Last edited:
AC130: Video footage of the Auto Focus in action
- Thread starter ELN614
- Start date
kwkeirstead
Well-known member
Many mysteries . . . I just finished 10 AF tests done under strict conditions of horizontal pan, same lighting.
I filmed the LCD panel this time so that everyone can see what is going on and soon as I have a rendering I will upload it to YouTube.
The way servos operate, you have a start phase that overcomes static friction, then, once the target is identified, you have a deceleration phase with the design option of docking gracefully or allowing overshoot that requires a reverse of direction.
I have not been able to figure out whether the AC160/130 lens functions differently for distant/near focus changes, compared to the functioning for small changes in focus.
Perhaps we could measure this by doing a fast pan from the distant object to the near object and observe lock-in time once we are at the near object. Then repeat with a very slow approach where the scenery is at all times in focus during displacement.
In any case, my sole observation following the 10 test runs is that if you are able to identify a high contrast area near/on your target and you cause the camera to pan through that area, you get faster focus lock-in.
The AREA focus option seems to be quite small in this camera which is a good thing in one sense that you can pick where you want your main focus but this of course works against you when there is motion because it is harder to maintain your position.
In my tests, as you will see, panning through the left edge of a lamp base (my near object) gives consistent "rapid" focus lock-in. Arriving at my target position from the right edge of the lamp base gives consistent "slow" or "no" lock-in.
Now, if one edge of a target allows rapid focus lock-in and another does not, bearing in mind the target in my testing is stationary, how does this map to the scenario where the same object is moving?
i.e. the contrast of the background scenery that accounted for part of the rapid lock-in in the static scenario is now changing
Once you have your object locked-in, will the camera retain its focus on the moving object?
See http://youtu.be/y41D0CNmbVw
I filmed the LCD panel this time so that everyone can see what is going on and soon as I have a rendering I will upload it to YouTube.
The way servos operate, you have a start phase that overcomes static friction, then, once the target is identified, you have a deceleration phase with the design option of docking gracefully or allowing overshoot that requires a reverse of direction.
I have not been able to figure out whether the AC160/130 lens functions differently for distant/near focus changes, compared to the functioning for small changes in focus.
Perhaps we could measure this by doing a fast pan from the distant object to the near object and observe lock-in time once we are at the near object. Then repeat with a very slow approach where the scenery is at all times in focus during displacement.
In any case, my sole observation following the 10 test runs is that if you are able to identify a high contrast area near/on your target and you cause the camera to pan through that area, you get faster focus lock-in.
The AREA focus option seems to be quite small in this camera which is a good thing in one sense that you can pick where you want your main focus but this of course works against you when there is motion because it is harder to maintain your position.
In my tests, as you will see, panning through the left edge of a lamp base (my near object) gives consistent "rapid" focus lock-in. Arriving at my target position from the right edge of the lamp base gives consistent "slow" or "no" lock-in.
Now, if one edge of a target allows rapid focus lock-in and another does not, bearing in mind the target in my testing is stationary, how does this map to the scenario where the same object is moving?
i.e. the contrast of the background scenery that accounted for part of the rapid lock-in in the static scenario is now changing
Once you have your object locked-in, will the camera retain its focus on the moving object?
See http://youtu.be/y41D0CNmbVw
Last edited:
Terry_Martin
Active member
ELN614
Your examples were very disturbing, and got me interested in this thread.
I could explain everything by saying is is bad software coding, but I give Panasonic too much respect to go there.
My theory, (and it just pure speculation from clues in this thread), is that there is a processor power limitation in the system that implements autofocus. Essentially it is a real time computer control system with data in, and control signals out. Now what happens if the computer is just barely capable of keeping up with the data load and something goes wrong, the output control signals either stall or glitch.
To me, the symptom of jumping off a locked down shot is a clue to unstable CPU hardware glitches. I doubt software errors would jump off good shots for no reason.
@ sewolla
In a HPX250 post, Jan also told me that the HPX250 uses the same hardware for focusing that the AC130/160 uses. I'm sure it is an honest comment, and probably refers to the lens/ imager part of the system. I once ran a large R&D department, and I doubt anyone in marketing/ sales really knew what we did beyond the spec sheets. I certainly never told them about our mistakes.
