#apertus IRC Channel Logs

Hello

Not yet

morning everyone!

good morning all!

Yep, I believe it is

btw, hello sasha-w!

20 minutes to go

Hello!

hello philippejadin!

Here is some pre-meeting reading material: http://blackbettycameras.com/first-camera/

when nofilmschool announced the exclusive release of a new cinema camera a lot of people thought it was us ;)

We can do better than that

http://nofilmschool.com/2013/08/new-digital-cinema-camera-exclusive/comment-page-1/#comment-442458

please check if anyone didnt get one and let me know

perfect

got it

got it as well, some ideas :

- allow customisable button labels

I think we should use DSLRs as a model, though we will have to adapt the particulars to cinema usage.

The thing with DSLRs (for pro photography purposes) is that they are designed as a tool that can be used mostly by feel.

ok let me make a list of topics to discuss directly in the google doc

Bertl: for example, let's say there are 4 rotary knobs, I'd like to be able to setup the camera to have one knob with iso, second with shutter, third with framerate, etc.... The knob could be around a small lcd that shows the function of each knob, and the current value of eah parameter (iso, fps, etc)

s

philippejadin, actually the dictator works exactly that way, look at the LCD in the last image

philippejadin: small LCDs are usually a bad choice, unless you add a backlight, in which case, you can go for OLED displays, but in any case, this requires lots of space, power and adds a lot of hardware

OTOH, if the knobs are nearby an existing LCD/TFT/OLED display, it can be use for dynamic labeling as well

in fact the current layout is good

although it adds mechanical stuff (which can easily break, again uses up valuable space and complicates things) I consider the concept of rotary buttons a good UI

anyone familiar with the thumbwheel on modern dslrs?

no, we want physical buttons

with physical knobs, once you are used to them, you don't even have to look at them, you can change a setting without looking at a menu. in documentary setup it's a must

philippejadin: yes, exactly

we should design a tool which can be operated without looking or moving one's hand from the grip

how do you know that you got the right setting? (without overlay in the viewfinder) or can this be assumed?

(or is the display on the side the viewfinder?)

Bertl, specific hardware: http://www.youtube.com/watch?v=y2mAKEdNUVg

arduino plus 320x240 2.8" touchscreen LCD plus a bunch of butons and knobs

se6astian: where are the turnable knobs?

the assistant is for example at the left of the camera, and can choose the iso on the camera, and have some feedback on screen

dmj_nova, no

se6astian: so no specific knob or device tested for this purpose yet?

Bertl, I bought a bunch of rotary encoders and knobs but didnt implement them yet

Bertl: if I need to see the exact value there's a small LCD on the top of my DSLR that tells me. Could completely ignore everything but the physical controls and that little LCD if I wanted

philippejadin: that link doesn't seem to work

dmj_nova: works here

short push : auto, long push (un)lock

Bertl: it's for color grading

ah, what are the red balls?

link just times out for me

Bertl: the top part (with knobs and oleds) reminds me of the dictator

I have a question that's off the current topic: What type of cabling will be used to connect the dictator to the axiom body? USB2, USB3?

se6astian: this "wooden handle" is it something we're married to or what's the deal with it?

wireless not so good, but can be done via usb as well, if the connection can handle a wireless dongle

philippejadin: do you use DSLRs/pro cinema cameras for video?

dmj_nova, the wooden handle was inspired by http://www.aaton.com/images/penelope5.jpg - if we add buttons controls to it it will be something new

dmj_nova: I already used quite some, I guess like others (PD170, Z1, EX3, Red one & scarlet, 5D, GH1)

what are the distinguishing ergonomic features of purpose cinema cameras

My favorite camera from an ergonomics perspective is the Arri Alexa

Mich3alG: why?

the claw, perfect :)

dmj_nova It feels like their old 35mm film cameras. It feels like an actual cinema camera. When you put it on your shoulder, it has weight that allows for smoother shots

se6astian: DSLRs are basically a claw for thumb and forefinger

I was more thinking of putting a button for each finger on the grip (which, of course, can be customized and disabled)

the thumb, being free to move can operate several functions if there is a distinct feel to each

and many more, the main question is more do we want that at all, or do we plan that as add-on ...

