I recently posted in dpreview about flash exposure differences I have experienced between the 5D and 5DMKII. I shot my first weddig yesterday with the 5DMKII and I had to dial up FEC much more on the MKII than the MKI to get decent exposures. It was also more hit and miss - I would dial up to get the exposure right, and then, in the same situation, same settings, it would be far too much and overexpose. Just all over the place .... but mostly in terms of the 5DMKII underexposing considerably. I have two MKII bodies and two MKI bodies and the same thing holds for all of them, with both the 580 EX and EXII.
I've tested a bit today, and here are some shots - nothing spectacular just living room shots, but they clearly show the difference. Anyone else seeing this same thing? This is much more of a problem to me than any white dots or vague banding.
Mike, you clearly don't know how to use a camera. There's no other rational explanation for it
I hope you get it sorted. I've never had that much variation in flash shots except when the flash wasn't fully charged or the hot shoe was coming loose (which happened on several cameras). It's hard to imagine what caused the problem in your sample photos.
CBDigital wrote:
I recently posted in dpreview about flash exposure differences I have experienced between the 5D and 5DMKII. I shot my first weddig yesterday with the 5DMKII and I had to dial up FEC much more on the MKII than the MKI to get decent exposures. It was also more hit and miss - I would dial up to get the exposure right, and then, in the same situation, same settings, it would be far too much and overexpose. Just all over the place .... but mostly in terms of the 5DMKII underexposing considerably. I have two MKII bodies and two MKI bodies and the same thing holds for all of them, with both the 580 EX and EXII.
I've tested a bit today, and here are some shots - nothing spectacular just living room shots, but they clearly show the difference. Anyone else seeing this same thing? This is much more of a problem to me than any white dots or vague banding.
www.mydigitalproofs.com/5DMKI.jpg
www.mydigitalproofs.com/5DMKII.jpg...Show more →
See my thread on just this sort of problem I also am using a 5D II.
What you need to realize is that evaluative keys off the brightest areas. If you look only at the brightest highlights, such as the radio dead center in the frame, you'll see that its not too different in the two shots, but clearing the mkII is evaluating the overall scene differently. Not too surprising considering that 5D was one of the first implementations of the Digic II processor and E-TTL II and processors are now up to Digic IV with lots of tweeks to the ETTL-II flash algorithms and new variables like highlight recover mode.
Each new camera I've ever owned has incorporated changes and has required testing to find its baselines for AF and exposure control and what techniques produce the best results. When getting my 20D I had to unlearn 30 years of manual flash habits in order to understand how to get reliable results with Canon flash. I'm finding my 50D does things differently than the 20D.
One of the things I learned was different about E-TTL II versus the conventional wisdom about metering in general is its ability to sort out specular reflections from highly reflective objects like the tinsel on the tree and the brass trim over the fireplace from textured highlights. If you aim the camera into a mirror with average flash you'll get no flash exposure at all because the bright bounced reflection kills the flash. But with evaluative the camera, by comparing zones, can figure out that the zone with the flash reflection is out of context with everything else reflecting light and discount it when calculating flash exposure. As a result you get a more normally exposed (but not necessarily perfect) shot of you standing in the mirror with a big blown spot over your camera.
Evaluative makes most of its decisions based on radar-like map of reflected pre-flash. With direct flash stuff that is closer reflects more light, making it possible to deduce -- by comparing the 35 zones to each other -- what is closest and most reflective -- the two key variables for correct exposure with flash. If two adjecent metering zones have similar levels of ambient, but one of the zones had much brighter pre-flash reflection the camera can deduce there is something much closer in second zone and adjust flash accordingly - if direct or direct / diffused flash (i.e. falling off front-to-back)
But when flash is bounced off the ceiling the distance between the source (now the spot on the ceiling) and everything in the room becomes more or less the same: the there is no flash fall off from front to back from the flash to evaluate. That tends to confuse evaluative mode which apparently incorporates the inverse-square law into its algorithyms and logical assumptions. The best clue, focus distance information from USM lenses, is only used when the flash is used directly on the camera. Focus distance has no bearing on exposure when flash is bounced because the light doesn't fall off predictably from the flash, it falls off from the point on the ceiling it hits.
Evaluative also tends to work much better in Av mode outdoors because in Av mode the exposure computer can compare the ambient reflection in each zone to its pre-flash reflection, getting more clues about the scene content and relative distances. That provides additional data on which to draw logical conclusions regarding scene content and distance to the flash. If you click on the WWW button and look in the Canon flash section there is test I did with high speed flash outdoors in Av mode which illustrates the type of situation evaluative is ideal for. But using evaluative indoors with Av can produce unpredicable results when AE is locked on a different area than where the camera is pointing when the pre-flash fires.