I shoot mainly music/ event video and 99% of it is with autofocus. My preference for best video when autofocus is needed is 1) HPX 250, 2)HMC 150, 3) HMC 40, and 4) XH A1.
ChrisHarding
Well-known member
Hi Terry
I had two AC-130's and had to on-sell them after 80 hours due to a lot more issue than just the autofocus. If it helps any, I found that even at wide angle and close to subjects, the AF display (set to feet) would run around almost out of control. I do weddings and I was filming people on the dance floor from maybe 15' away at the very most with a good on-camera light ( 6 x 3W PowerLEDs) on the camera and I noticed that the distance display would run around like crazy from anything from 3' right out to infinity!! Now with my current HMC82's (which have 1/4" chips) the display of distance is very stable and will vary up and down only a few counts (the HMC only displays 00 -99 and at 15' it would be in the 40-50 region) Same light on the camera too! I've probably clocked up 300 hours or more on the HMC's with never a focus issue and like music events, most times you have to be in auto as everything is going on at once. Both the HMC40, 80, 150 and the 130/160 shoot at 24mbps VBR so why would the processor need to work any harder in an AC-160 compared to say an HMC40/80/150 ..in fact the 150 has the same size chips and the 250 seems to have almost identical lens/chip specs to the 160!! 22X lens and 1/3"3MOS so it runs the same lens/imager setup and obviously you have no AF issues at all.
I would have thought that a company always mindful of R&D and design costs would have simply used the lens/electronics from the 250 unless the 130 and 160 are supplied with different electronics? Everthing to me seemed different and I don't know if you use audio in auto but if you use it in the 130/160 it doesn't control input levels but simply shaves the top of the waveform off to decrease it's amplitude (with not very good results) I have been on Panny gear since 1989! (Yep VHS -M4 cameras and right thru the range and audio has always been so good) If they are using a new audio board then maybe they are NOT using the same processor or electronics that would account for your theory.
What I fail to understand is that the HPX250 is almost identical to the 160 in specs APART from the recording media (AVCHD and DVCPRO) and I don't see any 250 owners crying about autofocus so what has changed ?? Surely purely for costing, Panasonic would have used the existing electronics for everything except the recording and not completely redesigned all the boards.... 250 boards would have already been in production and seem to me to be a logical way to launch an AVCHD version of the HPX250...the HMC80 and 40 are done that way and operate almost the same so what's so different to cause all these issues???
I'm sure the guys here are very grateful for your input
Chris
I had two AC-130's and had to on-sell them after 80 hours due to a lot more issue than just the autofocus. If it helps any, I found that even at wide angle and close to subjects, the AF display (set to feet) would run around almost out of control. I do weddings and I was filming people on the dance floor from maybe 15' away at the very most with a good on-camera light ( 6 x 3W PowerLEDs) on the camera and I noticed that the distance display would run around like crazy from anything from 3' right out to infinity!! Now with my current HMC82's (which have 1/4" chips) the display of distance is very stable and will vary up and down only a few counts (the HMC only displays 00 -99 and at 15' it would be in the 40-50 region) Same light on the camera too! I've probably clocked up 300 hours or more on the HMC's with never a focus issue and like music events, most times you have to be in auto as everything is going on at once. Both the HMC40, 80, 150 and the 130/160 shoot at 24mbps VBR so why would the processor need to work any harder in an AC-160 compared to say an HMC40/80/150 ..in fact the 150 has the same size chips and the 250 seems to have almost identical lens/chip specs to the 160!! 22X lens and 1/3"3MOS so it runs the same lens/imager setup and obviously you have no AF issues at all.
I would have thought that a company always mindful of R&D and design costs would have simply used the lens/electronics from the 250 unless the 130 and 160 are supplied with different electronics? Everthing to me seemed different and I don't know if you use audio in auto but if you use it in the 130/160 it doesn't control input levels but simply shaves the top of the waveform off to decrease it's amplitude (with not very good results) I have been on Panny gear since 1989! (Yep VHS -M4 cameras and right thru the range and audio has always been so good) If they are using a new audio board then maybe they are NOT using the same processor or electronics that would account for your theory.
