Wait a minute... GX85 is only 3-axis IBIS??

[oldracer]

i get what youre saying but here is my own explanation based on the discovery channel feature ... the cipa tests i saw (wobbling the camera, moving from side to side in fast jerky fashion, etc), do not account for this micro movements hence the seeming discrepancy.

I dunno. Here you are basically questioning whether the CIPA waveform is representative of real-world shooting. That could be, and while they do not describe their data acquisition protocols and do not provide detailed results, the comprehensiveness and detail in the spec leads me to believe that they would not base this whole complicated and expensive spec writing effort on sloppy or sparse camera motion data. So, again, I see no reason to buy your theory.

Re handling subject motion, that is another can of worms that probably can never be quantified because subject motion itself cannot be standardized. (Another issue here, somewhat peripheral to the multi-axis discussion, it that it is possible for manufacturers to "game" the CIPA spec in the same way that VW gamed the emissions testing. Since the CIPA waveforms and test setups are known, it is probably possible for a manufacturer to tune an algorithm set to suit. He could even go so far as to match VW, identifying a CIPA test run based on accelerometer data and selecting the tuned algorithms for use during the testing.)

 

[travelbug]

I dunno. Here you are basically questioning whether the CIPA waveform is representative of real-world shooting. That could be, and while they do not describe their data acquisition protocols and do not provide detailed results, the comprehensiveness and detail in the spec leads me to believe that they would not base this whole complicated and expensive spec writing effort on sloppy or sparse camera motion data. So, again, I see no reason to buy your theory.

Re handling subject motion, that is another can of worms that probably can never be quantified because subject motion itself cannot be standardized. (Another issue here, somewhat peripheral to the multi-axis discussion, it that it is possible for manufacturers to "game" the CIPA spec in the same way that VW gamed the emissions testing. Since the CIPA waveforms and test setups are known, it is probably possible for a manufacturer to tune an algorithm set to suit. He could even go so far as to match VW, identifying a CIPA test run based on accelerometer data and selecting the tuned algorithms for use during the testing.)

like i said, ive seen the tests featured in an episode of discovery. but it seems to me that youre more concerned with proving yourself right rather than finding genuine information on the subject. so ill just leave it at that.

 

[Turbofrog]

oldracer, I just don't get it. You're willing to buy into the absolute intellectual authority of CIPA (an organization primarily intended to facilitate business and collect statistics) and their infallible status as research scientists, but you're implicitly refuting the credibility or value of companies like Olympus which literally spend $100 million USD every year on R&D in their Imaging Business alone, and produce hundreds of patents. Panasonic has their hands in a lot of products, of course, but they spend $4.4 billion USD in R&D. You don't think these companies can accurately characterize the forces and moments on a camera, but CIPA can?

 

[tkbslc]

I'm sorry if I started this argument. I personally do not worry about 3 vs 5 axis IBIS, but I'm not going to claim it does nothing. I just don't really need anything beyond 2-3 stops for my own shooting, which is just enough to help skirt that 1/FL rule for those small windows between good light and no light.

I will argue that there are very limited scenarios where 5-6 stop stabilization is really more useful than 2-3 stop. So that's kind of what I was trying to say. If it has some stabilization and it works in video, I'm good. The number of stops is irrelevant to me as long as it isn't worse than the GX7 is now.

 

[Turbofrog]

I'm sorry if I started this argument. I personally do not worry about 3 vs 5 axis IBIS, but I'm not going to claim it does nothing. I just don't really need anything beyond 2-3 stops for my own shooting, which is just enough to help skirt that 1/FL rule for those small windows between good light and no light.

I will argue that there are very limited scenarios where 5-6 stop stabilization is really more useful than 2-3 stop. So that's kind of what I was trying to say. If it has some stabilization and it works in video, I'm good. The number of stops is irrelevant to me as long as it isn't worse than the GX7 is now.

Don't worry, you definitely didn't start this argument.

 

[oldracer]

oldracer, I just don't get it. ... spend $100 million USD every year on R&D ... You don't think these companies can accurately characterize the forces and moments on a camera, but CIPA can?

