kye

Thanks @IronFilm @Kisaha Are you perhaps talking about delivering audio with parts at -0dB vs capture of audio with clipping? My point was that if you clip a very small part of the signal (ie, just the extreme peaks) and you do it on capture (and then process it to be ok in post before delivery) that you're probably ok. Otherwise we'd have huge problems with signal-to-noise while capturing the entire dynamic range of uncompressed audio sources (assuming we didn't have a limiter before the capture device). I'm simply trying to be practical. We tend to recommend that the thing we're recording (a voice perhaps) doesn't go above, say, -12dB, and that we have a safety track at perhaps 20dB less than that again, which would give us 32dB headroom over the peaks in our target audio. If we were recording on-set sound of a narrative in a kitchen, perhaps, and someone puts the saucepan on the stove down a bit harder than they intended then we'd be screwed if our mantra was "no clipping under any circumstances ever". Even more if we're recording in public where transients can be significant. In reality, we accept that the transient clipped, it only did it for a few ms, the associated distortion (THD and IMD in this case) probably wasn't audible, and we just fix that moment in post and move on with our lives. We don't go to the lengths of recording in 24-bit and having a -60dB safety track to ensure that no transient never ever clips I guess in summary, the OP was talking about clipping and audibility, and my point was that you can have clipping without it being audible, and that's probably ok. In terms of your audio credentials, good stuff. One day we can talk about high quality audio reproduction

Agreed. It's the same in the "hifi" equipment that big box retailers push - they're great at the things that are easy to measure and explain to customers, but music is sorely lacking.

The review by Ken Rockwell is pretty amazing - he's not afraid to say what he thinks and doesn't pull punches but that is basically a rave review. Any those images!! What a spectacular looking lens ???? Thanks Robert - that's interesting feedback. I've heard that IBIS and OIS (and gimbals) can sometimes work against each other so it seems like it's a bit of a mine-field in a way. What camera body are you using? Having a combination that works really well would be fantastic for me

Yes and no. Technically, when it clips it clips, so you're right about that. However, if your audio looks like this: and you up the gain by just a tiny bit then it will clip, but only at one point and on only one transient. Will it be audible? Very very very unlikely. Let's understand what is happening when you clip a signal like this: The unclipped signal on the left appears to be a Sin wave, and assuming that it is, it will be a single audio frequency. The clipped signal on the right is approaching a Square wave, which has a base frequency but also has a series of odd harmonics - the difference in tone between an acoustic guitar and the kind of distortion that Metallica might use. So, here's the "quite a bit" part.. If you push a Sine wave into clipping by a very small amount then what you end up with is a base frequency at full volume and a series of odd harmonics that are very quiet in comparison. The harder you clip the signal the louder these odd harmonics will be. This is why a "little bit" of clipping only gives you a "little bit" of distortion and "quite a bit" of clipping gives you "quite a bit" of audible distortion. ie, there is a huge difference between these two: I hope this helps to explain. Audio is hugely complicated and it's easy to make simple statements but as with most things there's usually more to it.

Absolutely! I think I've also got a similar split between different uses - cameras for narrative vs run-and-gun vs hybrid all have different combinations of strengths and weaknesses but I'm looking for a non-standard combination I probably mis-spoke about RAW - I said "approaching RAW" but what I meant was nice 1080p - ie more than Canon currently delivers. Something like the C100 1080 would be fine I'm not waiting. I have lots of cameras presently and they get used My most recent purchase was a 12Tb HDD because I had run out of space on my Backup drive from all the footage I record It's definitely looking like that's the best option. I didn't think about speed boosters - good point. I also got the impression it was more of a MF camera than anything and speed boosters and the vagaries of camera-to-lens communication. When I think of it I imagine people just coming out of film school who have a basic lighting kit, a basic audio setup, and who would have been shooting with a T2i or a 5D back in the day but now can get a Pocket 2 and stop dreaming about owning a Red and just make spectacular films. Yes, there will be people who put a single prime on it and hand-hold it throughout their holiday and create magic, but I don't think it's going to be replacing the A6500 in that many setups. I have half-a-mind to teach myself to manually focus and then re-evaluate things by taking Autofocus off my list of requirements, but it adds another thing that can go wrong to the list. LOL about 400mm - you're right about that not being stealthy!! Of course, you're talking to someone who yesterday took my 700D with EF-S 55-250 lens to my kids football game, and using crop mode on ML was shooting at 1200mm equivalent, hand-held! Of course, that lens is almost a whole universe away from a white lens a foot and a half long hanging off the end of a camera!!

What did you find? I have a vague memory of reading something about the 24-70 being better, but maybe I'm not remembering correctly. 24-105 is a very flexible range, especially with the crop/zoom functions also available.

Thanks for mentioning those lenses, I hadn't come across them when I went through DPReview and B&H looking for stabilised fast zooms - I'll have to do a bit of research on them @andrgl @SR I know it looks like it's a utopian fantasy, but you'd be surprised. The XC10 and my iPhone8 each have 7 out of the 9 criteria, a 70D / ML / 18-35 combo would have 7 or 8, an A7III / 24-70 looks like it gets 8 and maybe 9 if the overheating isn't too bad, the GH5 gets 8 but misses AF... If Canon has a genuine attempt at good video IQ on their FF mirrorless camera then it might hit all of the criteria. If this list actually looks exhaustive to you then I'd encourage you to write your own list and compare - most people aren't aware of all their criteria.

