were done with switching and effects desks like those from grass valley and sony (amongst others), mirage didn't have a long life at quantel, there is almost no mention of it towards the end of the 80s it seems

[Willyarma]

Yeah, thers not much about it online.
(Those vids were mine BTW, when i first posted they got removed somehow so I went back and edited the post)
Ignoring the LUT size for the moment (which is a hurdle) a static image would be easy, but how on earth did they make it animate and turn etc.. was it hard coded in to the logic? I heard it was programmable but I have a hard time imagining a computer in the 1980s being able to calculate a 1024x1024 grid of dots at 50fps! Was it all pre calculated for every frame in a ginormous ram (seems crazy)? In my rotozoomer, all the fast stuff is hard coded into the logic of the FPGA, only the gradients and starting positions are changed externally to make it turn & zoom.

[james_s]

Hey that's pretty cool! Did you release the HDL for that? Looks like something that would be fun to play with.

[dexters_lab]

Yeah, thers not much about it online.
(Those vids were mine BTW, when i first posted they got removed somehow so I went back and edited the post)
Ignoring the LUT size for the moment (which is a hurdle) a static image would be easy, but how on earth did they make it animate and turn etc.. was it hard coded in to the logic? I heard it was programmable but I have a hard time imagining a computer in the 1980s being able to calculate a 1024x1024 grid of dots at 50fps! Was it all pre calculated for every frame in a ginormous ram (seems crazy)? In my rotozoomer, all the fast stuff is hard coded into the logic of the FPGA, only the gradients and starting positions are changed externally to make it turn & zoom.

i've been told the Mirage was a box that processed the video and had all the magic in but had a mini-computer that controlled it, i know Quantel used a lot of VAX stuff so could well have been a VAX. How the two communicated and how the computer calculated and stored all the 3D stuff i have no idea. If i ever happen upon details i will post them up.

[Willyarma]

Is there a pinout for the RAM store card?
Is the DRAM refresh circuitry on board?
It's got me wondering if it could be recreated with modern parts, it would be a doddle if its just address bus + data bus on the connector, I expect its a bit more complex than that though!

[dexters_lab]

Is there a pinout for the RAM store card?
Is the DRAM refresh circuitry on board?
It's got me wondering if it could be recreated with modern parts, it would be a doddle if its just address bus + data bus on the connector, I expect its a bit more complex than that though!

no, there is no pinout, i would have do make one... but  there is a picture on page 1 of this thread that shows a closeup of the bunch of logic at at the connector, that could be part of the refresh

We have looked at re-creating the entire ramcorder, it would actually be simpler to engineer it from the other way. The video in/out are SDI so that can easily be re-created then it's just a case of deciphering the rs232 protocol it uses. A tame fpga and a stick of old PC ram would suffice to replace the whole thing. I have wondered if the ramcorder tells the paintbox how many frames it has available which could potentially allow a replacement to have gigabytes of ram for a huge increase in video storage space.

I've never looked at the protocol though, i should get out my hantek LA and see what it says

[james_s]

It's pretty crazy to think of replacing that whole huge thing with a little board containing an FPGA and a stick of RAM.

[Willyarma]

Thats a much simpler and better idea!
I would love to see the protocol deciphered, i think it would make a great video.
I assume the SDI signal is 576i and not 576p?

I'm just reading Xilinx XAPP1076 - Implementing Triple-Rate SDI with Spartan-6 FPGA GTP Transceivers.... looks like SDI on FPGA is possible, its a big document though!

On a side note I remembered about some cheap HDMI to SDI and SDI to HDMI convertors that I saw on eBay (mainly used in CCTV) they might be handy for this system to use on a modern screen and to record etc.. but I dont think HDMI devices like 576i very much, so maybe not.

[Willyarma]

Hey that's pretty cool! Did you release the HDL for that? Looks like something that would be fun to play with.

Not for this one - at least not yet, it was so difficult to get right, one tiny change would throw out all the timing, I need to have another go with a different approach.

It's pretty crazy to think of replacing that whole huge thing with a little board containing an FPGA and a stick of RAM.

I know. I really like seeing moderen stuff made for old hardware and also the hacking and reverse engineering of the signals and protcols to get it to work is really interesting to me!

