basspig
Veteran
My perspective comes from the context of having used the PMW-EX1/3 cameras for several years, so bear that in mind as you read my comments.
The first few hours with this camera were a panic frenzy to try to understand the controls enough to get meaningful output. Even having spent days in advance of the purchase, studying the relatively useless manual (it really doesn't tell you WHAT a control or feature does, just "turn on such in such to activate the blah blah feature") we all know about that. Largely, the manual is a list of features. But.. it's also out of date with the new firmware and needs to be updated to reflect the current feature set.
As many of you (painfully) know, I have stringent audio requirements for my cameras. Every camera must pass a proof of performance before being put into service. I tested audio two ways. First, I bench-tested the camera, using a Tektronix SG-5010 Programmable Oscillator, with a balanced output, driving the input of the FS7's mic channels. I monitored the headphone output on a Tektronix 2465B oscilloscope and Tektronix AA-501 Distortion Analyzer. I discovered that the headphone output clips at less than 2 volts peak to peak, which is about 775mV RMS. High efficiency cans are a must.
On the sweep test, I set the Distortion Analyzer to read relative levels and set '0' at 1KHz. I have the oscillator programmed to sweep 20-20,000Hz and triggered the sweep. I noted that at 20Hz, the mic preamps through the headphone jack were down 7.4dB. A 1KHz square wave had quite a bit of tilt, albeit very little overshoot. The DAC and decimation filters indicate high quality performance. At this point, I suspected that the headphone amp was responsible for the bass rolloff. So, on to the RMAA test. Before I go further, I will relate my findings about gain-staging. I wanted to find out where the overload level is for this preamp. I set the Ref level to -40dB and ramped up the input signal from -40dBm, slowly, finding along the way that the headphone level had to be reduced several times as it was clipping and not the mic preamp. I got all the way to -3.95dBm until the mic preamp clipped. This is respectable performance. A condenser mic would have to be in front of something very loud to approach that level, and it probably would clip internally anyway. No worries about handling loud signals. Oh, and the dial setting for that level is what I would describe as '3.1', or a sliver past '3' on the dials. Below that, and you're not lowering the noise floor any further, but you are limiting the top end headroom. In other words, this is the optimum level (compare that the 2000 ISO being the optimum level for S-Log 3 where you get the full 14 F-stops).
In the studio, I connected the output of a MOTU 896 Mk III to the mic inputs of the FS7 and played the reference tones and then the RMAA test signal and recorded it. Downloaded to Premiere, exported a wav file and analyzed the result with RMAA. So the mic pre is the source of the rolloff, not the headphone amp.
To be sure, I also checked the Line Input mode. That was as flat as any decent studio DAC. More interesting, is the s/n and dynamic range are as good as many studio DACs. As such, a corrected EQ in post to lift 7dB @20Hz could work, as the noise floor is way down in the absurd range and some boost won't cause any electrical rumble. I'm pondering building a passive network to attenuate everything above 20Hz with a knee at 200Hz, by 7dB, so that I don't have to do this in post every time. Another possibility is to buy a phantom powered preamp that can mount on the camera. Another item hanging off this already immense camera. Okay, enough on audio.
Ergonomics.. the hand grip fits nicely, but as a shoulder cam, it's awkward. This camera will need to be rigged to make it comfortable and balanced. Needs a quick release plate and a shoulder pad. I have some articles by Allistair Chapman with some proven rigs that I'll put on my shopping list.
Operation: First day, I felt like a fish out of water. I don't know where ANYTHING is! Fumbling is the operative word. Good thing I didn't have a gig to do within the first week. This camera has a steep learning curve. There are some familiar things that EX users will recognize, but there are things that aren't the same--no picture profiles to be found. It appears that one saves and loads looks from SD card. But I have a lot of testing to do to find out just WHAT gets saved to the SD card.
Of course the big deal is S-Log. S-Log 2 is clean as a whistle and relatively easy to grade. S-Log 3 can bring out some noise if the exposure isn't through the roof (looks completely blown out in the VF), so the uninitiated will under expose and screw up the first few shots. I was surprised that some footage I shot today that looked like throw away, looked fine in post after applying a reverse curve.
Light sensitivity: A word: wow! Even with the F4 28-135 Sony G lens. Way more light sensitivity than the EX3 with a F 1.9 lens. I can't imagine an F.95 lens on this sensor. ND64 is barely enough with F8 aperture in daylight. Filming in low light, the details and the bokeh are both creamy and noise-free. Convenient feature over the EX line: shutter speed may be completely controlled without switching shutter to a special mode. If you're filming by moonlight, and you need more light without increasing the gain, it's easy to dial down the shutter speed (works in speed, not degrees mode) into subdivisions of the frame rate.
The 28-135 Lens: Okay, it IS a limited range, and is still tunnel vision on the 'wide' end. It's a slow lens being F4, too. Differential focus will not be as good as a F1.4 lens, obviously, nor will the light gathering (but light is no problem with this sensor! F11 indoors at night under LED lighting!) Yes, I'd like a wider lens. And one with more zoom. But this was okay on the outdoor soccer field and was perfect for filming kite flying.
