Hello everyone,
I would like to ask for advice from people with archival and VHS preservation experience.
First, I would like to sincerely thank all the experts at digitalFAQ for their advice over the years. Your comments helped me better understand archival workflows and helped me make decisions when buying some of the equipment I now own. I would especially like to thank lordsmurf for helping me obtain and sending me some of the drivers I needed for several capture cards.
My goal is not to capture and preserve the most of the video as possibl.. My main objective is long-term archival using a lossless codec, preserving maximum information for future restoration and post-processing.
The tape I am preserving is extremely important to me personally. It contains family memories, especially recordings of my family.
The original content was filmed in 1986 using a Sony Betamax camcorder in Hi-Fi Stereo at maximum recording quality. Unfortunately, the original Betamax master tapes were lost forever many years ago.
Around the year 2000, my father copied the Betamax recordings to consumer VHS tapes to preserve the memories. Because of this, the VHS tape I am capturing today is most likely a second or third generation analog copy recorded on a 6-hour EP/SLP VHS tape.
During the last two years I became passionate about analog video preservation and started building several archival capture systems.
My systems:
Windows XP Integral Edition, extremely stable
Windows 10 Enterprise LTSC 2019
Internet disabled and updates disabled on all capture systems
I mainly focused on PCIe capture systems because I found them stable and easier to work with mother computers and XP. I have no dropped frames. I did not focus on AGP systems.
I also became very interested in the ATI Theater 200 chip because of its reputation in archival capture workflows.
Main VCRs:
Panasonic AG-1970
Panasonic AG-1980 (I dont used because it needs some new capacitors)
JVC HR-S7600
All two perform surprisingly well with EP/SLP tapes.
The internal VCR TBC is not perfect, but I found that the Panasonic DMR-T6070 the T6070 corrects most of the remaining instability very effectively (this was the last professional device that Panasonic sold). In my experience, performs even better than the common ES10 workflow for difficult EP tapes Main processing and synchronization equipment:
Panasonic DMR-T6070
Panasonic DMR-ES10
Datavideo DVK-200
Leitch DPS-575 professional TBC/frame synchronizer
The DPS-575 is configured for S-Video output in my workflow.
Capture hardware:
ATI AIW X1900 (ATI Theater 200)
AJA Kona LHi
Matrox MXO2 Rack
I use vMix with the AJA Kona LHi very successfully under Windows 10. One thing I appreciate about the Kona LHi and the MXO2 Rack is that their frame synchronization keeps audio/video synchronization perfect during long captures (6 hours).
All workflows are fully S-Video based.
I capture using lossless HuffYUV codec, as often recommended for archival workflows.
I tested many workflows, but I selected four examples for comparison.
Image #1:
Windows 10
Panasonic AG-1970 → Panasonic DMR-T6070 → Datavideo DVK-200 → Matrox MXO2 → VirtualDub → HuffYUV AVI
Image #2:
Windows 10
Panasonic AG-1970 → Panasonic DMR-T6070 → AJA Kona LHi → vMix / AJA Control Room → HuffYUV AVI
Image #3:
Windows XP
Panasonic AG-1970 → Panasonic DMR-T6070 → Datavideo DVK-200 → ATI AIW X1900 → VirtualDub → HuffYUV AVI
Image #4:
Windows XP
Panasonic AG-1970 → Panasonic DMR-T6070 → Leitch DPS-575 → ATI AIW X1900 → VirtualDub → HuffYUV AVI
I post the screenshots below.
I would really appreciate opinions regarding :
detail preservation
filter inside devices for noise handling
black levels
which one preserves the most video structure and colors in the video for my family memories
One of the shoots (Aja card) shows more defined interlaced lines, the theater 200 has more noise and the MXO2 is a bit more soft (like there is some noise filter applied
It's impossible to judge anything accurately when you're pointing a camera the the screen rather than doing an actual screenshot by pressing PrtSc and pasting it into a file... or, better yet, using VirtualDub's ability to grab a frame directly from the video.
