Working with the PS Cam X35 – 750fps slow motion

The PS Cam X35 is a camera few people are aware of, even though it was for a long time among the more affordable cameras to shoot up to 750fps (1500fps via paid upgrade). It has been around since 2011 when it was introduced at IBC and still there’s hardly any information available about it. Let’s change that.

psx35For the day of religious liberty I thought it is appropriate to publish this clip I shot on the PS Cam X35 last year and share the experience. I will not call this writeup a “review” because it was my first attempt at high speed shooting and there are several unlucky things that got in the way. I will mention these in the hopes that others might be better prepared.

rf_setfoto_021So what’s this camera about?
Well I already mentioned that it was among the more affordable high speeders. That said at about $74,000 the price is still in the pro league and recently high speed has become a lot more affordable with the introduction of other new cameras like the Phantom Miro.

The other thing you should know about this camera is that additionally to highspeed it can record sync sound in rf_setfoto_098normal speed up to 60fps which makes this camera more versatile. In terms of slomo it can record 1-750fps @ 1920×1080 with a global shutter CMOS and does so by caching the frames.
You hit record and when the buffer is full after a few seconds it will stop. Next thing you do is playback the shot at normal speed and record that signal via the 10-bit uncompressed (4:2:2) 1,5G HD-SDI outputs.

For the recording purpose we had the Gemini 4:4:4 recorder, one of the single most annoying pieces of tech I’ve ever used. That had to go off my chest. I can not recommend the Gemini and will personally never touch one of these again.

rf_setfoto_061We need more light! This was a charity shoot with no money. I tried to spend as little as possible on lighting but apparently I miscalculated and we ended up with too little. This is one of the reasons why there’s just too much noise in all the footage which was a pity.
If you plan on shooting with this camera make sure you bring sufficient light or otherwise you’ll pay for it later.

rf_setfoto_003Obviously highspeed needs light, but apparently a camera like this needs even more light as the sensor is very unforgiving and has noise patterns in the blacks and dark areas. It’s wise to not dial up the ISO’s like on a normal speed camera. And one should keep in mind that the signal from the X35 was not RAW in the configuration we shot.

clipAnother thing about lighting and the X35 or highspeed in general is dynamic range. Here you only get about 9-11 stops of range which is a whole different way of working than with the high dynamic range cameras we’ve gotten used to. Even though we used very soft lighting fixtures and the darks were already drowning the highlights still clipped easily.
Another thing I noticed in terms of picture quality was the image density. It seems to me like there is either a native resolution lower than 1920×1080 or a kind of pixel smoothing going on. On the left you can see a 100% crop of an image with fine details. It seems like the details don’t get displayed correctly and result in something that looks like aliasing or something I’d call “pixel stepping”.

You can also see these phenomena in P+S’ own demo videos which can be found here.
The P+S knowledge base entry states that the native resolution of the sensor is full 1920×1080.

Now the last thing you should know:
Being a thorough guy I did a test shoot several days before the actual shoot. The camera produced a flat clip with the LOG90 gamma curve that I had chosen. The other alternative was PSlogC which I didn’t choose because it was too flat for my purpose.
The result of the white surface I filmed had little noise. Had I tested on a black background (as the subjects on the actual shoot) and color corrected that beforehand with the right LUT I would have noticed that I need to expose much stronger in order to get rid of the pattern noise in darker areas. Lesson learned.

I should mention that with the color correction LUT part I was very unlucky as I could not find a LUT for the camera. I contacted P+S Technik and asked them for the LUT for their camera. Unfortunately they only got back to me after the shoot was already over 4 days later. I called the day before the shoot, I called on the day of the shoot and I called on the day after the shoot which was when I could finally get someone to answer that e-mail I had sent days earlier. But it was already too late and the actual answer was that there is really no LUT for the camera, but they used gamma curves of other cameras. Here’s the official statement by P+S Technik:

Due to the diversity of post-production software we decided during the X35 development not to create a new lookup table for the logarithmic conversion of the image data. Instead of a new table we implemented several common lookup tables.
After some test we choose the Arri logC and the SI2k log90 curves (besides the regular Rec.709 and a linear curve).
To convert the logarithmic information into linear data again you can use the Arri respectively SI2k LUT’s.

I contacted the guys who make the SI2K, but they never replied either. Until this day I don’t have a LUT that will work with the camera’s LOG90.

All in all the result is pleasing enough at 720p, and the client is happy. For the next highspeed shoot I will get the camera tested earlier on and most importantly in similar conditions.
As for the X35 I’m a little disappointed with their support and the camera seems not very forgiving. If you treat it right though, and that might just be the case for other high speed cameras too, it seems like a realiable working tool that allowed a straight forward working process.

As I mentioned earlier this is not a review and I could only observe, not judge the performance of the camera as it was very different than normal speed cameras and I did not have other high speed cameras for comparison.

Thanks to our friends at Digirental in Vienna for their support. If you need cameragear in Vienna check them out.

For more information on products by P+S Technik go to:

The religious liberty clip was an ultra low budget project. This is an unbranded version of the original which can be found here: link

Video Credits:
Production – Comfilms
Director – Benjamin Hoffmann
Production Manager – Samuel Hoffmann
Unit Manager – Jeannette Kupper
Camera Assistant – Gerhard Weiner
Photography – Sebastian Wöber

Watch it on Vimeo

Leave a Comment

You are not subscribed to this post. Follow new comments

Login to comment

Jeremy January 20, 2014

Wonderful video. Lighting is excellent on the contrary. It didn’t need anymore light in my opinion.
I only imagined 30 times slower than real motion to be slower than this !