(because it needs to be connected somehow to the camera :)

I've not used pentax, Canon models (pro ones, not the rebels) have good ergonomics

Scroll down to the button layout section

I second that, red scarlet is pita

just to change the iso, if you don't have the touch screen connected, you have to navigate the menu with the four buttons

Mich3alG: like this? http://blog.planet5d.com/wp-content/uploads/Canon-C300-test-shoot-1.jpg

OK, so to summarize: I will create a concept for the wooden handle with buttons/controls and present it to you

dmj_nova: I've used Canons too, but something keeps making me go back to the pentax for its customizable, dual control dials. They allow me to control the camera exactly how I'm naturally inclined to

I would suggest we discuss what the concept shows already not what it does not show yet

can we have page numbers?

yup

great! thanks!

i.e. the most recent? sensor frontend design

check the headers

now page 4 :)

now page 3 again :)

added page breaks

(page 4 now seems to be left blank for notes)

hmm, not here...

philippejadin, what is the "sensor side" ?

and I'm not sure if pentax does it the same way, but all menus and relevant settings can be adjusted via the thumbwheel and the rotary wheel behind the shutter button

this is what I'd like to focus on for a moment

(note: we should find proper names for the parts)

no, I just want to discuss how that connects to the rest of the camera

with the recent advances in capabilities we can expect from boards, we can assume the camera and the recorder will be in the same box (even if this box is made of separate parts)

philippejadin: yes and no, basically we have at least three technical challenges to handle:

2) we need to move a lot of data from the sensor to the FPGA and from FPGA to the interfaces

heat is our biggest enemy :P

*water

we could make it more modular, a cube containing sensor and processing, with sata, sdi, hdmi, power

same for power

philippejadin: yup, but separating (some) parts gives the problem that we need to transport the data between them

se6astian: FPGA (zynq), voltage regulators, ImageSensor, serializers (IMHO in this order)

well, since we need sata interfaces anyway for the SSDs, those can be a separate unit

Bertl: do you think we'll need fans ?

philippejadin: maybe, maybe not, mostly depends on the design

for example, the metal block of the 'SFE' could be used to cool down the sensor (to some degree)

there could be some kind of cooling fins on it too

let me give a few numbers to get an idea what we have to handle:

Ideally the SSDs should be slot mounted like the Red cameras/ Blackmagic

I do like the SSD mounting shown in the doc

agreed

Bertl: so perhaps we can thermally isolate the two?

provide much more metal surface area to cooling the sensor?

another quite important question is: do we want to have different sensors?

sensor tech will advance faster than we must replace the FPGA

yes, at an undefined point in the future

we need to transport lots of data, which requires high speed communication, which in turn adds communication devices, which in turn draw power and generate heat

no 'internal' data juggling/transfer required here, all power/heat will be minimal

this will add communication overhead and the higher the speed, the more power is required and heat is generated

I know that this sounds like technical problems (and those are indeed technical problems) but the very much affect the design :)

(or even worse, with the sensor on one edge :)

I was thinking a T shaped system, with 2 boards

T shaped, like the weiss cam? sounds like a good idea

dmj_nova: yes, at least

More L than T in our case but yes ;)

I'd avoid anything that might make people think we're ripping off ideas from Weisscam

for example, the sensor CMV12k we are considering, has 67 differential lines

(and that is not considering the power supply lines)

that's a lot of space, as you can easily see on the prototype boards

socket or ribbon basically?

fmc connectors might exist in 90°

sasha-w: don't like their t-cam look much anyway

there is no 90° FMC connector

what densities, lane counts can we expect from standard B2B connectors?

what about a cable? or is that no good?

but we have another problem there, the electrical impedance

(which means we need specific lane widths and spacing (and ground connections)

I know that's somewhat technical again, but basically it means that we can not reduce the connector arbitrarily

basically, how thick can a "socket" be?

up to an inch @300MHz can be considered harmless (that's sensor to FPGA)

question to the 'camera users': would you mind if some (which?) parts of the camera get really hot?

Bertl: I'd say we should avoid any parts getting "hot"

I wouldn't want the handle or back getting hot though

well, that's why I'm asking, because, for example, we could cool down the sensor with a peltier element, which would heat up a heat spreader at the other end

(should be comfortable to hold, nothing should be painful to touch

Bertl: that said, if the SFE block gets hot it's fine

which brings the next question, what about cooling fins all over the FE case?

