Thanks for the videos and info, Zcooger.
Are the spikes "bearding"? Do you know what causes them? https://bavc.github.io/avaa/artifacts/bearding.html
+ Reply to Thread
Results 61 to 90 of 576
Using the same settings I get them on "overused" recording of multiburst pattern. I think middle frequencies aren't filtered enough but I may be wrong.
Without FFT display it's hard to guess the original correct recording frequency.
What do you need the composite output for? Man those samples look pretty stable with only two minor problems, There is white and black short horizontal lines that appear in the video section only I assume those were in the original video and there are some long horizontal white noise lines, it will be nice to compare to normal capture and post them on youtube.
Man that Canal+ scrambled video reminded me of the 90's and my Aston satellite receiver, those were the fun days if you know what I mean (emulator cards !!!).
I see that you posted normal video captures, so those short noise spikes are original.
Last edited by dellsam34; 2nd Dec 2019 at 10:52.
Personally I don't like to spam with links and videos any topics but there's regular capture of this momemt for easy comparison
I failed a bit with the tracking setting - forgot to set it to automatic and didn't have any monitor to see the picture.
Last edited by Zcooger; 2nd Dec 2019 at 11:15.
I saw the regular video samples and updated my post above, So any progress on the chrominance ?
Dropout compensation isn't properly implemented for VHS yet, so any dropouts become quite visible long noisy white lines like in e.g sample 3. The short white (or in a few cases black) spikes I think are mostly a result of the demodulation (transforming the FM signal frequencies into video levels, or voltage in the analog domain) process picking up the wrong frequency (though there could be some short dropouts in there too.) The luma video signal on the tape is FM-modulated, i.e high frequency wave = white, low frequency wave = black. It makes sense for there to be spikes/noise on a re-used tape as there may be more residual noise and wear from previous recordings that hasn't been entirely eliminated by the erase head, and a worn tape won't store the magnetick tracks as well as a new one. It's sort of the opposite of the "bearding" described in the link where spikes are generated in the modulation stage due to the modulator being overdriven (though I've seen VCRs creating black spikes on playback too).
The filtering and demodulation of the RF signal in vhs-decode needs more work to avoid picking up the wrong frequency. Adjusting the band pass filter currently used to filter waves of too high and low frequency has some impact, but may need something more complex. Demodulation seemed to be done with a somewhat different process in VCRs judging by the books I've read compared to how ld-decode implemented it, so I don't know if I need to change more than just the filtering of the RF signal before demodulating.
As for chroma, it's decoded to a degree now, but it still lacks the frequency and phase correction functionality found in VCRs to compensate for the unstable nature of the signal coming off the tape. Granted, we do TBC it, which makes up for it a little. There's also a bug that results in the chroma not being centered right which needs to be tracked down.
Sorry for complicated wall of text. Been thinking of making a blog post or something detailing the VHS decoding process in a bit more readable manner.
https://en.wikipedia.org/wiki/Differentiator - burst frequency are superimposed over DC bias - check CCIR18 signal for example in https://cdn.rohde-schwarz.com/pws/dl_downloads/dl_application/application_notes/7bm25/7BM25_0E.pdf - drop of bias level passed trough capacitors may be visible as pattern around burst.
I'm going to try and capture from my 1994 RCA VCR this week to post a sample.
Last edited by Titan_91; 8th Dec 2019 at 19:04.
Well, it died. The tracking circuit no longer works at all and the RF and composite output of the machine are crap. Guess I'll re-install the mod in my original VCR since the signal level is about on par with the Spider-Man sample and will be looking at Goodwill for an older/better one.
I want to get going with this right away, but I don't want to have to wait for a Domesday Duplicator to be built because nobody seems to be offering to do do. What other devices are up to the task for this?
Here's the Conexant CX2388 based card I bought.
I searched eBay for "Blackbird capture card" after doing some research. You want one with a CX2388 based broadcast decoder. This has the 28.6MSPS high speed ADC needed for FM signal sampling from a LaserDisc player or VCR. Check the item photos to verify the chip is on the card. Once you have one you will likely need a second PC with a PCI slot to run Linux on.
I got to thinking about how to cleanly use a VGA splitter/amp for capturing. I have a generic brand DVD/VCR combo that plays tapes pretty well. The DVD player tray sensor went bad so I disconnected it. The VCR side still works fine. So I might try to remove the DVD drive and use that space to put my VGA splitter. I can then install an RCA jack and just run a cable from the machine to my PC. Using a VGA splitter is detailed in the Google document.
Edit: I pulled the DVD drive. I think I found the RF test point along with the head switching test point. After checking the service manual this connects to pin 79 of the main video IC and is pre-amplified like on the Phillips VCR. I can use one of the pre-punched holes in the back of the case for the output.