I never said that. Certainly they can. What I said was that there is no objective, public, data that calls the CIPA results into question.

Oly, Panny, Pentax, and Kellogg's corn flakes all spend a lot on R&D, but that says nothing about what they are doing. It is entirely possible that the first company's introduction of multi-axis stabilization (Oly, right?) was simply a marketing decision, designed to get a leg up on the competition. It certainly sounds wonderful. That alone probably sold a lot of camera for them.

It would have taken R&D money to implement that decision but spending the money is no proof that the technology solved a real world problem. That proof would come from test data, which is currently (if it exists) not public. Then the competition had to respond, spending more R&D money. Together the manufacturers have now convinced most of the market that multi-axis stabilization is important, as evidenced by the post that started this thread. But there is no public data to support the herd's belief.

Hans Christian Andersen : The Emperor's New Clothes

As an OT example, one of Mazda's big R&D investments in developing the Miata/MX-5 was in getting exactly the right exhaust note. They recorded many sports car exhaust notes and did spectral analysis as they developed the final car's exhaust system. All of this cost a lot of money --- R&D money. But it was R&D money spent to support a marketing goal, not money spent for technical reasons. Along the same line, I think I remember reading an article about Oly spending R&D money to get exactly the right shutter click sound. And Oly has certainly spent a ton of R&D money on that goofy "Air 01" thing. Do you think they have a ton of data that says that the gadget solves some real-world problem? Of course they don't. They did it because they wanted to be first to market in a new category of camera, hoping that it would take off.

You mentioned patents. I hold several patents and hence know a little bit about patent law. It is not necessary to show that there is a real-world problem before one can file a patent to solve the asserted problem. Said another way, I could file and get a patent for a device to defend against asteroids circling the earth at an altitude of 5 miles, despite the fact such an orbit is physically impossible (due to kinetic energy loss/heating from atmospheric friction). So I am sure that there are several patents for multi-axis stabilization systems, but not even that is proof of a technical need.

 

[TonyG]

I never said that. Certainly they can. What I said was that there is no objective, public, data that calls the CIPA results into question.

Oly, Panny, Pentax, and Kellogg's corn flakes all spend a lot on R&D, but that says nothing about what they are doing. It is entirely possible that the first company's introduction of multi-axis stabilization (Oly, right?) was simply a marketing decision, designed to get a leg up on the competition. It certainly sounds wonderful. That alone probably sold a lot of camera for them.

It would have taken R&D money to implement that decision but spending the money is no proof that the technology solved a real world problem. That proof would come from test data, which is currently (if it exists) not public. Then the competition had to respond, spending more R&D money. Together the manufacturers have now convinced most of the market that multi-axis stabilization is important, as evidenced by the post that started this thread. But there is no public data to support the herd's belief.

Hans Christian Andersen : The Emperor's New Clothes

As an OT example, one of Mazda's big R&D investments in developing the Miata/MX-5 was in getting exactly the right exhaust note. They recorded many sports car exhaust notes and did spectral analysis as they developed the final car's exhaust system. All of this cost a lot of money --- R&D money. But it was R&D money spent to support a marketing goal, not money spent for technical reasons. Along the same line, I think I remember reading an article about Oly spending R&D money to get exactly the right shutter click sound. And Oly has certainly spent a ton of R&D money on that goofy "Air 01" thing. Do you think they have a ton of data that says that the gadget solves some real-world problem? Of course they don't. They did it because they wanted to be first to market in a new category of camera, hoping that it would take off.

You mentioned patents. I hold several patents and hence know a little bit about patent law. It is not necessary to show that there is a real-world problem before one can file a patent to solve the asserted problem. Said another way, I could file and get a patent for a device to defend against asteroids circling the earth at an altitude of 5 miles, despite the fact such an orbit is physically impossible (due to kinetic energy loss/heating from atmospheric friction). So I am sure that there are several patents for multi-axis stabilization systems, but not even that is proof of a technical need.