[dexters_lab]

right well i spent most of today with my LA on the serial link and made a bit of progress, i should have livestreamed it but didn't think about it at the time!

the link is RS232 at 9600bps 8/n/1, the protocol is simple, but more complex than i expected so it's work-in-progress. For example it takes 11 bytes from the paintbox to get the ramcorder to display a single frame. Even playing continuous video requires the paintbox to request every single frame in realtime.

the 323 frames limit looks to be hard coded into the paintbox software so probably no scope to expand it, i couldn't see anything that informs the paintbox the size of the storage.

the ramcorder can display and record not only individual frames but also individual fields of video, i spotted this in the protocol and verified it


to reply to your posts...

I assume the SDI signal is 576i and not 576p?

yes, the paintbox and ramcorder both are 576i. The interlace is entrenched in the paintbox design and i don't believe they ever offered progressive scan on this architecture.

looks like SDI on FPGA is possible, its a big document though!

we have SD-SDI running on a Cyclone 2 thanks to user OzOnE, again another unfinished project, maybe we'll brush the dust off it if i make more progress on the protocol

I would love to see the protocol deciphered, i think it would make a great video.

noted, once i make a bit more progress i will look at putting something together

On a side note I remembered about some cheap HDMI to SDI and SDI to HDMI convertors that I saw on eBay (mainly used in CCTV) they might be handy for this system to use on a modern screen and to record etc

thankfully i have a Blackmagic unit that will record SDI to a SSD as uncompressed video which can easily be converted on my PC. I also have a couple of really nice pro scalers and converters i picked up cheap off ebay, means i can convert pretty much any type of video in and out of the paintbox.

[dexters_lab]

if anyone is interested or maybe able to help?...

here is a little summary of what we found in the ramcorder protocol...

the paintbox sends a string of bytes to the ramcorder, the string can be variable length and includes a checksum, the replies back from the ramcorder are similar in they are variable length and have a checksum.

looking at this set of packets which happens when i enable the ramcorder option on the paintbox:
PB = Paintbox TX, RC= Ramcorder TX

Code: [Select]

PB: F7 53 09 50 08 12 00 00 00 43
RC: FC 41 01 07 08 12 00 00 00 04 01 43 09 4A 06
PB: F2 4B 0A 00 B9
RC: F2 41 01 04 C8
PB: F7 53 09 50 08 12 00 00 00 43
RC: FC 41 01 00 08 12 00 00 00 04 01 43 09 4A 0D
PB: F0 73 9D
RC: F5 41 01 00 00 05 00 00 C4
The first byte always has the upper 4 bits as 1, the lower 4 bits are the payload/string length excluding the first and last bytes. The last byte is a basic  checksum of the payload including the first byte.

So the first message from the Paintbox is:
FC 41 01 07 08 12 00 00 00 04 01 43 09 4A 06

FC = length 13 bytes
41 01 07 08 12 00 00 00 04 01 43 09 4A = string
06 = checksum of string including the first byte

Interestingly bytes 10 and 11 look like a 16 bit value 0x0143 which is 323, exactly the number of frames available in the ramcorder. So it could well be informing the paintbox of how many frames are available.

Reading/Writing frames & fields. Quantel call this buying and selling frames.

the following is the comms that are selling (writing) a frame to frame 1 of the ramcorder:

Code: [Select]

PB: FA 73 09 50 02 01 00 07 00 00 00 01 2F
RC: F5 41 01 00 05 00 01 C3
PB: F0 52 BE
RC: F2 41 01 00 CC
PB: F0 73 9D
RC: F5 41 01 00 05 00 01 C3
Selling a frame into frame 2 of the ramcorder:

Code: [Select]

PB: FA 73 09 05 02 01 00 07 00 02 00 03 2B
RC: F5 41 01 00 05 00 01 C3
PB: F0 52 BE
RC: F2 41 01 00 CC
PB: F0 73 9D
RC: F5 41 01 00 05 00 03 C1
Selling a frame into frame 5 of the ramcorder:

Code: [Select]

PB: FA 73 09 50 02 01 00 07 00 08 00 09 1F
RC: F5 41 01 00 05 00 07 BD
PB: F0 52 BE
RC: F2 41 01 00 CC
PB: F0 73 9D
RC: F5 41 01 00 05 00 09 BB
Buying frame 100 from the ramcorder:

Code: [Select]

PB: F8 50 02 00 00 07 00 C8 00 C9 1E
RC: F2 41 01 00 CC
PB: F0 73 9D
RC: F5 41 01 00 05 00 C9 FB
PB: F0 73 9D
RC: F5 41 01 00 05 00 C8 FC
Buying frame 103 from the ramcorder:

Code: [Select]

PB: F8 50 02 00 00 07 00 CC 00 CD 16
RC: F2 41 01 00 CC
PB: F0 73 9D
RC: F5 41 01 00 05 00 CD F7
PB: F0 73 9D
RC: F5 41 01 00 05 00 CC F8
As you can see this is not as simple as you would expect, i think the ramcorder is works more in video fields rather than whole frames and that the individual fields are bought and sold in separate operations?

