budget (usb) mic for classical guitar recording needed

The “front” of the microphone is with the microphone positioned vertically and the writing facing toward the sound source.
“Cardioid” is the most commonly used pick up pattern. It is generally a good choice unless you have a specific reason for using something else.

As expected the TC Electronic Konnekt produces a much cleaner sound than the T-Bone USB Micplug, but at a tiny fraction of the price of the Konnekt I think the Micplug did pretty well. The hiss from the Micplug is clearly audible when listening through headphones or monitors but not too obtrusive when listening on my hi-fi speakers.

From a totally personal point of view;
02-cardioid-front.flac had a shade too much “body” sound but has the lowest hiss level (big signal from the sound hole)
05-cardioid-top.flac sounded more balanced - the bass response of the microphone rolls off as you move off-axis and is compensating for the thing I didn’t like in 02.
15 and 18 show a lot more hiss than the previous two and seem a bit harsh and less rounded in the bass (obvious advantages to using a more expensive sound card).
There is also a bit of “DC-offset” in these last two - this can be removed using the Normalize effect in Audacity.

It might be worth trying the microphone with the front toward the 12th fret but closer than before. There’s no hard and fast rules about this. At the end of the day it’s what you like best, but experimentation will reveal a “sweet spot” that produces the (subjectively) best sound.

Thank you very much for sharing the results with us.

If you want to try the Noise Removal effect in Audacity 1.3.11, here’s a “trick” that may help to achieve some effective lowering of the noise level without introducing horrible side effects:

  1. Record a few seconds of “silence” (and of course record the guitar)
  2. Use “Normalize” to remove any DC offset.
  3. Amplify the “silence” by +10dB
  4. With the “silence” still selected, open the “Noise Removal” effect and press the “Get Noise Profile” button. (The noise removal effect will close)
  5. Select the guitar recording and open the “Noise Removal” effect again.
  6. Set the “Noise Reduction (dB)” setting to about 6 (other settings at default), then click the “OK” button.
  7. Trim off the “silence”.

The “trick” here is step 3 which in effect changes the noise detection threshold (in combination with a low Noise Reduction (dB) setting in step 5).
Without this trick (or a user setting for the noise detection threshold) you will probably hear the noise reduction cutting in and out.

The figures used here are a guide - tweaking the settings may give improved results.
The Noise Removal in Audacity 1.2.x is not up to the job at all.

<<<If you want to try the Noise Removal effect in Audacity 1.3.11, here’s a “trick” that may help to achieve some effective lowering of the noise level without introducing horrible side effects:>>>

This is me writing that down…


I forgot to check the dc offset, but I did try the noise reduction without that trick and the result was exactly that annoying cutting in and out… So I gave up on it… I shall try it again with that trick, thanks :wink:

As for my latest tests… I’m now trying it in a more vertical position with the front facing the guitar (not exactly vertical… with a small angle towards the floor). This way I can get more sound so I can lower the gain and therefore less noise.

I’m also doing some experiments with the mic’s low cut filter which might work well for reducing the amount of bass without any major “side effects”, as far as I can tell…

Also other important thing I found out was that increasing/decreasing the gain on the micplug preamp button has little effect on the noise level. On the other hand changing the gain level on the computer has a great influence on the noise level… so from now on I’m putting the micplug’s button gain level on max and reducing the gain level on computer.

As for linux compatibility I’ve just connected it to my linux pc and it’s working out-of-the-box. I’ve also bought a 10m xlr cable so I can put the mic in a different room than the computer. I’ll have more experiments on that tomorrow. As for the noise level on the Linux seems to be lower than on the Mac laptop, probably because I’m using lower gain on linux than on macos-x (about 50-60% on linux vs 75-85% on macos-x), don’t know why such difference… maybe because usb drivers are different or so… So eventhough my linux pc is much noisier than my laptop it seems that I get less recorded noise on it, but this still needs further testing…

One difference from recording in audacity in linux and macos-x is that on macos-x I can change the mic input level on audacity and on linux I can’t… it’s stuck to max and doesn’t let me change it, I have to adjust the mic input level on gnome’s control panel. Not that this bothers me… but might be of some interest for someone :slight_smile:

I didn’t upload any sample of the few tests we made using figure of 8 pattern but it did have a “brighter” sound than the other patterns. The sound level was a bit lower though so that means more gain and more noise so I didn’t dig much into it… but I might give it another try later once I get other things cleared out.