This link (https://petapixel.com/2016/10/22/iphone-7s-camera-parts-cost-26-9-5-phone/) says that the iPhone 7 camera cost about 10% of the cost of the phone, and this teardown of the iPhone X (http://www.techinsights.com/about-techinsights/overview/blog/apple-iphone-x-teardown/) says that the camera / image parts are about 11% of the cost of the phone. I'm not sure what that means considering that the iPhone 7 has two cameras and the iPhone X has three plus some other bits and pieces like the DOT projector but the iPhone X isn't that much more in terms of a percentage, but if we pessimistically assume that things will scale, it wouldn't be a far stretch to put a fourth camera with a longer lens in one, giving ~25mm, ~50mm, and maybe 100mm (or 150mm) equivalents. The one common application that's missing is the 'take pictures at my kids sports game', so although as film-makers we're picky about often obscure things, having a third longer lens is something that would be very useful on a weekly basis to most people on the planet. It would probably mean they'd have to turn the camera sideways and put a 45 degree mirror in there to get the length of the lens in, and it might require a bigger assembly to incorporate the IS requirements of a longer lens in there too, but longer focal lengths are typically used in better lighting conditions, meaning that the lens could be slower and the sensor smaller, so there's some ability to compromise there I think.

Currently the Canon APS-C cameras are probably eliminated due to no IS in wider aperture lenses, the Pocket 2 lacks AF-C, the GH5 can't focus well enough, the XH-1 battery life and AF issues, the C100 and Canon FF I believe only have AF points in the middle of the sensor. The remaining candidates are the A7III, but I'm waiting until the magical Canon FF camera and all the other 'just in time for xmas' cameras are revealed before moving forwards. The Pocket 4K looks like it will be a spectacular camera and I really wanted to want it, but it's just not aimed at me. If you're shooting in a situation where you can pause long enough to change lenses or you can tailor your shots to one focal length (eg, like street photographers often do with a single prime) then it would be just wonderful, but I want cameras to fit in to my style of shooting not the other way around. I shoot my home videos sometimes without even stopping walking etc, so things are often in motion because that's how life is and that's what I'm capturing. TBH it would be great if I could end up with a camera where lots of people use it and there's lots of support in terms of products and online discussions etc. I do have half-a-mind to pick up a second-hand Pocket 1 (are we calling it that now?) and get a single prime, like maybe a 35mm equivalent, and kind of have it as an 'art' setup to compliment whatever I get for my 'documenting' setup, but without IS, good low-light, zoom, or AF, it would only ever be an experimental camera. The Pocket 2 will suffer from the same compromise that the GH5 and all other cropped sensor cameras suffer from, you can get zooms and you can get shallow DOF but not both. The fastest zooms you get are F2.8 but that isn't as shallow as f2.8 on FF, so you're forced to choose between fast and flexible.

Currently I have two setups - 700D running ML and Sigma 18-35 f1.8, and XC10. The 700D takes lovely stills and nice shallow DOF but lacks image stabilisation and reliability, and the XC10 has great IQ, reliability, better low light, etc but the AF isn't great and it lacks shallow DOF. The XC10 is almost perfect but just doesn't give the look I've now come to realise that I want. My list of criteria is: Shallow DOF (equivalent to FF at F4) Looks nice at ISO 6400+ Walk-around lens available (at least 24-70) Reliable auto-focus with face detection Codec giving approaching-RAW 1080 or 4K Stabilisation good enough for hand-held work Overheating not an issue Low RS Reliable

Have you tried a different SD card? It looks like it might be a video file on the edge of being corrupt. Other things to try: unplug any external accessories that are attached if you have access to a second setup, try different lenses, different batteries, etc re-load firmware for camera, lenses, and batteries (not sure if these batteries have firmware but some do) Good luck!

Geoff is completely wrong. The numbers are absolutely critical. We have far too much dynamic range, resolution, reliability... and even colour!! I did an experiment on Instagram for a while of trying to capture the most evocative images, and luckily if you're suffering with too much image quality then don't worry - there's an app for that!! This is one of my most liked photos from the campaign, using an app called Tintype to simulate those glory days where images had soul and nostalgia! Sometimes I had to use two apps to also blur the middle of the frame as tintype didn't do that. IIRC tintype also had a shooting feature where you were limited to taking a photo every few seconds, it would apply the filters in random ways with random amounts, to simulate the unreliability of early processes. It was far from an exhaustive project but I did discover that the lower the image quality the more feeling the images seemed to have. I'm sure there are exceptions in this and I'm sure there's a sweet spot too. But, luckily for us the tools we have available in post are wonderful and can save your footage from "the prison of the pristine" (to paraphrase morpheus). (in the spirit of AND thinking, this post is completely serious, as well as being complete satire...)

