Perhaps we need to take a step back - Are you able to record 24-bit data on Linux with your setup?
And how would I determine this (from within Audacity)? Record silence and then try to find the smallest non-zero sample?
I can with arecord, but if I have to go basic like this I could just as well get me something like a Zoom H1n…
I’m currently hacking around in the source, trying to get the embedded PortAudio copy to provide the actual sample format along with the other stream info it already provides.
The important aspect of the sample format (from the point of view of sound quality), is the dynamic range. Can you achieve better than 96dB dynamic range?
Initial results aren’t very reassuring; I have yet to figure out how to figure out the exact settings being used by a stream. For now it seems I’m always getting 32-float or 32LE returned even for my external SoundBlaster which doesn’t support those.
“This is the Signal-to-Noise Ratio of the line out (output) of the device. The SNR is defined as the ratio of signal power to the noise power, often expressed in decibels. A ratio higher than 1:1 (greater than 0 dB) indicates more signal than noise.”
but unfortunately does not say what the recording input dynamic range is.
The input dynamic range is usually less than the line out dynamic range - if that was not the case for the Play! 3 then I would expect SoundBlaster to be shouting about how terrific the recording input dynamic range is, rather than missing out that specification entirely.
Unfortunately this means that PLAY! 3 is not capable of more than 16-bit dynamic range. Even if it is producing 24-bit samples, the bottom 8 samples are noise.
Anyway, before I run out and buy a better interface I’d still want to get answers about my standing questions…
Out of curiosity, do you know what the dynamic range is of the typical laptop (HDA Intel/ALC269VC) audio subsystem? Common wisdom says you get noticeably better sound playback quality by using an external interface so I presume the same is true for the mic input?
The maximum theoretical dynamic range for 16-bit data is usually quoted as 96 dB.
That’s based on the fact that the range of signed 16-bit is -32768 to +32767 (= 65535), which works out as 96.3295 dB.
The quick and dirty rule of thumb is “6dB dynamic range per bit”. So for 16-bit:
16 x 6 = 96
24 x 6 = 144
In practice, for bit formats of 16-bit or more, the analog noise floor is usually the limiting factor rather than the format.
Out of curiosity, do you know what the dynamic range is of the typical laptop (HDA Intel/ALC269VC) audio subsystem? Common wisdom says you get noticeably better sound playback quality by using an external interface so I presume the same is true for the mic input?
Actually, playback from a “regular soundcard” is often better than human hearing. If you are not hearing noise an interface is unlikely to sound better. (Distortion and frequency response are almost always better than human hearing so noise is the only thing to be concerned with.)
With recording, noise is again the main issue. If you are recording line-level signals a regular soundcard may be OK but if you are using microphones, regular soundcards are not compatible with stage/studio microphones (balanced XLR connection), nor are they compatible with a direct electric guitar connection. So for high-quality microphone recording you need a good mic and a proper audio interface.
For educational purposes: what’s the dynamic range of 16 bit?
With digital you get quantization noise (at -93 or -96dB I think). If you’re recording “at home” from an analog source your analog and acoustic noise is almost always worse than that. But you can do better than that in a soundproof studio and good equipment.
If you want to hear quantization noise, you can hear it if you make an 8-bit file. (At 16-bits or better it’s inaudible under any 'normal" listening conditions.) It’s like a “fuzz” that rides on top of the signal. Like analog noise it’s more noticeable with quiet sounds but unlike analog noise it goes-away completely when there is digital-silence.
It’s always been my understanding that built-in soundcards are more susceptible to pick up all kinds of interference from the other components on the motherboard.
If you’re recording “at home” from an analog source your analog and acoustic noise is almost always worse than that.
Noise on the input will not get worse if you sample it at a higher resolution, but if what I remember from signal treatment is correct it should be easier to clean that up properly at higher resolutions.
It’s always been my understanding that built-in soundcards are more susceptible to pick up all kinds of interference from the other components on the motherboard.
That’s true… The computer is a “nasty environment” for (analog) audio. But if you’re not hearing noise it’s not a problem.
BTW - USB power also tends to be noisy and sometimes and with a USB powered interface that noise gets-into the analog electronics. An interface with its own-separate power supply is “safer”.
but if what I remember from signal treatment is correct it should be easier to clean that up properly at higher resolutions.
Top
Some people believe that… I don’t. (I’m not a DSP expert but I know a little about it.)
If you have 24-bit hardware there’s no harm in recording at 24-bits. Pros record at 24-bits and you can record at a low-level while still retaining more than 16-bits of resolution. …You do loose resolution when you lower the recording level, but it’s not a usually a problem unless the level is VERY low. Even with a little loss of resolution it’s still better than human hearing, plus the usable resolution is usually limited by analog noise.
