In my own recording experience, I've been able to hear some distinct differences between the mics I've used, but it's often taken a few listens. As a result, I've picked a setup and just trusted that it was a better choice than other options. Since I just got a new round of equipment to upgrade my recording gear across the board, I realized there was a good opportunity to make a set of reference recordings to compare different setups. The room I used isn't anything special, the mic location is nothing revolutionary, and the content being played isn't beautiful music, but I was able to record that same single performance with eight different microphones, and seven of them were clustered in the same location. While there are plenty of more accurate, cleaner, or more controlled experiments that can be done to evaluate recording equipment, but there's a lot to be said to a direct sound comparison of a single event.

The setup:

To record I used:

A Behringer UFX1204 Mixer - recording individual channels over a USB connection. No equalizer or FX processing, gain set to make each input approximately the same volume level.

A Rode NT1 large cardioid condenser microphone connected to the UFX1204.

An MXL 2003 large cardioid condenser microphone connected to the UFX1204.

An Audio Technica u853r cardioid condenser connected to the UFX1204.

An Audio Technica AT853Rx cardioid condenser connected to the UFX1204.

A Zoom H4n portable recorder using its own internal microphones as well as both external inputs.

An Audio Technica u853r cardiod condenser connected to the H4n.

An Audio Technica AT853Rx cardioid condenser connected to the H4n.

The internal microphone array from my Surface Pro 4.

Every recording was made at 24 bit resolution and 48kHz sampling rate, though Audacity saves the .wav tracks as 16 bit PCM at 48kHz, and the video uploaded is encoded with H.264 which youtube also processes. The tracks available here are encoded as 320kbps VBR mp3s, which to my ears are indistinguishable from the wavs they were generated from.

Speaking was done about 4 feet in front of the microphones which are about 5.5 feet off the ground, but the Surface Pro 4 was off to the right some distance. The playing was done seated from about 15 feet directly in front of the cluster of microphones, with no music stand.

To make the cuts between microphones in the video mix, I aligned all the tracks, then muted all but one track, changing which was unmuted every few seconds for a direct comparison between mics mid-sound. The resulting set of tracks full of holes was then mixed down into a single stereo track. Because of the volume difference going from 8 recorded tracks mixed together into one, I boosted the single-source speaking portions by +12dB and the bassoon portions of the tracks by +4dB. The silence section was further boosted by +10dB to demonstrate the difference in noise floors without being eaten by noise in the playback devices or compression by the codec. Because I didn't leave enough silence in the original track, I copy and pasted several seconds to give two seconds of recorded silence per mic used. The individual tracks available here have had no gain adjustments and did not have extra length to the silent section.

The recordings:

Video which cuts between every microphone for immediate comparisons:

Individual audio tracks:

From UFX1204:

                         Rode NT1    

                         MXL 2003    

    Audio Technica u853r    

Audio Technica AT853Rx    

From H4n:

    Audio Technica u853r    

Audio Technica AT853Rx    

                  Internal mics    

From SP4:

          Internal mic array:    

Direct interface comparison: same mic on each interface

          u853r on UFX1204 then H4n, alternating:    

      AT853Rx on UFX1204 then H4n, alternating:    

Analysis:

First off, talking about sound is difficult, so I'll do what I can, but it may read more wishy-washy of a difference than it actually sounds.

In terms of interfaces, the UFX1204 is the clear winner. The H4n has much higher noise floor on its preamps and the SP4 masks the noisiness in software (explained later). It's worth noting that to compare the noise floor between interfaces, the last two interface comparison recordings of same mics have the silence portion of the recording boosted by +20dB. Before hearing the interface comparison tracks cut to specifically compare them, I heard a little difference but I found them to be roughly equivalent. After hearing the same kind of mic on each interface, there is a clear color difference between them in addition to the noise floor differences. This is best shown on the u853r comparison track, the H4n is clearly brighter in tone nearly across the range, and there are a few notes that don't sound nearly as full. The AT853Rx is a bit of a different case, and the mics probably have a little difference between them contributing to that, but the UFX1204 seems to be favored slightly in the midrange and definitely on the top end, whereas the lows sound just a bit better on the H4n - sort of the opposite of the lows on the u853r. There is slightly more variability in timbre over the range, and while the noise floor is louder on the H4n for the u853s, it sounds nearly indistinguishable on the AT853Rxs - probably a product of the noise reduction in the amplifying electronics used in the more recent u853r. There is difference in the preamps in terms of color, but I suspect the individual mic difference (especially between the AT853Rxs) and the slight position difference (10” or so) also contribute this, more testing and matched pairs of input mics could determine for sure.