What I fail to understand is that the HPX250 is almost identical to the 160 in specs APART from the recording media (AVCHD and DVCPRO) and I don't see any 250 owners crying about autofocus so what has changed ?? Surely purely for costing, Panasonic would have used the existing electronics for everything except the recording and not completely redesigned all the boards.... 250 boards would have already been in production and seem to me to be a logical way to launch an AVCHD version of the HPX250...the HMC80 and 40 are done that way and operate almost the same so what's so different to cause all these issues???
I'm sure the guys here are very grateful for your input
Chris
Last edited:
Terry_Martin
Active member
Hi Chris
Yes, I remember when you posted that you sold your 130's. That's when I started paying attention to the focus issue.
The hardest job of the camera cpu is to encode the raw video to the final format. Remember when HDV cameras were common and AVCHD was just emerging. AVCHD is harder to encode than HDV, and it took a while to develop the chips to real time process the compression.
I would confidently guess that the P2 data compression (and higher data rate) requires radically different processing chips than an AVCHD camera, and hence the HPX 250 back end processing electronics is vastly different than an AVCHD camera line, which actually needs a somewhat specialized chip for the AVCHD. I would again guess the P2 electronics to be perhaps more powerful. Another clue is that Panasonic retained the older batteries used in P2 cameras for the HPX250, and passed over the newer HMC 150 batteries. Sure looks like they were just sticking with their proven designs.
In terms of audio, I have determined the input electronics could be designed to be more robust, but works OK is proper signals are applied. I only record audio to do multi cam syncing, and use a Sound Devices 788T or PIX240 to record production audio.
Yes, I remember when you posted that you sold your 130's. That's when I started paying attention to the focus issue.
The hardest job of the camera cpu is to encode the raw video to the final format. Remember when HDV cameras were common and AVCHD was just emerging. AVCHD is harder to encode than HDV, and it took a while to develop the chips to real time process the compression.
I would confidently guess that the P2 data compression (and higher data rate) requires radically different processing chips than an AVCHD camera, and hence the HPX 250 back end processing electronics is vastly different than an AVCHD camera line, which actually needs a somewhat specialized chip for the AVCHD. I would again guess the P2 electronics to be perhaps more powerful. Another clue is that Panasonic retained the older batteries used in P2 cameras for the HPX250, and passed over the newer HMC 150 batteries. Sure looks like they were just sticking with their proven designs.
In terms of audio, I have determined the input electronics could be designed to be more robust, but works OK is proper signals are applied. I only record audio to do multi cam syncing, and use a Sound Devices 788T or PIX240 to record production audio.
Chris, I have seen a few comments on AF in the HPX forum, and sometimes they were referred to the AC130/160 board, because of all the research done on it here. I also don't think HPX users are as interested in using AF as we are here, so it may not be as big a deal to them.
But I also am wondering, if they both (HPX250/AC130, 160) come out of the same factory? I have actually very little AF issues with my 160.
That has caused me to wonder, as an old aerospace manufacturing guy, if lenses are all coming from the same source, and if they are all being "built to print." Maybe on the more "extreme" cases of AF failure, in units where it occurs more often, is it because the manufacturing processes allow for more deviations from print. What does the tollerance stack up on the lens assembly look like? You could in theory have a lens where yeah, it's assembled to print, but if dimensioning is consistently on the high side of the tollerance, that could compromise the function of the lens.
But I also am wondering, if they both (HPX250/AC130, 160) come out of the same factory? I have actually very little AF issues with my 160.
That has caused me to wonder, as an old aerospace manufacturing guy, if lenses are all coming from the same source, and if they are all being "built to print." Maybe on the more "extreme" cases of AF failure, in units where it occurs more often, is it because the manufacturing processes allow for more deviations from print. What does the tollerance stack up on the lens assembly look like? You could in theory have a lens where yeah, it's assembled to print, but if dimensioning is consistently on the high side of the tollerance, that could compromise the function of the lens.
ChrisHarding
Well-known member
That's quite interesting ! I see that Terry does say he shoots music/event video and 99% is on autofocus and his choice of camera puts the HPX250 up front so if it behaved as the 160 does I'm sure he would not be using it?