\
The "critical thing"s are NOT digital…
If you are recording acoustically you need a “good” acoustic environment. A soundproof/sound absorbing studio or a music hall with good acoustics and no outside (or inside) noise.
You need a good performance, and a good instrument if it’s a musical recording.
You need a good microphone and good mic placement.
You need a good preamp (often built-into the interface).
USB power also tends to be noisy and sometimes and with a USB powered interface that noise gets-into the analog electronics. An interface with its own-separate power supply is “safer”.
And that brings us to the “Yeti Curse.” You get into the worst troubles when both the microphone and the computer cut corners at the same time. The computer produces sloppy or ratty USB connections because filtering and regulating is expensive. The microphone is sensitive to sloppy USB connections because filtering and regulating is expensive.
The problem is so popular that we designed a filter specifically for it. Mosquito Killer. It deletes common tones from your work. There is no easy way out of that. Change the computer.
Or stop using the computer. There’s no shortage of people using the Zoom series of sound recorders including me.
And that brings us to the pros using 24-bit. Yes, but they don’t do it on a computer or system they got at Best Buy, and they’re doing it in a soundproof room with no echoes.
I can’t seem to find any specs for the SNR or dynamic range of the H1n; its Tascam equivalent seems to be 93dB so a “fake” 24 bit solution that’s also very sensitive to radio interference. I’ve been looking at the Olympus LS-P1 too, at least it has a metal housing, but again I’ve not yet found an indication of its SNR or dynamic range.
I did find some info about the iPhone quality: 17 bits effective via the TRRS jack, or 16.5 via the lighthing2jack dongle (iPhone6 and probably for playback). At least I could use it as a “cheap” alternative.
My microphone is powered through the preamp which has its own powersupply that I connect via a filter. As to my recording environment; a wood-paneled attic room in an old house with heavy limestone walls in a tiny village where it’s usually dead quiet. With a radial LDC at 40cm max of the source I hardly hear my Mac’s fans in a recording (I do hear them myself because they will spin up when Audacity is recording). It’s not a pro studio but it’s good enough for me.
The H1n is a sound recorder, not a voice recorder. It doesn’t have the high pass filter and loudness processing permanently burned in (they’re available) and it will produce WAV sound files. It’s not MP3-only.
You can run it from wall power.
There is a common error with these. No, you can’t use it as a USB microphone. It will do that, but that puts you in the camp of the people recording USB connection noises.
Are you voice recording? I’m not sure we ever hit that in the blizzard of technical specifications.
very sensitive to radio interference
In general, RF radiation isn’t a big deal unless you’re trying to record in an intense radio field. I was once called in to trouble-shoot a noisy movie shoot. Turns out the movie company got a good deal on a studio located here.
There’s a rule for that. You should not be able to tell your computer is on by listening. Kiss of Death. Effect > Noise Reduction can get rid of gentle fan noises, but it can’t do anything about them if they change speed.
USB cable maximum reliable length is 6 feet (2M).
The audiobook people have a different noise source everybody forgets. I read from paper, but if you read from your phone…
Yes, I totally turned off the ringer and notifications, etc. That’s screen radiation. iPads do it too.
Nope. And if I were I’d probably treat the voice as another instrument (just like I try to sing with my instruments ^^) and apply whatever filtering is necessary downstream. Except for pop screen and family.
Entrée micro 1 (XLR) : dynamic range : 104 dB (A-weighted) SNR : 104 dB (+1.0 dBu, A-weighted) - Entrée Line 1 (Jack TRS 6.35 mm) : SNR : 101 dB (1 kHz, +4 dBu, A-weighted) - Entrée instrument 2 (Jack TRS 6.35 mm) : dynamic range : 100 dB (A-weighted) SNR : 102 dB (+ 4 dBu, A-weighted) Idéalement un préampli micro est à branché sur la connectique combo XLR/Jack (Input 1)
BTW, I’m not certain where in practice Audacity uses the bit depth quality setting in practice. If you just use the app as a recorder it will always ask for 32 bit float, regardless of what the input device is capable of. I still haven’t figured out how to obtain the actual parameters being used by ALSA (there’s a function to fetch them but it returns an error for me), but I know how to do this on Mac. And I get “32 bit float” even if the device is not capable of that. So apparently there’s a conversion taking place inside the CoreAudio framework already. Fortunately a trivial and non-lossy one in this case.
Listening to the recording or listening in your workplace? I can usually hardly tell if mine are just on by listening, but I typically can tell if they’re working (for me) or not