I’ll keep posting my findings and I’ll soon upload some more samples of my recordings on linux, which seems more promising than the ones on macos-x. BTW the samples I uploaded before were all recorded on audacity for windows at my friend’s pc, not sure which version. I’m currently using version 1.3.11 both on macos-x and linux.

I’ve just added two more samples:

21-cardi-front-linux.flac - mic facing the guitar, about 20cm away, +/- in front of the 14th fret

22-cardi-front-linux-lowcut.flac - same as above, except with the mic’s lowcut switch on

Both were recorded on linux and using the t-bone micplug. No normalize nor noise reduction… (getting late here and I’m quite tired, so I’ll play with that tomorrow…)

And I’m getting tired of playing the same chords over and over again… I’ll have to play something different for my next tests :stuck_out_tongue:

Forget the low-cut switch. It’s turning the show into a ukuklele. Even the flat one is a little gutless even though the fret work comes through loud and clear. Can you get a little closer to the tone hole? Don’t go crazy. We don’t want it to start getting boomy.

You can post as many musical notes as you want. If the tunes get too big to post directly on the forum, Private Mail me.


I see I’m joining the movie late. You’re already posting to a separate server. I hear the hiss problem, too. I like the tone of the third one back from the last one posted…if that makes any sense.

Did anybody ask you if you were in Audacity 1.2? Noise Reduction in 1.2 wasn’t a finished product, and no, it doesn’t have to sound funny if you use the one in 1.3.


This may be a job for the Kepex Simulacrum that Steve wrote a while ago. I gotta dig that out.


yes I have my own place to put them, no worries or limitations on that.

Did anybody ask you if you were in Audacity 1.2? Noise Reduction in 1.2 wasn’t a finished product, and no, it doesn’t have to sound funny if you use the one in 1.3.

I’m using 1.3.11 both on mac and linux now. I tried steve’s trick for NR and it did work, no more cut in and out…

I’ll try later to record some more samples with the mic closer to the hole.

<<<I’ll try later to record some more samples with the mic closer to the hole.>>>

I don’t know if I can stand this close brush with double entendre. I once tried to find a fastener in my supervisor’s desk. I announced to the whole office that I was rooting around in David’s drawers looking for a good screw. I found it, too. It was the fourth screw necessary to secure a repaired power supply in a PC case. He has a collection of good screws.

In his drawers.

It will be interesting to hear that latest effort. My impressions so far have been grab sessions and not in the killer sound system in the quiet living room.

Nobody would complain if you whipped out six or eight finished 3-minute pieces and posted them – strictly for testing, of course…


Well. I’m exhausted.

I could not find any setting of Noise Reduction that didn’t remove that crisp, wispy delicacy of the fretwork, so I threw in the towel and installed Noise Gate®.

http://www.kozco.com/tech/audacity/clips/21-cardi-front-linux-flat.wav (before)

http://www.kozco.com/tech/audacity/clips/21-cardi-front-linux-keypex.wav (after)

The silence at the beginning and the first two or three notes should give it to you.

20Hz, -12.8dB Reduction, -40 Threshold, Attack/Decay 50.

Kepex was a famous maker of stand-alone noise gates billions of years ago.


Oh, yes. The sample I got had significant DC offset which I removed. I’m not sure what that would have done to the Noise Gate.


I’m sitting in front of the weapons grade sound system and this version seems to have the best tonal balance. I’m not sure extra body-bass is going to help. It will just be a little different.

If you post any more clips, include something with more silences at both ends and a sample of lower notes. The fingering will give us the high frequencies OK. Those are grand and Noise Gate seems to be keeping them that way.

This tool gets killed when you have a long, slow decay in the last note. I’m interested in listening to one of those.


I think I’m getting punchy…



The big downside of Noise Removal.

Since bgravato has a Linux machine:
Gnome Wave Cleaner (available on Ubuntu through Synaptic) http://gwc.sourceforge.net/ has noise removal that is superior to the Noise Removal effect in Audacity. For best results you will probably need to tweak the settings and it has a couple of limitations. I think it only supports WAV format and it does not support “32bit float” WAV files (though 32 bit WAV files are OK). The WAV file must not contain metadata.

Of course it is better to have a really clean recording that does not need noise reduction, but these little techniques can be helpful for those of us that have less than perfect set-ups - it’s a balancing act between how much noise to remove and how much damage that causes.

I won’t complain if you post a full length piece of music. We usually ask for short audio samples as it saves time (and bandwidth) in helping people to solve technical problems, but this is definitely pleasure rather than work :wink: A few more seconds lead in/out of “silence” (noise only) would be nice so that we can play with noise removal more easily.