So just as a screwdriver isn't a good choice for pounding a nail, ETTL evaluative mode isn't the best choice for all flash jobs, especially when bouncing and lighting up the room like an overcast day. You are bouncing to average out the lighting, so logically averaging would also be a better choice for metering bounced flash. In a static situation like that where everything in the room is exposed to the same amount of light off the ceiling using M would be an even better choice. That way if someone walked in front of the camera in a white shirt it wouldn't skew the TTL based expose one way or the other.
I find evalutive works better, both with single and dual flash, when used directly or with reflection/diffusers which bounce the light forward. The more the light falls off front to back relative to the flash(es) and the more overall contrast the scene has, the better evaluative does at getting the highlights exposed correctly in the technical sense. But the catch is that exposing the highlights correctly often results in midtones which look too dark compared to what is seen by eye with the ambient light. That's due to the fact the eye has a longer range than the camera sensor and the flash light is usually coming from different directions.
Each camera is what it is and you can't change it, so the best approach is to understand how it works. Its a machine and will react the same to similar inputs. You just need to understand how it evaluates those inputs to control output. If you were to conduct that and other tests similar to those in the other Canon tutorials on my site you'll get a better understanding of how the camera evaluates different situations and which flash / metering mode will work best. I stick to evaluative in situations where I know it will work. In situations where I know it will not work I use M. In either case I rely on the camera feedback in the over-exposure warning to monitor and control highlight exposure.
Try testing with direct flash, no highlight recover engaged on the mkII, with scenes with and without specular refections.
Elías Seguí wrote:
580 EX is not for 5D MKII try 580 EX II
Elías Seguí
um, that is not true - the 580 mk1 should work with the 5dm2.
Anyway, I ran across a different problem and was wondering if I was just having a bad day or something else was going on. The last cocktail hour I shot I was getting wildly fluctuating flash results until I switched my 5D from spot metering to evaluative. AFAIK the flash metering is keyed by the custom function and not the camera metering mode, but that was not how it seemed.
Anybody have any conclusive documentation on this?
If it's any consolation, I have experienced the very same thing. The first event I've shot (only had the camera a week) was a family reunion and I consistently noticed I had to bump up FEC on the MK II using a 580 (v.1). I've shot a lot of events (usually weddings) using 5D's and know that I'm used to the way it exposes/meters when paired with a 580, but I kept looking at the 5D II's exposure settings and scratching my head while shooting the family reunion.
It's not a deal breaker, it just means I'm having to adjust the way I normally do things. I expect that one camera to the next is going to meter different (especially between different versions of digic), but I've never experienced such a large variance.
Shane Parker wrote:
If it's any consolation, I have experienced the very same thing. The first event I've shot (only had the camera a week) was a family reunion and I consistently noticed I had to bump up FEC on the MK II using a 580 (v.1). I've shot a lot of events (usually weddings) using 5D's and know that I'm used to the way it exposes/meters when paired with a 580, but I kept looking at the 5D II's exposure settings and scratching my head while shooting the family reunion.
It's not a deal breaker, it just means I'm having to adjust the way I normally do things. I expect that one camera to the next is going to meter different (especially between different versions of digic), but I've never experienced such a large variance. ...Show more →
Ditto... Except mine was over Christmas and with a 580EX II.
I have had a similar experience in using the 580 EX flash on the 5D Mark II. I have to crank the flash exposure compensation up by +2/3 to +1 to get a unclipped histogram (on the shadow side). My previous camera was a 20D and I didn't really have that problem.
I had assumed it was because I didn't have the 580 EX II.
Perhaps they did this because of the 14 bit and better high iso performance over the old 5D. My experience of shooting the 1Ds3 tells me that the second shot should push nicely in DPP, while preserving highlights, where not much can be rescued if it burns out. Maybe not. I don't like or use ETTL anyway, so this is just a suggestion.
Lynwood Cowan wrote:
I have had a similar experience in using the 580 EX flash on the 5D Mark II. I have to crank the flash exposure compensation up by +2/3 to +1 to get a unclipped histogram (on the shadow side). My previous camera was a 20D and I didn't really have that problem.
I had assumed it was because I didn't have the 580 EX II.
You don't want expose digital for the shadows. Expose of the highlights, add fill for the shadows to bring them up to the level the sensor can record them with detail.
