Two comments leap out from here - if you put the values all at 33% on the channel mixer and select monochrome, you're keeping the values of R, G and B at the same equal balance - you might as well just make it greyscale, that's all it's doing - the real power and flexibility of this method is when you mix the balance of the colours - you need to end up with a set of values that approximates to 100 - so red @ 20, green @ 42 and blue @ 38 would give a grey scale with different tonal values for diffrent areas of the image, but overall, the same level of darkness and lightness.

If your three totals make more than 100 your image will end up lighter overall, if they're less than 100, your image will be darker overall, so you can use this creatively too - so as you can see it's very powerful. My favourite starting combination is red 28, green 41 and red 31 - to which I then add some contrast and a very large radius of USM. I also have other combinations for particular types of image - and if they don't work, I just move the sliders until I like the result.

The second thing was Arn's comment about using it with an HSL adjustment layer is to have two layers on top of your image - an HSL adjustment layer, then a channel mixer layer. By tweaking the colour in the original by using the hue slider in the HSL layer to change the colours, you can get very different results in the monochrome - the two together are infinitely flexible.

The method Audiophiling linked to is known as the Russel Brown HSL adjustment technique and is a variation of the same thing - it's another I use extensively too - Arn's suggestion was to combine this with a channel mixer to get the power of both methods together. In his suggestion, he was subsitituting the basic desaturation layer in the tutorial linked to - the second HSL with saturation at -100 - with a channel mixer layer to make it even more powerful and flexible - this is something I do with tricky images too.

I very often boost the saturation in a colour image before doing a monochrome conversion - although that sounds counter-intuitive - because the more you make distinctions between the colours, the more tonal distinction there will be when black and white - it can really bring out detail in a flat monochrome image.

Some images simply work better with one type of conversion and others with another, it depends entirely on the colours and nature of your original and the result you want. You'll find that all of us that love doing black and whites will have an arsenal of techniques we use regularly - there's no wrong or right way to do it, but more to play with than you can ever need. Actually there is a wrong way - using a simple grey scale or de-saturation without anything else - it's throwing away so much creative potential.

But the core of it is to manipulate the colours of the original so that they change the tones of the resulting black and white until you get what you want - it's mind-boggling really.

Either take it at high ISO - or add grain later. I tend to do the latter when I have images that look lumpy when finsihed - I do a lot of high ISO work with my live music stuff and some images that might have been underexposed for example can end up with uneven noise - where even grain looks nicer, so if that's the case, I just make the grain a feature by adding some - it's also useful if the image wasn't that sharp as you can sharpen quite hard before adding grain and it can rescue a poor image.