How to record "old timey" music?

[clepsydrae]

Originally Posted by drtedtan

The benefit of using a lower sample rate is less latency (delay) in hearing what you play when you are using a keyboard to control a software synthesizer and so forth.

When the OP said "I had my driver set for 256k sample rate and 3ms buffer length", OP used the term "sample rate" but (as I'm sure drtedtan knows) he was erroneously referring to something like "number of samples in the buffer", as 256k is obviously not the sample rate. If the sample rate was 88.2k, then 256 samples is 256/88200*1000 = 2.9024943ms. OP: often when setting buffer sizes the UI will specify it as number of samples, or milliseconds that that number of samples results in at a given sample rate, or both.

drtedtan is right: the buffer size and latency have nothing to do with audio quality, apart from possible dropout issues. For what you're doing, you might as well set the "buffer size" aka "samples in the buffer" aka "latency" to the max, just to minimize risk of dropouts.

In terms of the actual "sample rate", if you were in fact recording at 88.2k or something, you might as well set that to 44.1k and forget about it. Opinions vary on that issue, but I think you'll hear no difference between 44.1k and anything higher.

It won't affect the recorded quality, but you will hear pops/clicks in the playback if you set this too low. Note that this is on playback only; these pops and clicks aren't actually recorded into the wav file.

If you set the latency too low, you will get dropouts in the recorded WAV file as well.

[wavedrone]

I'll second the suggestion of using a ribbon mic. If you want to try a cheap one, Musicians Friend has the MXL R40 on sale for 69 bucks.

http://www.musiciansfriend.com/ribbon-microphones/mxl-r40-ribbon-microphone?source=3WWRWXGP&gclid=CM2d39u788ECFbA7M godWTsAyg&kwid=productads-plaid^83572311147-sku^H69305000000000@ADL4MF-adType^PLA-device^c-adid^53736456387

I own a couple of these mics, and they are pretty decent. They are about the only MXL offering that I have used that I would recommend.

[hzandbits]

Hi - while I don't play banjo myself, I do have a little experience recording clawhammer banjo. The guy I recorded played rather quietly (with his fingertips), which kept the piercing transients in check and sort of allowed more of the "head" sound to be heard. Gave a nice hollow sound, with a touch of...you know the sound you get playing with brushes on a coated snaredum head? To me, it seems very much a matter of playing technique - and micing technique. Getting those two factors right (and in that order) should get you most of the way there. Processing after the fact...well...

One time, we re-amped a whole mix through an old RCA BX74 (cheaper and darker sounding than the famous 44) so cheap ribbons may help.

[SaulT]

Opinions vary on that issue, but I think you'll hear no difference between 44.1k and anything higher.

Personally, I feel that if you can go for a higher sample rate, then do it. It makes less of a difference if you're just recording things, but once you start processing, especially when you're talking about gain or compression, the higher sampling rates give you more headroom to avoid aliasing.

6. Watch those transients - they may not show on meters as they are big and fast!(?)

The specs on traditional PPMs (5-20 ms) and certainly the specs on VU meters (in the hundreds of ms) give response times that are too slow to track sharp transients, so that's probably what they're talking about.

I set the recording levels at around -18. I read that in an article about recording in 24bits .

The most important thing is to not clip. Record high enough to stay well above the noise floor, and give yourself enough headroom to avoid clipping by a few dB's. I personally aim for around -15 and peak around -6 to -3 at the absolute highest, because I'm trying to stay as high above that noise floor as possible.


So I'm seeing a few common suggestions -

* change your mic position
* change your playing style/playing tone
* change your mic
* try some sort of ad hoc room treatment
* add some form of input compression

The first three have been covered, but I would like you to consider the last two as well. Throwing up some blankets, whether on the walls or by making a mic tent/tunnel, is a very cheap and non-permanent form of room treatment. It should allow you to put more distance between the banjo and the mic without negative consequence, and distance is a great form of natural compression. Even a pop filter might be useful to you.

