Skew-Off: EX1 vs. prototype HPX300

B

Barry_Green

Moderator
Yesterday was an interesting day. I got to go to the Panasonic headquarters in NJ to settle the question of relative skew/wobble/jello between the HPX300 and the EX1. Not only did I get a chance to put both cameras through their paces, but I also had a rather notable audience, including Jan Crittenden (the Product Line Manager for the HPX300 in the USA) and several of the engineers from Japan!

They had built a side-by-side rig for some other shooting purposes; we pressed it into service for the first phase of the test, but I wasn't happy with the idea that a slight parallax error could "skew" the results, so at the end of the test I bolted a Bogen 501 head on top of the HPX300 and mounted the EX1 to that, and got the critical test measurements. With that rig I could ensure that there was no parallax error side-to-side, that each camera was centered horizontally and I balanced them such that one was pointing slightly upwards and one slightly downwards towards the same target.
Cams-Over-Under.jpg


The target was about 25 feet away, which further eliminates parallax.

I set the cameras as similarly as I could get them, with an emphasis on removing as many variables as possible. Accordingly, I turned detail off both of them, I turned off the lens image stabilizer in the EX1, and I set the field of view to be as identical as I could get it so that the relative image magnification would be the same. The tests were conducted near the telephoto end of the lens; the EX1 was at about 75mm and the HPX300 was set to match its field of view.

I also set the shutter manually, to the same value on both. I used 1/60th for normal-view testing, and 1/1000 for super-crisp to remove motion blur from the tests. I also did one test at 12fps with 1/12 shutter.

I put them through their paces in the following combinations:
1080/60i @ 1/1000
1080/60i @ 1/60
1080/24p @ 1/60
1080/24p @ 1/1000

720/60p @ 1/1000
720/60p @ 1/60
720/24p @ 1/60
720/24p @ 1/1000
720/12p @ 1/12

The test consisted of rapid panning past a Fujinon backfocus chart. The contrast of the white left page against the black-background backfocus chart on the right page made for an excellent, easily-definable and easily-measurable skew angle.
1080-60i-thumbnail.jpg


For each test, we sync'd the cameras up with a good old-fashioned hand clap so that in post I could extract identical frames for comparison.

Okay, so -- blather, blather, blather, methodology and settings, blah blah -- what's the results?

Very, very interesting results indeed. Very. In fact, I was surprised and somewhat shocked, and so were the engineers.

So, here's the lowdown: in 60i, and in 60P, at any shutter angle, the EX1 and the HPX300 are within spittin' distance of each other. When there was any sort of noticeable difference, the HPX300 actually had LESS skew, about 27% less on average. This usually showed up in 720/24p and 720/12p.

BUT -- and, as Pee Wee Herman said, "everyone has a big BUT" -- there's one mode where the HPX300 was dramatically more skew-y. And that's 1080/24p. And it wasn't a minor difference, it was a big difference, with the HPX showing an average of 75% more skew. And, since 1080/24p is such a popular mode, this is why people have been reporting that the HPX300 shows more jello than the EX1 or EX3. In all other circumstances the HPX300 should be less skewy and less jello-y, but in 1080/24p mode it's much moreso.

I was surprised at the difference, but so were the engineers. Why would 1080/60i performance be nearly identical, yet 1080/24p performance be so different? Or, for another perhaps more intriguing question, why would the HPX300 show 27% less skew in 720/24p, but 75% more skew in 1080/24p? How could they be so different?

I don't have the answer, but I did get a commitment from the Japanese engineers that they are committed to studying the situation. They saw it, they saw the difference, I gave them copies of the footage to take back with them, and I have all confidence that they will be able to address it. And, it is my understanding that if a firmware update does become available, every HPX300 camera will be eligible for it.

For those who want the pedantic details, a table follows which gives the measured skew difference, in pixels, between the two cameras, in all modes. The skew was calculated by using PhotoShop's "marquee" tool to encircle the edge of the skewed triangle, from start to finish, and having PhotoShop's "info" window report the width of the marquee window.
Measure.jpg


(keep in mind motion blur and interlacing made it a little harder to measure the 1/60 and 60i frames pixel-perfectly)

(also, the frames were extracted from any old portion of the video, so there's no guarantee that the speed was the same between any two tests, but the speed is guaranteed identical between cameras within each test)

(in the following table, smaller numbers mean less skew; underlined statements are links to the side-by-side pictures)
1080/60i @ 1/1000: EX1 112, HPX300 119; HPX300 shows 6% more skew*
1080/60i @ 1/60: EX1 160, HPX300 146; HPX300 shows 9.5% less skew

1080/24p @ 1/60: EX1 115, HPX300 185; HPX300 shows 62% more skew*
1080/24p @ 1/1000: EX1 59, HPX300 111; HPX300 shows 88% more skew

720/60p @ 1/1000: EX1 144, HPX300 145; same amount of skew
720/60P @ 1/60: EX1 94, HPX300 77; HPX300 shows 19% less skew

720/24p @ 1/60: EX1 39, HPX300 27; HPX300 shows 31% less skew
720/24p @ 1/1000: EX1 93, HPX300 72; HPX300 shows 22% less skew*

720/12p @ 1/12: EX1 38, HPX300 27; HPX300 shows 30% less skew

* = tests conducted with over/under for eliminating any potential parallax error.
 
Last edited:
Very weird that the HPX300 would show less skew in 720p/24p than the EX, but then be so much worse than the EX1 in 1080/24p mode...

weird...

Later,
Jason
 
Two inquiries:

1. All shot with DVCProHD? If that is correct, would AVCIntra make any difference?

2. How preproduction was the HPX300? Jan has indicated that later or the latest version has attenuated the skew over earlier versions?
 