You can see two counters, in the immediate example above for buying frame 103: bytes 6 to 9 seem to be frame/field counters. 0x00CC is 204 and 0x00cd is 205. If we assume the counter runs from 0-322 (rather than 1-323) then 204 could be frame 103 field 1 and 205 would be frame 103 field 2.

It doesn't all make sense much but that's where we are at.

[helius]

Yes, from the capture it does use field addresses. What's more: It seems that "selling" involves a command prefix to indicate that a descriptor will be used for writing, and "buying" just uses a bare descriptor. Look at

Code: [Select]

PB: FA 73 09 50 02 01 00 07 00 08 00 09 1Fvs

Code: [Select]

PB: F8 50 02 00 00 07 00 C8 00 C9 1E
Maybe 0x50 refers to the field format, 50Hz or PAL? Next byte is 0x02, which could mean number of fields in the descriptor. 0x01 could be direction (r/w) or something else. 0x07 possibly is a version or maybe means which port to use or a signal encoding.
All the replies with code 0x41 appear to be "Success" status, I think there would be something different if you requested a nonexisting field descriptor. 0x41 is ASCII 'A'.

[Willyarma]

I can see lots of patterns in it.

Looks like when selling frames the initial command is describing 2 fields, then the following is some kind of ack about the 2 fields with the 1/50 sec time difference maybe?
I'm not sure if its possible but having timestamps might help figure out if thats true.

How are you capturing this, is it using an RS232 cable or with your logic analyser? Because im puzzled about the line that reads F0 51 BE when selling frame 5, that line is F0 52 BE on the others, have we got a bit error there or is it correct?

Whilst on-site a few years ago I needed to view serial data as hex so threw together a quick VB program to capture serial and display it hex editor style, its been quite useful although I never polished it up. Is that any use to you?

The captures are good, but could do with more.
Can we see the same frame numbers bought and sold, and does it also do just fields too?

Oh and seeing the total number of frames available is very interesting indeed!

[dexters_lab]

thanks for the comments everyone 

helius certainly made me think about how the packets are formed

i am capturing with a Hantek 4032L logic analyser, the serial decode is a bit crap and there are sometimes errors from it and from me making typos so there could well be errors but i thought i checked all the checksums, but the 51 / 52 sounds like a typo from me!

It should also be noted there could well be some things that maybe legacy, there were devices around at the time that could read and write individual frames of video at the time that may share a similar protocol to the ramcorder. For example the write (sell) frame command seems to be saying 'seek to frame xxx' and then wait for an acknowledge, the next two commands are the write field 1 and then write field 2 of that frame. This could be allied to tape based systems.

Following that and picking up on Willyarma's comments about timing... yes i have omitted the timing info and there is for sure some interesting timing between the responses, i will try and detail these in some suitable form!

There are many many ways within the paintbox that generate buys and sells or generate other comms between the paintbox and ramcorder so it's hard to turn it all into a sensible post that you can understand without confusing things.

It really would be amazing to understand how it works, because the video in and out is SDI it could mean creating a replacement for the ramcorder with extended capacity a distinct possibility.

more info on the way...

[Willyarma]

I can see similarities between buying and selling, it looks like selling includes buying, like a combo.
When youre using the machine and selling a frame does it buy that same frame at the same time?

[dexters_lab]

here are some more captures...

i really need to work out a slicker way of getting this data, i have to manually type these across and work out the timing from the Hantek software which is utterly crap!

These first two are using the handheld quantel 'grip' which does a sell/buy/increment operation in one button press. Typically used for rotoscoping operations.

