Low noise usb soundcard for sleeptalk recording


I often sleeptalk, so does my girlfriend.

So I made this program wich can record only when there is sleeptalk.
It filters the signal betwen 300-4000Hz and removes the ambiant noise for voice detection (not during the recording)

However my sound card is super noisy, thus the ambiant noise is sometimes too elevated to detect anything.
I tried the same microphone at work and there is less noise (still too much tho).
Problem with the noise is that it mask whispers and very soft speak.

I’m not looking for a 500$ setup, something around 100$ max, USB for portability (I could use it with a rapsberry pi and a battery).

Right now I’m using a desktop computer built in soundcard with a 5$ mic.

Example of recording : https://drive.google.com/open?id=0B9DQrxKN9o9kdEdaRVl4b0s1NzQ

Any recomandations?

I doubt that a USB mic will be very suitable as USB mics (especially less expensive ones) tend to be optimised for medium to loud voice, and often struggle with quieter voices (and worse for whispers).

At around $100 your budget is just about sufficient for a separate mic and USB pre-amp. At the low price range end of USB mic pre-amps, there’s a couple of models by Behringer that we’ve had good reports about. The choice of microphone is tricky to get within budget. I would be drawn to the Rode NT1A because of its low noise, but that will push you well over budget. You may need to trawl through spec sheets and price lists - the most important things you will be looking for are:

  • High sensitivity
  • Low noise
  • Cardioid or Hyper-cardioid (directional) pick-up pattern

“High sensitivity” is probably the most important for this job.

I’d suggest looking at some of the budget studio brands rather than “unbranded” ultra-cheap off ebay stuff.
For example, Superlux, Samson, Behringer …


So first, the microphone, I just did some research.

So I need a sensitive microphone with a low equivalent noise.
A -33dB is more sensitive than -55dB (I would have assume the oposite)
A microphone with a xdB equivalent noise means that it’s as if I was running a perfect microphone in a room with a xdB noise source.

so we are going for a microphone with a sensitivity in dB as close as possible as 0 and an equivalent noise smaller than 6dB (because I want to record wispers, and it’s around 10dB)

Second the preamp :

This I have no idea, I’m guessing some of them add some bad noise to the signal, some does not.
A quick google search told me that this one had low noise levels : https://www.amazon.fr/gp/product/B00SAV0VP0/ref=ox_sc_sfl_title_3?ie=UTF8&psc=1&smid=A35ATNU92FJ7LH

Unfortunately both noise and sensitivity measurement (specifications) are quoted in many different and incompatible forms. Unless you are thoroughly familiar with dBA, V/P, mV@dB SPL, and all the rest, then you may need to rely on reviews. For example, if you see that there are lots of people ask "how do I make my voice louder with … microphone, then it could be an indication that the mic is not very sensitive. On the other hand, it could be an indication that the mic is very inexpensive and is used by a lot of people that are inexperienced in recording.

If you live in Europe, then another brand that is definitely worth considering is “t-bone”, which is an “own brand” for Thomann.de
I wouldn’t recommend their cheapest mics unless cost is everything, but their mid-range mics are still very low cost and in my opinion, excellent value for money. I own several t-bone mics.

After reading this page everything became clear : https://geoffthegreygeek.com/microphone-sensitivity/

Rode NT1-A, sensitivity [dB re 1V/ Pa] -31.9, sensitivity [mV/ Pa]: 25, equivalent noise 5dB, 199€
T-Bone SC 300 : sensitivity [dB re 1V/ Pa] -34, sensitivity [mV/ Pa]: 19.9, equivalent noise 13.5dB, 35€
T-Bone SC 450 USB : sensitivity [dB re 1V/ Pa] -32.7, sensitivity [mV/ Pa]: 23.3, equivalent noise 10.3dB, 88€

Indication the sensitivity in mV was useless but it was just to show that my data was reliable (more sensitive mic = more mV per Pascal).

I spent my whole afternoon comparing mics, I’m really interested by the T-Bone SC 450 USB, because it’s USB I won’t need a preamp, it’s really sensitive and it’s equivalent noise is as small as a leaf falling.
For far cheaper I can get the SC 300 which is somewhat equivalent but I’ll need a pre-amp.

I read a lot of “usb mic are evil, get a regular one + a preamp” but I can’t get a good reason.

Do you think I should get the SC 450 USB?

Comparaison of several mics (sensitivity and equivalent noise) for those interested : https://ibb.co/juQaMF

Yes. But it can’t be taken in isolation. If a mic is massively sensitive, and huge power handling (both of which are generally ‘good’) but also has a huge amount of “self noise”, then the sensitivity and power handling are in vain. On the other hand, a mic with modest sensitivity and extremely low self noise may be very good for recording quiet sounds, provided that the mic is used with a really low noise pre-amp.
Note that the most expensive of the three that you list, the Rode, has pretty good sensitivity and extremely low self noise.
Everyone wants a sensitive mic with very low noise, good power handling (max SPL), and smooth frequency response, for under $100. If you find that mic, buy three for me :wink:

Many of the cheaper USB mics have fixed gain. “Gain” is what determines how big the digital signal is for a given audio input level.
Mic pre-amps (including USB mic pre-amps) and some USB mics, have adjustable gain.
The digital signal has an absolute maximum level of 0 dB. If you try to push the level above 0 dB, all that happens is that the peaks of the waveform are clipped off (nasty distortion). Because of this, USB mics with fixed gain are factory set to give 0 dB for very loud sound, (so that normal level sound will not clip). The downside of fixed gain is twofold:

  1. If you try to record very very loud sound, it will clip and there’s nothing you can do about it.
  2. If you try to record quiet sound, your cannot make use of the full dynamic range, and have a quiet signal stuck down in the digital noise floor.

The benefit of variable gain is that the signal level from the mic capsule can be amplified to a level that makes best use of the analog to digital converter. So in case (1), you can reduce the gain and prevent clipping, and in case (2) you can increase the gain and improve the signal to noise ratio.

Another limitation of USB mics is that they are (usually) powered by the 5v USB supply. There are several problems with this:

  1. USB power supplies are often noisy, so it’s quite common to get a high pitched whistle in the recording.
  2. Other components in the computer are often noisy, and because there is little isolation between noisy components and the USB supply, it’s common for USB mics to pick up interference from mice, hard drives, video cards, fans and other components.
  3. Traditionally, condenser mics are powered from 48v. USB power is 5v. There are often poor compromises required to get a condenser mic to work from only 5 v.

USB leads tend to become unreliable if they are too long. That means that with a USB mic you are shackled to within a couple of meters of the computer fans.
Good quality microphone cable can work happily at 12 m or more.

USB mics are not evil, but the do have limitations, and those limitations are “final”. There’s no upgrade path for a USB mic because it’s all-in-one. The main benefits of USB mics are that they are compact, convenient, and often quite inexpensive.

I have a T-Bone SC 450 (non-USB). I like it. Very good value for money in my opinion. I think it was the first large diaphragm condenser that I bought, and it still works fine.

Thanks you. I get it now.

Well in that case I’m going to try the usb one (because it’s more convienient to use/transport), if it does not fit my needs (electrical noise from the computer masking my whispers, quiet sound being clipped) then I’ll return it and get the analog one with the pre-amp.

Cheap, fairly silent, very sensitive, the Superlux Hi10:


And maybe easier to use in a bedroom :sunglasses:

mmmh, 25dB equivalent noise.
I prefer to go with the 35$ mic which has a 13.5dB equivalent noise :stuck_out_tongue:

But thanks.