Removing DC off-set should be done before noise removal or gating as DC off-set messes up the threshold settings (the noise level in these samples is less than the DC off-set).
On one of the samples that I’ve just tested, the “silence” never goes below -37.6 dB, but after apply DC off-set correction it drops 20dB lower.

I’d forgotten all about that little beasty.
<<“Noise gates” are blunt instruments for bashing low level noise between tracks or audio clips.>> I like that description :smiley:
Given what’s been said about DC off-set and gating, perhaps I should add a sub-sonic filter to the gate (which will remove DC before gating).

Sorry I’ve been a lil busy in the last couple of days (and still am), but I’ll try to record and upload new samples today.

Thanks for the help, tips and knowledge sharing.

<<<Given what’s been said about DC off-set and gating, perhaps I should add a sub-sonic filter to the gate (which will remove DC before gating).>>>

Perhaps you can find an easier way to find the noise floor as in the Noise Reduction profile technique. “See this sample? This is noise!”

The bouncing lights sound meters and dB timeline are almost useless in finding the noise level.

You could use a by guess and by golly technique, too. Play the whole piece and assume the lowest part of the performance is noise.

I found the operating values of Noise Gate for this performance by total spit-on-the-hand, close your eyes and point guess work.

This tool suddenly bubbled up because I have a very serious show where I couldn’t get rid of the background noise in the live performance. The capture session was a success, but the wideband background noise is too high and I know I’m going to need to do it again.


That is possible.

I can think of a few alternative methods of doing this - take your pick and I’ll have a go:

  1. A small plug-in for “Find Peak/RMS level” - The Peak/RMS values are displayed as text. (could also have an option for ignoring DC off-set.)
  2. As above but part of the “Noise Gate” effect.
  3. As (2) but the noise level is saved as temporary data and is automatically used when the Noise Gate plug is used a second time.

Pros. and Cons.

  1. This may have application other than just with gating. It should be simple to implement. Having to locate a second plug-in effect is perhaps inconvenient. Such a plug-in should really be in the Analyze menu which may be overlooked by users.

  2. Easier for users to find than option (1) but makes the interface more complicated.

  3. This would be cool, but not without potential problems. Just as the “Noise Removal” effect confuses many users (“I clicked to get profile and nothing happened”), selecting “Get Noise Level” would apparently do nothing. What action should happen if the Noise Gate is applied without first analysing the noise level? If a different plug in was used after getting the noise level and before applying the Noise Gate, some other plug-ins could delete the temporary data.

Some other considerations:
i) The actual “noise floor” may not necessarily be the best level to set the Noise Gate threshold. In most cases the threshold should ideally be “a little bit” (technical term) above the noise floor.
ii) The noise floor can be measured by calling up the “Amplify” effect.
iii) It is not currently possible with Nyquist plug-ins to output a result and keep the interface open, so just like with Noise Removal it would be necessary to run the effect twice - First to get the noise level and then a second time to apply the effect.

Finding the minimum (or maximum) level in a track using Nyquist can be rather slow and memory intensive, though it should not be a problem with tracks of 5 minutes or less. It becomes a problem when dealing with whole albums on machines with little RAM. It could also be messed up if there were gaps (white space or total silence) in the selected region, though it could theoretically be programmed to ignore total silence.

I generally call up the “Amplify” effect with a noise region selected to determine the noise floor, then spit on the hand and guess a little bit higher.

I start to feel like I’m going back to my signal analysis classes :stuck_out_tongue:

I’d also like to try to connect the mic directly to my soundcard line in (it’s an Asus Xonar Essence STX), but I need to solve the phantom power problem first…

One option is to get a XLR->Jack interface that will provide phantom power such as the Art Tube MP, which is not expensive and seems to have good reviews…

Another option is to build my own DIY interface… I’m an electronics engineer so I got the skills, just not sure if I have the time/patience for it :slight_smile:

If I was to count the number of DIY projects I started and never finished… most didn’t even get out of the paper… actually most don’t even reach the paper… they just get lost somewhere inside my brain :stuck_out_tongue:

<<<I generally call up the “Amplify” effect with a noise region selected to determine the noise floor, then spit on the hand and guess a little bit higher.>>>

I missed that one. I was trying to do it with the existing timeline tools. And yes. You can’t use the value. You have to go a bit higher.

I need to think about this a bit. It’s a philosophy question in addition to a programming task.

There was a Keypex joke. Many people set the attack/release too tight (if it was even adjustable) and it would sheer off the very front and back of each word. There were announcers who could “talk Keypex” in normal speech around the water cooler and everybody in the group knew exactly what they were doing.