Ir illuminators

I have lots of cameras in my current design (solve problem once, apply
many times!).

Most instances require Ir illuminators (I am trying to standardize on
a single device "used appropriately" instead of different devices
for the different applications).

I'd like to save power on the illuminators.

One approach is to drive the entire "illuminator" with different
current levels to get varying intensity output.

Another approach is to drive varying *numbers* of emitters
(at fixed drive levels).

[In each case, closing the loop with video feedback to ensure
I'm not over-illuminating the scene]

Suggestions as to which will give me most *effective* control
vs. power dissipated?

mandag den 11. september 2023 kl. 23.50.46 UTC+2 skrev Don Y:
> I have lots of cameras in my current design (solve problem once, apply 
> many times!). 
> 
> Most instances require Ir illuminators (I am trying to standardize on 
> a single device "used appropriately" instead of different devices 
> for the different applications). 
> 
> I'd like to save power on the illuminators. 
> 
> One approach is to drive the entire "illuminator" with different 
> current levels to get varying intensity output. 
> 
> Another approach is to drive varying *numbers* of emitters 
> (at fixed drive levels). 
> 
> [In each case, closing the loop with video feedback to ensure 
> I'm not over-illuminating the scene] 
> 
> Suggestions as to which will give me most *effective* control 
> vs. power dissipated?

if you are trying to save power you'll need a switching regulator so it is 
trivial to control the output level and controlling the level instead of 
number won't change the light pattern

On 9/11/2023 3:11 PM, Lasse Langwadt Christensen wrote:
> mandag den 11. september 2023 kl. 23.50.46 UTC+2 skrev Don Y:
>> I have lots of cameras in my current design (solve problem once, apply
>> many times!).
>>
>> Most instances require Ir illuminators (I am trying to standardize on
>> a single device "used appropriately" instead of different devices
>> for the different applications).
>>
>> I'd like to save power on the illuminators.
>>
>> One approach is to drive the entire "illuminator" with different
>> current levels to get varying intensity output.
>>
>> Another approach is to drive varying *numbers* of emitters
>> (at fixed drive levels).
>>
>> [In each case, closing the loop with video feedback to ensure
>> I'm not over-illuminating the scene]
>>
>> Suggestions as to which will give me most *effective* control
>> vs. power dissipated?
> 
> if you are trying to save power you'll need a switching regulator so it is
> trivial to control the output level and controlling the level instead of
> number won't change the light pattern

There are a lot of emitters (in many designs) so I suspect putting every
other or third, fourth, etc. in a different string -- still spread out
around the lens -- won't appreciably change the light pattern for the
sorts of scenes where lower illumination is acceptable.

I may opt to excite the illuminator with a higher potential
to get more compliance (some of the cameras I've torn down
have as few as three emitters in a string -- with a dozen
or more strings!  This seems like a lot is being lost in the
ballast)

I think I may have to hack together a prototype that
lets me control individual emitters so I can see what the
effect on illumination will be.  That might make the driver
choice more obvious...

(the units I've examined drive all of the emitters simultaneously,
regardless of how wired; so, PCB layout is the driving factor,
not "dispersal of light")

On 11/09/2023 22:50, Don Y wrote:
> I have lots of cameras in my current design (solve problem once, apply
> many times!).
> 
> Most instances require Ir illuminators (I am trying to standardize on
> a single device "used appropriately" instead of different devices
> for the different applications).
> 
> I'd like to save power on the illuminators.

Decide what signal to noise in the video image is acceptable for your 
application and use only that much artificial light.
> 
> One approach is to drive the entire "illuminator" with different
> current levels to get varying intensity output.
> 
> Another approach is to drive varying *numbers* of emitters
> (at fixed drive levels).

If you get down to 1 or 2 LEDs then that is a whole stop difference in 
exposure, but once you go up to 2 or more LEDs you can get exposure 
right to the nearest half stop which is good enough for all practical 
purposes (even back in the days of conventional film).

You effectively want a sequence of LEDs lit up along the lines of:

2 3 4 6 8 11 16 22 32

just like the old f stops on a classical camera.

Beware that if they are too far off axis the lighting profile might 
change as you switch more distant ones on and off.