I used the Rode NT1 as my sound reference in the comparisons of mics, because according to the spec sheets, it has the most even frequency response (by a decent margin) over the range and has the lowest noise floor. While I'll compare mics to each other sometimes, if I just make a claim about their sound, it's referenced to the NT1 and my own perspective of the sound.

The MXL2003 felt like it had slightly less depth of sound – a difficult thing to describe and something better felt than really heard – but I found that positioning my head between my speakers while listening, there was a fullness of sound especially in the lower register of the bassoon that was present in the NT1 more than the others, including the MXL 2003. I don't think this is an artifact of picking up poorly because it is a similar sound to playing a full note, and I don't think it's because of volume because the Rode ended up being somewhat softer in the mix. Comparing the MXL to the Rode, I think the MXL sounds very slightly “thinner” in sound, it has a slight ring over top (not high harmonics, but almost reverb) of the sound that isn't present on all mics and doesn't sound correct to my ears, and it has a slightly raspier lower register tone than the Rode. Remembering from comparing this mic to the AT853Rx in previous recording sessions – it has a warmer general sound that I associate with large diaphragm condensers, but it has a less even frequency response, where certain tones stand out where they aren't louder in person. The MXL 2003 has the highest noise floor of the discrete mics used, which isn't so surprising given that its specs are about in line with that and it's in a price class lower than the others.

The u853r seemed to have a little less sound presence on the lowest end of the spectrum and generally sounded more even than the MXL 2003. It also had a little bit of that same 'ringing' sound over everything..... not exactly reverb, but there seemed to be some artificial resonance, or perhaps it picked up more of the room than the NT1 – not that the room is particularly live or good. I don't think the ringing was a good addition to the sound in terms of accuracy, but it didn't bother me a lot in terms of tone and sound color. The AT853Rx is effectively a couple generations earlier version of the u853r, so unsurprisingly, it was a little worse in most categories. The sound overall had a slightly brighter, almost raspy quality – mids were more “abrasive” than I thought they should and certainly more than the NT1. Oddly, this mic also sounded the most even – I think there's some kind of dampening that I'm hearing that takes the edge off of many of the louder sounds, so going up a chromatic scale, for example, each note sounds slightly more similar to the surrounding ones, but I don't necessarily think this is accurate to live sound. The bassoon is directional enough and has enough inbuilt differences in resonance that I think this evenness doesn't quite represent what was played. In the same way the MXL2003 felt like it had less sound depth than the Rode NT1, the AT853Rx feels less full than the u853r. The AT853Rx has a slightly higher noise floor than the u853r, but both seemed to be fairly good performers.

The internal mics on the H4n sound fairly clear, but their low end seems very slightly lacking and they are overall just a little bright or metallic sounding. I think these mics have the highest noise floor (and are attached to the higher noise floor preamp) and are slightly uneven across the range – certain notes stand out more than they probably do in person. With the H4n's preamps and matched Audio Technica mics, I did notice some slight color differences aside from the preamp noise. I felt the u853r had slightly less fullness and presence of sound on the H4n than it did through the UFX1204, and the AT853Rx seemed to have a slightly brighter tenor range on the H4n than the mixer. These could be slight differences in the same mic, there's enough of those characteristics to all of the mics the H4n was recording, I'm comfortable saying that the H4n itself probably had a hand in that sound difference.

The Surface Pro 4's internal mics did a lot better than many integrated and cheapy headset mics could do, but they were dramatically lacking when compared to actual recording equipment. The low end and midrange sounded very muddy – up almost until the top range of notes played – and some of it almost sounds physically obstructed – like there's a layer of something between the mic element and the incoming sound. There was also much worse response in the low end, not only a rolloff, but simply not having good performance below say 100-150Hz. Nice to have as a comparison, but avoid recording on it. I know from previous experiments that it's SPL tolerance is much lower, too, trying to do some simple dubbing while recording from my music stand, I could not lower the input gain to a point where it wouldn't clip playing over a middle dynamic. As a contrast, I had the AT853Rxs within about 5” of the tone holes of the bassoon as a previous project and had no problem getting usable sound of them.

