212 lines
9.6 KiB
Markdown
212 lines
9.6 KiB
Markdown
Converting VHS and DV to Modern Formats - Part 1
|
|
|
|
Over the past 10 years I've been meaning to convert my family's VHS tapes to a
|
|
modern format. Originally that would have been DVD, but as it seems that DVD
|
|
and Blu-Ray would have a limited lifespan, I've opted to go directly to modern
|
|
encoding formats.
|
|
|
|
---
|
|
|
|
This will be a multi-part series stepping through all the challenges with
|
|
converting these formats using Linux. This post, focuses around getting Video
|
|
Grabber's (USB Dongles) to work under Linux.
|
|
|
|
---
|
|
|
|
In order to get VHS content to a modern video format, you will need a compatible
|
|
USB capture device, a VHS Player and a PC.
|
|
|
|
The process should be fairly straight forward, but there were a number of issues
|
|
which made the task difficult to achieve without compromise.
|
|
|
|
Conceptually it should be as follows:
|
|
|
|
1. Connect VHS Player to USB Capture Device
|
|
2. Connect USB Capture Device to PC
|
|
3. Install-\>Launch Capture software
|
|
4. Press Record in the Software
|
|
5. Press Play on the VHS player
|
|
|
|
Step 2 is where it get's tricky on Linux. While there are a significant number
|
|
of capture devices supported on Linux, it is still *luck-of-the-draw* when
|
|
purchasing a device with Linux in mind.
|
|
|
|
From what I have found many devices are rebranded. E.g. Four same branded
|
|
devices may contain 4 different chipsets. Maybe only 2 of them contain the Linux
|
|
supported chipset. To make matters worse, vendors don't typically list the
|
|
chipset on their site or packaging.
|
|
|
|
The following is a description of my journey to get a couple of different
|
|
capture devices to work on Linux. Your mileage may vary.
|
|
|
|
In my case, I had two USB capture devices. One was very old *Pinnacle Dazzle*
|
|
obtained from my Dad that he was throwing out and the other was purchased at
|
|
[Aldi](https://www.aldi.com.au/) by my Mother-In-Law. This one had no label on
|
|
the device, but the box stated:
|
|
*Bauhn DVD Maker*
|
|
|
|
### Dazzle
|
|
|
|
I started with the Dazzle. This unit looked very old. Sometimes with Linux, Old=
|
|
Good. As devices age, the likelihood that some enthusiastic Linux hacker will
|
|
add driver support goes up (I don't actually know this and cannot prove it, but
|
|
you probably can't disprove it either, so there).
|
|
|
|
So, I plugged it in and typed `lsusb`
|
|
|
|
Bus 003 Device 007: ID 2304:021d Pinnacle Systems, Inc. Dazzle DVC130
|
|
|
|
*Figure 1. Content snipped for brevity*
|
|
|
|
The important part of the `lsusb` output is the ID `2304:021d`
|
|
|
|
We now know what this beast is. Let's check if the kernel has already recognized
|
|
it. You can usually do this by checking your kernel messages and see if a driver
|
|
module loaded and told you it registered anything. Therefore: `dmesg`
|
|
|
|
[14842.638559] usb 3-11.3: new high-speed USB device number 7 using xhci_hcd
|
|
[14842.714905] usb 3-11.3: New USB device found, idVendor=2304, idProduct=021d, bcdDevice= 0.00
|
|
[14842.714907] usb 3-11.3: New USB device strings: Mfr=1, Product=2, SerialNumber=0
|
|
[14842.714908] usb 3-11.3: Product: DVC 130
|
|
[14842.714909] usb 3-11.3: Manufacturer: Pinnacle Systems, Inc.
|
|
[14843.570702] usb 3-11.2: reset low-speed USB device number 5 using xhci_hcd
|
|
|
|
*Figure 2*
|
|
|
|
The output shows the `idVendor` and `idProduct` match the ID from the output of
|
|
`lsusb` in *Figure 1*
|
|
|
|
Essentially, the kernel is reporting the USB device, but the absence of messages
|
|
from drivers other than `usb` are not a good sign.
|
|
|
|
We know by the output of `dmesg` that a kernel module has not loaded. It is time
|
|
to check that the kernel modules we need are indeed compiled and available.
|
|
Luckily for me, I'm running [Gentoo](https://gentoo.org) so always have the
|
|
kernel source at hand. With video capture devices, these parts of the kernel are
|
|
referred to as the Video For Linux subsystem.
|
|
_(Technically, Video for Linux 2)_
|
|
|
|
These kernel modules can be found under: Drivers-\>Media-\>USB.
|
|
|
|
Device Drivers --->
|
|
<*> Multimedia support --->
|
|
[*] Analog TV support
|
|
[*] Media USB Adapters --->
|
|
<M> USB video devices based on Nogatech NT1003/1004/1005
|
|
<M> STK1160 USB video capture support
|
|
<M> WIS GO7007 MPEG encoder support
|
|
<M> WIS GO7007 USB support
|
|
<M> WIS GO7007 Loader support
|
|
<M> Conexant cx231xx USB video capture support
|
|
<M> Empia EM28xx USB devices support
|
|
|
|
*Figure 3 shows kernel options to enable video capture*
|
|
|
|
After enabling all the relevant looking ones, I compiled them all and unplugged
|
|
and re plugged my device, then checked `dmesg` again. Still no luck.
