Current status of ld-decode / vhs-decode (true "backup" of RF signals)

[Sharc]

Unfortunately the Hi-Fi audio is affected. I just haven't played around with this enough to try fixing it. After adjusting video head-switching, the VCR automatically switches to the adjustment for Hi-Fi audio head-switching. Ideally, an oscilloscope would be used (which I don't have).

Maybe something like this:
https://www.digikey.com/en/products/detail/analog-devices-inc/ADALM2000/7019661
https://wiki.analog.com/university/tools/m2k/scopy
You can try out the scopy GUI without the HW to check if it meets your purpose for an oscilloscope.

[Titan_91]

I think this is the problem with the sample that I posted earlier for cvbs-decode.


[Attachment 64827 - Click to enlarge]

Maybe some filter can help to remove this. What do You think?

The only way to fix that is to isolate the source of the RF noise. Are you using a well shielded composite video cable? Is there a wifi router, wifi adapter, or phone nearby that could interfere?

[Brad]

Unfortunately the Hi-Fi audio is affected. I just haven't played around with this enough to try fixing it. After adjusting video head-switching, the VCR automatically switches to the adjustment for Hi-Fi audio head-switching. Ideally, an oscilloscope would be used (which I don't have).

Maybe something like this:
https://www.digikey.com/en/products/detail/analog-devices-inc/ADALM2000/7019661
https://wiki.analog.com/university/tools/m2k/scopy
You can try out the scopy GUI without the HW to check if it meets your purpose for an oscilloscope.

That's an intriguing USB oscilloscope option. For some reason I've never seen anyone else mention it! Thanks for the links, Sharc.

[Titan_91]

Has anyone tried vhs-decode on an M1 Mac running Asahi Linux? I'm curious about the performance. These processors are very competitive to x86 running Mac OS. With Linux that performance is considerably better yet.

[nogginvid]

Unfortunately the Hi-Fi audio is affected. I just haven't played around with this enough to try fixing it. After adjusting video head-switching, the VCR automatically switches to the adjustment for Hi-Fi audio head-switching. Ideally, an oscilloscope would be used (which I don't have).

Maybe something like this:
https://www.digikey.com/en/products/detail/analog-devices-inc/ADALM2000/7019661
https://wiki.analog.com/university/tools/m2k/scopy
You can try out the scopy GUI without the HW to check if it meets your purpose for an oscilloscope.

The normal issue with PC Digital Scopes is that they don't usually support continuous, lossless, sample streaming (some of the £££ Picoscope USB 3.0 models do), and instead sample to local RAM on the Scope interface, for slower off-loading to the PC for display, so they sample a chunk of signals (based on the trigger mode used), then ignore a bunch etc.

If that model does support continuous sampling streaming at ~30MS/s for long periods of time with no dropped samples - it could be interesting. But every other £200 model we've seen is based around the buffered burst sampling system.

[Sharc]

Unfortunately the Hi-Fi audio is affected. I just haven't played around with this enough to try fixing it. After adjusting video head-switching, the VCR automatically switches to the adjustment for Hi-Fi audio head-switching. Ideally, an oscilloscope would be used (which I don't have).

Maybe something like this:
https://www.digikey.com/en/products/detail/analog-devices-inc/ADALM2000/7019661
https://wiki.analog.com/university/tools/m2k/scopy
You can try out the scopy GUI without the HW to check if it meets your purpose for an oscilloscope.

The normal issue with PC Digital Scopes is that they don't usually support continuous, lossless, sample streaming (some of the £££ Picoscope USB 3.0 models do), and instead sample to local RAM on the Scope interface, for slower off-loading to the PC for display, so they sample a chunk of signals (based on the trigger mode used), then ignore a bunch etc.

If that model does support continuous sampling streaming at ~30MS/s for long periods of time with no dropped samples - it could be interesting. But every other £200 model we've seen is based around the buffered burst sampling system.

I don't know its limits for continuous sampling. The memory depth can be selected in predefined steps depending on time base setting and ranges from 1600 samples to 800k samples @100MS/s. I have not verified all settings, as the continuous lossless sample streaming over longer periods has not been an issue for my applications so far.

[Titan_91]

Here's a good analogy for impedence matching. If you're ever driving down the highway and roll down your driver side window, the air may oscillate inside the cabin. There is negative pressure drawing air in following by positive pressure pushing it out. This is a reflection of air pressure. But you if you roll down your passenger window the same amount, the oscillation stops.