Hmm, so the cooling fins could actually help the FE case to slot into the body

possible, yes, so from a technical PoV, I'd make the entire FE a big heat spreader with lots of cooling fins

on the red scarlet, (like here http://www.thefilmmakersworkshop.com/wp-content/uploads/2012/05/Scarlet-X-1.png) it works like that : there is a fan at the front that takes fresh air and exhausts it at the top (on the back of the camera).

yeah, with the eyepiece running along the side of the camera. The camera could be positioned on your shoulder and still allow you to look through an eyepiece

sasha-w: though I imagine an electronic viewfinder would be more practical for a cinema camera

Let's avoid fans and air intake if possible

because that will improve image quality and reduce noise

Bertl: can a peltier function passively as a regular heatsink?

as it will generate electricity instead :)

ah

dmj_nova: Optical viewfinder makes things onset go a bit quicker. There's no reliance on battery power and you can set up shots in a fraction of the time.With that said electronic viewfinders allow for peaking and exposure levels to be visible...

philippejadin: that sounds really awful

or the fan starts while you shoot after a while, and everyone looks at the camera asking, what's going on :-)

sasha-w: no, I think it's internal temp

more likely on the cpu

if the temperature is 5° different then last calibration, you have to clibrate it again

sensor is at around 30° or something

maybe it's the price to pay for higher dynamic range?

yeah, well, the heat has to go somewhere

but maybe we can take a green approach and actually recover some of the heat as energy

this is where peltiers could come into play as heatspreader

i.e. the difference between internal temperature and external temperature could generate electricity

yeah, that's probably the downsinde, peltiers are somewhat heavy

what about cooling fins all over the camera (except the display and handle part of course)

the bottom must be very sturdy

to attach it to a tripod

Bertl: my guess is that we can probably not actually have fins on the outside of the body

well, that would be bad from the thermal PoV, as a smooth surface has little heat spreading capability

Bertl: the camera body being the handle and everything else

so much bigger than the SFE block

in fact, we need to know how big the sensor frontend and the other boards will be, and how they will connect to each other

the original idea was to have two board in sandwich, roughtly 7 by 7 cm iirc

I guess bigger = easier to cool

a low speed fan could be used

but the bigger the fan air input, the more risk with dust and everything

or we could use a box in a box model ?

hmm

priority is image quality, so if a fan is needed, so will be :-/

this couples cooling to the body design, not the SFE block

philippejadin: well, we need both image quality and good ergonomics so it's usable.

the fan could be an add-on attached to the top of the camera?

i.e. you decide if you want the better image quality or the silent cam?

if we want to compete with other cameras, I think we'll want the ideal sensor cooling all the time

which probably requires a peltier for the sensor anyways (but just one in the right place)

we won't get it to -30 of course without significant sacrifice

of course not

everything below 5°C is a nightmare electronics wise

or said otherwie, I guess they designed the sensor to perform best at room temp

Bertl: why not put the peltier in the body?

philippejadin: couldn't you just calibrate the sensor for each temperature and adjust profiles automatically based on reported temp?

dmj_nova: it's just that we'll probably discover that we have good image quality at 20°, decent at 30° and awfull at 60°

you cannot "calibrate" above some temperature, you'll get higher noise in the blacks

at which point you *need* something like peltier

I think in red user guide, they tell about calibration when you work at lower temperatures than room (not higher)

sasha-w: depending on the batteries we have we can hope to have 90w batteries (pretty big stuff already). It means that we would use one battery like that in 1:30 hour just for cooling (Bertl correct me if I'm wrong)

philippejadin: depends on the actual cooling, the peltier is quite efficient, so for let's say 5W heat generated, we need maybe 8-10W to cool it down

and it can provide heat to sensor as well (in theory)

philippejadin: so I'd guess you could store a range of calibrated profiles and use the one that matches the sensor's current temperature best

1.8" 2.5" ?

1.8"

Bertl: We'll need to

quite reasonable

(the power there depends on what we actually do)

so we need to cool and supply 30W of heat/power

luckily not all is dissipated as heat

cooling 3-5 watts is not a big issue

with that being the case, I expect RED's cooling issues have much more to do with processing power than sensor temp

I gotta go, I hope you'll end up with interesting proposals :-)

philippejadin: thanks for the input

really helps having pro design feedback

Bertl: you mean the whole unit...hmm...that's much bigger than the light sensitive region

ok should we close todays meeting, I have to start cooking dinner ;)

I will summarize everything into the google doc and email everyone

thanks to everyone for participating!