Sebastian Wöber January 20, 2014

Hi Jeremy,
Thanks. 750fps is only 5 times slower than real time, not thirty. 1500fps would be 6 times slower, etc. It’s exponential…

Jeremy January 20, 2014

Not sure Sébastian !
considering 25fps is real time :
750fps would be 30 times slower. And 1500fps, 60x slower !
Am I missing something ? Or is it you ?

Sebastian Wöber January 20, 2014

Jeremy, to make motion 1 time slower you need twice the framerate. So for each step you have to double framerate. There is no noticable difference in slow motion between 750fps or 774fps (750+24), you will have to double it to achieve the same effect as going from 24 to 48.

1 time: 24–>48
2 times: 48–>96
3 times: 96–>192
4 times: 192–>384
5 times: 384–>768

Think of ISO or shutter speed, it’s the same game.

Jeremy January 20, 2014

Sebastian, I know you must be knowing what you are talking about. But, what
you’re saying is still complete nonsense for me ! And I don’t see any similarity with ISO there. It’s simple math !
Maybe the word “time” doesn’t have the same meaning for you and for me, or for this camera !

48 -> 24 : “2” times etc …

1 second captured at 750 frames per second, played back at normal speed, say 25 frames per second, will take 30 seconds to play. So a 750fps video is viewed at 1/30th of real life speed when it’s played back at 25fps. 30 times slowed down.
Do you agree ?

Sorry but I just want to make sure I’m not crazy !

Sebastian Wöber January 20, 2014

A 30 second clip is 2 times longer than a 15 second clip. Not 15 times longer.

Sebastian Wöber January 20, 2014

Think of speed as light with a certain wavelength of Pi and then calculate the square root of that minus the wavelength of your current time. Make sense now?

Jeremy January 21, 2014

You’re kidding me right ? ;-)

If not, I’m sorry but you’re wrong !

Nothing exponential here.

750fps is 750/25=30 times slower than real time (25fps).

No need to do complicated calculation, it’s simply linear.

Jeremy January 20, 2014

No doubt about it

David Lerner February 9, 2014

In 10 seconds at 30fps there are 300 frames.

In 10 seconds at 750 fps there are 7500 frames.

7500 frames /300 frames = 25X

So if you film 10 seconds at 750fps (7500 frames) and you display 7500 frames at 30fps you will have 250 seconds of video.
It’s linear.

Eric Darling January 21, 2014

If you play back a 240 fps clip in a 24 fps sequence, it will take 10 times longer than it would for the same amount of “real time” captured natively at 24 fps.

Put another way, if that clip contained 10 seconds of actual time, it would take 10 seconds to play if recorded @ 24 fps, but at 240 fps, it would take 240 seconds to play back.

We’re not talking about different amounts of time at the same frame rate. We’re talking about the same amount of time captured at different frame rates. Those are not the same thing. Sorry, Sebastian, I have to side with Jeremy on this one.

Roel Roelf January 21, 2014

The first person pictured would NEVER accept article 18 of the Universal Declaration of Human Rights. Odd, very odd.

Sebastian Wöber January 21, 2014

Do you know the person?

Roel January 21, 2014

Do you konw Quran 2:191

Edin Gacic Reply
Edin February 25, 2014

roeal that is not correct aspect ratio and you are wrong about Islam and Qur’an in general.

Francis Carver January 22, 2014


For this??????????????

Don’t know about that — but will you take $1K for it? It being a rare white elephant and all that?

Joel Richards Reply
Joel Richards February 1, 2014

Great video and write up. RE: slow down’s I think the difference is between perceived and actual slow motion. What Sebastian says makes sense from a perceptual stand point but in absolute terms playing back 750fps at 24fps is clearly more than 5X.

I think he knows that (he presumably ingested the footage and shot it so he’s aware that 1sec of 750fps footage takes more than 5sec to play back @ 24fps) so he’s clearly making a point about the perception of slow motion.

Sebastian Wöber February 1, 2014

I suppose I’m talking perception. I’m aware that 5 times 1 second is 5 seconds.

Edin Gacic Reply
Edin February 25, 2014

it is more then 5x in playback speed but not in motion of the actual recorded footage.

Jeremy February 5, 2014

You are defending the indefensible.
why on earth the perception of slow motion should be different than the actual slow motion ?! And where does he get this exponential thing from ?!

No offense, but what Sébastian said was simply absurd. I just expected him to say Mea Culpa.

Michael Radeck Reply
Michael Radeck October 25, 2014

\”Another thing I noticed in terms of picture quality was the image density. It seems to me like there is either a native resolution lower than 1920×1080 or a kind of pixel smoothing going on. On the left you can see a 100% crop of an image with fine details. It seems like the details don’t get displayed correctly and result in something that looks like aliasing or something I’d call “pixel stepping”. You can also see these phenomena in P+S’ own demo videos which can be found here. The P+S knowledge base entry states that the native resolution of the sensor is full 1920×1080.\” the answer to this problem is simple: the resolution of this sensor is not 1920×1080 by rgb it is a bayerpattern so the picture resolution is only 960 green pixels so not more then SD-Resolution this sensor is fare away from full HD Picture resolution Full HD has 6.3 Megapixel, a 1920×1080 Sensor has only 2.1 Megapixel!