I wonder if all video cameras actually come out of the Osaka factory or does Panasonic have facilities elsewhere where cameras are built and maybe also designed...I have never owned an HPX so I cannot do a comparison but I did many years ago own two MD10000's which were the PAL versions of the old DVC20 and having seen both the build quality of the PAL model seemed completely different ...the MD was a lot cheaper and very plastic but, shucks, it still shot great video in the SD days!!! I wouldn't have a problem with the origin of my product (as long as it was a Panasonic and worked like one) but if some parts of the 160 are sourced from places other than the 250 supplier then the electronics (and processor board) might also be totally different too?? Then again the 130 and 160 is a completely different animal to the 1/4" chip shoulder mount camera and it might have been just too different for me? I learnt to live with the AF BTW!! It's still pretty good at wide and more often than not I could switch to manual. My BIG issue was the autoiris that doesn't kick in shutter speed as the light changes so for a solo operator with 2 cameras I couldn't baby it enough. That was my problem not Panasonics as the cam worked the way it was supposed to!
Terry?? Have to done any sort of test (or could you?) shifting from one object to another in "lowish" light at full zoom and does the HPX250 do the same slow focus or focus lockup that the 160 tends to do??? I'm sure that Rod would very much appreciate a result if you had enough time to do it..That would confirm that AF at full zoom on the 160 is or is not a problem. If the HPX also locks up focus that would mean it's definately a big lens/imager/processor issue.
Chris
I wonder if all video cameras actually come out of the Osaka factory or does Panasonic have facilities elsewhere where cameras are built and maybe also designed...I have never owned an HPX so I cannot do a comparison but I did many years ago own two MD10000's which were the PAL versions of the old DVC20 and having seen both the build quality of the PAL model seemed completely different ...the MD was a lot cheaper and very plastic but, shucks, it still shot great video in the SD days!!! I wouldn't have a problem with the origin of my product (as long as it was a Panasonic and worked like one) but if some parts of the 160 are sourced from places other than the 250 supplier then the electronics (and processor board) might also be totally different too?? Then again the 130 and 160 is a completely different animal to the 1/4" chip shoulder mount camera and it might have been just too different for me? I learnt to live with the AF BTW!! It's still pretty good at wide and more often than not I could switch to manual. My BIG issue was the autoiris that doesn't kick in shutter speed as the light changes so for a solo operator with 2 cameras I couldn't baby it enough. That was my problem not Panasonics as the cam worked the way it was supposed to!
Terry?? Have to done any sort of test (or could you?) shifting from one object to another in "lowish" light at full zoom and does the HPX250 do the same slow focus or focus lockup that the 160 tends to do??? I'm sure that Rod would very much appreciate a result if you had enough time to do it..That would confirm that AF at full zoom on the 160 is or is not a problem. If the HPX also locks up focus that would mean it's definately a big lens/imager/processor issue.
Chris
kwkeirstead
Well-known member
Looking forward to feedback re my http://youtu.be/y41D0CNmbVw
Can anyone explain why approaching the near object one way gives reasonably acceptable focus lock-in whereas approaching the near object another way gives consistently poor focus lock-in or no focus lock-in?
And, secondly, why do we see consistent quick focus lock-in at the distant object?
The "fast" focus lock-in at the distant object may be the result of that object having just the right contrast that the camera "likes" - so how about I repeat the testing by switching the positions of the two objects?
Can anyone explain why approaching the near object one way gives reasonably acceptable focus lock-in whereas approaching the near object another way gives consistently poor focus lock-in or no focus lock-in?
And, secondly, why do we see consistent quick focus lock-in at the distant object?
The "fast" focus lock-in at the distant object may be the result of that object having just the right contrast that the camera "likes" - so how about I repeat the testing by switching the positions of the two objects?
Last edited:
kwkeirstead
Well-known member
Today, after repeating the testing, I am able to get consistent quick lock-in at both the near and far objects.
It seems it's all about contrast as Bernie and others have many times pointed out.
When I approach the near object from the left we get fast lock-in because the adjacent background has no contrast of its own. Put the area box on this background and the camera never settles down. Overlap the area box, partly on the no-contrast area and partly on the near object, and you get instant lock-in.
Approach the near object from the right where the adjacent background has lots of contrast and it seems there is no way for the camera to decide what to focus on. The near object itself is brass. The way light falls on it, it looks like it has undulations but I suspect that it is not contrasty enough to give rapid focus lock-in
It seems it's all about contrast as Bernie and others have many times pointed out.