> [In each case, closing the loop with video feedback to ensure
> I'm not over-illuminating the scene]
> 
> Suggestions as to which will give me most *effective* control
> vs. power dissipated?

Using the least amount of illumination you can get away with.

-- 
Martin Brown

On 9/12/2023 12:28 AM, Martin Brown wrote:
> On 11/09/2023 22:50, Don Y wrote:
>> I have lots of cameras in my current design (solve problem once, apply
>> many times!).
>>
>> Most instances require Ir illuminators (I am trying to standardize on
>> a single device "used appropriately" instead of different devices
>> for the different applications).
>>
>> I'd like to save power on the illuminators.
> 
> Decide what signal to noise in the video image is acceptable for your 
> application and use only that much artificial light.

I'm trying to come up with *one* camera that I can "adjust"
to suit different scenes.

>> One approach is to drive the entire "illuminator" with different
>> current levels to get varying intensity output.
>>
>> Another approach is to drive varying *numbers* of emitters
>> (at fixed drive levels).
> 
> If you get down to 1 or 2 LEDs then that is a whole stop difference in 
> exposure, but once you go up to 2 or more LEDs you can get exposure right to 
> the nearest half stop which is good enough for all practical purposes (even 
> back in the days of conventional film).
> 
> You effectively want a sequence of LEDs lit up along the lines of:
> 
> 2 3 4 6 8 11 16 22 32
> 
> just like the old f stops on a classical camera.

The cameras I've found (so far) tend to have a lot of emitters
(the one I toredown tonight had ~40).  I can afford to move the
emitters *out* of the camera (and disable any that are internal)
but still need the Ir filter, inside, to be operable.

> Beware that if they are too far off axis the lighting profile might change as 
> you switch more distant ones on and off.

Yes but most seem to be arranged in concentric rings.
I suspect the large numbers are intended just to boost range.

>> [In each case, closing the loop with video feedback to ensure
>> I'm not over-illuminating the scene]
>>
>> Suggestions as to which will give me most *effective* control
>> vs. power dissipated?
> 
> Using the least amount of illumination you can get away with.

The question is whether that can be obtained by running a bunch
of emitters at a low current *or* a fewer number at a higher
current.

(I've already taken care of the case of using the camera
at an effective lower frame rate for scenes that aren't
changing -- much.  But, there's only so much I can do
in software to juggle the power budget...)

On 2023-09-12, Don Y wrote:
> On 9/12/2023 12:28 AM, Martin Brown wrote:
>> [...]
>> Beware that if they are too far off axis the lighting profile might
>> change as you switch more distant ones on and off.
>
> Yes but most seem to be arranged in concentric rings.
> I suspect the large numbers are intended just to boost range.

The multitude of LEDs in the ring(s) are to ensure the scene is
illuminated clearly.  That is, they're less to increase range, and more
to combat things like multiple shadows or other visual artifacts
without resorting to a diffuser.

-- 
|_|O|_|
|_|_|O| Github: https://github.com/dpurgert
|O|O|O| PGP: DDAB 23FB 19FA 7D85 1CC1  E067 6D65 70E5 4CE7 2860

On 12/09/2023 09:32, Don Y wrote:
> On 9/12/2023 12:28 AM, Martin Brown wrote:
>> On 11/09/2023 22:50, Don Y wrote:
>>> I have lots of cameras in my current design (solve problem once, apply
>>> many times!).
>>>
>>> Most instances require Ir illuminators (I am trying to standardize on
>>> a single device "used appropriately" instead of different devices
>>> for the different applications).
>>>
>>> I'd like to save power on the illuminators.
>>
>> Decide what signal to noise in the video image is acceptable for your 
>> application and use only that much artificial light.
> 
> I'm trying to come up with *one* camera that I can "adjust"
> to suit different scenes.

Even so your lowest power budget will still be when you have just enough 
illumination to do the task at hand (and no more than that).

>> You effectively want a sequence of LEDs lit up along the lines of:
>>
>> 2 3 4 6 8 11 16 22 32
>>
>> just like the old f stops on a classical camera.
> 
> The cameras I've found (so far) tend to have a lot of emitters
> (the one I toredown tonight had ~40).  I can afford to move the
> emitters *out* of the camera (and disable any that are internal)
> but still need the Ir filter, inside, to be operable.
> 
>> Beware that if they are too far off axis the lighting profile might 
>> change as you switch more distant ones on and off.
> 
> Yes but most seem to be arranged in concentric rings.
> I suspect the large numbers are intended just to boost range.