Thanks for sharing. Out of curiosity, why the 16-35 over the 24-70? I thought your style might have been more aligned with the longer focal lengths (but I could be wrong). There's always the combination of crop mode and clear-image zoom to take the 35 to 79mm equivalent, seemingly without much loss in quality. I was assuming I would get the 24-70 but then realised that a wider lens might be useful so now am torn.

I used to write electronic music and my experience was that you can clip quite a bit before it becomes audible, but that doesn't mean that you should. I second what @IronFilm says above of using a conservative gain value and using a safety track - that's how I try to record my audio. You'll find that things will seem quiet, but that's why you use compressors and limiters in post, which gives you much greater control over how things end up in the final mix.

This video talks a little about settings, but he didn't seem to change them that much, so I'm not sure if it will help: The reason I mention this YT video is that he shot a wedding with this camera and used AF on a gimbal with face detection and was really impressed, so he should know some stuff.. This is a snippet of that AF:

Actually, there are lots of excellent robotic solutions to the problem of mobility. This from 2006 (or earlier): Which evolved into a prototype that could wrap around things and then climb vertically, and then also evolved to have multiple legs etc. https://techcrunch.com/2017/07/06/how-carnegie-mellons-snake-robot-became-the-multi-legged-snake-monster/ this from 2010 - a robot that balances on a ball: and other approaches too: http://www.sciencemag.org/news/2016/01/new-space-robot-would-hop-not-drive-across-other-worlds but this is the problem - we don't want robots to solve our problems, we want them to be like us, to be familiar. Which is why it's difficult to program them to do human things, like simulate a larger aperture in recorded video from multiple cameras, which is optimising an aesthetic experience that they never have and won't ever have unless we program them to have it.

Wow, that seemed a bit hit and miss @jonpais. I wonder if there's a setting somewhere in there that's derailing the process? Nice images though, the overall look and bokeh seems nice

I'd imagine that the Xume would attach to a magnet pretty well, so if you were to attach a couple of magnets somewhere then you could pull the filter off the lens and stick it on the magnets and install the second one pretty quickly. Of you could just have a kind of holster on a belt or strap somewhere handy. They're very simple to operate, if a little pricey as @Phil A correctly mentioned.

OMG.. please no! CAN does not mean SHOULD!!!!

In case anyone is interested, here's a good argument for and against Canon making an amazing mirrorless camera.. As usual there's lots of facts included.

Just shoot 4K RAW on the H6D-100C. Completely silent, very wide, large sensor for a nice cinematic look, and 8MP RAW stills are fine.

Maybe they go upmarket and have the A9S, include all the fancy features, give a bigger offering from the A7SII, and because it's A9 series they can charge more for it? In a way having the A7III is kind of a duplicate for the A9 but at a lower level - so an A9S above the A7SII makes sense.

You underestimate our desire for self-preservation!!

It depends on your frame of reference. For me, 'can do everything' includes wild-life one second and a wide panorama the next without time to change lenses, and coming from a world that thinks that a 700D is a 'big camera' a 5D or C100 are so absurdly large that people openly stare with mouths agape. Also, I'd like image quality approaching 4K compressed or 1080 RAW. We're in a difficult time - cameras have enough capability and accessibility to give us grand ideas, media consumption is high enough quality that we know nice images when we see them, but the tech isn't quite there yet. I know BM is concentrating on the slow/planned styles of film-making, but they're so insanely good at delivering awesomeness I'm just jealous I'm not in their target camera market!!

You're right about thinking at a human level. Robots building robots is definitely going to be a thing, and just like any other exponential process it's going to take far more time making basically zero progress than we think, and then bang - out of nowhere it will take off. Another problem is that robots are so different that getting smarter doesn't mean anything to us if they get smart in ways we don't appreciate or like. You might have heard about the Facebook AI robots inventing their own language. They did that because the programmers didn't incentivise them to speak proper English, so they went off the reservation. That is a pretty straight-forward example, but what separates computers inventing their own shorthand (which we don't like) and teenagers doing the same thing (which is inherently human), let alone how we would respond to an AI that made no sense because it was speaking to teenagers! There's a great quote from the snopes article about that language thing: Link: https://www.snopes.com/fact-check/facebook-ai-developed-own-language/ It's an interesting article talking about lots of general stuff to do with AI etc, but the quote is interesting because we have stupidly high standards about doing things that are useful to humans. In a sense, if AIs were training humans they might say something like: I don't understand these humans, they can't accurately remember a f*cking thing, can't even add a thousand numbers a second (which any $0.02 chip can do!), but they manage to self-repair, resist rust, and develop virus counter-measures without even trying!! They will be capable of amazing things, but if you ask a robot to help enhance the photographs you took you're more likely to get back an image that has been optimised so that every n-th pixel has its colour values rounded to the n-th prime number, or simply made to be all black because it's easier to store. The idea they would simulate a wider aperture in an optical system in a 3D environment is far more alien to them than the other examples I mention because they don't have these things, or if they did, then they'd optimise them to have a deep depth-of-field because that's better for object recognition algorithms and accurate recording and cataloging of the world is more valuable than walking around half-blind or recording huge amounts of data to then throw away a lot of it afterwards.