When I approach the near object from the left we get fast lock-in because the adjacent background has no contrast of its own. Put the area box on this background and the camera never settles down. Overlap the area box, partly on the no-contrast area and partly on the near object, and you get instant lock-in.
Approach the near object from the right where the adjacent background has lots of contrast and it seems there is no way for the camera to decide what to focus on. The near object itself is brass. The way light falls on it, it looks like it has undulations but I suspect that it is not contrasty enough to give rapid focus lock-in
Barry_Green
Moderator
Regarding the "laws of physics", I'm surprised to see it be dismissed so quickly, as it is absolutely a factor.
The cameras that came before, the HMC150 and DVC30 etc that people want to compare it to, had a comparatively much smaller lens, and a much lower resolution chip. The AC160/HPX250 lens is massive. It's a lot of glass to shuffle around.
Furthermore, it has a much much sharper chip which means there is far less room for error. So you have to move a heavier weight, to a more precise location -- sorry, that's a harder job than moving a small weight to a more ballpark location, no matter how you slice it.
Now, if you want to make the task impossible, shoot at a slower frame rate while you're at it. Shoot 24p, on a huge 82mm-wide 22x zoom lens, on a true 1920x1080 HD chipset, on a chip that has shallower DOF than the DVC30 or other smaller-chip camcorders, and you can (hopefully) see that it's a much more difficult task than shooting 60i on a 1/4" standard-def 10x zoom lens (like the DVC30 was).
The reason the "laws of physics" discussion was brought into play was because people were comparing apples with oranges, without understanding the magnitude of difference that is involved. Imagine trying to park a Smart car in a normal parking spot -- there's plenty of room all around, it's spacious, and the smart car is lightweight and nimble and agile and has a magnificent turning radius (Bob Diaz showed how he could turn 180 degrees in his garage without ever leaving the garage). It's a pretty easy task.
Now, in comparison, imagine trying to park a Hummer H1 in that same parking spot. It's a lot harder. The H1 is much bigger physically, and weighs 8,000 pounds -- it's harder to move it around with precision. And because it's physically bigger, the margin of error for fitting it within the space is a lot tighter, it's a lot more difficult to make sure that every bit of the H1 is contained within the given parking space. And now, imagine adding the complexities of a slower frame rate (such as 24p) to the equation -- let's presume that instead of being able to see out the windshield of the Hummer, you had to rely on a 1/2-second delayed camera feed, so every time you moved the car you saw the results displayed 1/2 second later. How easy would that task be? Not so easy.
In combination, that's the difference that I was trying to illustrate by pointing out that the AC160/AC130/HPX250 lens is a whole different ball of wax as compared to the cameras it was being compared to (a small-chip DVC30 and a lower-res, smaller-lens HMC150). You cannot reasonably expect the same performance from a standard-def, 1/4", 10x super-small 37mm zoom lens on a palmcorder, and a massive 82mm 22x hunk of glass strapped to super-high-res high-def chips.
The HMC150 is a more appropriate comparison, but even then the HMC150's lens represents probably 1/2 to 1/3 as much glass as what's in the AC series, and the chips just aren't as sharp. The more weight there is, the more effort it takes to move it and the less precise that movement will be and the more difficult it will be to reverse direction on that weight. Shuffling around a huge 22x lens is a bigger task than shuffling down a smaller 12x lens.
Finally, Terry, I appreciate you trying to help, I just want to keep things in the realm of the realistic. For example -- the AC160, AC130, and HPX250 all have the same lens as far as I can tell. Hardware seems identical, and when I had an AC160 and HPX250 side by side, as near as I can tell they perform identically. And you say that your HPX250 works fine. Yet there are folks clinging on your words as an authoritative response as to why their AC cameras are "broken" or whatever. My take on it is -- it's all in the customer's expectations, and I think some folks are expecting more from the system than what it can realistically deliver. I think you, being content with your HPX250's performance, probably have realistic expectations.
kwkeirstead's testing is revealing what we already know -- you have to have adequate contrast for the lens to know what to lock onto. Period. That's ALWAYS been the #1 rule in autofocus and has always been the way all these cameras work, you can read my statements on that back in the DVX Book published in 2004, this is nothing new. In order for autofocus to work, you need contrast, and you need frequent updates. A flat-contrast scene shot at 24p or 30p is going to result in terrible autofocus performance. A high-contrast scene shot in 60i/60p is going to result in the best autofocus performance the system can deliver.