David S. said:
Two inquiries:

1. All shot with DVCProHD? If that is correct, would AVCIntra make any difference?
Click to expand...
All shot with AVC-Intra. We did try DVCPRO-HD for one test and it was identical to Intra. We were also outputting the image live to identical monitors, so the skew was being seen live as well as in the recorded footage.

2. How preproduction was the HPX300? Jan has indicated that later or the latest version has attenuated the skew over earlier versions?
Click to expand...
This one was running the latest firmware and is the latest model.
 
David S. said:
Two inquiries:

1. All shot with DVCProHD? If that is correct, would AVCIntra make any difference?
Click to expand...


These were done in AVC-Intra.


2. How preproduction was the HPX300? Jan has indicated that later or the latest version has attenuated the skew over earlier versions?
Click to expand...

I had mentioned that the current firmware did indeed improve the cameras' performance. You assumed I meant skew, I meant better color and bettter S/N. Keep in mind it is only in the 1080P/24PN or in DVCPRO HD 1080i/ 24P-24PA that the image had this issue. The other modes we same or maybe even better.

We did find that the Infra-red issue on the HPX300 was basically not an issue with the HPX, unlike the counterpart.

Best,

Jan
 
I'm guessing the test footage that Shipsides posted was shot in 1080/24p because it skewed like a muthuh, whereas the sample Central Park footage wasn't so bad, artifacty as it was.
 
As I read these results, I interpret them in a couple of way:

1. On balance, the 300 does not have as significant a skew than the EX1 and/or EX3 under similar settings and shooting environment.

2. The engineers are aware of the one instance where the 300's skew exceeded the other cams, ie. 1080/24p.

Correct or augment these interpretations as needed.

But what I would still like to see is native footage on reasonable pans and other everyday shooting environments.

We are seriously looking at this camera, but the absence of sample footage where skew could be an issue is what I'd like to evaluate.

Thanks though
 
1. It appears to be, on balance, the same or better than the EX1 in terms of skew performance, with the one big exception of 1080/24p mode.

2. They are most definitely aware of it now, yes.
 
nice test, hope they will solve the problem in 24p mode before the cam comes available. Do you think the result between the hpx and ex will be the same when panning ?
 
Very interesting findings.
I doubted there's any other company in the world that has such a transparent discussion about a product that hasn't even hit the market yet.
Very impressive, that's one of the main reasons I'm such a fan.
Thanks Barry, Jan and others involved.
 
hope they will solve the problem in 24p mode before the cam comes available
Click to expand...
I don't know if that's likely; the cam is due within the next couple of weeks or so, and any changes would have to be thoroughly tested before being certified ready to go.

I would instead expect that whether it's updated before shipping or not, if they find a way to speed up the skew performance, they'll make it available to all owners. That is what I understood them to say.


Cees Mutsaers said:
Do you think the result between the hpx and ex will be the same when panning ?
Click to expand...
Well, as it stands now, the panning results will be as good, or better, on the HPX in any mode except 1080/24p.

If they can provide an update that speeds up the 1080/24p scan rate, then yes, you should be looking at something that's as good or better than the EX1 in all ways in terms of skew performance.

But if they can't update it, then the HPX300 would have significantly worse skew performance in 1080/24p (but only in that mode).

We will have to wait and see what the engineers come up with.
 
Can not imagine that if sony got this under controll why pana can't. At least you showed them the area they have some work to do and i am sure they will manage and even surpass the ex.
 
Can not imagine that if sony got this under controll why pana can't. At least you showed them the area they have some work to do and i am sure they will manage and even surpass the ex.
Click to expand...
That's my feeling too.
 
Cees Mutsaers said:
Does increase of the scan rate also reduce the change on partial exposure due to flash lights?
Click to expand...
I didn't test that portion, but I can't fathom why flash performance would be any different than skew performance, since it's caused by the same issue (scan rate). Faster scan rates will show more flash and less dark, slower scan rates will show less flash and more dark.
 
You'd think the Panasonic might have done a test like this (side-by-side w/ competitor) during development.

(I say this as someone chomping at the bit to purchase one of these, BTW)
 
Kenn Christenson said:
You'd think the Panasonic might have done a test like this (side-by-side w/ competitor) during development.

(I say this as someone chomping at the bit to purchase one of these, BTW)
Click to expand...

Keep in mind that there are so many things, there are 6 codecs in this camera all of those with the various frame rates have to be tested for failsafe operation, with 24P, 24PA , Standard Def, High Def, 50 HZ, 60HZ. there is a ton of stuff, plus new technologies like the 3-MOS, the LCOS VF and the LCD. Add to that the additional Proxy for the 60 modes, the wireless slot in on a camera of this size with a full set up for the Dual receiver from Lectrosonics. Oh and don't forget the mapping an correction on the XT17 CAC lens. They have been busy. ;-) They see this now and will study to figure out the problem.

Always easier to see where we should have looked in hindsight.

Best,

Jan


Best,
 
About that fast pan footage?

That certainly would help a bunch of people to decide.

I sound like a voice crying in the wilderness, but I'm one of those people who wants to evalaute footage, not charts.

Not trying to be a pest, but some native footage shot with this concern in mind might go a long way to sell a bunch of hardware

best :)
 
Part of the reason that folks like me and Philip Bloom and Art Aldrich and Adam Wilt and others get to see these cameras in their pre-production phase is to find stuff like this. There's internal testing, which is the first phase; then there's external testing. We're the external testing phase. And we will do stuff to them that the engineers might not anticipate.
 
You must log in or register to reply here.
Back
Top