.
Want to know what's GOOD about this lens? The image quality. It's simply the BEST lens I own at this time. This is the first lens that doesn't have any visible chromatic aberration. And it's sharp from corner to corner. The ultimate test for CA is filming up through the trees to overcast sky. Other lenses are horrid in this regard, with 2-3 pixels of red/green fringing. Not this lens. I had to look again and again. It's not there. The images are in a different league from the Fujinon lenses, which themselves are very good.
S&Q: As my Premiere CS6 won't do XAVC, I fortunately have a Atomos Shogun that writes a compatible file. I mostly record on the Shogun off the HDMI port. For S&Q, I record the high frame rate to XAVC and then play it back at 24FPS to the Shogun, where it gets recorded and converted to ProRes. That works. The YT clip above has insets of video at 180FPS played at 23.976. From what I can tell, the image quality is very good.
There are numerous mysteries yet to uncover. One is how to get the scopes to display. An article I was reading tonight said that they only work when LUTs are active, but I don't know which LUTs or where they need to be applied to get past 'cannot proceed' when the scopes attempt to be accessed. Another mystery is what replaced Picture Profiles, and what type of file stores that kind of information to the SD cards.
I need to get some wide angle lenses for indoor shooting. I'm looking at Rokinon lenses from 10 to 14mm. Many of excellent reviews by their owners. I'll also look at some cine primes. And I really could use a zoom that goes wider, but is as clean and free of distortions as this Sony G lens. What an amazing lens.
Rails, matte box, quick release, shoulder rig.. all necessary items. As configured from stock, there's no proper way to set this thing down on a flat surface. And that grip handle has to be removed and re mounted in the opposite direction to slide the cam onto a tripod head. These things can be fixed with accessories, so I'm not too concerned.
My decision to buy a Shogun recorder last March proved to be the right one, as it's been the savior that lets me get 2160P into the editor. NVENC plugin and nVidia GPU with 1536 CUDA cores makes it possible to work with the footage and to render out 2160P videos for Youtube in reasonable time. I can't imagine doing the renders in software.
In conclusion, I think the FS7 is a keeper. The picture is amazing. Especially if the scene is halfway decently exposed. The images on my DCI 4K monitor look like you can just reach through the bezel and touch them. The dynamic range is such that I find myself turning the monitor backlight all the way up to 100, where the forest floor is nicely lit, and when the camera pans up to the sky, you have to squint, it's so bright. Yup, the way to enjoy some of those 14 F-stops is to turn the backlight all the way up, so the darker parts of the scene come out of the black. It's amazingly clean, still. And I'm sure the shooting can be optimized way further than my early experimental shots.
The first few hours with this camera were a panic frenzy to try to understand the controls enough to get meaningful output. Even having spent days in advance of the purchase, studying the relatively useless manual (it really doesn't tell you WHAT a control or feature does, just "turn on such in such to activate the blah blah feature") we all know about that. Largely, the manual is a list of features. But.. it's also out of date with the new firmware and needs to be updated to reflect the current feature set.
As many of you (painfully) know, I have stringent audio requirements for my cameras. Every camera must pass a proof of performance before being put into service. I tested audio two ways. First, I bench-tested the camera, using a Tektronix SG-5010 Programmable Oscillator, with a balanced output, driving the input of the FS7's mic channels. I monitored the headphone output on a Tektronix 2465B oscilloscope and Tektronix AA-501 Distortion Analyzer. I discovered that the headphone output clips at less than 2 volts peak to peak, which is about 775mV RMS. High efficiency cans are a must.
On the sweep test, I set the Distortion Analyzer to read relative levels and set '0' at 1KHz. I have the oscillator programmed to sweep 20-20,000Hz and triggered the sweep. I noted that at 20Hz, the mic preamps through the headphone jack were down 7.4dB. A 1KHz square wave had quite a bit of tilt, albeit very little overshoot. The DAC and decimation filters indicate high quality performance. At this point, I suspected that the headphone amp was responsible for the bass rolloff. So, on to the RMAA test. Before I go further, I will relate my findings about gain-staging. I wanted to find out where the overload level is for this preamp. I set the Ref level to -40dB and ramped up the input signal from -40dBm, slowly, finding along the way that the headphone level had to be reduced several times as it was clipping and not the mic preamp. I got all the way to -3.95dBm until the mic preamp clipped. This is respectable performance. A condenser mic would have to be in front of something very loud to approach that level, and it probably would clip internally anyway. No worries about handling loud signals. Oh, and the dial setting for that level is what I would describe as '3.1', or a sliver past '3' on the dials. Below that, and you're not lowering the noise floor any further, but you are limiting the top end headroom. In other words, this is the optimum level (compare that the 2000 ISO being the optimum level for S-Log 3 where you get the full 14 F-stops).
In the studio, I connected the output of a MOTU 896 Mk III to the mic inputs of the FS7 and played the reference tones and then the RMAA test signal and recorded it. Downloaded to Premiere, exported a wav file and analyzed the result with RMAA. So the mic pre is the source of the rolloff, not the headphone amp.