It's impossible to judge anything accurately when you're pointing a camera the the screen rather than doing an actual screenshot by pressing PrtSc and pasting it into a file... or, better yet, using VirtualDub's ability to grab a frame directly from the video.
The AI reported that the bottom-left image had the highest measured edge information (sharpness score 636.1) and was the only image reaching full 0255 luminance range.
Can I get some expert advice on whether this methodology makes sense for evaluating VHS archival captures, and whether these results indicate more actual detail or simply more noise?
Top-left and Bottom-right have similar range, but Bottom-left reaches the full 0255 extremes.
You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community.
You must be logged in to view this content; either login or register for the forum. The attached screen shots, before/after images, photos and graphics are created/posted for the benefit of site members. And you are invited to join our digital media community.
You'll want to crop out the head switching noise and the borders when analyzing color since those areas may artificially introduce white or black levels that aren't actually a part of the picture.
Capturing in RGB makes it hard to get an accurate read on how complete the luma info is since values above 240 and below 16 get cut off, but it looks like the AJA and Matrox are blowing out whites. The ATI captures appear to be in rage.
Matrox also has some spikes in the chroma waveform which I think suggests some kind of filtering or compression. Interesting that it also doesn't seem to show the same tape flaws (Tracking error? that rolling static about 75% of the way through.) the other captures do.
The following users thank Feedbucket for this useful post:
goznevi (05-30-2026)
You'll want to crop out the head switching noise and the borders when analyzing color since those areas may artificially introduce white or black levels that aren't actually a part of the picture.
Capturing in RGB makes it hard to get an accurate read on how complete the luma info is since values above 240 and below 16 get cut off, but it looks like the AJA and Matrox are blowing out whites. The ATI captures appear to be in rage.
Matrox also has some spikes in the chroma waveform which I think suggests some kind of filtering or compression. Interesting that it also doesn't seem to show the same tape flaws (Tracking error? that rolling static about 75% of the way through.) the other captures do.
Thanks for the comments Feedbucket, I think I understand what you mean after seeing the video comparation.
From the analysis you did, it seems that the Theater 200 outperforms both the Matrox and the AJA in terms of preserving more information. I was hoping the AJA and Matrox would provide a better or equivalent capture without any filtering, and unfortunately there is no way to disable it. The main advantage of those cards for me is that they work under Windows 10, and I can use vmix and that they are solid with video audio sync.
Thank you very much for taking the time to do the analysis.
Regarding the RGB comment, what would be the best way to analyze still images or video captures? The PNG still frames be appropriate, or is there a better format for this type of comparison?
As for the tracking issue, that part puzzles me. I dont know whether I accidentally used a combination of equipment that reduced or eliminated it, or if I simply got lucky. Its the same tape being played over and over again, and I havent been able to reproduce the tracking problem consistently. 🙂
I will continue analyzing the videos and post additional samples and results as I go. Hopefully, once I have gathered enough data, I can put together a more complete conclusion that may help other readers who are considering the Matrox or AJA cards for VHS archival capture.
By the way, I havent looked at the test.avs file yetsomething new to learn. 👍
How did you create the 4-video comparison and histograms? Its a great way to compare captures.
Could you recommend a guide or resource to learn this workflow?
Capturing in RGB makes it hard to get an accurate read on how complete the luma info is…
Thank you, Feedbucket, for your comments.
I actually found how to correct the tracking issue. I manually adjusted the tracking on the AG-1970 until I found the sweet spot, and the problem disappeared.
Since your post, I have also been learning Avisynth and the differences between RGB and YUV. I realized that some of the clips I had previously exported were not in the proper format for analysis. I have now recreated all the test clips from the original HuffYUV lossless captures and generated new PNG/TIFF frame exports. I also applied the Avisynth script you shared.
The new results are quite interesting. At this point, I think we may have found a contender to the ATI AIW Theater 200, the AJA LHi. The Matrox appear to apply some form of processing or noise reduction and missing the interlaced line structure. The Theater 200,
Quote:
Originally Posted by Lordsmurf
’s long-time favorite card
, still appears to retain slightly more detail, but I would appreciate an expert opinion since I am still learning and don’t yet fully trust my own conclusions.
What I do understand is that the goal is to extract as much of the original signal as possible without applying any processing or filtering. Once a clean master capture has been created, restoration can always be performed later, as
Quote:
Originally Posted by Sanlyn
has recommended in a pas post
, while preserving the original master.
I will continue sharing my findings and learning as I go. My goal is to identify the capture card that best preserves these family VHS tapes from 1986 for long-term archival purposes.
Now that I have fixed the video clip in YUV, I have generated these new black-and-white image exports.
I'm still having a hard time deciding between these captures and don't have the expertise to properly evaluate them.
Please comment on which capture appears to preserve more detail and whether the wider luma range is actually beneficial for archival preservation? Any observations would be greatly appreciated.
Also, if you know of any tools or software that could help, I would be interested in learning about them.
I think your observations with respect to the different captures have been pretty spot on, and it looks like you have the levels sorted out. None of these look particularly terrible - mostly just a difference in the level of noise (and detail) reduction, it seems.
Don't want to push you down a rabbit hole, but other things I like to look at in AviSynth when evaluating capture quality are the individual chroma channels (UToY and VToY) and how things might look deinterlaced (a simple Bob just to quickly visualize) frame by frame in sequence - helpful for identifying interference patterns and other oddities.
Was the Matrox sample rescaled? It looks weird when deinterlaced, and I remember it being 4px taller than the other samples. I also noticed in your earlier upload that the field order on the Windows 10 captures was bottom field first while the XPs were top field first, but I don't think that matters too much as long as it's internally consistent.
I think your observations with respect to the different captures have been pretty spot on, and it looks like you have the levels sorted out. None of these look particularly terrible - mostly just a difference in the level of noise (and detail) reduction, it seems.
Don't want to push you down a rabbit hole, but other things I like to look at in AviSynth when evaluating capture quality are the individual chroma channels (UToY and VToY) and how things might look deinterlaced (a simple Bob just to quickly visualize) frame by frame in sequence - helpful for identifying interference patterns and other oddities.
Was the Matrox sample rescaled? It looks weird when deinterlaced, and I remember it being 4px taller than the other samples. I also noticed in your earlier upload that the field order on the Windows 10 captures was bottom field first while the XPs were top field first, but I don't think that matters too much as long as it's internally consistent.
Thank you for the feedback and suggestions.
The Matrox and AJA captures were done on Windows 10 using vMix. Perhaps that is why they are being captured as Bottom Field First, while the ATI X1900 captures were done on Windows XP with VirtualDub and appear to be Top Field First. I really don’t know. Does each capture card use its own strategy, or is this determined by the driver, software, or codec?
Regarding the Matrox samples, it was not rescaled and yes it does look like it was scaled or deinterlaced. I will explain, the first one was captured using Matrox’s MTXTools capture utility, configured for 720x486 uncompressed capture, and then re-encoded with HuffYUV. The second Matrox sample was captured with vMix at 720x480. As far as I know, it was not intentionally rescaled, but vMix does not expose all of the Matrox DirectShow filter settings and will not allow me to select 720x486. The MXO2 supports both hardware MPEG capture and regular uncompressed capture, so I am wondering if vMix may be accessing the MPEG capture path, which could explain some of the differences. The capture quality in vMix does not seem to be as good as with MTXTools.
vMix can capture AVI files and use the HuffYUV codec. I think this is an excellent alternative to VirtualDub. So far, with the AJA card, it is working like a charm.
As for the rabbit hole, no worries at all. For the last two years I have been comparing hardware, drivers, capture chains, operating systems, and settings in an effort to learn more about analog video hardware. I have now reached the point where I want to go deeper into the analysis, and your suggestions are exactly the kind of direction I was looking for. I did not realize how powerful AviSynth could be as an analysis tool.
Thank you again for taking the time to share your observations. I will follow your suggestions, analyze the UToY and VToY channels, inspect the captures with Bob deinterlacing frame by frame, and report back with the results.