|
|
|
|
This is the point I recommend most people give up. I myself am not a big fan of
|
|
giving up. So what do I do? I bust out my trusty screw driver and pop that
|
|
sucker open to see whats _inside_. Under the magnifying glass I read out and
|
|
google all the names on the chips. One of them catches my eye:
|
|
|
|
*go7007*
|
|
|
|
Well would you look at that. There is a module for this chipset. Perhaps all we
|
|
need to do, is _teach_ it to use my USB device. I start exploring the `.c` files
|
|
to find which one of them contains definitions looking like USB device and
|
|
product IDs. I began poking around in the source files looking for a `struct`
|
|
recording all the supported USB Product and Device ID's.
|
|
|
|
Here is the `struct` I found in `go7007-usb.c`
|
|
|
|
static const struct usb_device_id go7007_usb_id_table[] = {
|
|
{
|
|
.match_flags = USB_DEVICE_ID_MATCH_DEVICE_AND_VERSION |
|
|
USB_DEVICE_ID_MATCH_INT_INFO,
|
|
.idVendor = 0x0eb1, /* Vendor ID of WIS Technologies */
|
|
.idProduct = 0x7007, /* Product ID of GO7007SB chip */
|
|
.bcdDevice_lo = 0x200, /* Revision number of XMen */
|
|
.bcdDevice_hi = 0x200,
|
|
.bInterfaceClass = 255,
|
|
.bInterfaceSubClass = 0,
|
|
.bInterfaceProtocol = 255,
|
|
.driver_info = (kernel_ulong_t)GO7007_BOARDID_XMEN,
|
|
|
|
},
|
|
{}
|
|
};
|
|
|
|
*Figure 4*
|
|
|
|
The figure above is one device in the `struct,` so in `vim` of course, I yank
|
|
yank a bit of text here and there and try to add support for my own USB device.
|
|
The main thing to get right, is to substitute the `idVendor` and `idProduct`
|
|
from the `struct` with the ID's we discovered in *Figure 1 and Figure 2*.
|
|
|
|
For other interfaces of the `struct` I'm mostly guessing, but hoping I can just
|
|
copy some of the other devices, recompile and it might work. But if it doesn't
|
|
work, change things around a bit, like some of the other cards and try again.
|
|
|
|
If done correctly, the kernel module should load automatically when the device
|
|
is plugged in.
|
|
|
|
Sadly, I tried many combinations, caused a couple of kernel panics and sometimes
|
|
the card would load and I could get to see a blue screen when viewing the
|
|
device, but no content :( I was also able to setup a USB trace and see that the
|
|
driver was successfully uploading the firmware to the device.
|
|
|
|
In the end, I gave up on this device.
|
|
|
|
PS, I tried to capture using this device from a Windows VM. The device was so
|
|
old it would only work on 32-bit Windows 7 or older. While I was able to get
|
|
a driver, I could not find the original special software required to capture
|
|
and regular DirectShow capture to VirtualDub or OBS didn't work either.
|
|
|
|
### Bauhn DVD Maker
|
|
|
|
This devices seemed a tad more modern. It has cables for capturing Component
|
|
and Composite. It came with a CD with drivers and software for Windows. Other
|
|
than that, there was little evidence online of this device. No-one mentioned
|
|
this model and certainly not in relation to Linux. I tried the obvious tests
|
|
(lsusb, dmesg) but didn't spend too much time trying to make it work for Linux.
|
|
|
|
On my Windows 7 VM though, I did have some luck. I got the software working
|
|
and was able to capture some VHS tapes. Although the capture was successful
|
|
there was no flexibility with the file format. The files came out in MPEG 2
|
|
(essentially DVD type files), which are okay, but if I wanted these files to
|
|
be online, I would need to re-encode them. That would mean a potential
|
|
degradation in quality. I mean, come on! Haven't these videos degraded in
|
|
quality enough!
|
|
|
|
What I wanted to do, was for the videos to be captured in a near lossless format
|
|
and then re-encode to 2 files. 1 for showing on the internets (so that means
|
|
smallish file size, optimized for streaming and compatibility) and 2 for
|
|
archival purposes (use the most forward leaning tech available large close to
|
|
lossless).
|
|
|
|
My thinking is that future generations will be having to convert historical
|
|
videos to another format and I want to make a large lossless file available
|
|
for that purpose.
|
|
|
|
I decided to crack open the Bauhn DVD maker and see what makes it tick.
|
|
|
|
|
|
|
|
Okay, I know how to do this, google all the little numbers and words. It didn't
|
|
take long to discover the chip and the existing Linux module for this baby.
|
|
Like the `go7007`, the `cx231xx` module supports many devices that use this
|
|
chip. Again, I had a crack at setting up the .c driver and adding in the Product
|
|
and Device ID's to make it work. And guess what?
|
|
|
|
*Boom* It worked. And that, my friends is joy. I tested using VLC --> Media -->
|
|
Open Capture Device..., then click the 'Video device name' dropdown box and
|
|
choose /dev/video0
|
|
|
|
You can download my [patch file](files/cx231xx.patch) against kernel 4.19.44
|
|
|
|
Now onto the task of actually doing something with the video files which I will
|
|
cover in my next post.
|
|
|
|
Tags: linux, ffmpeg, gentoo, kernel
|