Probably to act much like a ring flash in close up and macro work to 
avoid there being any distinct shadows in the field of illumination. 
That may not always be what you want - off axis illumination casts 
better sharp shadows that highlight targets in the field of view.

>>> [In each case, closing the loop with video feedback to ensure
>>> I'm not over-illuminating the scene]
>>>
>>> Suggestions as to which will give me most *effective* control
>>> vs. power dissipated?
>>
>> Using the least amount of illumination you can get away with.
> 
> The question is whether that can be obtained by running a bunch
> of emitters at a low current *or* a fewer number at a higher
> current.

I suspect it will be very similar until you start pushing the power LEDs 
quite hard. Conversion efficiency only falls off gradually with 
increasing current until you get fairly close to Imax.

I'd expect several emitters at lower current to be very slightly more 
efficient but I doubt it is by enough to make more than 1% saving.

> (I've already taken care of the case of using the camera
> at an effective lower frame rate for scenes that aren't
> changing -- much.  But, there's only so much I can do
> in software to juggle the power budget...)

You might want to take a look at QCUIAG who are the (amateur astronomy) 
masters at tweaking standard webcams for ultra low light use.

http://www.qcuiag.org.uk

Although they mostly want longer exposures and low noise sensors and are 
very price sensitive. Beginners destroy two or three when learning...


-- 
Martin Brown

On 9/12/2023 2:29 AM, Dan Purgert wrote:
> On 2023-09-12, Don Y wrote:
>> On 9/12/2023 12:28 AM, Martin Brown wrote:
>>> [...]
>>> Beware that if they are too far off axis the lighting profile might
>>> change as you switch more distant ones on and off.
>>
>> Yes but most seem to be arranged in concentric rings.
>> I suspect the large numbers are intended just to boost range.
> 
> The multitude of LEDs in the ring(s) are to ensure the scene is
> illuminated clearly.  That is, they're less to increase range, and more
> to combat things like multiple shadows or other visual artifacts
> without resorting to a diffuser.

That;s not what I'm seeing.  If I use, for example, every other
group of three emitters in the outer ring, coupled with the
alternate groups of three emitters in the inner ring, I get
uniform illumination -- but not as much distance if I turn
ALL of them on, simultaneously.

So, when the objective is near enough, I can easily drop
half of the power used in those "extra" emitters and
keep the same NEAR scene fidelity as if all were on.

On 9/12/2023 7:46 AM, Martin Brown wrote:
> On 12/09/2023 09:32, Don Y wrote:
>> On 9/12/2023 12:28 AM, Martin Brown wrote:
>>> On 11/09/2023 22:50, Don Y wrote:
>>>> I have lots of cameras in my current design (solve problem once, apply
>>>> many times!).
>>>>
>>>> Most instances require Ir illuminators (I am trying to standardize on
>>>> a single device "used appropriately" instead of different devices
>>>> for the different applications).
>>>>
>>>> I'd like to save power on the illuminators.
>>>
>>> Decide what signal to noise in the video image is acceptable for your 
>>> application and use only that much artificial light.
>>
>> I'm trying to come up with *one* camera that I can "adjust"
>> to suit different scenes.
> 
> Even so your lowest power budget will still be when you have just enough 
> illumination to do the task at hand (and no more than that).

Yes, but will it matter if I drive 4 emitters at the compass points
to 100% vs. 16 emitters (evenly filling in the gaps) at 25%?  Or,
by extension, if I drive ALL of the emitters at a set of N different
power levels, will this give the same results as driving N subsets
at on/off levels?

