Manual Focus Accuracy with Ec-L Cross Split Screen

[orly_andico]

I just installed a modified Ec-L focusing screen from focusingscreen.com in my new old Canon 5D.

In the past I had a jinfinance screen on my 40D and I noticed that the screen calibration was off compared to autofocus, but I had low expectations because the jinfinance screen was cheap. I didn't try to shim the screen with tape. Besides the 40D had Live View so if I cared enough I could MF with Live View.

Now the 5D doesn't have Live View. I notice that the Ec-L agrees with autofocus quite well (so the screen is correctly calibrated). However, with fast long lenses like my 135mm f2 L wide-open, autofocus consistently produces sharper results than manual focus, e.g. if I AF, the screen correctly indicates focus, but if I manual-focus, the screen (may) indicate focus, but the results at 100% are less sharp than if I use AF.

It's not calibration because the screen agrees with the AF results (and this is a Canon screen, rather expensive and I assume it is properly calibrated).

Or is it that the Ec-L is not accurate enough? (it is a cross-split screen) or maybe it will only be accurate enough if I use an eyepiece magnifier?

Basically only about half my MF images are as sharp as when I use AF. Of course with my old MF lenses, there's no point of comparison (because I can't fine-focus using Live View).

[orly_andico]

Did a bit more digging.. the screen is FRONT focusing ever-so-slightly.

This is obvious at macro distances, where the screen indicates focus but the actual focus is perhaps 1cm behind the indicated.

Because of the front focusing, I can't add shims to correct. Maybe it's within spec for the screen?

[orly_andico]

Oddly enough, when I did the "ruler trick" (screen is indicating focus at the "4")



Seems that it's dead on, or at least within 2-3 mm (minimum focusing distance, Sigma 50mm macro).