Sell frame 85, buy frame 86

Code: [Select]

  0MS   PB: FA 73 09 50 02 01 00 07 00 A8 00 A9 DF
 23MS   RC: F5 41 01 00 05 00 AD 17
 32MS   PB: F0 52 BE
 39MS   RC: F2 41 01 00 CC
 99MS   PB: F0 73 9D
108MS   RC: F5 41 01 00 05 00 A9 1B
117MS   PB: F8 50 01 00 00 07 00 AA 00 AB 5A
136MS   RC: F2 41 01 00 CC
199MS   PB: F0 73 9D
208MS   RC: F5 41 01 00 05 00 AA 1A
260MS   PB: F0 73 9D
268MS   RC: F5 41 01 00 05 00 AB 19
Sell frame 86, buy frame 87

Code: [Select]

  0MS   PB: FA 73 09 50 02 01 00 07 00 AA 00 AB DB
 23MS   RC: F5 41 01 00 05 00 A9 1B
 33MS   PB: F0 52 BE
 39MS   RC: F2 41 01 00 CC
 99MS   PB: F0 73 9D
108MS   RC: F5 41 01 00 05 00 AB 19
117MS   PB: F8 50 02 00 00 07 00 AC 00 AD 56
136MS   RC: F2 41 01 00 CC
199MS   PB: F0 73 9D
208MS   RC: F5 41 01 00 05 00 AC 18
260MS   PB: F0 73 9D
268MS   RC: F5 41 01 00 05 00 AD 17
Buying & selling individual fields:

Buy frame 91 field 1

Code: [Select]

  0MS   PB: F8 50 02 20 00 07 00 B3 00 B3 29
 20MS   RC: F2 41 01 00 CC
 80MS   PB: F0 73 9D
 87MS   RC: F5 41 01 00 05 00 B3 11
140MS   PB: F0 73 9D
148MS   RC: F5 41 01 00 05 00 B3 11
Buy frame 91 field 2

Code: [Select]

  0MS   PB: F8 50 02 20 00 07 00 B4 00 B4 27
 17MS   RC: F2 41 01 00 CC
 79MS   PB: F0 73 9D
 88MS   RC: F5 41 01 00 05 00 B4 10
140MS   PB: F0 73 9D
148MS   RC: F5 41 01 00 05 00 B4 10
Sell frame 250 field 1

Code: [Select]

  0MS   PB: FA 73 09 50 02 21 00 07 01 F2 01 F2 2A
 24MS   RC: F5 41 01 00 05 00 AB 19
 33MS   PB: F0 52 BE
 39MS   RC: F2 41 01 00 CC
 99MS   PB: F0 73 9D
108MS   RC: F5 41 01 00 05 01 F2 D1
Sell frame 250 field 2

Code: [Select]

  0MS   PB: FA 73 09 50 02 21 00 07 01 F3 01 F3 28
 24MS   RC: F5 41 01 00 05 01 F2 D1
 33MS   PB: F0 52 BE
 39MS   RC: F2 41 01 00 CC
 99MS   PB: F0 73 9D
108MS   RC: F5 41 01 00 05 01 F3 D0

[dexters_lab]

Looking at those latest data grabs... eg this buy part:

PB: F8 50 02 00 00 07 00 AC 00 AD 56
this is like the paintbox saying, select field 172 (frame 87 field 1) from the store to output to field 1 and select field 173 (frame 87 field 2) from the store and output to field 2

RC: F2 4A 01 00 CC
ramcorder acknowledges

PB: F0 73 9D
paintbox says display field 1

RC: F5 4A 01 00 05 00 AC 18
ramcorder acknowledges request and paintbox grabs field 1

PB: F0 73 9D
paintbox says display field 2

RC: F5 41 01 00 05 00 AD 17
ramcorder acknowledges request and paintbox grabs field 2


...so in the case of only buying a single field, the request is the same but the paintbox asks for both fields to be the same like here in this case with 0x00B3 in both field counters:
PB: F8 50 02 20 00 07 00 B3 00 B3 29

[Willyarma]

Awesome!
Thanks for the captures!
Is it RS232 or RS485?
If its RS232 then I would use 2 serial ports on my PC and tap off the transmit lines from the equipment leaving PC transmit disconnected so as not to mash the transmitters together!
I would gladly write a little VB program to capture the data for you if you need it!

[dexters_lab]

Awesome!
Thanks for the captures!
Is it RS232 or RS485?
If its RS232 then I would use 2 serial ports on my PC and tap off the transmit lines from the equipment leaving PC transmit disconnected so as not to mash the transmitters together!
I would gladly write a little VB program to capture the data for you if you need it!

it's rs232
it's a nice offer but i only have one USB->RS232 adaptor!

EDIT: just ordered another USB -> RS232 off ebay

An app that could concatenate the TX from the Paintbox and Ramcorder together would be perfect

i should also point out that OzOnE will be posting up shortly, he's been helping with this too and has the EPROM from the ramcorder in ADAPro for disassembly

[Willyarma]

I've done the GUI, I'll finish the actual code tonight.