Numerous experienced users and reputable review sites tell us that IBIS makes a noticible difference. I believe them. I notice a difference with the primtive 2 axis IBIS in my Pen mini. Having 5 stops of image stabilization would be wonderful for maco work.

 

[piggsy]

I think at some point it would be useful to go back and unpack some assumptions ? What is the argument exactly ?

CIPA: pitch and yaw are the two dominant forces at 20x focal length

This really tells us nothing much - certainly not that 2 axis stabilisation is all that is required or that there is no benefit from 5 axis stabilisation. I mean - just at the basic gut test - you would assume that the number of axis is just one element in how a stabilisation system responds (eg sample frequency of movement detection, maximum momentum it can compensate for on the sensor in x amount of time, etc)

IIRC the on-sensor stabilisation on dual IS mode on the GX8 at least reverts to correcting for vertical/lateral/roll anyhow? I think the A7-2 works pretty similarly to this as well.

Panasonic GX8 Review - GX8 Overview

Subscribe to see EXIF info for this image (if available)

 

[Turbofrog]

I never said that. Certainly they can. What I said was that there is no objective, public, data that calls the CIPA results into question.

Oly, Panny, Pentax, and Kellogg's corn flakes all spend a lot on R&D, but that says nothing about what they are doing. It is entirely possible that the first company's introduction of multi-axis stabilization (Oly, right?) was simply a marketing decision, designed to get a leg up on the competition. It certainly sounds wonderful. That alone probably sold a lot of camera for them.

It would have taken R&D money to implement that decision but spending the money is no proof that the technology solved a real world problem. That proof would come from test data, which is currently (if it exists) not public. Then the competition had to respond, spending more R&D money. Together the manufacturers have now convinced most of the market that multi-axis stabilization is important, as evidenced by the post that started this thread. But there is no public data to support the herd's belief.

So despite CIPA's testing that confirms that cameras that use the 5-axis systems happen to perform better than 2- or 3-axis systems, "there's no public data to support the herd's belief." Just total coincidence, and the E-M5 II would still have 5 stops of CIPA effectiveness, the A7r II would still have 4.5 stops, and the GX85 would still have 4 stops even if they ripped the other axes out of their system. Except that when they did do that for the E-PL7 and E-M10 (no X, Y shift), the CIPA effectiveness went down to 3.5 stops...

 

[oldracer]

So despite CIPA's testing that confirms that cameras that use the 5-axis systems happen to perform better than 2- or 3-axis systems, "there's no public data to support the herd's belief." Just total coincidence, and the E-M5 II would still have 5 stops of CIPA effectiveness, the A7r II would still have 4.5 stops, and the GX85 would still have 4 stops even if they ripped the other axes out of their system. Except that when they did do that for the E-PL7 and E-M10 (no X, Y shift), the CIPA effectiveness went down to 3.5 stops...

Wow. Have you even read the spec. ?!!?

Per the CIPA spec the test waveforms are limited to pitch and roll movements. So if a system that claims to compensate for other movements is tested correctly using the CIPA protocol, those extra compensation modes are not exercised. If said system performs better, it can only be because it performs better in the two axes being tested. And yes, the correct conclusion is that your example systems would post the same test results if the other axes were "ripped out" -- because the effect of compensation in those other axes is not being tested.

Weighing a salami tells you nothing about its diameter or its flavor.

As @piggsy correctly points out, "just at the basic gut test - you would assume that the number of axis is just one element in how a stabilisation system responds ... " It is not just possible, it is likely that newer camera systems have improved performance in the two axes that CIPA tests. Optimizing the camera for the tests, even to the point of cheating, is an absolutely predictable marketing response. So, it's not "total coincidence" at all. It's the expected result.

 

[tkbslc]

You guys realize you are arguing the hypothetical at this point, right?

 

[Ross the fiddler]

Actually, I am being an empiricist, looking at the factual data. It is you who are being a theorist, hypothesizing about what might be and defending your hypotheses without presenting any supporting data.

You might be right, as I have said. But we don't know. The data we do have point to stabilization in more than the pitch and yaw axes being unnecessary. I don't know why that is so upsetting to you.

