Enumerating VHS's horizontal resolution

aths

Newcomer
Anyone with a video recorder and a bit of time, please do me a favour. Get a freeze image and count its "pixels" (distinctive different luminance elements). Because to do so for an entire scanline would be tedious, it is best to encapsule a small portion of the image, such as a TV logo or part of a text (with a small font to discern all "pixels") and then count the projected pixels in perspective to the TV screen width

Please count carefully and do an exact measure of the TV screen width as well as the part you enumbered. I need some most exact numbers, not guesses.

If you have an CRT-TV, multiply your total "pixel" number by 1,07 to incorporate the overscan.

Thank you in advance.
 
At best, I believe VHS is 240 pixels per scan line, with about ~480 scan lines. It's pretty terrible.
 
Since analog tv signal is not composed of discrete pixels, the task of counting them is not meaningful.

I slightly more sensible question is how many pixels you need to correctly sample an optimal quality VHS signal. Counting imaginary pixels doesn't give you the answer though.
 
I slightly more sensible question is how many pixels you need to correctly sample an optimal quality VHS signal. Counting imaginary pixels doesn't give you the answer though.

According to the almighty wiki, VHS has a video bandwidth of about 3 MHz; by applying the familiar Nyquist theorem, we get that a sample rate of at least 6 MHz is needed. The timings of the VHS signal are basically the same as those of NTSC/PAL (or whatever TV system it is played back to), giving an active video period of about 52.6 us per scanline, of which about 7% is usually lost to overscan. This gives a total of 6 * 52.6 * (1-0.07) = 294 pixels.

The actual number appears to be lower, though; the wiki article says ~240 pixels. I suspect that the discrepancy has to do with the way color information is stored; the 3 MHz bandwidth of VHS appears to be split between luma and chroma information, which would in case leave approximately ~2.3 MHz of bandwidth to luma.
 
Please forget wikipedia, because if it comes to AV technologoy, wikipedia often cites sources which I believe are misinterpreted. This is especially true if it comes to the resolution of analogue video images (and also is valid for the english as well as the german edition of wikipedia.)

Since VHS is an analogue system, we got no real pixels, that is true. However we got a smallest size of detail, or a maximum amount of (luminance) details per scanline. I doubt that VHS is limited to 120 vertical lines alternating black and white (making 240 "pixels"). I already counted that for myself =) but I need independent results from you.

It is extremely difficult to get real information about VHS. I already bought (and read) a book about video recorder technology which made me understand how the loading system works and what the heads are for, but it lets a lot of questions open about the maximum possible image quality.
 
Last edited by a moderator:
surely what matters is the horizontal resolution of the tv the video was desined to display on (pal + ntsc are slightly diiferent iirc)
since vhs was designed to only display on tv's it would be pointless designing vhs do have a higher res than a tv
 
surely what matters is the horizontal resolution of the tv the video was desined to display on (pal + ntsc are slightly diiferent iirc)
since vhs was designed to only display on tv's it would be pointless designing vhs do have a higher res than a tv

As I understand it, VHS actually has substantially lower resolution than standard broadcast TV.

Also, VHS apparently also uses a frequency modulation scheme for the luma signal, which puts a rather sharp upper bound on the attainable resolution - very much unlike what is the case with unmodulated analog media (like the audio LP).
 
We could *discuss* on and on ... or somebody just counts the number of "pixels" (luminance details) of a scanline. I already did but I want to get a confirmation about my figures.
 
"I already did but I want to get a confirmation about my figures."

and by any chance did it turn out to be the same as the amount of horizontal pixels on your t.v ????

1440 when i watch it on a tv with a horizontal res of 1440
if i watch it on an anlog tv im guessing it would be about 768 the same res as pal analog tv
if i play a vhs through my capture card its displayed at 768x576
 
the 3 MHz bandwidth of VHS appears to be split between luma and chroma information, which would in case leave approximately ~2.3 MHz of bandwidth to luma.

well yes, it's the point of NTSC, PAL and SECAM. that gives B/W compatibility and allows to spend much less bandwith on chroma.
 
"I already did but I want to get a confirmation about my figures."

and by any chance did it turn out to be the same as the amount of horizontal pixels on your t.v ????

1440 when i watch it on a tv with a horizontal res of 1440
if i watch it on an anlog tv im guessing it would be about 768 the same res as pal analog tv
if i play a vhs through my capture card its displayed at 768x576
I don't want you to count your tv's picture elements but the VHS's freeze image number of luminance details per scanline. Since VHS's horizontal resolution is much lower even than common CRT's resolution, that task is actually possible.

Analog PAL does not have 768 pixels per scanline. Since it is analog, there are no pixels anyway, but if you count smallest details as "pixel" it comes out to an equivalent of about 540 "pixels". Since most common CRTs have less than 600 RGB elements per scanline, you cannot display DVD images with full resolution (of 704 or 720 pixels per scanline) and therefore cannot count the DVD's actual resolution on a TV freeze image. VHS however got such a low resolution that you can count the number of "pixels" per scanline.
 
Last edited by a moderator:
and how would i count the pixels other than diplsying a vhs on my tv
and when i do it fills the screen hence uses all the pixels of my tv for display
hence my point it depends on what your displaying it on
in my case pal tv which is 720 x 486 i beleive (i got it a bit wrong before :D )
 
I don't want you to count your tv's picture elements but the VHS's freeze image number of luminance details per scanline. Since VHS's horizontal resolution is much lower even than common CRT's resolution, that task is actually possible.
How are you going to do that? How is anyone to see the difference in color between two ajacent pixels that look the same?

And how would you cope with a resolution that isn't an integer fraction of the total amount of pixels on the screen? How is someone supposed to determine what part of each pixels on the sceen is what part of the original pixel? And what about overscan?