To be sure, I also checked the Line Input mode. That was as flat as any decent studio DAC. More interesting, is the s/n and dynamic range are as good as many studio DACs. As such, a corrected EQ in post to lift 7dB @20Hz could work, as the noise floor is way down in the absurd range and some boost won't cause any electrical rumble. I'm pondering building a passive network to attenuate everything above 20Hz with a knee at 200Hz, by 7dB, so that I don't have to do this in post every time. Another possibility is to buy a phantom powered preamp that can mount on the camera. Another item hanging off this already immense camera. Okay, enough on audio.
Ergonomics.. the hand grip fits nicely, but as a shoulder cam, it's awkward. This camera will need to be rigged to make it comfortable and balanced. Needs a quick release plate and a shoulder pad. I have some articles by Allistair Chapman with some proven rigs that I'll put on my shopping list.
Operation: First day, I felt like a fish out of water. I don't know where ANYTHING is! Fumbling is the operative word. Good thing I didn't have a gig to do within the first week. This camera has a steep learning curve. There are some familiar things that EX users will recognize, but there are things that aren't the same--no picture profiles to be found. It appears that one saves and loads looks from SD card. But I have a lot of testing to do to find out just WHAT gets saved to the SD card.
Of course the big deal is S-Log. S-Log 2 is clean as a whistle and relatively easy to grade. S-Log 3 can bring out some noise if the exposure isn't through the roof (looks completely blown out in the VF), so the uninitiated will under expose and screw up the first few shots. I was surprised that some footage I shot today that looked like throw away, looked fine in post after applying a reverse curve.
Light sensitivity: A word: wow! Even with the F4 28-135 Sony G lens. Way more light sensitivity than the EX3 with a F 1.9 lens. I can't imagine an F.95 lens on this sensor. ND64 is barely enough with F8 aperture in daylight. Filming in low light, the details and the bokeh are both creamy and noise-free. Convenient feature over the EX line: shutter speed may be completely controlled without switching shutter to a special mode. If you're filming by moonlight, and you need more light without increasing the gain, it's easy to dial down the shutter speed (works in speed, not degrees mode) into subdivisions of the frame rate.
The 28-135 Lens: Okay, it IS a limited range, and is still tunnel vision on the 'wide' end. It's a slow lens being F4, too. Differential focus will not be as good as a F1.4 lens, obviously, nor will the light gathering (but light is no problem with this sensor! F11 indoors at night under LED lighting!) Yes, I'd like a wider lens. And one with more zoom. But this was okay on the outdoor soccer field and was perfect for filming kite flying.
Want to know what's GOOD about this lens? The image quality. It's simply the BEST lens I own at this time. This is the first lens that doesn't have any visible chromatic aberration. And it's sharp from corner to corner. The ultimate test for CA is filming up through the trees to overcast sky. Other lenses are horrid in this regard, with 2-3 pixels of red/green fringing. Not this lens. I had to look again and again. It's not there. The images are in a different league from the Fujinon lenses, which themselves are very good.
S&Q: As my Premiere CS6 won't do XAVC, I fortunately have a Atomos Shogun that writes a compatible file. I mostly record on the Shogun off the HDMI port. For S&Q, I record the high frame rate to XAVC and then play it back at 24FPS to the Shogun, where it gets recorded and converted to ProRes. That works. The YT clip above has insets of video at 180FPS played at 23.976. From what I can tell, the image quality is very good.
There are numerous mysteries yet to uncover. One is how to get the scopes to display. An article I was reading tonight said that they only work when LUTs are active, but I don't know which LUTs or where they need to be applied to get past 'cannot proceed' when the scopes attempt to be accessed. Another mystery is what replaced Picture Profiles, and what type of file stores that kind of information to the SD cards.
I need to get some wide angle lenses for indoor shooting. I'm looking at Rokinon lenses from 10 to 14mm. Many of excellent reviews by their owners. I'll also look at some cine primes. And I really could use a zoom that goes wider, but is as clean and free of distortions as this Sony G lens. What an amazing lens.
Rails, matte box, quick release, shoulder rig.. all necessary items. As configured from stock, there's no proper way to set this thing down on a flat surface. And that grip handle has to be removed and re mounted in the opposite direction to slide the cam onto a tripod head. These things can be fixed with accessories, so I'm not too concerned.
My decision to buy a Shogun recorder last March proved to be the right one, as it's been the savior that lets me get 2160P into the editor. NVENC plugin and nVidia GPU with 1536 CUDA cores makes it possible to work with the footage and to render out 2160P videos for Youtube in reasonable time. I can't imagine doing the renders in software.
In conclusion, I think the FS7 is a keeper. The picture is amazing. Especially if the scene is halfway decently exposed. The images on my DCI 4K monitor look like you can just reach through the bezel and touch them. The dynamic range is such that I find myself turning the monitor backlight all the way up to 100, where the forest floor is nicely lit, and when the camera pans up to the sky, you have to squint, it's so bright. Yup, the way to enjoy some of those 14 F-stops is to turn the backlight all the way up, so the darker parts of the scene come out of the black. It's amazingly clean, still. And I'm sure the shooting can be optimized way further than my early experimental shots.
