Pixel Doubling Issue on 8bit Larger Screen Cheapo DSO

[Mechatrommer]

i was studying the topic... i came accross the relationship between screen size, and dso measurement resolution. on rigol its lesser than speced 8bit, ie around 7.6bit only when transferred to pc, eqv = 200 variation of byte value. so in order to plot the point without pixel doubling on screen larger than 200 pixels height, i'll get interleaving just as in 1.png and 2.png, more clearly in 2 (ie with infinite persistence on), it means from what i understand, there is useless band in it (the interleaved black area), we either can interpolate the value in between (just like vector graph, ie line between point) and trust it, or else, dont trust it.

now, i think this is the reason why some other cheap/low class dso with larger screen like Uni-T will implement pixel doubling (i read the discussion and complaint about it, but i dont own it), just as simulated in 3.png and 4.png, it will create ugly jaggy graph, but nice continuos no interleaving space.

pls give your opinion, i would like to hear some aspect like... do you still think that implementing drawing without pixel doubling on larger screen is better? (screen resolution > data resolution) pls consider what i said about useless band, i think interpolation (vector line) will be needed to fill the gap to simulate nice continuos thin lined graph, but actually its not the real measurement. and what really bad about pixel doubling other than the ugliness? (i mean on the display) etc etc. pls give your idea. thanx

[alm]

Interpolation is required for the Nyquist-Shannon sampling theorem to apply (how else do you reconstruct a continuous signal from discrete samples?), so I consider proper interpolation essential except in some specialized cases. With interpolation, there's no need to revert to ugly hacks like pixel doubling, which is actually a very primitive form of interpolation.

[Mechatrommer]

thanx, thats a good idea alm. anyone else? or any many?

pls note as well, that this study is done in dot mode, ie the real sampled value points plotted (scaled) on the larger screen, no interpolation (sinc or even vector, no, not used)

a case example is say we have N=600 samples available at one time with ideal 8bit value resolution (0-255) if we want to plot two consecutive samples s0 and s1, with value say 0 and 1... if we provide a screen with WxH = 600 pixels * 256 pixels, then we can plot both s0 and s1 as single point/pixel with coordinate (0,0) and (1,1) (base-0 indexing) which is continuous to the eye. but if we provide a screen with double resolution, say 1200 * 512, in order to fill the whole screen "evenly" we need to plot (2*0, 2*0) and (2*1, 2*1) ie (0,0) and (2,2) which not continuous if plotted as single pixel.

in other word, think of the difference between...
[dynamic range of 256, resolution 1 pixel (or byte)]
[dynamic range of 512, resolution 2 pixel (or byte)] ie one single value will lie in between 2 pixel, but not both (without pixel doubling)

ok, we can say... use sinc or vector to make it continuous, but when we turn persistence to infinity, then we will end up with figure 4.png. but if we disable sinc or vector with single pixel and double sized screen... in dot mode we will get 2.png. thats my point. about pixel doubling in dot mode.

ps: maybe this is not a good discussion? more on graphics and programming side, not ee but i can see users complaining, but i think this issue is not clear to them. need more understanding and better technique to solve this dynamic range and resolution issue when plotted (dot mode) on pixelated screen area. we may leave it interleaved to maintain single pixel, or else better technique than pixel doubling. maybe no solution other than sincx or vector.

or maybe this is the reason why rigol do not allow the "real" single pixel dot plot when in run mode, to avoid interleaved effect as in 2.png if infinite persistence is ON. i tried some setting in rigol (dot mode + persistence) but didnt get the interleaved effect, so i'm suspecting rigol also do some level of pixel doubling but not obvious. my explanation is pixel doubling only worstly done on Y axis, but not very clear on X axis since rigol screen is around 640 width x 480 height pixels compared to 600 data points with 256 dynamic range (200 actually) so X axis doubling at factor 640/600 = 1.07 which alot lesser than Y axis at factor 480/256=1.9 which is very close to "doubling" (to be more accurate 480/200=2.4) will look closely into this, this is my asumption base on rough observation so far.

i wonder if high end dso will do the same (8bit larger screen) except they do it pretty nice including antialiasing and blurring effect (ie they put more effort on graphics FW engine fanciness) i understand engineers deal with real data, not graphics fabulous, but engineers are human too, and they like pretty stuffs. thats why maybe dave went WOW! looking at mega thousands dollar dso in his video when he went to renexas exhibition iirc... fancy graphics!