The wind is the RF signal, and your driver and passenger side windows are the input and output impedence. If the resistance on both ends doesn't match, you will get oscillations (ringing) and reflections, and non-ideal voltage/pressure on the output end.

[Titan_91]

Here's another one-off VHS exclusive pickup of mine, I'll capture and upload the RF and decoded video to Archive.org soon.

[oln]

I pushed an update that changes some things a bit recently.

Disabled brightness clamping as it seems to be prone to causing flicker on some tapes, especially macrovision (and the associated -noclamp option). The old behaviour can be turned on by using

--clamp

Also tweaked the filtering for vhs a bit (doubled up the 3 (NTSC) / 3.5mhz (PAL) low-pass filter), which should help reduce ringing (and slightly soften the image) a tad.

There is also a lot of work being done on ld-analyse at the moment. It can now load both chroma and luma at the same time to preview color + image (done by mixing them at the moment so won't look 100% like the final output fron running the gen_chroma_vid.sh script.) Others are also working on adding a vectorscope and some other tweaks.

[Titan_91]

Also tweaked the filtering for vhs a bit (doubled up the 3 (NTSC) / 3.5mhz (PAL) low-pass filter), which should help reduce ringing (and slightly soften the image) a tad.

There is also a lot of work being done on ld-analyse at the moment. It can now load both chroma and luma at the same time to preview color + image (done by mixing them at the moment so won't look 100% like the final output fron running the gen_chroma_vid.sh script.) Others are also working on adding a vectorscope and some other tweaks.

Interesting! I'll try that and mess with the sl flag again to see if I can get a bit more sharpness without the ringing. The McDonald's video has both FMV and animated scenes which will help me with analysis.

And glad to see work done on ld-analyse after not much done with it the last few years.

[Titan_91]

I'm having an issue building the latest commit.

install -d "/usr/local/bin"
install -m755 ld-ldf-reader "/usr/local/bin"
if [ -z "" ]; then \
	python3 setup.py install --prefix="/usr/local"; \
else \
	python3 setup.py install --root="" --prefix="/usr/local"; \
fi
Traceback (most recent call last):
  File "setup.py", line 4, in <module>
    from Cython.Build import cythonize
ModuleNotFoundError: No module named 'Cython'
make: *** [Makefile:55: install-python] Error 1

So I installed the C-Extensions for Python using

pip install Cython

Which was successful, but I'm still getting the error when building.

[oln]

right I should have also mentioned that, cython is now a dependency.

Not sure why it wouldn't work after installing Cython via pip, does it work if you install like this (provided you are on ubuntu/debian or derived)

sudo apt install cython3

?

[Titan_91]

right I should have also mentioned that, cython is now a dependency.

Not sure why it wouldn't work after installing Cython via pip, does it work if you install like this (provided you are on ubuntu/debian or derived)

sudo apt install cython3

?

Yes, that worked. Thanks. I'm on Linux Mint, which is Ubuntu based. I think Cython was installed for my base Python version, which is really old. So I'm guessing I'll need to use python3 in my command line from now on. Which I was doing a couple of years ago anyway.

[Titan_91]

As expected, the double filtering is a bit softer. I personally don't mind the ringing as it seems others have removed it in post production. For my personal use, the extra sharpness is preferred. Is there a flag or adjustment to the code I can make to disable the double filter, while still keeping the brightness clamping turned off? The brightness change is most welcomed. And the performance is better!

[oln]

Not yet, can add it though. For now you can temporarily comment/delete the lines that add it in vhsdecode/process.py to disable it.

[Titan_91]

Thanks again. I'm not trying to be critical, just making observations. I realize getting this process even closer to VHS spec is getting more tricky. We appreciate your time and effort.

I'd say this project's main weakness at the moment is the ringing for sure. And I'm still not sure what causing it. But it doesn't bother me that much.

[Titan_91]

I put this on a very silly pedestal on Archive.org if anyone is interested.

https://archive.org/details/scaredsilly-rfcapture

[Titan_91]

Someone was asking about the software process I follow.

