Skip to main content

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Request for diagnosis - what went wrong with that recording (Read 3704 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Request for diagnosis - what went wrong with that recording

Welcome

I need to instruct my friend how to modify her recording settings so that there are no cracks and noise as in her original recording:

http://dl.dropbox.com/u/30854535/Samples3.mp3

In order to instruct her, I need to know what is wrong with it. There seem to be no clippings in the audio. 

All I know is that she recorded it with omnidirectinoal mic, saved in wav at 44kHz.

I converted it to mp3 for space economy reasons but there is no audible difference between wav and mp3.

I would appreciate your input very much.

Best regards,

Greg

Request for diagnosis - what went wrong with that recording

Reply #1
Recorded with an omnidirectional mic...into what?  Onboard sound on a PC/laptop/Mac? Outboard USB audio interface? Standalone digital recorder?

It definitely sounds like some kind of digital clocking issue - you're right, audio levels are fine - but without knowing how it was recorded, it's difficult to say what the cause might be.
"Not sure what the question is, but the answer is probably no."

Request for diagnosis - what went wrong with that recording

Reply #2
A major problem is missing samples. This results is the cracking sounds. It is rather common from a computer that has not been set up for real time capture.

The input signal comes constantly from the analogue world. The computer is doing too many things while the recording is being made. Computers multi-task, they mostly switch between different operations much too quickly for a human to notice. But some things take more time and/or more resources. The audio input keeps coming but the computer's attention is elsewhere. When the computer comes back to the recording it just picks up where ever the real time input is at the moment. Anything that came in between the last time the computer was paying attention and now, when it is paying attention again, is lost forever.

This is an over-simplification. There is some buffering. A certain number of samples will be automatically captured in RAM that has been set aside for that purpose, even though the CPU is now doing something else,  then lickety-split put into place in the recorded file when the recording program gets its next time slice. Buffer size can be adjusted in the recording program. However, on computers that have not optimized for audio, some samples will be lost. As I wrote above, this is rather common. Professionals tend to make as sure as possible that nothing else is going on while recording.

Some common things are anti-virus, and other anti-malicious software programs, automatic updates, task scheduling, and network traffic. I have read that proper recording on a computer that has wireless capabilities turned on is sometimes possible, but do not expect it to be easy. There are various websites that have instructions about how to optimize for audio. I have not needed to reference any of them for a long time, so can't easily point you on, but a search should turn up several.

Another thing is this recording is probably a low quality microphone preamp. There is considerable hiss that probably originates there. That can be compensated for somewhat, but not greatly, by where the gain is set. The higher the gain, the more preamp noise. This must be balanced against too low a signal level from turning the preamp gain down too low. Cheap microphones create more noise too. Some microphones can be concentrated on the speaker and will reject many extraneous noises.

After that the recording environment gets to be important. External noise and sound reflections can make for a poor sounding result. One has to built a proper area and/or learn where to place the speaker and microphone for best acoustics.

Request for diagnosis - what went wrong with that recording

Reply #3
I converted it to mp3 for space economy reasons but there is no audible difference between wav and mp3.

One tip I would give is NOT to convert to MP3 before posting. The highest frequencies are removed and what remain are smeared in the time domain. Although this is not audible, it does affect the fine detail of the waveform when inspected in an editor.

Digital dropouts due to resource starvation in the recording PC nearly always result in instant (single-sample) spikes or steps in the waveform, but because of the MP3 coding there is no chance of seeing this. So while I *suspect* this crackling may be due to some kind of digital dropout, there is no way to be sure.

The distortion only happens when the lady is actually speaking - not during the gaps. So it may be that some analogue circuitry (or the microphone itself) is being overloaded. That said, digital dropouts tend not to be so audible on low level signals, so it could be that they are present but not noticable.

Repost the original WAV and we'll have a better chance of diagnosing the problem.

Request for diagnosis - what went wrong with that recording

Reply #4
I converted it to mp3 for space economy reasons but there is no audible difference between wav and mp3.

Repost the original WAV and we'll have a better chance of diagnosing the problem.


Sorry for mp3, here is wav http://dl.dropbox.com/u/30854535/wavsample.wav

Digital dropouts due to resource starvation in the recording PC nearly always result in instant (single-sample) spikes or steps in the waveform, but because of the MP3 coding there is no chance of seeing this. So while I *suspect* this crackling may be due to some kind of digital dropout, there is no way to be sure.

The distortion only happens when the lady is actually speaking - not during the gaps. So it may be that some analogue circuitry (or the microphone itself) is being overloaded. That said, digital dropouts tend not to be so audible on low level signals, so it could be that they are present but not noticable.


Exactly, when speaking or even speaking some particular sounds. Now I'm thinking that these cracks just might have been bursts of air into the mic which might have been too close to mouth or of too poor quality.

Regards, Greg

Request for diagnosis - what went wrong with that recording

Reply #5
But it's not—if you look at the waveform, there are step changes that really only point to chunks of audio having been deleted somehow, in the digital domain.

