Confused about light loss on "progressive" cams

[John Trent]

I've read conflicting things about this, so could someone set me straight.

Progressive scan costs 1 stop of light and has more noise. So, since the Canon A1/G1/H1 and Sony FX1/Z1 aren't "officially" progressive they don't have these downsides, (when F mode and Cineframes are employed) correct?

And the FX7 and V1 are said to be progressive but use interlaced chips, so.... they do or don't have the stop loss and noise when shooting progressive? Thanks.

[AuditoryVisuals]

Where'd you hear that the FX7 was progressive? Where'd you hear the V1U uses interlaced sensors? 1080i60 has better compression than 1080p60 so that might be your compression artifacts that you hear about.

Where did you hear about all this?

[magichristopher]

The reason there is light loss in progressive is because the CCD or CMOS get sampled less frequently, 1/2 or less to be more precise. For 60i it is sampled 60 times a second for 30p or 24p it is 30-24 times a second.

This is why, in a way that the Canons have better sensitivity in lo light than true progressive.

I am glad that they implemented the frame mode, considering the quality is as good as the true progressive camcorders in the same price range.

[Barry_Green]

Interlaced scanning does produce about a 6dB benefit in noise and sensitivity. That's the only way those cams could produce a viewable image with pixels that small. People gripe about the interlaced nature of the chips, yet they want their 1440x1080 too; you couldn't have any sort of practical camera if you didn't combine those two!

FX7 and V1U are both progressive scan systems. And even though Sony says that the light-gathering surface of each pixel on the 1/4" CMOS FX7/V1U is about the same size as the light-gathering surface of each pixel on the 1/3" FX1/Z1U, the FX7/V1U is about a stop slower than the FX1/Z1U. Which, under comparable-CCD conditions, is exactly what you'd expect. But that was not a foregone conclusion because these cams use CMOS instead of CCD, so there was no way to accurately predict, but according to Jon Fordham's review on DVInfo.net he says that the V1U is about a stop slower than the Z1, so it all basically adds up.

[John Trent]

Can't read the Fordham review, it's down. Dvinfo is at work censoring again, I know from personal experience about that.

I read before that the DVX is about as sensitive as the Sony VX/PD series but as soon as it is put in Progressive mode it loses one stop of light and adds noise. That's the trade-off. Is this how it works? If so, is the V1 one stop less sensitive in interlace (than the FX1/Z1 because of the 1/4 chips) and as soon as Progressive is employed it loses yet another stop, making it two stops less sensitive than the Sony FX1/Z1?

[Barry_Green]

Well, yes and no. I'd have to test to find out, but I would suspect that that's probably true that there's no sensitivity boost when going to interlace on the V1. It's true of the HVX (no sensitivity gain when going to 60i mode) so I would guess (and it's only a guess at this point!) that the V1 would similarly not gain any sensitivity when going to interlace.

However, no it won't be two stops slower than a Z1 because the Z1 is always interlaced, thus it is always the same "speed". And a V1 is always progressive (I ASSume) so it should always be the same "speed". So a V1 in interlace or progressive should be the same speed at all times, and a Z1 should be the same speed at all times, and therefore a V1 should always be about 1 stop slower (according to Fordham's report) than a Z1, regardless of progressive or interlace setting on either.

[Barry_Green]

I read before that the DVX is about as sensitive as the Sony VX/PD series but as soon as it is put in Progressive mode it loses one stop of light and adds noise. That's the trade-off. Is this how it works?

Well, yes and no. That would be a fairly accurate way to describe the effect, but it's actually looking at it backwards.

The appropriate way to say it would be that when you put it in interlace mode it gains a stop, and the noise is reduced. You don't actually "lose" a stop in progressive, you get the raw chip sensitivity, but in interlace the chip gets scanned twice and the row-pair summation results in averaging those scans together to get a 6dB gain in brightness and it also blends the noise together for a 6dB reduction in visible noise.

[Emanuel]

Can't read the Fordham review, it's down. Dvinfo is at work censoring again, I know from personal experience about that.
Bad fortune.

[FatDaddy]

Is then the Canon at 24F actually have the ability to be more sensitive because it is not true progressive?

[Barry_Green]

Is then the Canon at 24F actually have the ability to be more sensitive because it is not true progressive?
It's more sensitive than it would have been had it been true progressive.

Right now it's almost a stop slower than the HVX, and equivalently noisy. Had it been true progressive it'd be two stops slower and twice as noisy. That's about what you'd expect when the pixels are 1/3 the size. Cramming that many pixels into the surface area that they did lowers light sensitivity. By using interlace instead of progressive they were able to recover a stop's worth of sensitivity and cut the noise down.

[magichristopher]

Really, a stop slower?

I thought that it was better in low light, isn't the XLH1 better than the hvx?

Users have reported that it is brighter than the FX1 at 12db with more colors and detail, but considerably more noise, and the fx1 is about equal in brightness at 18db.

Also at 0db, and open, they appear to be neck and neck; I thought that the HVX was 1 stop slower than the fx1. Am I wrong on this?

[Barry_Green]

"better in low light" is not a scientific, quantifiable measurement. ISO sensitivity is. But even then, it should be sensitivity at a certain noise level. And color response at low light levels varies too.

The FX1 rates at about 160 ISO, the XLH1 at about 200, the HVX at 320. At zero gain, the XLH1 is about 1/4 stop more sensitive than the FX1. But the noise levels on the XLH1 and HVX are about equal, whereas the FX1 is a lot quieter. So you can gain up the FX1 to about 6dB and get comparable sensitivity and comparable noise. The XLH1, gained up, will be more "sensitive" but a lot noisier.

