A significant Difference between Lightmeter and Camera

[xpatUSA]

I'm still playing with my Christmas "present" - a Gossen Lightmeter, today with the 5-degree spot-metering function. I notice quite a difference between the readings even though my cameras have a 5 or 6 degree spot size. They are both fixed focal length, so no variation in spot size. For example, the Gossen recommends 1/1000 sec, the camera recommends 1/400 sec. That's a big difference to my pedantic mind and is not explained by differences in K-factors, nor by different ISO speed definitions ...

Help! Anybody know why I can get over an EV difference with the same light, same subject, same time, same ISO, same f-number? Have I missed something during my long journey through Exposure 101?

[Manfred M]

A couple of thoughts Ted.

The some Gossen meters used a Cadmium Sulphide metering cell, which had a "memory". Shine a bright light on it and it would take some time (measured in seconds) to get rid of the memory effect. This could fool the meter. I remember some issue with my father's Gossen meter from the 60s or 70s. I remember Gossen switched to Silicon Blue cells at one point to compensate for this, but these cells were quite battery hungry. Forgetting to switch off the meter could result in having to replace the batteries the next time one needed to use just a few days later.

The other thought is that the older meters were effectively monochrome, whereas the newer ones in many cameras actually measure the R, G and B wavelengths separately. I believe they use a standard mix (~30, 60, 10) to calculate the exposure, whereas the older meters did not and just did a pure luminance measurement . Potentially another source of error should the light source use not match the RGB values from the camera?

[xpatUSA]

A couple of thoughts Ted.

The some Gossen meters used a Cadmium Sulphide metering cell, which had a "memory". Shine a bright light on it and it would take some time (measured in seconds) to get rid of the memory effect. This could fool the meter. I remember some issue with my father's Gossen meter from the 60s or 70s. I remember Gossen switched to Silicon Blue cells at one point to compensate for this, but these cells were quite battery hungry. Forgetting to switch off the meter could result in having to replace the batteries the next time one needed to use just a few days later.

Thanks for the advice, Manfred. The Gossen is actually the fancy one I got brand new for Christmas, model Starlite 2. Manual says it has "two 'Sbc' silicon photodiodes, color-corrected". The model is very similar to the Sekonic L-758D in operation and specification. No coincidence, I'm sure ...

The other thought is that the older meters were effectively monochrome, whereas the newer ones in many cameras actually measure the R, G and B wavelengths separately. I believe they use a standard mix (~30, 60, 10) to calculate the exposure, whereas the older meters did not and just did a pure luminance measurement . Potentially another source of error should the light source use not match the RGB values from the camera?

Both cameras I tested today are mirrorless, so the metering should indeed be of luminance, as you suggest. But modern meters, of which my Gossen is one, do also have a luminance response often claiming 5% or so adherance to the CIE luminous efficacy curve.

Sorry if my post was misleading ...

[mugge]

I think we need some small clarification here regarding 'Christmas present.' Special consideration on the word 'present' need serious clarification in particular

[xpatUSA]

I think we need some small clarification here regarding 'Christmas present.' Special consideration on the word 'present' need serious clarification in particular

Yes, you'll notice I later said "I got for Christmas", and yes it was gift from me to me ... thanks, me.

Tomorrow, I might compare my three other cameras with the Gossen ... One them is a DSLR, for what that's worth ...

I still think that over an EV is too much difference and is worth investigating further, eh?

[Manfred M]

I still think that over an EV is too much difference and is worth investigating further, eh?

Yes, but not unheard of.

One thing the photographers at the studio I frequently shoot at have noticed that the flash meter (both the studio's Polaris 2 and my Sekonic L-358) require Canon shooters to open up about 1 EV versus Nikon shooters, regardless of camera model or lens being used. The readings are bang on with the Nikons (and Sonys).

[xpatUSA]

Originally Posted by xpatUSA I still think that over an EV is too much difference and is worth investigating further, eh?

Yes, but not unheard of.

Grump! Not what I wanted to hear ...

One thing the photographers at the studio I frequently shoot at have noticed that the flash meter (both the studio's Polaris 2 and my Sekonic L-358) require Canon shooters to open up about 1 EV versus Nikon shooters, regardless of camera model or lens being used. The readings are bang on with the Nikons (and Sonys).

Thanks for that note, though. Further investigation needed here, methinks - with a bit more rigor such as nice constant artificial light and the trusty Kodak R27 card.

