Hello, I'm new posting here but read and search for help from time to time. I have a question that I haven't been able to find an answer to.

I do a lot of realtime audio captures with MP3 compression using programs like Messer and RecAllPro. Usually it's high quality 'line-in' material that I capture, with frequency content up to 20+ kHz. I've confimed this by analyzing uncompressed captures. In Messer I usually use Lame 3.91 but have used the newer versions as well.

I have noticed that no matter what bitrate I capture at, or what Lame version I use, if I analyze the frequency spectrum of my captured files there is a strong cutoff at 16 kHz. Occasionally there will be sproadic stuff up to about 20 kHz, but this is very few and far between. Like I said before, I know that my original content has response up to 20 kHz.

Is this something inherent with Lame DLLs? or maybe part of the Layer 3 spec? If anyone has any solutions or comments I would appreciate to hear them. I was hoping to at least capture some consistant signals up to 20 kHz, even if the response is slightly rolled off.

It is normal, the weakest sounds above 16 kHz are removed in order to reduce the filesize compared to an uncompressed format. By the way, there are sounds removed at all frequencies, but since there are some other left below 16 kHz, you don't notice on your graph, that there is in fact something missing everywhere.

Try with an higher bitrate for example 192 or 256... usually the lowpass of 16KHz is used when using a 128 bitrate... It is also possible to have a 128br MP3 without lowpass (by using the -k parameter when encoding) but its quality will be much lower as regards the most important range of frequency a human can hear...

Try lame dll that only encodes to --alt-preset standard. Download here. You can get similar versions that only encode to --alt-preset fast standard, --alt-preset extreme and fast extreme at rarewares too.

Thanks for the responses. I'll try that link you provided.

Tigre,

Thanks for the link you provided in response to my Messer 16 kHz cut off question. It appears Lame.dll v3.90.3 is fairly steady out to 18.5 kHz and then rolls off sharp down to 19 kHz.

It also appears that CBR is only selectable at 320 and 256 kbps. Anything lower than this gives me a VBR file around 220 kbps. If I choose VBR mode, then I have more control over the output bitrate and can choose something lower.

I will use this version in Messer for high quality capture, and switch to 3.91 or one of the others if I need a low CBR.

Thanks!

You hearing is also acting in a similar way. Remember that music is for your ears, not for your eyes on a spectrogram.

dvjones' last post merged from seperate thread.
----------------------------------------------------------------
the point about my recommendation is that Messer uses lame.dll but can't use --(alt-)presets. The modified .dlls from rarewares (should) only encode to one lame (alt)preset, no matter what the program calling them, tells them to do. So with the dll you've tried, you get --alt-preset standard. For details about lowpass vs quality have a look at the sticky thread "recommended lame settings" and use the search function, please. You don't want mp3s with content up to 22.05kHz.

In this case, the FAQ is easier than the search function, with the link to the Why not using graphs to compare codecs ? thread.

Tigre and Pio,

Thank you for the help with this issue. In general I agree, I can't really hear any difference between a 16 kHz and a 18.5 kHz response. But from a semi-scientific point of view it is an interesting point to discuss. The response of the modified 3.90.3 Lame dll interested me as well since I've only used the regular distributions prior to last night.

Thanks again.

Gabriel,

Thanks for your response. I can't tell the difference between my 20+ kHz source and my new v3.90.3 Lame dll captures. Analyzing the spectrum of the captures is always an interesting and fun task in any case.