It's less useful than a straight sound comparison, but I noticed some visual differences in the visualization in Audacity, and I figured I'd highlight a few. This pictures the chromatic scale section of the recording. The color coding on the left shows you the mics on the same recording hardware and then each pair of matching mics on different hardware. The red boxes show some visual differences between mics that should be the same, bur which are on different interfaces (and I think individual differences and interface differences contribute to this) and the green boxes point out some other inconsistencies. I tried to look for things that did the opposite of other tracks – where one mic displays very wide and another much narrower – and while most of it is the same, there are some interesting differences that you can probably hear if you match up the timestamps. Probably the most interesting of the green boxed sections is right at the end of the scal, on the second or third to last note. Both large diaphragm condensers respond to it well (as wide as the final note), but the only other that does is the AT853Rx on the H4n.... not the one on the UFX1204 and not the other mics. Not sure what it means, but it's an interesting finding. It's important to note that this visualization does not necessarily show volume – different notes will respond differently and have a different look – but often big differences in look can still indicate some kind of meaningful difference of sound.

Results:

Taking in everything, I think my upgrades (from the H4n, AT853Rxs, and MXL2003 to the UFX1204, NT1, and u853rs) are worthwhile, and I think every new component performs better than the last. To my ears, the NT1 sounds the closest to live sound and the most even, and it has a very low noise floor. I think the u853r takes second place but doesn't quite have the same performance of the NT1, even if the color it adds isn't a bad one. The MXL 2003 has a nice sound, which I associate with large diaphragm condensers, but doesn't sound as refined in terms of tone and has a few nodes where it responds outside of where it should. I think the MXL 2003 colors the sound the most strongly of the discrete mics, even if it's generally in a desirable direction. The AT853Rxs seem to be generally unobjectionable, but slightly worse performers than the updated u853r. There is some aspect to them, or at least, some of their individual mics, which makes me want to put them about on the same level as the u853r in terms of overall sound quality, but it's not consistent between both mics and the noise floor and low end response does favor the u853r. The H4n has decent internal mics that are a little too bright and a preamp that is noisy enough to be audible when you turn the volume up. It's great for smaller time portable recording, but it does not stand up to a proper set of equipment. You shouldn't be expecting professional quality sound from any integrated computer microphone.

Curiosities:

There were a couple of interesting things I found in recording which I felt were worth documenting. First, in setting up the recording, Audacity only supports recording one stereo channel at a time – even with multiple instances of the program. Proper DAW software will do many more and the UFX1204 will let you record from any channel individually over USB or primary/aux mixes, but proper software costs something, and especially without experience with them and needing something while setting up, I wasn't willing to throw down the $100 or so it would take to get a good one (at least yet). I ended up using KRISTAL Audio Engine (which is still available for free for personal, noncommercial use), which can be a little finicky, but which properly supports multichannel recording. From the generated file, you have to “bounce” each track individually to export them as wav files unless you want to do all of your editing and mixing in the program.

Timing between tracks was a surprising issue to encounter, but it was apparent in a couple of ways. One, in KRISTAL, I recorded the SP4 mics and the UFX1204 inputs, and there was a slight delay visible in the waveforms – the SP4 mic was behind. I believe this is because of the audio software the input was using, as the UFX1204 used the ASIO interface while the SP4 mics were on MME. Manually aligning the track fixed that timing issue. Perhaps the more interesting one, though, can be heard in this mix of the final spoken portion of the recording:

Basically what you are hearing is the timing difference between the UFX1204/SP4 USB timing crystal and the one built into the Zoom H4n. Including setup, testing, and redos for spoken portions, the total recording made was about 12.5 minutes, and in that time the slight difference in frequency exactness of the timers between recording setups added up and manifested in a delay. For my purposes it didn't matter, the difference is tiny so the cuts don't sound problematic and I could just mute the H4n mics in the mix for a clean sounding ending, but it could be an issue with larger scale recording setups or longer tracks. Make sure you record everything for your mix on one device, to be sure.

Another interesting bit is the noise floor on the SP4 built in mics. Given their environment and the more budget nature of them vs. the semi-professional and professional equipment used for other interfaces, they by all means should have the worst noise floor of all of the recordings, but they are actually dead silent (basically no movement when zooming in on the track in a visualizer). My guess is that there is a software check that looks for a low-sound point that's just over the system's noise floor, and if the sound level drops below, it simply mutes the channel in software. This gives you a simulated zero level noise floor when everything is quiet, but this does not give you the same performance when actually listening to something.

August 11, 2016