You can put a Sony Z1 on 36dB Super-Hyper Gain, and get it to practically see in the dark (36dB = about 16x as bright as 0dB) but would one then call that "better in low light"? Maybe, if the goal is to get any manner of picture whatsoever, but maybe not if the goal is to get a usable low-noise picture.

I haven't used an H1 since the DV shootout, and we didn't specifically test for "low light performance" but we did test for sensitivity and noise, and found it to be about 2/3 stop slower, with comparable noise level, to the HVX. About 1/4 stop faster, and much noisier, than the Z1. About 1/2 stop slower, but noticeably cleaner/quieter, than the JVC.

But I wouldn't call that a comprehensive "low light" test.

Here's the thing: each manufacturer gets so many "points" to assign to the various categories, but it all has to add up to the same thing. Canon apparently decided that smaller pixels for more resolving power was their goal, so they had to compromise noise performance and sensitivity; a progressive system would have been horrifically noisy and slow so they gave up true progressive in an engineering tradeoff to keep the noise and sensitivity at reasonable levels. Sony apparently decided that a noise-free image was their main goal, so they gave up progressive and sensitivity (and dynamic range), but got maximum noise-free images. Panasonic and JVC apparently both chose progressive 24P as their main goal. JVC must have decided that their secondary goal was maximum pixel count, so JVC suffers on noise and split-screen issues but gets good resolution marks and surprising sensitivity. Panasonic got the best sensitivity and comparable noise performance (even though it's a progressive system!) by compromising on the # of pixels on the chip. But it's all a zero-sum game, there is no magic formula there, each manufacturer starts with the same basic clay to build from (i.e., modern CCD technology) and molds it one way or the other to extract whatever their preferred image quality performance is. Add performance in one area, and you necessarily have to take it away in another area. There's only so many "points" you can allocate overall.

Besides biasing towards sensitivity, the only other things I can think of for a manufacturer to do to improve "low light" performance are to use a bigger-diameter lens (hence why the FX1 went to 72mm from the VX2000's 58mm, and why the HVX went to 82mm from the DVX's 72mm) or to juice up the internal gain/noise floor. And that's probably why we're hearing complaints from early XHA1 adopters about noise, because on the A1 they actually went with a smaller-diameter lens than on the XLH1. So if they want comparable sensitivity, they'd have to juice up the "internal gain" (i.e., raise the noise floor).

[Ralph Oshiro]

Kinda off-topic a bit, but . . .

Okay, now I have a question that I just can't figure out . . . Why do I LOSE a stop when I go from "shutter off" (which I assume is a 1/60th shutter) to 1/48th shutter on my DSR450 when shooting in progressive? Like when I need more light, I just turn my 1/48th shutter off, and voila--I gain a whole stop!

[Barry_Green]

Because on bigger broadcast cameras, "shutter off" means NO SHUTTER. So in 24p, shutter off = 1/24.

On the DVX/HVX it doesn't work that way; on the DVX/HVX "shutter off" = "default shutter speed", typically 1/50, 1/60, or 1/48. But on broadcast cameras it typically means no shutter at all.

So when you go from "shutter off" to 1/48, you should lose exactly one stop, as you're going from 1/24 to 1/48.

[Noel Evans]

And that's probably why we're hearing complaints from early XHA1 adopters about noise, because on the A1 they actually went with a smaller-diameter lens than on the XLH1. So if they want comparable sensitivity, they'd have to juice up the "internal gain" (i.e., raise the noise floor).

Actually its been found those with issues are not manually operating their cameras and thus using the auto gain control, which is causing the increased noise. Now of course I will to some degree defend the camera I purchased, but putting that aside totally, I am not suffering any noise issues whatsover. So the probably why in this case is inaccurate.

[Elton]

And that's probably why we're hearing complaints from early XHA1 adopters about noise, because on the A1 they actually went with a smaller-diameter lens than on the XLH1. (i.e., raise the noise floor).

Just a minor correction...but I'm pretty sure the XH lens is 72mm also.

On another note, isn't what we're talking about that really differentiates the Canon's and the HVX is the subjective "quality" of the noise, not necessarily a measure of actual noise level? I've always felt that the H1 and now the XH exhibit a more fine luma noise, wheras the HVX seems to have more of a chroma noise which has a totally different feel to it.

It comes down to an aesthetic choice, and deciding which camera has a gain switch you're less afraid to use when you have no other alternative.

[magichristopher]

Looking over at this thread:
http://www.dvinfo.net/conf/showthread.php?t=79138

It appears that with the coring down the xha1 is a significantly better picture than the HVX with cleaner detail and better dynamic range. This is just from my eye, but the HVX looks soft and saturated, lacking good color seperation. From a subjectvive perpective it feels like a watercolor painting ( I know that is a weird analogy) but the way that the HVX handles detail is unusual to my eye.

I do like the warmness of the HVX and the palette, but everything else looks inferior based on the comparison.

[magichristopher]

I just looked at it again on a better monitor, and WOW, it is night and day the differences in quality in low light. I am sure that the HVX could be configured differently in the menus, but that wouldn't give it a criper look without introducing gain. The XHA1 looks really nice clean and natural.

[John Trent]

Thanks for clarifying everything, Barry. I agree with your summation of the companies' camera philosophies.