[dje]

Further investigation needed here, methinks - with a bit more rigor such as nice constant artificial light and the trusty Kodak R27 card.

Yes I think so Ted.

One thing that might be interesting is to take a shot of the gray card with each shutter speed and compare the raw histograms and jpeg histograms. The jpeg "spike" should sit around a DV of 118 I believe for the SOS definition of ISO.(18% grey sRGB gamma adjusted). But who knows how the camera manufacturers define ISO and I wonder if you know how the light meter defines ISO?

In a mirrorless camera, I assume they use the main sensor for metering which would mean that they are working with RGB values. Presumably the variance between R G and B channel responses must be allowed for and also an assumption must be made about the spectral characteristic of the light source.

Dave

[xpatUSA]

Yes I think so Ted.

One thing that might be interesting is to take a shot of the gray card with each shutter speed and compare the raw histograms and jpeg histograms. The jpeg "spike" should sit around a DV of 118 I believe for the SOS definition of ISO.(18% grey sRGB gamma adjusted). But who knows how the camera manufacturers define ISO and I wonder if you know how the light meter defines ISO?

Thanks, Dave. Yes, I've done that test quite recently on one of the cameras, using the Foveon middle layer which is close-ish to the CIE human efficacy spectral response. The raw comes out low, i.e. neither 18% (nor 12.7%) but the sRGB spike is centered at about 122/255. Didn't do it at a range of shutter speeds though ... I was in aperture priority so as to allow the mid-gray recommendation. Have also used the white side of the card, very little difference of course.

I don't know K for the Gossen Starlite 2 but the latitude offered by standards is certainly not 1+ EV!

In a mirrorless camera, I assume they use the main sensor for metering which would mean that they are working with RGB values. Presumably the variance between R G and B channel responses must be allowed for and also an assumption must be made about the spectral characteristic of the light source.

Yes, there is a live luminosity histogram so there is indeed a continuous conversion from sensor space to luminance, Method not published; After Sigma bought Foveon, that kind of information dried up. For SOS, the ISO standard says "The linearized luminance signal shall be formed from the linearized RGB signals using the equation: Y = (2 125/10 000 R) + (7 154/10 000 G) + (721/10 000 B)" which you will no doubt recognize if you ignore ISO's stupid use of integer fractional format.

The now defunct CIPA DC-004 (originator of SOS) says the same thing and clearly mandates the use of Daylight D55. Like, every amateur photographer has a few laboratory-grade D55 bulbs lying around somewhere ... LOL.

Not entirely relevant but this might be of interest:

https://www.dpreview.com/forums/post/63520843
.

[dje]

Thanks, Dave. Yes, I've done that test quite recently on one of the cameras, using the Foveon middle layer which is close-ish to the CIE human efficacy spectral response. The raw comes out low, i.e. neither 18% (nor 12.7%) but the sRGB spike is centered at about 122/255. Didn't do it at a range of shutter speeds though ... I was in aperture priority so as to allow the mid-gray recommendation.

Ted I've just done a similar test with my Sony A7. I got similar results to you, RAW green channel was well under 18% (Can't give you the exact figure until I confirm what the Black Level is). The sRGB values were around 115.

If your camera metering seems to be giving about the right exposure according to SOS, then it makes me think there is something funny going on with the external meter.

Dave

[xpatUSA]

Ted I've just done a similar test with my Sony A7. I got similar results to you, RAW green channel was well under 18% (Can't give you the exact figure until I confirm what the Black Level is). The sRGB values were around 115.

Thanks for the quick test. Mine came out at about 9 per cent, relative to the green layer Saturation level.

If your camera metering seems to be giving about the right exposure according to SOS, then it makes me think there is something funny going on with the external meter.

With two cameras showing a similar result, I am inclined to agree - in spite of the low confidence level. At my age, there's always the possibility of finger-trouble too.

Still dark here. I intend to try later, weather permitting.

[xpatUSA]

I've just tried the Gossen and the two Sigma cameras indoors under "warm" LED lighting and, at f/2.8, spot-metering the 8x10 Kodak R27 gray card from about 4ft away, I get 1/8 sec for all three. Good enough for "Govermint Work", eh?

As to what happened outside to prompt this thread, I remain wary of all the possibilities for error mentioned, and have learned yet again that pointing the camera and shooting doesn't always result in perfection!

Thanks to all who responded so far.