Request for diagnosis - what went wrong with that recording

Reply #6
But it's not—if you look at the waveform, there are step changes that really only point to chunks of audio having been deleted somehow, in the digital domain.
I agree. Also note that the missing samples pattern is very irregular, almost like varispeed. This is definitely not an analog problem and most likely not a sample rate mismatch because IME the missing samples would be periodic. Also a (too) slow hard drive is unlikely since I would expect larger data blocks to be missing.
My guess is that the problem should be located between the output of the soundcard and the input of the hard drive. Plenty of experts for that on HA (I have to bow out).

Request for diagnosis - what went wrong with that recording

Reply #7
Not sure, but here are a few things that she could try:
- Disable any 'Spread Spectrum' settings in the BIOS settings.
- Check out all the power options in your BIOS and set everything for maximum performance. Create a new Power Plan in Windows and do the same. Set this Power Plan as your default. Check out every setting under 'Advanced Power Settings'
- Use this tool to see if there are any latency issues: http://www.thesycon.de/deu/latency_check.shtml
- Make sure she isn't overclocking
- Update Sound drivers + disable all 'enhancements'. If she is using Windows Vista/7 and the soundcard supports the Universal Audio Architecture (UAA) standard then sometimes using the generic drivers built-in to Windows Vista/7 works better.
- If it is a multicore CPU then try unparking it in Windows: http://www.coderbag.com/Programming-C/Disa...Parking-Utility
- Make sure no anti-virus/ant-malware programs are interfering with the recording.

Request for diagnosis - what went wrong with that recording

Reply #8
So...  You need to minimize multitasking.  Close all applications except your recording application, and you may have to turn-off your anti-virus and check for stuff running in the background.  (You can't eliminate all multitasking because the operating system is always dong stuff like checking for mouse/keyboard input and updating the display, and a bunch of other background stuff.)

Next, you can try increasing the latency/buffer.  You should be able to do that from your recording application.  Here is some information about how to do it with Audacity.  (You are actually changing settings in the driver/hardware, but you can usually access those settings through the application.)

She's already doing this, but it also helps to record at lower data rates (such as 16-bit, 48kHz) and avoid 24/96, especially if recording multi-channel.

If none of that helps, she might need a new soundcard/interface or a different computer.

Most of what you find on the Net about latency is all about reducing latency.  This is for people who are monitoring themselves though the computer while recording.    If you are not monitoring yourself in real-time, or if your monitoring bypasses the computer latency is of no consequence and a bigger buffer/more latency is better.

Request for diagnosis - what went wrong with that recording

Reply #9
A great many of the missing samples gaps in a recording done on a computer with too much background activity are not "single sample" but multiple samples missing. The waveform can be seen to make a sudden step from where it should have gone in the next sample to something quite different. It may, for instance just have been starting down from a positive peak and the next captured sample is again nearing the positive peak -- a dozen samples or more are missing. This is easy to observe with test signal input where every sample is predictable.

The same problem happens with digital input, as in an S/PDIF transfer. Under the right conditions, by comparing the source and result, it can be seen that several positive/negative cycles were lost while the OS was busy doing something else.

Request for diagnosis - what went wrong with that recording

Reply #10
I had similar dropouts the first time I used Audacity to capture from the USB connection of a Bose Tonematch on-stage mixer/effects device. As soon as I changed the project rate to the native 48000 samples/second instead of 44100, it worked fine, so I suspect resampling on the fly was necessary at unsupported rates.
Dynamic – the artist formerly known as DickD

 

Request for diagnosis - what went wrong with that recording

Reply #11
A major problem is missing samples. This results is the cracking sounds. It is rather common from a computer that has not been set up for real time capture.


After that the recording environment gets to be important. External noise and sound reflections can make for a poor sounding result. One has to built a proper area and/or learn where to place the speaker and microphone for best acoustics.


Assuming you're right that would have to be one incredibly lame computer. I was recording analog audio flawlessly into a PC in '98 with a Pentium 166. This does assume the machine isn't clogged up with unneeded software or viruses.

Looking at the waveforms in Audition I agree that it looks like missing samples but may in fact be repeated sequences because of write cache problems. A couple of ways to test this would require a properly operating computer to generate predictable tone bursts and capture on the bad PC. Another way would be to use a 'Y' cord and capture on _both_ computers simultaneously and compare he waveforms. I'm betting on configuration problems and may be a trivial as needing a newer driver.

I agree the environment is important for good recordings but what is appearing in the waveform absolutely can NOT be caused by the environment.


Request for diagnosis - what went wrong with that recording

Reply #12
Computing requirements are not very great for recording, only how they are utilized. I did more than 700 LP transfers on an original AMD 133MHz K6 , with 128Meg of RAM, running Windows 95. It would record 24/96 with no difficulty (Audiophile 2496 soundcard), but it was optimized for audio. My daughter wanted to do some audio on her modern laptop, which has wireless internet, and too many other things running. It was hopeless.

The external environment will not effect continuity of capture but will strongly impact the overall quality.