As an engineer I am surprised you can't see the benefits of sensing all 5 axis' for a correlated compensation in the sensor, but then maybe you are like some classical musicians I know that can only play from music & cannot improvise & play by ear as well.

I think you need to look past what can be measured on the bench & accept what you see in from actual real life applications. As mentioned above, IBIS can only compensate for a certain amount of movement & that is why Olympus decided to add optical stabilisation to their 300mm f4 lens using the Sync IS system (dual syncing of IBIS & optical) to cater for the larger movements.

 

[gr6825]

Wow - classical music burn! This thread is getting good. Oldracer, can't you do math?? 5 is better than 2 bro

 

[Halaking]

We need a new topic! 5 axis is so old...

 

[oldracer]

... Oldracer, can't you do math?? 5 is better than 2 bro

Actually, in project management lingo that is referred to as "creeping elegance." After the project is planned and kicked off, someone successfully argues "Look, there are all kinds of 'benefits of sensing all 5 axis' for a correlated compensation' It's intuitive! Let's do it!" without regard to project requirements or real-world needs. Or maybe the argument is "We can sell another ten thousand widgets if we offer the 'benefits of sensing all 5 axis' for a correlated compensation.' Slam dunk and it will take the competition two years to catch up! Let's do it!"

The existence of specific features in a design is not proof that that was any technical need or technical justification for them.

 

[piggsy]

As an engineer I am surprised you can't see the benefits of sensing all 5 axis' for a correlated compensation in the sensor, but then maybe you are like some classical musicians I know that can only play from music & cannot improvise & play by ear as well.

I think you need to look past what can be measured on the bench & accept what you see in from actual real life applications. As mentioned above, IBIS can only compensate for a certain amount of movement & that is why Olympus decided to add optical stabilisation to their 300mm f4 lens using the Sync IS system (dual syncing of IBIS & optical) to cater for the larger movements.

I think it comes down to what's being captured and what's being measured. The CIPA terminology seems very carefully chosen and is very specifically about a certain focal length. Pitch and yaw motions when just a milliradian's difference can result in masses of pixel shift during the exposure are pretty important (what is it, 10cm at 100m?). You could also make the case for lateral and vertical motions being very important for run and gun video to filter out big up/down/side to side impulses from the camera being at the end of an arm used to balance and to filter footsteps, or to hold sub-mm perfect framing on macro, or roll being important to compensate for rotational motion from hitting the shutter at the edge of the camera, or whatever. All of those things are going to be more or less noticeable depending on what's being shot at what magnification and at what length.

I don't think the CIPA statement makes any particular claim about any technology or operation of any device other than their own test and test conditions.

 

[tkbslc]

Schrodinger's IBIS - it is both an ineffective marketing hoax and fabulously effective until such time as one opens up their laptop and observes the images on your computer.

 

[peppermonkey]

What in the world is everyone arguing about? Look, I could care less about 3 or 5 axis stabilization... But what I do know is that Oly's IBIS in video, regardless of 5 or 3 or two billion axis stabilization, is better than Panasonic's (though gx80/85 may contend). If I needed stabilisation in video, and don't want a rig, I'll get an Oly or possibly the new GX80/85. Whether the 5 axis is better than 2 or 3 axis is besides the point since whatever stabilization in the 5 axis IBIS is just better than what is in the 2 axis IBIS.

Sent from my E5823 using Tapatalk

 

[agentlossing]

That's pretty much where I'm at too. All indications in the real world have pointed to 5-axis being the better image stabilization, at least as implemented by Olympus. Is that just correlation not causation because and Olympus just has better hardware? Maybe. But it gives this new GX85 a fighting chance at least speculatively.

 

[Steven Norquist]

Put a non-stabilized lens on the GX8 and you will see that it has sub-par stabilization compared to Sony and Olympus cameras.
Only with a stabilized lens and the in body stabilization can the GX8 approach the in body only stabilization of Olympus.
And it still is not as good.
I like the GX8 but I wont give up the superior Olympus stabilization for it.
I will never buy a non-stabilized camera again.