Best Patterns to test VCRs/Capture Cards/TBCs with?

[aramkolt]

Hello all,

I've been working on getting around to finally comparing different capture methods, VCRs, capture cards, and TBCs against each other. For the actual content to be used as the test material, I'll be using test patterns as it is known what they are supposed to look like.

I'll eventually use actual video samples as well, but for now I need to keep the captures short so they don't take up too much space so I can post them all for others to scrutinize.

The testing will involve capturing from the digital source, a first gen VHS tape, a second gen VHS tape, and possibly a third gen VHS tape. Passing the patterns through a TBC alone will also show to what degree each TBC or passthough device harms the patterns as well as there will be no timebase errors for the TBC to need to correct, but it'll still undergo an analog to digital and then back to analog conversion which is likely not "visually lossless" for all passthrough devices.

For anyone curious, the method is to output using a BMD studio 3G monitor which can output interlaced SD via either SDI or HDMI. From there, will use something like the Brighteye 16 digital to analog converter which can output S-Video or composite from the interlaced SD SDI. Since the Brighteye has a proc amp, I can fine tune the output using a waveform monitor and vectorscope to make sure that levels are as they should be coming out of the Brighteye.

So my question is - What patterns would people suggest using that are absent from the attached video? Basically I'm trying to use patterns that will quickly identify different capturing flaws.

[pandy]

This depends on your experience and skills - i see on this movie most of US (NTSC) patterns used by industry - there is few patterns dedicated subjective quality testing and some (bigger group) patterns that can be used to perform measurement - few patterns are combo type so aggregate many tests in single screen - those are most commonly used in industry to measure objective quality. Also seem DaVinci offer waveform monitor (software oscilloscope) so in theory you can capture and perform measurements. It is not clear to me what you understand by "different capturing flaws" - those flaws can be subjective ones or objective ones - objective ones require tools and some electronics knowledge to perform objective quality verification.

For example Zoneplates perform function similar to the 1D sweep but in 2D space - X and Y (Horizontal and Vertical) direction - Zoneplates are used to verify quality of for example deinterlacers, 3D denoisers, scalers etc
Static patterns are mostly used to perform objective measurements, color bar to perform color conversion accuracy, some patterns are used to perform phase measurements (Chrominace/Luminance gain/delay) important from CVBS perspective etc.

For electronics knowledge and meaning of some signals you can see in patterns google for "tektronix ntsc pal standard high definition video measurements" - in those PDF's you will get serious foundation of video signal measure basics.

[lollo]

In addition to the test patterns you showed I would add a test including moving horizontal, vertical and especially diagonal lines, to check the intra frame TB correction of the VCRs, and integrity across frames of the output signal.
I often use a video like this: https://www.youtube.com/watch?v=Tec6g_Qjh1Q

Remember that test patterns check is a necessary but not sufficient condition. Do also tests with real video, covering different cases.

[Sharc]

How would one transfer such digital testpatterns reliably onto a VHS master tape so one can test the entire signal flow from Tape->VHS player->...whatever ...-> capture device? Most capturing problems have their origin in the analog signal domain.

Edit:
Also, for color test patterns it should be clear IMO what the colors are, for example by indicating the RGB values of the bars or squares. Otherwise we just see "some" colors without a reference. Something like this I mean (the 160,160,160 should be 120,120,120 btw)

[Attachment 83347 - Click to enlarge]

https://en.wikipedia.org/wiki/ColorChecker

[pandy]

How would one transfer such digital testpatterns reliably onto a VHS master tape so one can test the entire signal flow from Tape->VHS player->...whatever ...-> capture device? Most capturing problems have their origin in the analog signal domain.

You can verify your source and eventually recorder too - but you can safely assume that modern sources like DVD player equipped with digital video encoder achieved state close to reference video generators from end of 80's.

I pointed in different thread that if someone can afford (place and perhaps budget) then you cheaply acquire state of the art video analyzer from Tektronix - VM700 - they are on ebay from 100$ - with VM700 you can measure video almost every aspect.