Consider some form of compression in front of the interface. It won't save you from a signal that's overall too hot, but a limiter with a nice fast attack and a ratio somewhere north of 6:1 is going to help those sharp transients and keep you from having to go through and manually correct each and every clip. Ugh, I can't imagine.

[clepsydrae]

Originally Posted by SaulT

the higher sampling rates give you more headroom to avoid aliasing.

Could you clarify what you're saying here? Thanks!

[SaulT]

Could you clarify what you're saying here? Thanks!

TLDR; it helps keep your high-end clean if you distort or compress a lot inside the computer with plugins, and on second thought, probably doesn't apply to you at all.

...

The highest possible frequency you can represent is the Nyquist, or sampling rate/2. Whenever you distort something, you create harmonics, or multiples of that distorted frequency. Basically, anytime you create an "edge" or a "corner" in your waveform you are distorting it, and generating harmonics. Signals that are generated above the Nyquist limit are "folded back" into your frequency spectrum.

Smooth compression generates fewer and lower amplitude harmonics. Hard clipping generates more with higher amplitude.

Let us imagine a guitar that is distorted with a plugin. Its' highest natural frequency is probably around 6 kHz. Let us imagine a sampling rate of 44.1 kHz. The Nyquist frequency is 22.05 kHz, or just above what we can hear (max 20 kHz, and most people can't hear above 17 or so).

6 kHz (fundamental)... below Nyquist (22.05), so we're safe
1st harmonic = 12 kHz. Safe.
2nd harmonic = 18 kHz. Safe.
3rd harmonic = 24 kHz, 22.05-(24-22.05) = 20.1 kHz. Technically we are still safe.
4th harmonic = 30 kHz, and since 22.05-(30-22.05)=14.1 kHz, it may be audible
5th harmonic = 36 kHz, folded back to 8.1 kHz, and it's in very audible range

Let us imagine that we are processing our signal at 88.1 kHz. We wouldn't start to fold back until the 7th harmonic, and it won't be potentially audible until around the 11th or so.... but it takes a lot of distortion to generate that many harmonics!

The Big Question - does it matter?

Only if you're generating a lot of high-frequency harmonics, and you don't get those unless you're distorting/compressing/limiting/using gain a lot, or are doing something like generating square waves.

If I'm just recording and not using very little compression, 48 kHz is probably fine. If I'm distorting or using heavy compression, then I want to use either a higher sample rate or oversampling (which is almost the same thing).

It is entirely likely that 48 kHz is plenty for you - you're recording acoustic with not much gain or compression. It is unlikely that you'll need more than that.

I'm sorry that I brought up such a technical topic that, on further thought, probably doesn't apply to you.

[clepsydrae]

Thanks - I'm not the OP, so we are surely OT at this point, as you said, so I'll keep it short unless OP is interested:

You're saying that since ITB aggressive compression (not to mention intentional distortion) can add high-frequency components, one should use higher sample rates to prevent poorly-written plugins from creating aliased artifacts in the audible range. Good to know about.

I gotta wonder how often this happens with compressors in normal operating conditions, though... which is probably your point in saying that it probably doesn't apply to OP. I.e. do you have any examples of free compressors that exhibit audible aliasing at 44.1 that someone could hear, in comparison to a higher rate?

[drtedtan]

Originally Posted by clepsydrae

If you set the latency too low, you will get dropouts in the recorded WAV file as well.

I was thinking of the tracks being played back when I posted that, but you are right - drop outs in the track being recorded would be recorded in the resulting wav file as well. Good point.

Originally Posted by clepsydrae

I gotta wonder how often this happens with compressors in normal operating conditions, though... which is probably your point in saying that it probably doesn't apply to OP. I.e. do you have any examples of free compressors that exhibit audible aliasing at 44.1 that someone could hear, in comparison to a higher rate?

I think he meant that these artifacts wouldn't be audible with compression, but would be in cases with lots of distortion being applied ITB, e.g. with a high gain guitar amp model in an amp sim or a fuzz pedal, etc.