Glow effect - not seamless?

Merion

Hi, me again

I just noticed a weird threshold with the glow effect. The strength field makes it seem that I can set a percentage value that gets translated to seamless levels of glow. But I noticed a hard jump at 12.5% strength (see below).

Is this always like this? Or did I run into a weird corner case? And are there any workaround to achive something in between?

Lorelei

Yeah, i've found this problem often, too. I've tried multiple different settings to no avail.

Loopysue

Its more of a steeped scale than a sliding scale.

I quite often reduce the strength to 0, set the blur radius, and then duplicate the effect as many times as necessary to get just the right intensity.

Merion

Good to know, and thanks for the workaround Sue!

jslayton

Because of how the percentage that you enter is treated internally, the answer to your question is "no, there are no ways to get more gradations between 12.4 and 12.5 for strength."

As Sue suggests, using a value of 0 for the strength and ignoring it thereafter might be your best choice. A value of 0 makes the effect ignore the rather idiosyncratic Strength processing part of the algorithm.

For the extremely technically-minded, Strength is the number of high bits of alpha (opacity) to preserve in the image (0 to 8), mapped as a percentage from 0 to 100. There are 8 bits of alpha, so percentages 0 to 12.4999 map to the special "no effect" value, 12.5 to 24.999 preserves only the alpha high bit, 25-37.499 preserves the high two bits of alpha, and so on. For added confusion, because it's a number of bits, each range should have twice as many gradations as the previous one.

Merion

Thank you, that makes total sense - well, at least until the "added confusion" part. It didn't click for me that 12.5 is of course 1/8th of 100 percent and thus it could be something bit-related.

jslayton

The precise expression would be "alphaout = (((alphain >> 8) << (uint_8)(strength / 100.0 * 8.0)) & 0xFF000000) << 8;" if you're into C-style programming and put alphain into the upper 8 bits of a 32-bit word.

Merion

Interesting! Unfortunately, I never managed to learn more than basic java.

Wyvern

As a non-technical-specialist, would that not mean it would be better to set the Strength rating to a series of numerical values that actually represented these non-linear step-changes? If I'm working with a percentage, I'm expecting a continual sliding scaling to happen - percent is 100 separate, equal, values, after all.

However, it's nice to know I'm not alone in sometimes having a battle with this feature, and that there's even a reason why it isn't behaving as the terms would suggest

jslayton

From the reports here, the biggest difficulty with the "Strength" setting is that it's entered as an apparently-linear value from 0 to 100, but the result of that percentage is very much different than that: it's both discontinuous and non-linear. Combined with the special value of 0 (ignore strength), it's very confusing.

Monsen

Posted By: WyvernAs a non-technical-specialist, would that not mean it would be better to set the Strength rating to a series of numerical values that actually represented these non-linear step-changes?

It actually used to be represented in the dialog as values of 0-8. I think it was changed to percent because it is a bit more intuitive to people what it represents, although as has been pointed out here, this also have issues.

jslayton

Posted By: MonsenIt actually used to be represented in the dialog as values of 0-8. I think it was changed to percent because it is a bit more intuitive to people what it represents, although as has been pointed out here, this also have issues.

It would be nice if such things were kept in a reasonable space and done with a slider (even a popup one) to avoid the mystery dead spots in the range. Labeling the slider with ranges like "maximum", "a little", "more", "a lot" (or more appropriate values) would be helpful, I think.

Monsen

Well, you have the power to do so

Shessar

Light bulb!

This has been another of the enlightening (and useful) posts on the forums. I love when jslayton drops in to educate us.

Wyvern

Posted By: Monsen

Posted By: WyvernAs a non-technical-specialist, would that not mean it would be better to set the Strength rating to a series of numerical values that actually represented these non-linear step-changes?

It actually used to be represented in the dialog as values of 0-8. I think it was changed to percent because it is a bit more intuitive to people what it represents, although as has been pointed out here, this also have issues.

Thanks Monsen. I thought that had been the case, but couldn't recall if it was this or another Effect I'd seen it in.

Maybe a little graphic diagram and/or a text line note above or beside the Strength-value box would help?

Or do we just make the sort-of in-joke, When is a percent not a percent? When it's a Glow Strength value...