Try this... open an image as a smart object. Float a new layer using (PC) Ctrl alt shift e
Go back to the smart object and edit however you want - the changes will not be apparent in the new layer. As soon as you have a pixel bearing layer, as opposed to an adjustment layer, any changes under that layer, smart object or otherwise, will not be passed into the pixel bearing layer. Of course you can let some "leak" into that layer by changing the opacity, but that's a different issue.
Here's an example in my work flow. I open the image then create a shadow highlights layer as a smart object (I could just as easily open the image as a smart object from the beginning). Eventually, after 4 or 5 intervening adjustment layers, I want to apply a "local contrast" usm of 20 50 0 to the image, to do that I need to have a pixel layer against which to apply the USM. I float a new layer to the top of the stack and apply the USM to that layer. At that point any changes to the underlying smart objects (or adjustment layers) will have no impact on the top layer. I've seen fairly complex workflows predicated on the requirement to float a new consolidated layer several times during the editing process (eg: Vincent Versace's, Welcome to OZ). If I recall correctly he refers to that useful shortcut as "the claw"
If I'm missing a trick, I hope someone will point me in the right direction...
[{POST_SNAPBACK}][/a]
ahh... I got ya.
I was handling those issues in a similar, equally, well, inelegant way. (I don't think there is really an "elegant" solution for that...) but then I just dove into the deep water, I guess. I do all my USM work on each visible part of the layers, as Smart Objects... they then are "Smart Filters", right, which you can go back and amend later.
I don't really do any global sharpening, just on the spots that need it, so I can hit each area selectively, masking the filter and such... all my cloning and healing likewise.
You can start with a "spotted" Smart Object and just keep jamming copies of it up. (Layer>Smart Objects>New Smart Object via Copy, if memory serves... I've made a keyboard shortcut for it, so don't do it manually anymore.) It will copy all the attributes, so you can take advantage of that...
As long as you can live totally in the Smart Object, you're ok. Here's a peek at the madness that is my workflow... [a href=\"http://www.teddillard.com/2008/08/smart-objects-gone-wild.html]http://www.teddillard.com/2008/08/smart-ob...-gone-wild.html[/url]
...you will, if you look carefully, see a couple of rasterized layers there, too.