Yep, most surprising here is how good FR-2LE is.

@Vitaliy_Kiselev

Fostex makes good stuff. The DC-R302 doesn't get much love, but I think the front end is very clean compared to any of the other camera-mount sound recorders. Its not SD build quality, but what is?

So the question is, how good is the comparison, and how useful? As far as how good it is, on the same site, using what appears to be more scientific methods, the SD 722 has a measured ein of 128 instead of 130 http://www.avisoft.com/test/noisefloors.pdf Since the theoretical limit of noise is in any case 129.5, any figure over 129.5 should be regarded with a raised eyebrow. Meaning it is BS. "Noise floor below the theoretical limit!" Wow.

So the answer is, it isn't good--that is, the tests aren't something that have a useful margin of error.

The second question, how useful--useful for real work--depends on on how well the units mesh with particular mics, whether you will get a bit less noise with a standalone preamp, how "good" the sound is and whether the noise exceeds the ambient noise floor.

Will you get a bit less noise with a standalone pre like a grace audio? Yes. A very small difference.

Will you get a slightly better sound? Yes, small difference. Note that moving your microphone one foot will make a bigger difference than buying an expensive box. Really.

Are many of the boxes tested above quieter than a typical noise floor (i.e., the noise figure is irrelevant, since the noise will be completely covered)? Yuppers, they are quieter than the noise floor. The mic makes more of a difference than the pre, especially the impedance.

Which one sounds the best? That's subjective. The Fostex is good enough for professional audio work, that's for sure, whether you like the sound is subjective. Most people will buy the wrong one, don't ask me why.

What is missing? What is missing is that many modern systems use a hybrid design in which the AD is coupled directly to the pre. In this case, not only is it impossible to measure the pre noise, even if you could it is not a relevant figure.

Why does the gain knob sometimes get really, really noisy when you turn it up "high"? This is a basic part of the whole preamp BS. You do your research, you read the specs, you buy it, you plug in your mic, you crank the volume and the top end is garbage. And we aren't talking "1dB" of garbage, we are talking dumpster level garbage in the sound, not just noise but weird harmonic colorations, plasticky sounds. Well, that's because a lot of these figures are BS. Shocker. BS in a budget box? Yup. And not just in the budget box. I tested a top brand once--a $4,000 unit--that was junk.

Any pres you can really crank? Absolutely, the digital pres in some of the RME products (Micstasy, FF400), just turn them up. Some of the high end pres have "real gain" as well. You can also pair what you have with a high output mic like the Sennheiser MKH 40, which will give you not only a lower noise floor but an extra 6dB of gain. Again, you won't need it on a live set because the noise floor will be greater than the noise in your gear.

How is noise measured? http://209.68.17.232/appnotes-a/mmpnoise.html Is the "1dB" figure important? Yes, and most pres neither have it nor advertise it. Not only that, if you call them up they would not be able to explain it.

Is there a true audiophile solution that won't break the bank? Actually, there is. You can buy a Grace Stereo pre, a Lavry AD11 or an RME Digital pre and use it as the Front End for a Tascam DR 680. You will get two audiophile channels plus the rock steady clock for your track timing. The Lavry has a real, workable soft-limiter if you need such things. You can try it for a year, and if you don't like it when you sell the pre on eBay you will get your money back, or maybe even more than you paid for it. Cheap for what you get.

That graph is not SNR vs. gain. It's EIN vs. gain.

Hate to beat a dead horse, but a single gain setting can't possibly give the best SNR for all input signals. If you have clipping at that gain setting, a lower gain setting is surely better. If the recording is very quiet with lots of quantization noise at that gain setting, a higher gain setting is surely better. Internal noise is just one source of noise, and not the only one that affects the SNR.

The one gain setting minimizes the internal noise and maximizes the size of the recorder's dynamic range. But maximizing the recorder's dynamic range without regard to the input level only helps you if you can freely shift the level of the input signal to fit the recorder's range, with an external attenuator or pre-amplifier. If you have a high quality external pre-amplifier, then yes, set the H1's gain to 37 and don't touch it: just use the pre-amp's gain control.