Does the AC160/HPX250/AC130 perform as tack-sharp and quick as the HMC150? No they do not. Sorry, they don't. Hauling around 2x to 3x as much glass, and having to move it to the precision demanded by the higher-resolution chips, has made a system that isn't AS good as the HMC150 was. It still works reasonably well, especially when combined with the area focus, and when used in situations where autofocus can be expected to perform at its best -- high contrast and high frame rate.
What I would be interested to see is a comparison between comparable lens systems from different manufacturers, to see how the AC/HPX lens holds up compared to its competition. Is there any camera directly comparable? The XF300 has the chips to match, but its lens is 18x, not 22x. Sony has a 20x on the NX5U but that's 1-megapixel chips and a 72mm filter diameter so it's not as much glass...
As far as I know Jan said she compared the AC160 against the Canon and Sony equivalents and found that they had comparable issues. I haven't compared them and have no firsthand testimony about that. I googled "NX5U autofocus" and found a few threads:
http://www.dvinfo.net/forum/sony-nxcam-hxr-nx5u/504351-focus-issue-nx5u.html
http://www.dvinfo.net/forum/sony-nxcam-hxr-nx5u/497215-nx5u-major-focus-issue.html
http://www.dvinfo.net/forum/sony-nxcam-hxr-nx5u/495423-nx5u-focusing-still-issue.html
http://forums.creativecow.net/thread/298/1197
The cameras that came before, the HMC150 and DVC30 etc that people want to compare it to, had a comparatively much smaller lens, and a much lower resolution chip. The AC160/HPX250 lens is massive. It's a lot of glass to shuffle around.
Furthermore, it has a much much sharper chip which means there is far less room for error. So you have to move a heavier weight, to a more precise location -- sorry, that's a harder job than moving a small weight to a more ballpark location, no matter how you slice it.
Now, if you want to make the task impossible, shoot at a slower frame rate while you're at it. Shoot 24p, on a huge 82mm-wide 22x zoom lens, on a true 1920x1080 HD chipset, on a chip that has shallower DOF than the DVC30 or other smaller-chip camcorders, and you can (hopefully) see that it's a much more difficult task than shooting 60i on a 1/4" standard-def 10x zoom lens (like the DVC30 was).
The reason the "laws of physics" discussion was brought into play was because people were comparing apples with oranges, without understanding the magnitude of difference that is involved. Imagine trying to park a Smart car in a normal parking spot -- there's plenty of room all around, it's spacious, and the smart car is lightweight and nimble and agile and has a magnificent turning radius (Bob Diaz showed how he could turn 180 degrees in his garage without ever leaving the garage). It's a pretty easy task.
Now, in comparison, imagine trying to park a Hummer H1 in that same parking spot. It's a lot harder. The H1 is much bigger physically, and weighs 8,000 pounds -- it's harder to move it around with precision. And because it's physically bigger, the margin of error for fitting it within the space is a lot tighter, it's a lot more difficult to make sure that every bit of the H1 is contained within the given parking space. And now, imagine adding the complexities of a slower frame rate (such as 24p) to the equation -- let's presume that instead of being able to see out the windshield of the Hummer, you had to rely on a 1/2-second delayed camera feed, so every time you moved the car you saw the results displayed 1/2 second later. How easy would that task be? Not so easy.
In combination, that's the difference that I was trying to illustrate by pointing out that the AC160/AC130/HPX250 lens is a whole different ball of wax as compared to the cameras it was being compared to (a small-chip DVC30 and a lower-res, smaller-lens HMC150). You cannot reasonably expect the same performance from a standard-def, 1/4", 10x super-small 37mm zoom lens on a palmcorder, and a massive 82mm 22x hunk of glass strapped to super-high-res high-def chips.
The HMC150 is a more appropriate comparison, but even then the HMC150's lens represents probably 1/2 to 1/3 as much glass as what's in the AC series, and the chips just aren't as sharp. The more weight there is, the more effort it takes to move it and the less precise that movement will be and the more difficult it will be to reverse direction on that weight. Shuffling around a huge 22x lens is a bigger task than shuffling down a smaller 12x lens.