Edit:
Also, for color test patterns it should be clear IMO what the colors are, for example by indicating the RGB values of the bars or squares. Otherwise we just see "some" colors without a reference. Something like this I mean (the 160,160,160 should be 120,120,120 btw)

[Attachment 83347 - Click to enlarge]

https://en.wikipedia.org/wiki/ColorChecker

Color bars (also other patterns) are well defined even standardized so such color checker is rather good for camera - check https://www.itu.int/rec/R-REC-BT.801-1-199510-W/en - there is everything - sample values are provided.

[lollo]

All the patterns test are nice, but the final experiment should be capturing “real” VHS/S-VHS tape contents.

[pandy]

True - but the main purpose of the test patterns is to expose eventual problems affecting objective quality - there is no possible to perform good capture with faulty equipment.

[Sharc]

How would one transfer such digital testpatterns reliably onto a VHS master tape so one can test the entire signal flow from Tape->VHS player->...whatever ...-> capture device? Most capturing problems have their origin in the analog signal domain.

You can verify your source and eventually recorder too - but you can safely assume that modern sources like DVD player equipped with digital video encoder achieved state close to reference video generators from end of 80's.

OK. So you suggest to use legacy DVD players which still have Composite, S-Video and Component outputs to convert digital test patterns to analog video signals suitable for testing VHS recorders, passthrough devices, capture cards, USB capture dongles etc.

[dellsam34]

Some devices I have can generate color bars, ramp, black screen, with audio tone and output via an analog out, devices such as capture devices and Betacam decks, there are also some digibeta tapes recorded with such materials if anyone interested I can make some recordings if you pay shipping, I got a lot of tapes to get rid of anyway. These are the only professionally calibrated sources that can be trusted for calibrating ingest workflows, Digital files has to go thru a HDMI to analog or some sort of consumer digital to analog converter that is not guaranteed to meet the standard specifications, Don't get me wrong, some DVD players from back in the day are good but being consumer grade devices the tolerances are too loose.

[aramkolt]

Thanks for the feedback! I agree that real video will eventually need to be used, but that test patterns can give some great initial feedback. Some of the patterns really do show the same things as some of the others, so I'll have to whittle those down. Part of the reason I have the variety is that I'm not sure if there will be certain ones that VHS playback and capture cards will struggle with specifically.

A standalone waveform monitor looking at the output of the BMD 3G SDI monitor gives the same waveforms and vectors as what I recorded, so I am confident that is giving the appropriate colorspace output as recorded. I have verified that the setup can produce a range of 0IRE to 110IRE which is useful for evaluating automatic gain control and clipping whereas I don't know if DVD is allowed to encode "illegal" levels, guessing not.

Most of those patterns did come from real pattern generators and those that didn't have SDI output were converted with a BE75 with the proc amp tweaked on it to get the levels exactly right (as shown by the SDI vectorscope/waveform monitor) when recording via the SDI output. Given that they were all patterns with known vectorscope targets and luma values, it was really easy to see where the levels needed to be as I was recording them. Even pattern generators vary a bit on the luma and chroma. My guess is they are all calibrated to send signals through different lengths of cable where the signal will drop some over distance and I'm using all relatively short cables (2ft usually), so levels might be a bit hot from that alone.

Oddly, when recording the SDI stream into ProRes444 with the BMD Ultrasdudio Recorder 3G, the colors were way off just using Blackmagic media express as the capture program. I always thought SDI being captured was basically just storing digital data and sampled at a different frequency, but apparently not. Feeding that same SDI into an AJA Ki Pro preserved the colors perfectly in ProRes422HQ. The KiPro can be used as a playback device as well, but it only has component and composite out for analog (no S-Video). You could go the same route where you convert the SD SDI output it has back to any analog output with the Brighteye 16 and tweak the proc amp as you look at a vectorscope/waveform monitor to ensure that the levels going into the recording VCR are accurate.

I did think about using a DVD player as a source with the same patterns, but you're then limited to 4:2:0, limited bitrate, harder to change patterns if you don't like one, limited color depth, and probably not able to display illegal values (like 0IRE or 110IRE in NTSC) and no guarantee that luma output is correct (just based on what I can see on the waveform monitor on my basic Sony Blu Ray player). You could probably use that plus a proc amp to adjust levels, the SDI route just seems more direct, gives a much higher bitrate, and at least 4:2:2 10 bit color, etc.

Now how faithfully that known accurate signal gets recorded onto a VHS tape, that's a little harder to prove. Vectorscopes/Waveforms from VHS tapes usually aren't very pretty compared to pattern generators haha.

I do have a working VM700T as well - I'm not exactly clear on how I'll use that with this testing, but would be very open to suggestions. I've seen a YouTube video describing some of the functions and I have played around with it a bit. I was originally thinking that you'd look at the various characteristics before and after a "passthrough device" like a TBC or DMR-ES10 to see to what degree the different stats change. Mainly I planned to use it for jitter measurements as that would actually show you to what degree any type of TBC reduces signal jitter and it does have measurements for both frame and line jitter. For some of the tests, it needs to be fed specific patterns so it can tell you how accurately that pattern is being produced. So some of the more funky patterns are there in the test patterns to be used with VM700T specifically.

I haven't tried dropping higher resolution videos into the Davinci timeline, but I assume it can scale them. The trick is they might need to be 29.97 or 59.94fps I think.

[dellsam34]

SDI is just a transmission line, if the device is set to process the video then the input and output will be different, We've seen it in my other thread where I forgot to click sync and the output became blurry, This is true for any capture device, For example if you slide a proc amp item from its neutral position you instantly activate the processing, But some processing are necessary for the good look of the picture, some can destroy the picture if not chosen wisely "I know, I feel embarrassed".

[pandy]

OK. So you suggest to use legacy DVD players which still have Composite, S-Video and Component outputs to convert digital test patterns to analog video signals suitable for testing VHS recorders, passthrough devices, capture cards, USB capture dongles etc.

Generally speaking yes, modern digital encoders providing decent signal quality - frequently outperforming mid 80's professional broadcast equipment - lack of 4:2:2 should be not a problem from objective quality measurement as most of test signals can be fine in 4:2:0 space.

[pandy]

I did think about using a DVD player as a source with the same patterns, but you're then limited to 4:2:0, limited bitrate, harder to change patterns if you don't like one, limited color depth, and probably not able to display illegal values (like 0IRE or 110IRE in NTSC) and no guarantee that luma output is correct (just based on what I can see on the waveform monitor on my basic Sony Blu Ray player). You could probably use that plus a proc amp to adjust levels, the SDI route just seems more direct, gives a much higher bitrate, and at least 4:2:2 10 bit color, etc.

Now how faithfully that known accurate signal gets recorded onto a VHS tape, that's a little harder to prove. Vectorscopes/Waveforms from VHS tapes usually aren't very pretty compared to pattern generators haha.

I do have a working VM700T as well - I'm not exactly clear on how I'll use that with this testing, but would be very open to suggestions. I've seen a YouTube video describing some of the functions and I have played around with it a bit. I was originally thinking that you'd look at the various characteristics before and after a "passthrough device" like a TBC or DMR-ES10 to see to what degree the different stats change. Mainly I planned to use it for jitter measurements as that would actually show you to what degree any type of TBC reduces signal jitter and it does have measurements for both frame and line jitter. For some of the tests, it needs to be fed specific patterns so it can tell you how accurately that pattern is being produced. So some of the more funky patterns are there in the test patterns to be used with VM700T specifically.

4:2:0 is not limitation for most of synthetic test signals as they are rather large block or even whole screen static - 4:2:0 may be somehow limitation for spatial dithering if your goal is to simulate 10 in 8 bit quantization depth but still not an big issue IMHO.

limited bitrate is not a problem as most of the test can be single frame i.e. 15Mbps also navigation can be quick if they are made like chapters.

luma level should be OK as this is part of design specification i.e. accuracy is quite high and most of the digital encoders are based on 10..12 bit DAC's anyway.

limited quantization depth can be overcome by using spatial and temporal dithering - VM700 can work in average mode (32 lines maximum) so you can easily increase accuracy of measurement even in noisy environment - dither is noise intentionally added to signal to remove nonlinear quantization error - internally VM700 use 10 bit ADC with 12 bit DAC and 6 bit dither pattern so it can be very precise and accurate in measured figures - if you pursuing maximum accuracy then you need to use special 75Ohm terminators - they can be costly when compared to present VM700 prices.

Nowadays it is possible to just draw - create pattern as bitmap - convert it to encoded still (I frame) ad that's it - it can be turned to useful signal (not necessarily in DVD - this can be also some USB media player equipped with S-Video/CVBS/component output. Having VM700 you can verify signal quality and later use it as source to objectively verify accuracy of your capture - almost complete signal chain can be verified.

[Bogilein]

Here you found my PAL test file with which I test the resolution (720x576) and see whether the hardware used covers the full range 0-255 or only to a limited extent.

https://forum.videohelp.com/threads/413492-Missing-pixels-with-s-video-vs-firewire-cap...e2#post2725268

The playback device plays a major role, I have tested DVD players, Blu-Rayers and over 10 DVD recorders and only the Pioneer DVD recorder DVR 630 was able to play the test file with a resolution of 720x576 and the full range 0-255. Some could only playback the resolution over the S-video output at 704x576 with black bars on the left and right or cut off the full range.

[dellsam34]

As I said in pot#9, I'm not against the source file or source frame, I'm questioning the conversion of that source to analog, and as Bogilein confirmed that out of 10 players only one was up to standard, and I'm assuming those were from back in the day, not the modern players with lousy composite out or chinese HDMI to composite toys.

[pandy]

Here you found my PAL test file with which I test the resolution (720x576) and see whether the hardware used covers the full range 0-255 or only to a limited extent.

Problem with your test file is bandwidth - seem you are not limiting bandwidth - every step (transision) or pulse shall be bandlimited - usually this is 250ns Bessel like lowpass filter or in BT.801 https://www.itu.int/rec/R-REC-BT.801-1-199510-W/en digital shaping filter with Blackman window (formulae is provided on page 2) - Blackman window is better than ancient Bessel response - also easier to calculate in digital framework (made lot of test signals purely in spreadsheet).
If you feed not limited bandwidth signals into your device then some misbehavior can occur (particularly in filters - some of them may manifest oscillations visible as ringing, also all codecs with frequency domain so MPEG, JPEG AVC etc may manifest Gibbs phenomenon https://en.wikipedia.org/wiki/Gibbs_phenomenon?useskin=vector ).

[pandy]

As I said in pot#9, I'm not against the source file or source frame, I'm questioning the conversion of that source to analog, and as Bogilein confirmed that out of 10 players only one was up to standard, and I'm assuming those were from back in the day, not the modern players with lousy composite out or chinese HDMI to composite toys.

Trust me - digital encoders such as https://www.analog.com/en/products/adv7391.html#documentation outperform most of professional equipment from beginning 2000's - digital to analog conversion is not a problem. Of course there can be some simplifications or poorer quality so i've wrote earlier that typical consumer level digital encoder (in DVD, media players etc) usually outperform broadcast equipment from 80's/90's.
They implement everything in digital form, analog part is significantly reduced and simplified so less likely to introduce issues common to complex filters used 30 - 40 years ago.

Check page 11 in datasheet - SNR in flat field is 75dB where common consumer equipment will have something around 42..46dB, even on ramp signal this is 58dB (and ramp signal in digital world is the most noisy part) i.e. over 9.5 bit quality.

[dellsam34]

What DVD players you have tested that outperform professional equipment? Any model #'s. As far as I know pro digital to analog equipment are still being made, Snell & Wilcox and similar devices under Grass Valley ownership were still being made till 2016, Ensemble Designs BrightEye digital to analog converters are still being made, Not sure about Aja devices, So I'm curious what DVD players outperform these devices and other rack mount devices that still in production.

[Sharc]

Here you found my PAL test file with which I test the resolution (720x576) and see whether the hardware used covers the full range 0-255 or only to a limited extent.

Problem with your test file is bandwidth - seem you are not limiting bandwidth - every step (transision) or pulse shall be bandlimited - usually this is 250ns Bessel like lowpass filter or in BT.801 https://www.itu.int/rec/R-REC-BT.801-1-199510-W/en digital shaping filter with Blackman window (formulae is provided on page 2) - Blackman window is better than ancient Bessel response - also easier to calculate in digital framework (made lot of test signals purely in spreadsheet).
If you feed not limited bandwidth signals into your device then some misbehavior can occur (particularly in filters - some of them may manifest oscillations visible as ringing, also all codecs with frequency domain so MPEG, JPEG AVC etc may manifest Gibbs phenomenon https://en.wikipedia.org/wiki/Gibbs_phenomenon?useskin=vector ).

A simple sledge-hammer approach to bandwidth limit such signals would be to downscale-upscale the sharp original test patterns. The scaling includes lowpass filtering to prevent aliasing.

[Sharc]

Here a color reproduction test of 2 capture cards with their default settings. The numbers represent the original RGB values.
Left: Original
Center: Device 1
Right: Device 2
At a first glance all 3 may look ~identical. Now take a color picker and compare the squares, and you will see the differences.

Added: Same interleaved. Makes the color shifts better visible.

[Bogilein]

Here you found my PAL test file with which I test the resolution (720x576) and see whether the hardware used covers the full range 0-255 or only to a limited extent.

Problem with your test file is bandwidth - seem you are not limiting bandwidth - every step (transision) or pulse shall be bandlimited - usually this is 250ns Bessel like lowpass filter or in BT.801 https://www.itu.int/rec/R-REC-BT.801-1-199510-W/en digital shaping filter with Blackman window (formulae is provided on page 2) - Blackman window is better than ancient Bessel response - also easier to calculate in digital framework (made lot of test signals purely in spreadsheet).
If you feed not limited bandwidth signals into your device then some misbehavior can occur (particularly in filters - some of them may manifest oscillations visible as ringing, also all codecs with frequency domain so MPEG, JPEG AVC etc may manifest Gibbs phenomenon https://en.wikipedia.org/wiki/Gibbs_phenomenon?useskin=vector ).

What effect does this have on the result if I want to test the resolution of 720x576 of the devices used and whether the devices used can reproduce the full range of 0-255 or only to a limited extent?

[pandy]

What DVD players you have tested that outperform professional equipment? Any model #'s. As far as I know pro digital to analog equipment are still being made, Snell & Wilcox and similar devices under Grass Valley ownership were still being made till 2016, Ensemble Designs BrightEye digital to analog converters are still being made, Not sure about Aja devices, So I'm curious what DVD players outperform these devices and other rack mount devices that still in production.

Tested various devices based on Broadcom, Intel, STM digital encoders and most of them provide sufficient margin of electrical parameters to compete with broadcast equipment (not only but also some signal generators) made in 80's. this was my point - after 30 years consumer technology matched or outperformed professional technology from 80's past century.

[pandy]

A simple sledge-hammer approach to bandwidth limit such signals would be to downscale-upscale the sharp original test patterns. The scaling includes lowpass filtering to prevent aliasing.

You can also apply Gaussian blur or similar lowpass filer but problem is in definition and standardization - when you perform measurements then you need to follow standard so whenever you take Tektronix, Rohde & Schwarz or Snell & Wilcox and similar companies test signal sources then you will find exactly same sample values as for example defined in BT.801.

When you measure something you need to follow accepted rules.

[pandy]

What effect does this have on the result if I want to test the resolution of 720x576 of the devices used and whether the devices used can reproduce the full range of 0-255 or only to a limited extent?

There is many side effects possible - simply pointed difference between artistic expression of some pattern and real test pattern - for example sweep or multiburst is sine signal not set of black/white stripes...
Using compression you unavoidable introduce additional signals not present in your source - this is important when you measuring and not important when you just watching pictures on screen.

[Bogilein]

Can you be a little more specific with regard to my test file, preferably with a practical example of the effects this has on the resolution and e.g. the luminance range? My aim is to understand whether I have misjudged my test results.

[Sharc]

..... for example sweep or multiburst is sine signal not set of black/white stripes...

Once we are at it here a sine-shaped resolution sweep (no stripes) I composed some time ago for testing horizontal resolution e.g. of VHS recorders.
No warranty for absolute correctness though.

[pandy]

Can you be a little more specific with regard to my test file, preferably with a practical example of the effects this has on the resolution and e.g. the luminance range? My aim is to understand whether I have misjudged my test results.

Sorry i can't support you on this - i can provide one of few main issues with too fast or not bandlimited signals - for example read https://www.rohde-schwarz.com/pl/applications/incorrect-colours-in-the-rgb-display-of-...280-15800.html

[pandy]

..... for example sweep or multiburst is sine signal not set of black/white stripes...