Any suggestions to add to this idea?
Last edited by Titan_91; 15th Dec 2019 at 16:59.
Connecting the output to my scope, my DA appears boosts the signal level by a factor of 2. I'm feeding the video and Hi-Fi RF through the DA on the blue and red pins and will install output jacks for each. I wish I could capture both simultaneously but will need to do 2 playbacks of a tape to capture its audio RF. Looking in Audacity, the level is better but not as good as vaporeon800 is getting with his S-VHS machine. Could this be a cheap vs. expensive machine characteristic? Vaporeon800 can you explain your signal path, I.E. did you use an amplifier like I did? Any impedence matching done?
On top is a commercial copy of A Bug's Life from me, on bottom is vaporeon800's commercial copy of Spider-Man.
Also for funsies, I did this in AviSynth's RawSource filter and Gimp. Carefully eyeballing the "image" it looks like the school hallway fight scene with Peter Parker in the foreground. Notice the chroma phase inversion during the head switch pattern near the bottom of the field, particularly in the colorburst and Peter's shirt. Also, I counted the number of chroma cycles per line which is about 40 as expected. Pretty awesome to actually have the ability to visualize all this.
Last edited by Titan_91; 15th Dec 2019 at 18:14.
I didn't use an amplifier. I just connected directly to the test point. It was years ago, so I'm not sure exactly which cables I used.
On newer VCRs, the manufacturers managed to stuff the head amp in the same IC stuffed the head amp in the same IC as the rest of the video decoding, so the test point, at least on the JVC deck I tested, was the unamplified one, which made the signal level a bit too low for my cxadc setup to work. It's nice that a VGA amp helps, though maybe that will make it easier to use a newer deck. Older decks usually had a separate head amplifier IC, typically on a separate PCB (but some newer ones have it moved to the main PCB) so it's easier to tap an amplified signal.
Guessing the the combo Titan_91 is using (which looks like a modern LG mech?) has unamplified RF at the test pint, while the one vaporeon used may have been tapped from an amplified rf signal?
This is the best I'm getting right now:
The video IC on my Philips machine has an RF output on pin 79 after an internal amplifier in IC301 according to the schematic. I'm using a Zenith XBV342 combo player. Looking at the service manual for a very similar Zenith XBV243, the RF test point comes in on the same pin 79 on the same IC301 in the schematic. However, I'm not seeing an amplifier on this pin in the schematic. I compared the ICs between the two machines and they are not the same.
Could I maybe daisy chain a TV RF amplifier to the output of my VGA amp?
I tried two VHF/UHF amplifiers, neither passed the signal through (only noise). Even turning off the FM trap the signal seems to get filtered out. Any suggestions for low power amplifiers in the 3MHz-5MHz band I can try or do I need to look for a similar late '90s Sony machine like Oln has with a strong FM output?
Last edited by Titan_91; 25th Dec 2019 at 18:31.
What about components used in bluetooth tech.? (bluetooth 4.2 use the 5Mhz UHF band), some BT modules go up to 100 mW (20 dBm) (class 1), most constructors use class 2 modules (2.5mW / 4dBm). Source wikipedia
edit: sorry,that's 5Ghz for bluetooth..totally irrelevant
Thanks for the suggestion. I looked on eBay for a lower power 100mw or less 80m/HF band amplifier but can’t find one. All the HAM oriented amplifiers are of course much higher output power. Maybe I can use a composite or component distribution amp? I’ll try a composite one and report back.
Last edited by Titan_91; 26th Dec 2019 at 12:50.
Any slightly older VCR (i.e that has a head amp separate from the main chip so you can get amplified RF out) ought to work, I know zcooger here has been capturing from some panasonic VCRs.
NE592 is good for such things, general fast opamp should work but there are dedicated video opamps for such tasks (gain up to 100 with bandwidth few tens of MHz)
I found a 1999 Zenith VRB4215. While this machine has a single board, the RF test point is actually amplified. Looking at a scope the other Zenith combo unit I had gave me about 20mV peak to peak from the test point. This one puts out about 200mV which is excellent. I set the level to 26 to avoid clipping. But the signal looks and sounds completely off in Audacity. With the Panasonic Spider-Man sample the carrier signal can be clearly heard at 44100Hz. But my sample is "noisy" and I can barely hear the actual carrier. RawSource shows a decent image but there's a strong DC offset (above about 128 in the file). On the scope, the envelope looks perfect using a well shielded RCA cable.
Last edited by Titan_91; 28th Dec 2019 at 22:11.
I finally have something usable. I went back to the XBV342 with VGA preamp and tried two composite distribution amps. One didn't amplify the signal, but the older Radio Shack 15-1172 did. This brought the level up to the Spider-Man sample which should be a usable level for vhs-decode when the software is ready for NTSC. This is daisy chaining a VGA DA with a video DA. It's a pretty clean result compared to the VRB4215 player, I think that one just has a noisy/unisolated test point that I didn't notice on the scope. I would still like to use that machine though if anyone has any suggestions. I moved the ground thinking there was a loop causing noise from surrounding circuits, but this made no difference.
Last edited by Titan_91; 29th Dec 2019 at 09:45.
Had to convert it to .wav so not sure if I did it right, but here's a spectrogram view of that first sample, the one with the DC offset:
[Attachment 51245 - Click to enlarge]
[Attachment 51246 - Click to enlarge]
Comparing that to the Spiderman capture posted earlier in the thread, there's definitely something off about it:
[Attachment 51247 - Click to enlarge]
[Attachment 51248 - Click to enlarge]
If anyone else wants me to put their transfers through this for a looksee, let me know - might reveal a few things.
Thanks for the spectrum views. Here's a final sample from me at level 31. Oln do you think this signal level will be ok for vhs-decode? My peak level is 161/255 and trough is 100/255. Also, it shifts a bit at the head switching points. Should I adjust the alignment? The output picture and sound on the machine are perfect.
Sample from unamplified test point using VGA amp plus video distribution amp together:
Last edited by Titan_91; 29th Dec 2019 at 09:55.
Looks good - slightly stronger than my latest cxadc capture from my sony - which I have successfully decoded from.
If the normal picture is good, don't mess with alignment. Bad alignment shows up as a non-flat waveform, typically it gets lower at the start and/or end of each field. It's usually illustrated in VCR service manuals.
I don't know what causes the offset at head switch points. I've seen that on my scope on some decks when aligning though, so presume it's normal. maybe the head switch signal is added as an offset into the test rf or something to e.g help to see what part of the signal is from what head without needing two scope inputs.
Also, if you haven't already, with the newest version of cxadc you can up the sample rate and/or bit depth a little.
What do you use for those nice spectrums? I can plot spectrum with pyplot in the decoding code but it ends up being very low resolution, not as nice as yours.
Thanks for the feedback. From a storage space perspective do you think there would be any advantage to using 35.8MSPS as opposed to 28.6MSPS for VHS? With 4.8MHz being the highest frequency component, the difference would be about 7.5 samples per cycle vs. 6 samples per cycle. This would be helpful for high bandwidth laserdisc captures with 3.58MHz chroma signals. I know your early captures were using 28.6.
Last edited by Titan_91; 29th Dec 2019 at 12:22.
I am constantly using 35MHz mode with CXADC for Panasonic NV-FS200 & NV-HD630 capture. There's always some headroom for noise reduction. The more the better even if the advantage is 7MHz - for FM modulated VHS the more the better. I recognize you are using Izotope RX7 for spectrograms (probably?).
Ok, I'll bump it up since your YouTube videos look amazing. Here's a 12 minute Arthur cartoon "extra" tape from 1996 that only has 1 crappy copy on YouTube. The capture went very well. I built the latest CXADC module and will capture this again at 35.8MSPS. This one is about 20GB. With storage being so inexpensive and this being a nearly 8 year old Dell PC, there's really no reason to not go for the higher sample rate.
For future reference, I started when I pressed play. This captured the initial white noise and base carrier signal before the video came in.
I stopped when the VCR blue screened even though the carrier was still present, telling me the video signal had ended.
Fortunately there are only a few spots of slight signal reduction on the tape, and very few dropouts. The worst dropouts I could find:
Last edited by Titan_91; 29th Dec 2019 at 15:35.
I loaded the module with tenxfsc turned on and it seemed to work in a couple of test captures I did. According to dd it's still 28.6 doing another full 12 minute capture. Anything I seem to be doing wrong?
user@mintiplex:~/Downloads/cxadc-linux3-master$ sudo insmod cxadc.ko vmux=2 tenxfsc=1 user@mintiplex:~/Downloads/cxadc-linux3-master$ sudo mknod /dev/cxadc c 126 0 user@mintiplex:~/Downloads/cxadc-linux3-master$ dd bs=1M if=/dev/cxadc of=/home/user/Desktop/test.raw ^C118+0 records in 117+0 records out 122683392 bytes (123 MB, 117 MiB) copied, 3.75639 s, 32.7 MB/s user@mintiplex:~/Downloads/cxadc-linux3-master$ dd bs=1M if=/dev/cxadc of=/home/user/Desktop/test.raw ^C286+0 records in 285+0 records out 298844160 bytes (299 MB, 285 MiB) copied, 8.94297 s, 33.4 MB/s