Finally, Terry, I appreciate you trying to help, I just want to keep things in the realm of the realistic. For example -- the AC160, AC130, and HPX250 all have the same lens as far as I can tell. Hardware seems identical, and when I had an AC160 and HPX250 side by side, as near as I can tell they perform identically. And you say that your HPX250 works fine. Yet there are folks clinging on your words as an authoritative response as to why their AC cameras are "broken" or whatever. My take on it is -- it's all in the customer's expectations, and I think some folks are expecting more from the system than what it can realistically deliver. I think you, being content with your HPX250's performance, probably have realistic expectations.
kwkeirstead's testing is revealing what we already know -- you have to have adequate contrast for the lens to know what to lock onto. Period. That's ALWAYS been the #1 rule in autofocus and has always been the way all these cameras work, you can read my statements on that back in the DVX Book published in 2004, this is nothing new. In order for autofocus to work, you need contrast, and you need frequent updates. A flat-contrast scene shot at 24p or 30p is going to result in terrible autofocus performance. A high-contrast scene shot in 60i/60p is going to result in the best autofocus performance the system can deliver.
Does the AC160/HPX250/AC130 perform as tack-sharp and quick as the HMC150? No they do not. Sorry, they don't. Hauling around 2x to 3x as much glass, and having to move it to the precision demanded by the higher-resolution chips, has made a system that isn't AS good as the HMC150 was. It still works reasonably well, especially when combined with the area focus, and when used in situations where autofocus can be expected to perform at its best -- high contrast and high frame rate.
What I would be interested to see is a comparison between comparable lens systems from different manufacturers, to see how the AC/HPX lens holds up compared to its competition. Is there any camera directly comparable? The XF300 has the chips to match, but its lens is 18x, not 22x. Sony has a 20x on the NX5U but that's 1-megapixel chips and a 72mm filter diameter so it's not as much glass...
As far as I know Jan said she compared the AC160 against the Canon and Sony equivalents and found that they had comparable issues. I haven't compared them and have no firsthand testimony about that. I googled "NX5U autofocus" and found a few threads:
http://www.dvinfo.net/forum/sony-nxcam-hxr-nx5u/504351-focus-issue-nx5u.html
http://www.dvinfo.net/forum/sony-nxcam-hxr-nx5u/497215-nx5u-major-focus-issue.html
http://www.dvinfo.net/forum/sony-nxcam-hxr-nx5u/495423-nx5u-focusing-still-issue.html
http://forums.creativecow.net/thread/298/1197
Mike Warren
Well-known member
I am quite certain that this is the crux of the matter.
Typically, AC160/130 users are coming up to it from smaller cameras and they naturally expect it to work better in every respect. Conversely, the majority of HPX250 users are coming down to it from much more expensive cameras and not only expect it will not be as good in every respect, but are also used to focusing manually. These people are pleasantly surprised with the sharpness of the images and just accept some of the other shortcomings because it's so much cheaper than the cameras they're used to.
As an electronics and software engineer, I understand some of the compromises Panasonic had to deal with in these cameras. No complex electronic equipment can be made perfect, even when price is not a major factor, but for this sort of product price is very important. The competition is hot and the market not all that large.
Panasonic could have put a larger focus motor in it, which would have helped, but would still not have approached the focus performance of small chip cameras, and it would have added substantially to the cost. It's not just the cost of the motor. It would have flowed through to many other aspects of the design. A larger motor would have required larger gearing and the sliding parts of the focus mechanism would have needed to be redesigned to make sure they don't twist and bind due to the increased force. All this would have made the camera larger and heavier in addition to being more expensive, things that cost sales.
It wouldn't surprise me to learn that the actual performance in the final product didn't meet the theoretical performance. Panasonic would likely only have discovered this very late in the design process. This sort of thing happens all the time in the electronics design world. The options are to make a costly redesign, or to find a way to fix it in the software. For something like a lens assembly on a video camera, a complete redesign would have made the product nonviable. Someone on the design team would have had to make that call and decided that it could be made passable by tweaking the software.
These tests I've seen posted have produced results close to what I would expect, all things considered. There is no doubt to me that the algorithm could be better, but I doubt the processing power is available.