Once we are at it here a sine-shaped resolution sweep (no stripes) I composed some time ago for testing horizontal resolution e.g. of VHS recorders.
No warranty for absolute correctness though.

Normally you can use for this multiburst or sine sweep - i like also sinx/x testing signal as it will provide side to bandwidth also information about group delay (VM700 has this measurement) and it is fast so from engineer perspective as much as possible information in shortest possible time.

Your test pattern is fine but it take too much time - using oscilloscope or spectrum analyzer (frequently embedded in modern digital oscilloscope) you will get this information in single line - also you can objectively quantify (measure) it.

And once again - it was my feedback not critics - testing video equipment from over 20 years - also producing testing streams and signals.

[Sharc]

..... for example sweep or multiburst is sine signal not set of black/white stripes...

Once we are at it here a sine-shaped resolution sweep (no stripes) I composed some time ago for testing horizontal resolution e.g. of VHS recorders.
No warranty for absolute correctness though.

Normally you can use for this multiburst or sine sweep - i like also sinx/x testing signal as it will provide side to bandwidth also information about group delay (VM700 has this measurement) and it is fast so from engineer perspective as much as possible information in shortest possible time.

Your test pattern is fine but it take too much time - using oscilloscope or spectrum analyzer (frequently embedded in modern digital oscilloscope) you will get this information in single line - also you can objectively quantify (measure) it.

And once again - it was my feedback not critics - testing video equipment from over 20 years - also producing testing streams and signals.

Understood, and your feedback is appreciated. Professional testing according to standards and following precisely defined test procedures makes sense for many reasons, no doubt.
Real world is not standardized though, and one may be more interested in the effective milage per gallon of one's car than in the (usually optimistic) figures obtained from standardized test procedures and conditions. So one may deliberately want to use non-standard home-brewed test patterns with say excessive bandwidth/sharpness to demonstrate ringing and Gibb's effect and to compare how different devices behave under same conditions. Even simple and basic tests can reveal a lot.
On a sidenote, I was surprised to find that a well-know capture dongle seems to present a poor termination to the C wire of S-video. All but the expected 75 Ohms. I may do some more specific tests later to investigate its impact (exact color reproduction, signal reflection etc.) although I have some doubt how relevant it is in view of 'VHS color quality' .....

[aramkolt]

This is all really awesome information that I would have never thought about and I definitely appreciate the feedback, thank you all. Many of you certainly have a better technical grasp on limitations and how to get the most useful information out of such testing.

I do plan to put out as high of a bitrate DVD ISO (for free of course and posted in this thread once everything is finalized) preserving the interlacing of the original source files from the pattern generators so that people can use them to test out their own equipment and possibly post captures to show where maybe one device excels or doesn't, or perhaps to even show if my own eventual posted results are repeatable in other environments. I'll of course post a comparison of what the DVD does compared to the higher bitrate Davinci->SDI 3G Monitor->BE16(S-Video out)version and I might also see what the captures look like off of a DV output device like an ADVC-110 as well. An actual DV file could be recorded to a DV tape and then played back from a DV outputting device which should give pretty decent color accuracy I would think since it's a type of digital source, so I'll test that too.

As for the bandwidth stuff, the BE16 will be essentially converting to S-Video right at the end of the chain, so won't all of the bandwidth be limited to what S-Video is capable of at that step? Once it goes onto a first gen tape, it'll then be limited to what VHS can do etc? My question is - how can "too much bandwidth" for the source negatively impact the test if it's being standardized to S-Video at the end?

I will say the hardest pattern to obtain was the Snell and Wilcox SW2 pattern which requires a working TPG21 or TPG20 and they don't seem to pop up often. Unfortunately, I found the SW2 pattern only can be displayed when in NTSC/D2 mode or PAL/D2 mode. I always thought that "D1 and D2" were forms of SDI, but turns out D1 is the digital component and D2 is digital composite (I think).

I haven't been able to find much information about "D2 capture" as I'd rather be able to capture the digital version of the output from the TPG21.

Anyone know how I'd go about capturing a "D2" signal?