The sad fact is a digital sensor can't record the same range of tone perceived by eye in flat light, let alone off-axis light which creates even darker shadows. If exposure is not keyed to highlight detail its lost forever... So in the technical sense correct exposure is when all non-specular highlights are recorded below 255 in all channels on a 0-255 scale.
The Catch-22 with digital is that in most situations in ambient light or single flash (especially when used off axis) a correctly exposed digital file out of camera will look too dark PERCEPTUALLY and lack shadow detail.
Boosting exposure as you are doing isn't the best solution to that perceptual problem. Better approaches would be:
1) Record the highlight detail in the camera then open the mid-tone values in post-processing via the middle slider in Levels, a Curves adjustment, or the method I most often use selective lightening using a masked screen adjustment layer.
2) Use dual flash. This actually addresses the root cause of the problem -- scene contrast exceeding sensor range - by effectively reducing the range of the scene within the range of the flash to point where the lame-ass sensor can record all of it.
Think of it this way conceptually. Neutral fill over the camera is used to lift the shadows up to the point above the noise threshold of the sensor where detail in the shadows can be seen. If you where to look at a file exposed that way with just fill it would look way too dark in the midtones and highlights. But in a two-light scenario revealing those areas is the task of the key light.
When the key light is added OVER NEUTRAL FILL is doesn't hit the shadow areas the fill has lifted. They remain the same -- above noise where detail is revealed. So its just a matter of adding more and more off-axis key light until the point where the brightest textured areas the key light hits are about 1/3 stop below clipping. If you want lighter shadows (i.e. a lower lighting ratio) you simply start the process with more fill light.
Fill is always represented by the constant "1" in the portrait lighting ratio convention because when that convention was established it was assumed the fill light was even on both sides of the face and of a sufficient level to reveal the desired amount of shadow details. The highlight value represents how much brighter the highlights appear RELATIVE TO THE EVENLY FILLED SHADOWS.
H:S
1:1 Even neutral / axis fill on both sides of face
1:0 Overlapping key light of same incident strength
===
2:1 Reflected (perceptual) lighting ratio... highlights overlapping fill reflect 2X more light
When key light is 2x brighter than fill (1-stop) in absolute incident intensity the ratio is 3:1
H:S
1:1 Even neutral / axis fill on both sides of face
2:0 Overlapping key light 2x brighter than fill
===
3:1 Reflected (perceptual) lighting ratio... highlights reflect 3X more light
Fill over the lens - on a camera bracket is ideal - doesn't make the lighting "artificial" unless its overdone. When used skillfully it simply raises the shadows up into the range where the shadow detail can be recorded...
I used fill flash for all of the shots above, but hopefully not to the degree where it is noticed much if at all. I simply understand the limits of the recording medium and that in most situations fill on a bracket over my camera, usually with a diffuser, is the simplest solution to that basic limitation. Apart from in the studio I never shoot without a bracket and flash on the camera. I may not always use it, but its always there to lift the shadow detail out of the muck when it is needed...
My wife, as usual when I test stuff, was just ignoring me as I set the off camera light on the IV stand to short light her face: simply a matter of putting it 45 degrees from the nose and about 30 degrees above the eye line. Then because the fill is always on the camera bracket it was simply a matter of walking over to the shadow side, pointing and shooting. Very flattering lighting with no loss of shadow detail created with basic equipment with the most basic of set-ups: a single flash off camera. But the key to making EVERYTHING work was having the second flash on the bracket for neutral fill.
Try it, you'll like the result you'll get and you will never again worry about how much noise there is in the shadows.
Exposure with two flashes is also actually much simpler with Canon ETTL because you are not trying to make one flash do two jobs: put detail in the entire scene and create flattering lighting, In a two light scenario you simply dial in the desired ratio then using the over-exposure warning as a guide ride herd on the highlights with FEC. Whether than takes 0, +1, or +3 is irrelevant because is an arbitrary override of camera metering. Deviation from 0 simply means its a scene that is difficult to evaluate automatically.
Neither approach, indoors or out is the most convenient, but I feel the effort is worth it.
I don't mind having to adjust the flash by using FEC so it's not really a deal buster by any means.
At first, I was concerned that one of the advantages of shooting at higher iso - better flash battery life and faster recharges - would be lost by having to crank up the FEC. But..duh.. then it occured to me that FEC really does not fit into that equation at all, it will take a certain amount of flash light to expose properly in a certain situation and it will always take less flash light as the ISO goes up. The proper exposure could be set correctly by the camera or by the user, it really does not matter.