>>> You effectively want a sequence of LEDs lit up along the lines of:
>>>
>>> 2 3 4 6 8 11 16 22 32
>>>
>>> just like the old f stops on a classical camera.
>>
>> The cameras I've found (so far) tend to have a lot of emitters
>> (the one I toredown tonight had ~40).  I can afford to move the
>> emitters *out* of the camera (and disable any that are internal)
>> but still need the Ir filter, inside, to be operable.
>>
>>> Beware that if they are too far off axis the lighting profile might change 
>>> as you switch more distant ones on and off.
>>
>> Yes but most seem to be arranged in concentric rings.
>> I suspect the large numbers are intended just to boost range.
> 
> Probably to act much like a ring flash in close up and macro work to avoid 
> there being any distinct shadows in the field of illumination. That may not 
> always be what you want - off axis illumination casts better sharp shadows that 
> highlight targets in the field of view.

I've found "Ir Illuminators" (i.e., no camera).  They are just large
arrays of emitters arranged in whichever configuration they will
fit in the enclosure.  No attention to geometry.  So, "more is
better" seems to be the operative premis.

>>>> [In each case, closing the loop with video feedback to ensure
>>>> I'm not over-illuminating the scene]
>>>>
>>>> Suggestions as to which will give me most *effective* control
>>>> vs. power dissipated?
>>>
>>> Using the least amount of illumination you can get away with.
>>
>> The question is whether that can be obtained by running a bunch
>> of emitters at a low current *or* a fewer number at a higher
>> current.
> 
> I suspect it will be very similar until you start pushing the power LEDs quite 
> hard. Conversion efficiency only falls off gradually with increasing current 
> until you get fairly close to Imax.

What about at the *low* end?  Are they as (luminous) efficient
at 10% drive as at 90%?  Or 50%?

The OS (which manages resources in any system) has to be able
to manage the power budget of the different tasks in the
system.  It would be nice if this was a simple calculation
(it's already NP-complete -- NP-Hard?)

> I'd expect several emitters at lower current to be very slightly more efficient 
> but I doubt it is by enough to make more than 1% saving.

But, does that extend to the LOWEST end of the range?
I.e., is there a point where I am driving 40 emitters
at a level that *exceeds* the requirements that could
be met with just 20?  I.e., I'm just heating extra die
and not getting anything useful from them?

>> (I've already taken care of the case of using the camera
>> at an effective lower frame rate for scenes that aren't
>> changing -- much.  But, there's only so much I can do
>> in software to juggle the power budget...)
> 
> You might want to take a look at QCUIAG who are the (amateur astronomy) masters 
> at tweaking standard webcams for ultra low light use.
> 
> http://www.qcuiag.org.uk
> 
> Although they mostly want longer exposures and low noise sensors and are very 
> price sensitive. Beginners destroy two or three when learning...

I'm not as concerned over how to achieve low light (as in
many cases I will be pushing a lot of light from the emitters
but not as much as is *available*).  My focus is entirely on power
(and the efficiency consequences that come with that).

But, I'll have a look.  There might be some assertions that
i can rely on to guide my efforts (I'd hate to throw together
a design only to discover it's only operated at extremes)

On 12/09/2023 18:44, Don Y wrote:
> On 9/12/2023 7:46 AM, Martin Brown wrote:

>> I suspect it will be very similar until you start pushing the power 
>> LEDs quite hard. Conversion efficiency only falls off gradually with 
>> increasing current until you get fairly close to Imax.
> 
> What about at the *low* end?  Are they as (luminous) efficient
> at 10% drive as at 90%?  Or 50%?

They seem to be linear right down to the point where you can barely see 
them which for a high efficiency modern LED is somewhere around 1-10uA 
once your eyes are dark adapted. I'd be surprised if IR LEDs were any 
different (I found a datasheet for a suitable one).

It only becomes non-linear and lower efficiency at >Imax/4. eg

https://www.mouser.co.uk/datasheet/2/917/downloaddatafile-2853791.pdf

See the graph of luminous flux vs I_f log log graphs can hide a 
multitude of sins but that one is die straight for I_f < 1A.

>> I'd expect several emitters at lower current to be very slightly more 
>> efficient but I doubt it is by enough to make more than 1% saving.
> 
> But, does that extend to the LOWEST end of the range?
> I.e., is there a point where I am driving 40 emitters
> at a level that *exceeds* the requirements that could
> be met with just 20?&nbsp; I.e., I'm just heating extra die
> and not getting anything useful from them?

In general I think they are so close to linear output unless over driven 
quite hard that it is linear to all intents and purposes all the way 
down to zero intensity.


-- 
Martin Brown