I'd say we'll probably need a few more to get a concept ready

let me visualize the concept before we meet again

good to hear. Ok then, bye for now :)

bye!

ok, will do

I'm curious about the peltier cooling process. So it draws power from the battery to cool? The way you described it, heat generated by the SFE and CPU hardware could be converted into electrical energy and routed elsewhere

i.e. if you have two metals attached to eachother, and a temperature difference between the sides

So what Philippe says is correct, and that it will draw power from the batteries?

I see, and use power from the battery to cool certain areas. Clever

between a higher inner temperature and a cold outside temperature it could create electricity

and the conversion of heat energy into electrical energy would reduce the temp (as heat is being converted into another form of energy)

or if used the other way round, to generate electricity from a temperature difference

I see. Thanks for helping me understand this better :)

)

(or generator, when heat flows through it)

where the green parts are the PCBs, the brown is the interconnect, yellow is the sensor, blue the peltier and orange the FPGA/etc

this could be easily extended by an active fan on the top blowing 'up'

the tiny black bars would be inward faced fins from the heat spreaders on peltier and FPGAs

btw red epic cooling: http://www.extremetech.com/wp-content/uploads/2012/01/red-x-mysterium-sensor-behind-heatsink.jpg

Reds own image sensor

i.c.

and it seems that global shutter is preferable (artefact wise) whenever something moves or the camera turns

would also avoid bizarre banding when shooting LEDs and CFLs

*pipelined*

going through the datasheet, I wonder what the actual LVDS clock rates are

Mbps which results in an input LVDS_CLK_P/N of 300MHz. The minimum LVDS_CLK_P/N frequency is 300MHz for 12

but a little further in the datasheet it says:

30MHz in 10 bit mode with 64 output channels, register 82[15:8] is 23 and register 85 is 128. This will result in a FOT =

do I understand that correctly that CLK_IN is not the LVDS clock?

so it's probably something like "pixel clock"

I just added a cooling proposal to the google doc

s/I drawed/I've drawn/

what do you think of the copper pipes?

but I doubt that we will have the space for this config

you see the brown blocks? those are interconnects

http://www.ixbt.com/mainboard/asus/a8n32-sli-deluxe/board.jpg

but I guess its something else in the core of the pipes

condenses and flows back on the other

that means, that the frontend will be roughly 140mm long, yes?

i.e. vastly different lane/wire lengths

I placed the PCB mainly because the SSDs have their SATA interfaces at the backs and need to plug in to somewhere

maybe 3 PCBs then? 2 in sandwich and one for external interfaces?

for the sata part, I wouldn worry about the orientation

and the disks could also plug in from behind, which IMHO would be a better design anyway, as they are 'optional'

(might even allow for larger/cheaper SSDs)

okay

maybe we can have a riser-card concept?

with this as pump

doesn't help if you cannot dissipate the heat :)

and rather messy :)

I always wondered why you would want to do that

IHMO it would be way better (from the mechanical and electrical PoV) if you adjust the lens mount not the sensor

well, let's make a pro/con list

~ 250 connections need to be flexible

tilting the lens mount means:

if the adjustments are only in two directions, it can be simply solved by a spherical joint

i.e. basically a sphere sector directly touching another shere

we only need a few mm or a few degrees of freedom for calibration, maybe that would be quite simple to do....

note that they do not need to be very large to give good stability

get a ball/sphere, cut off the top

now take the lens mount and cut out a sphere (actually the same sphere)

yes but how to screw them together?

screws have to be perpendicular to the sphere

I'll do a drawing if that helps

starting a movie now

note that this is illustrative only, i.e. to make it work you want to use the equatorial region i.e. the base has to be either larger with the lens/front inside or somewhat smallerwith the lens tube covering the sperical section almost completely

so the screws from the outside hold everything in place just by applying pressure?

can be improved by some kind of rubber between the joints

but will that be enough to hold a lens that weighs several kilos?

FFD?

http://en.wikipedia.org/wiki/Flange_focal_distance

0.01mm accuracy is common for FFDs

thats why the lens mount has to be so precisely milled

can be done in the 10^-6m range

but it's just a concept, I haven't seen anything like this before

also, we could make the entire stack tiltable if the tilt is somewhere in the 0.01mm range

anyway, going to bed, I will think more about it tomorrow and draw up something