I have been wanting to purchase a camera in this class for some time, but for me video is just a hobby now days and I can afford to be picky. Unfortunately, 1/3" cameras fit right in the price range of what I want to spend, but I really think the 1/3" size is not the best in terms of performance. That's mostly because of optics. The 1/4 cameras have such a low diffraction limit that iris is pretty much useless, so an automatic variable ND filter is a requirement. By the time you get to 3/4" chips there is a bit of a usable aperture range so a manually switched set of NDs makes sense. 1/3" chip cameras have the worst of both these, especially when coupled with long lenses that have to have small maximum aperture to keep the price and size to sensible figures. At full telephoto, wide open, you are already coming close to the diffraction limit.
Mike Warren
Well-known member
Yes. The camera's focus algorithm does not see objects. It sees patches of varying tones and looks for sudden changes from one level to another.
Does someone want to send me a camera for a week? I'll do some proper scientific testing.
Mike Warren
Well-known member
This is the best test I've seen so far and I think can see exactly what's happening.
Firstly, understand that there absolutely must be a delay in the feedback loop of the auto focus or it would continuously hunt.
Now, both the orange and the curtains are low contrast subjects, very difficult subjects for the algorithm to work with. As you pan to the orange, because the whole image is completely out of focus, the camera doesn't know if it's seeing a low contrast subject or an out of focus one. At some point (decided by the programmer) the camera must chose. What is happening here is an edge condition where sometimes the camera makes the decision that it needs to be focused and sometimes that it's just a low contrast scene.
I'm pretty confident I could make that camera auto focus reliably in your same situation by getting a higher contrast edge into the focus zone for long enough to let the camera know something needs to be done.
It's a case of learning how the camera behaves and working with it, and not expecting it to learn your way of doing things.
It's not perfect.
Mike Warren
Well-known member
BTW, I do understand why people are complaining about this. If anyone here is interested in a detailed breakdown of what I understand to be happening on these recent tests, let me know. It will take me a couple of days as I will need to make a bunch of screen grabs.
Barry let me ask you this:
With your extensive detailed use & knowledge and the ins & outs of this camera; exactly how long should it take for the 130/160 to auto focus on an object outside? there must be some time frame.
How massive is this glass? 1/4" 1/2" I assume it's a convex lens.
Did my videos I posted look-- "in the ball park" and just need a little refining? To me not a single thing was close to being in focus for what seemed like 5 very long seconds. As you can see, I was outside with lots and lots of contrast.
Here I shall post one of them again for ease (I have more footage I just have not had a chance to edit them together and upload)
Mike it seems you have all the answers. So did the video I shot outside have too much contrast?
Mike Warren
Well-known member
I'm not Barry, but let me give you my take:
There needs to be a delay to minimize hunting.
The camera is seeing a lot of movement in the frame and is delaying the focus until it sees some stability. After a time it decides that it's not going to get any more stable and performs a focus action.
I think the algorithm could be better to cope with this sort of scenario, but that may be at the expense of seriously degraded performance at other times.
Personally, I'd rather it took too long to focus rather than hunt.
But I am curious about one thing. What was happening just prior to the start of this clip? You are fairly wide and yet quite out of focus.
There needs to be a delay to minimize hunting.
The camera is seeing a lot of movement in the frame and is delaying the focus until it sees some stability. After a time it decides that it's not going to get any more stable and performs a focus action.
I think the algorithm could be better to cope with this sort of scenario, but that may be at the expense of seriously degraded performance at other times.
Personally, I'd rather it took too long to focus rather than hunt.
But I am curious about one thing. What was happening just prior to the start of this clip? You are fairly wide and yet quite out of focus.
Mike, I don't know how much of this thread you have read, so I would mention again that my test video has been investigated by Panasonic here in NZ and they have now passed it on to Panasonic Japan. I am now awaiting feedback from them (we can all speculate until the cows come home but they will ultimately give us the official explanation for this, and if my camera is faulty, will fix it.
by the way, my shots of the curtains covered 50-110 ire and included vertical and horizontal lines of contrast on my WFM on my edit suite, and the orange in the bowl covered 0 to 100,with the orange showing something around 50-100 with specular highlight so actually quite a range of albeit subtle contrasts. For illumination and other exposure data see my earlier post, but in summary, fast enough and light enough to fall well within any limitations of this camera, meeting Barry's suggested best results settings.
Update comment: The orange and curtains were not moving!
Last edited: