mostly what he said. if the concern is FR, then almost anything will work fine and some choices will simply be more convenient or offer closer repeatability. a FR is about measuring a fairly loud signal, that you can rapidly learn to set according to your mic and room noise. and once that is done, you compare 2 of your headphones by measuring their relative FR on your rig. that will almost always work fine and give you a pretty accurate data on the relative variation between the 2 headphones at a given frequency. a 5$ mic and a cardboard box can give you similar and valuable informations.
@GREQ mentioning modding is a very good example of something you can really benefit with pretty much any measurement rig.
now, if you expect any given rig to tell you what is neutral, forget it, that won't happen. the first hurdle being that there is no neutral for a headphone. on your head, you're getting something different from the same headphone on my head. and in the event that our eardrums get the same FR, chances are that we will still interpret them differently based on our own HRTF. even the guys making the most expensive dummy heads understand that, so at best what they can aim for is a curve that's close to an average of many humans. and some such dummy heads are so famous and massively used in the industry and in research that their calibration has kind of become the de facto idea of neutral. so others(like me sometimes), try to copy that and their various compensation curves, but objectively speaking, it's both bad and fine. bad because those references usually mean nothing to our rig acoustically speaking. and fine because measurements are arbitrary stuff by nature, even the choice of a unit was an arbitrary decision made for convenience or whatever. my point isn't to judge, only to point out the reality of the situation so that you are conscious that just because we end up with a graph and apply some compensation to it, we do not suddenly become the holders of the truth about neutral.
if you aim for other measurements, THD for example is going to manifest some 40dB or lower(hopefully lower), below the test signal on most headphones. so already we need something to accurately capture air pressures that much smaller compared to the initial signal that would be captured for a FR graph. obviously some expensive stuff might be more sensitive, have lower self noise, etc. really cheap stuff won't give very significant graphs unless the distos are really loud. we also have to deal with ambient noises, which is a real issue for me but has nothing to do with money(well in a way it does, I could do with a little isolation box like they have at Head-fi's headquarters ^_^).
I mostly fool around with IEMs and the E.A.R.S turned out to be a pile of crap for that purpose(half my IEMs simply won't fit on that BS pinna, the IEMs like etymotic used with deep insertion can't be measured in situation because there isn't enough "ear canal", etc). so I haven't used the E.A.R.S a lot beyond enjoying a new toy when I got it and testing stuff out of pure curiosity with my headphone. for most stuff, I get better results with a sub 100$ mic. but certainly the E.A.R.S are very convenient compared to trying to use books, cardboard, sponges, and really anything I can put my hands on to try and have a mic hold in place in front of a headphone in a stable and repeatable way
. different uses will bring up the qualities and flaws of each solutions IMO. and to get back to "neutral", out of 4 cheapo makeshift rigs(one being the E.A.R.S), I get 4 different FR graphs. once I apply a compensation to make them all show the same one for a given headphone, I do get more consistency with other headphones, but they still had a tendency to show inherent differences(caused I guess by placement, pressure on the pads, difference in the fake ear canal length, etc). and with IEMs it's much much worst and I can't get any pretend consistency between my crap rigs for more than maybe one out of 3 IEM at best. the variations in the coupler changing how much resonance and at which frequency it will occur, cause obvious and hard to avoid variations.
but as I said at the beginning, if we put aside that fantasy of having a copy of some professional rig for a hundred bucks, FR graphs work fine for me even on the crappiest and cheapest stuff. I don't mod much but I EQ a lot, and I can clearly know how much the FR really changed when applying some EQ to the test signals. the amplitude variations are accurate from one measurement to the next, it is objectively good. just not good for everything.
for headphones on the E.A.R.S, I tried to add a layer of stuff between the ear blocks and the metal, to increase the ear canal length, but also to get a bigger "head" as I find the original spacing a little small compared to how a headphone holds on my own head. but I couldn't achieve a good enough seal between the surfaces in a way that was reversible, so I gave that up. then I thought that maybe I could at least move the mic capsules back inside and use some plastic tube or whatever, that's when I pulled out a contact on one of the capsules(the soldering job is almost as bad as what I can do, and I'm crap at it). so after those 2 failures, I lost my motivation to turn that device into something more practical for myself. maybe I'll get back to it someday.
2- If I have to build a custom measuring gears, do I need a dedicted mic, amp, dsp, etc or I can just buy the mic and use my computer to do the measurment?
again, for FR graphs, DAC and AMP choices are pretty much irrelevant. for other stuff, usually it still doesn't matter, but if you start using a 4%THD tube amp to measure headphone distortions, that might not go great ^_^. so I'd say it usually isn't important. again if you fall into IEM measurements at some points, then the amp becomes a little more important as its output impedance or maybe some protective caps could significantly affect the FR of some IEMs. with headphones, they tend to have more even impedance response and higher impedance values, so any potential impact is minimized.