Another factor at work is that the 5DMKII is more accurate in it's ISO ratings; ISO 400 is really ISO 400. On the 5DMKII, ISO 400 really is more like ISO 500. This should not affect the ETTL flash settings in a big way but it does mean that you cannot set a 5D and 5DMKII to the same natural light settings and get the same results. SInce ETTL is through the lens, this should not have an effect when shooting ETTL, but if the flash is used in M mode, then the 5DMKII will produce a darker image than the 5D unless the flash power is bumped up to adjust for the difference in Real ISO.
I have experimented with two light setups at weddings but I have not had great results, mostly because I never shoot in the same exact location for very long. I work for myself and it is tough to get things setup, quickly tested, and then move on to the next spot.
But there are two situations where I have used 2 lights and I have gotten it to work "just OK but not really good"
Situation 1 is where the off camera flash is behind the subject providing a bit of rim light. And situation 2 is where I have a very dark background (indoors) and want to put a bit of light in there so it does not look like we are in a cave.
I have also had very little success using two or more flashes with Canon's system in ETTL. I have always had to put the second flash in M mode and trigger it with a pocketwizard.
A great post for the end of the year. One of the many reasons I joined FM was for folks who put in time for quality material like this. I think your praying mantis shot demonstrates your point the best. I can see noiseless details in his body, while the highlights on his face are just below blowing out.
Having shots hundreds of weddings and events with the 5D mkI I always thought that the flash metering was one of the worst features of the camera. It is a constant effort to get consistent flash exposures. Overexpose flash for whites the more it fills the frame the more exposure comp is necessary this is great at weddings where the bride is in a big white dress. Black suites require underexposing the flash. Shinny dresses require more compensation than flatter fabrics etc. Canon flash metering is supposed to use distance info from the lens and the pre-flash is supposed to determine the reflectance of the subject this seems like it should give the camera the information for pretty consistent flash metering but in practice its really a poor system.
Recently a good friend who is getting back into weddings after a break of about a decade came with me to a couple weddings and was shooting with a couple of Nikon d300's and I was amazed at how he could just point his camera at a subject indoors or out black suits or white dress's and get really accurate ambient and flash metering. I have been anxiously awaiting a comprehensive review of the new 5D to hear how the camera performed in situations similar to the way I tend to use it, so I'm very disappointed to hear that many of the flash problems from the previous generation continue.
I'm really not thrilled with the idea of storing and manipulating 21 megapixel files, I don't care about live view or video. Sensor cleaning and a bigger LCD are appealing as are better high asa performance. Now to hear about the flash metering still being underwhelming I'm not that excited about buying one.
The design concept of metering off the viewfinder, which necessitates all the zone and pre-flash nonsense, is the root cause of the exposure problem. At the point when the mirror and optical viewfinder are eliminated there will be a huge improvement in automatic exposure in large sensor cameras. The x-sync limit will also be a thing of the past.
A $200 P&S digital does a better job of automatic exposure than a $3,000 DSLR because the P&S which only has "live view" is continually dumping and displaying the sensor. That means it can evaluate every single pixel 30 or more times a second to determine if its under-exposed and adjust the camera settings in a real-time control loop.
My camera previous to the 20D was a 5MP Minolta D7Hi which had a EVF. Most of the scenic and bug shots I posted above were shot with it. It has a live histogram in the viewfinder and no flash sync limit. The reason I wound replacing my trust Vivitar 285HV flashes for the 580ex was because that camera, and its lack of x-sync limit spoiled me: I bought the 580ex primarily for the high-speed sync.
So if Canon produced a 1.6 or larger sensor camera with an EVF I'd buy it in a heartbeat. It would no doubt be simpler to make and much less expensive. Granted an EVF is an acquired taste, but they've been the norm for video since the inception of the camcorder.
I have to say that I have noticed the same thing with the 5DII vs I. I have the 580EX. I was going to try my Metz 54MZ-4 and see if I get the same differences. I did not have a lot of time to try a lot of different settings. I will say that the "underexposure" is regardless of whether or not I use a bounce or direct flash. Mark
ONe thing is that there is a custom function set of choices to change the flash from average to evaluative. It is only supposed to work with the II versions flashes. In a few short tests the average seems to work a bit better even with the I version flash. However, only tired a couple of examples. I would be curious how much difference the II flash makes.
You don't want expose digital for the shadows. Expose of the highlights, add fill for the shadows to bring them up to the level the sensor can record them with detail.