Hardware process is here if you want to look back:
https://forum.videohelp.com/threads/394168-Current-status-of-ld-decode-vhs-decode-(tru...28#post2645340

1. Use older capture PC to capture with dd to internal hard drive using BS=1M parameters for a 1MB block size with Conexant PCI card and CXADC. Nothing fancy.
2. Copy the resulting .r8 file using a large flash drive to my main PC.
3. Run VHSresamp40to16.grc in GNU Radio Companion to low pass filter and resample the overclocked 40MSPS .r8 file to a 16MSPS .r8 file (https://forum.videohelp.com/threads/394168-Current-status-of-ld-decode-vhs-decode-(tru...29#post2653660).
4. Compress the resampled file with FLAC using flac --best --sample-rate=16000 --sign=unsigned --channels=1 --endian=little --bps=8 -f capture-resampled.r8
5. Verify the .FLAC file's integrity using flac -t capture-resampled.flac
6. If that's successful, change vhsdecode/format_defs/vhs.py to reflect rfparams["deemph_gain"] = 12 and RFParams_NTSC_VHS["video_bpf_low"] = 1500000.
7. Also change process.py to reflect #if self.options.double_lpf: and #self.Filters["FVideo"] *= filter_video_lpf. This disables the double RF filtering to preserve sharpness.
8. Decode the result using ./vhs-decode -n -t 15 -f 16 --recheck_phase capture-resampled.flac capture-resampled.
9. Then rasterize the result to a lossless FFV1 compressed MKV file using ./gen_chroma_vid_ntsc.sh capture-resampled.

Finally, if it's personal home videos I store the FLAC files and decoded video to hard drive, LTO-5 tape, and M-Disc Blu-ray so they are good for many years. If it's retail, I upload the FLAC and video files to Archive.org and just put a personal copy on a hard drive.

[harrypm]

Has anyone tried vhs-decode on an M1 Mac running Asahi Linux? I'm curious about the performance. These processors are very competitive to x86 running Mac OS. With Linux that performance is considerably better yet.

5fps - M1

6-7fps - M1 Max (top model 4TB NVME)

Observed under a Ubuntu 20.04 LTS VM

[harrypm]

Hello, Everybody not on the discord!

I'm Harry (UK) I have been updating the VHS-Decode & CXADC readme's and wiki's these have been updated weekly since December 2021 now, also promoted the project where I could.

I know some of you have probably not known about these updated things, but I have strived to clean up the black hole of lack of basic information and clear information on the surface this includes ground-up guides, which I have been working on this year for people who don't even know what a soldering iron, flux etc really is for example.

There is a shared drive dedicated and mirrored for all RF Decoding related things and a simple/clear standard naming system this unifies contributions and test captures for both reference and history of the project.

I made the vhs-decode/tape-decode subreddits ahead of time, I'm writing this thanks to Titan_91 as I did not really use them but I'll start mirroring some content over there to help with discoverability, one rule is keep to current information unless stated clearly non-current

I have also re-worked the DdD documentation for anyone to practically go and drag and drop on fabrication companies like PCBWay it was lacking clear instructions and an actual BOM file we have had 3 users including me have ordered like this without issue now, sadly the DE0-Nano is still very out of stock or inflated.

Also, side note VHS-C tapes captured with the DdD 40msps and FLAC compressed are around 1-1.1GB Per min even less with current CXADC 8-bit 20msps this means an 30/45/90min tape can comfortably fit alongside a viewable and other documentation on a 100GB M-Disk for solid archives witch for 10GBP per disk at the only amazon US seller not a bad deal, (Anyone put LD on M-Disk yet though? )

I probably won't be very active on here that much but if anyone wants to get my direct attention feel free to email me at: harry@opcomedia.com

[Titan_91]

Hey Harry! It's funny that I discovered your 3 month old Reddit page that you forgot about. I'm big on documentation and love what you're doing to round out the knowledge on this project. That being said, I'm not interested in creating a Google account or Discord account. One is data privacy, another is this particular thread is a permanent paper trail of everything up to this point. I'm making sure each page has Archive.org snapshots. It's obviously not current, and there are ways to export conversations from Discord to have a paper trail there. But this much historical information from years past that is publicly accessible without needing an account is still relevant, and it's all in one place without being fragmented across conversations. But the same can be said for the GitHub wiki page which is looking great.

That's my opinion based on the strengths and weaknesses of each medium. I'm pretty much done uploading test samples anyway, as I've reached the apex of my workflow and am satisfied with it. Anything more I want to preserve will go up on my Archive.org account, linked a few posts back.