It's even impossible to do on an old, really analog B/W television. Simply because there are no pixels at all, not in the source signal, and not on the output device.

It cannot be done. Listen to Arjan.
 
How are you going to do that? How is anyone to see the difference in color between two ajacent pixels that look the same?

And how would you cope with a resolution that isn't an integer fraction of the total amount of pixels on the screen? How is someone supposed to determine what part of each pixels on the sceen is what part of the original pixel? And what about overscan?

It's even impossible to do on an old, really analog B/W television. Simply because there are no pixels at all, not in the source signal, and not on the output device.

It cannot be done. Listen to Arjan.
I done it and got to a number of about 288 (while I expected 240 at that time) which looks to be close to the real number.

Overscan can be incorporated with assuming of having 7% overscan. To hide some geometry issues, I configured my TV picture to have a bit more overscan, though.

You cannot detect sharp "pixels" because VHS don't gives you that. But if you have a TV set able to produce images sharp enough, you can see that VHS's details are big enough that you can differentiate them from the CRT's RGB elements.

There is a problem, though, which I don't really figured out yet. It's limit frequency vs. amplitude. I made my observation with a fairly high contrast part of the image. I think (but I don't know for sure) that you can higher frequency at lower amplitude in your VHS picture.

However, there is another approach to get the VHS "effective pixel" count. Assuming VHS's video bandwidth to be 3 MHz while PAL got 5 MHz (or VHS with 3.2 and PAL with 5.5 MHz, I don't know the exact numbers; the sources are not clear about it) its safe to say that VHS delivers about 60% of the PAL's image quality. Knowing that you can have only about 540 "pixels" (or ~ 270 "Hz per scanline") it comes out to ~ 324 "pixels" (~ 160 "Hz per scanline) for VHS.
 
Last edited by a moderator:
The "correct" way to estimate the maximum frequency of VHS is to have a signal generator which can generate gradually higher frequencies, and a very high quality sampler which samples the results. Now, you use the sampler to do frequency analysis, and you can find the maximum frequency without alias.

For example, suppose that you put in a 1MHz sine wave, and in the sampled results you find the amplitude peaked at, say, 2MHz, then you know there's alias and 1MHz is already out of the maximum frequency the system allowed.
 
The "correct" way to estimate the maximum frequency of VHS is to have a signal generator which can generate gradually higher frequencies, and a very high quality sampler which samples the results. Now, you use the sampler to do frequency analysis, and you can find the maximum frequency without alias.

For example, suppose that you put in a 1MHz sine wave, and in the sampled results you find the amplitude peaked at, say, 2MHz, then you know there's alias and 1MHz is already out of the maximum frequency the system allowed.
That is (of course) true. However I don't managed it to connect my Laptop to my VCR. Then I would just record some pictures with different line densitiy. Due to some filters inside the VHS recorder, you should get no aliasing, but a blurry image if you get over the limit frequency.
 
We could *discuss* on and on ... or somebody just counts the number of "pixels" (luminance details) of a scanline. I already did but I want to get a confirmation about my figures.

There is no such thing like pixels in the analogue signal. Neither anything like resolution.
 
There is no such thing like pixels in the analogue signal. Neither anything like resolution.

While the concept of a "pixel" cannot directly be applied to an analogue signal, the concept of "resolution" most certainly can. Rather than defining "resolution" in terms of pixels, you instead just define "resolution" in terms of the smallest distinguishable feature that can appear in the signal. Such a definition lends itself perfectly well to use with VHS, as the analogue bandwidth of the VHS signal is very sharply limited (thus preventing features from becoming arbitrarily small).
 
There is no such thing like pixels in the analogue signal. Neither anything like resolution.
Since you cannot have infinite detail with a bandwidth-limited signal, "resolution" applies.

Also a form of "pixels". Lets assume we have a maximum of 160 periods per scanline, meaning 160 times SIN(0..2PI), or SIN(0..320PI). Then it looks somewhat like 320 pixels (since we have a black-white-alternating image with 160x black and 160x white.) However, if we actually want to resample the signal on a pixel raster, we could get anything between black-white-alternation and pure gray due to a phase shift. To have a high quality sampling I would propose to sample with 640 pixels. But I am not asking how to resample the signal.

I have the reverse approach: With 320 pixels, you can have 160 sine periods at max. Therefore we can consider 160 sine periods as 320 "pixels". (Not real pixels, that is why I write "pixels".)

Due to the analoge nature, we have of course no sharp pixels but sine waves only. With an oversampling of the video signal plus a sharpening filter however it is possible to get an output on screen which looks more like a periodical rectangular function and shows discernible "pixels".
 
Last edited by a moderator:
Since you cannot have infinite detail with a bandwidth-limited signal, "resolution" applies.

Also a form of "pixels". Lets assume we have a maximum of 160 periods per scanline, meaning 160 times SIN(0..2PI), or SIN(0..320PI). Then it looks somewhat like 320 pixels (since we have a black-white-alternating image with 160x black and 160x white.) However, if we actually want to resample the signal on a pixel raster, we could get anything between black-white-alternation and pure gray due to a phase shift. To have a high quality sampling I would propose to sample with 640 pixels. But I am not asking how to resample the signal.

I have the reverse approach: With 320 pixels, you can have 160 sine periods at max. Therefore we can consider 160 sine periods as 320 "pixels". (Not real pixels, that is why I write "pixels".)

Due to the analoge nature, we have of course no sharp pixels but sine waves only. With an oversampling of the video signal plus a sharpening filter however it is possible to get an output on screen which looks more like a periodical rectangular function and shows discernible "pixels".
But you need a pattern generator, color filters and an analog monitor (without pixels) to be able to do that.
 
Back
Top