The mode for listening to music has evolved over the years. Nowadays, people simply turn to their desktop computers, laptops and even more compact mobile devices to listen to or make music. The earlier formats, such as cassette tapes and CDs or compact discs are no match for the number of songs you can now save in your music library.

Compact discs revolutionized the music industry. The new format gave way to better sound quality and much simpler recording. Compared to cassettes, compact discs are more convenient for both listeners and music makers.

While compact discs are still popular nowadays, the demand for music availability online is growing exponentially. To match the demands of modern listeners, developers recently made a lot of new algorithms to make music available in several formats. These formats vary on sound quality so choosing the right format of music is important.

Although several formats are now available, the MPEG1 Audio Layer3, most commonly known as the MP3 is the crowd favorite. Several programs are available in encoding formats like this. Comparisons between programs encoding to MP3 are available online but the best program available so far is LAME. LAME is a software project that is available for free, and comes with no restrictions on licensing.

However, the MP3’s popularity is now being challenged by the existence of Ogg Vorbis. The Beta 3 version of Ogg Vorbis was launched in the summer of 2000 and is also popular among users. The year 2001 kicked off with the launch of the new versions of the encoders: LAME 3.88 beta and Ogg Vorbis 1.0 beta 4. Both are very different from their former versions and now we will be comparing them side by side.

process6_01s
process6_02s
When it comes to bitrate, the Ogg Vorbis (OGG) is not intended for constant bitrate coding similar to that of the MP3.

process6_03s
Variable bitrate or VBR is used for the six pre-installed modes including 112, 128, 160, 192, 256 and 350.

process6_04s
In addition to this, aside from the speed which can be altered by the user, the OGG does not allow manipulations on its parameters unlike the LAME which gives the user freedom to control most of its parameters. With this, we will now compare both encoders based on the recommended coding modes by the developers. However, this rule will be skipped just for lowpass filter, meaning the frequency signals below definitive levels, allowing the LAME to specify the frequencies to be avoided to prevent cuts.

process6_05s
Just recently, LAME added an average bitrate or ABR; it is already included in the OGG for a while now. Developers said that at the same bitrate, using the ABR mode for coding must be better than what is commonly used. In connection with this, sample codes in constant bitrate and variable bitrate will be used in estimating the quality of LAME.

process6_06s

Format limitations

Pauses are inevitable in coding nonstop albums into MP3 because files constantly need to pause both at the start and at the end of tracks. LAME corrects the beginning of files while files in Ogg Vorbis completely overlap with the original.  Similar to CD-DA format, LAME code files limited to 44100 Hz sampling frequency while OGG can code files at 48000 Hz.

process6_07s
process6_10s
process6_08s
process6_09s

Sound quality

The purpose of coding is to reach maximum sound quality. For this reason, maximum parameters were set for both of the coders. The LAME was set in the mode of 320 Kbit and sound range of up to 22 KHz and highest quality (-q0) and allows the coder to set other parameters. For OGG, the mode was also set at the highest quality that is 350 Kbit but unlike in the, other parameters cannot be regulated.

process6_11s
Results show that both of the coders’ psycho acoustic models have been through a lot of changes which can easily be identified when decoded samples were analyzed.

process6_12s
Significant changes were noted on high frequency processing. Before, LAME drops the complete range of up to 22 KHz in the 320 Kbit mode but these frequencies can now pass the psycho acoustic model. This change can be observed on sonogram.

process6_13s
Graphs are provided to compare both the original and decoded samples.

High frequency processing is quite unusual for the highest quality coding mode. Check the samples then compare them to the original recording.

process6_35s
Interestingly, the sound is too similar with the original recording, making it difficult to differentiate the three compositions from one another. Both of the coders’ sounding are quite identical to the original recording when maximum parameters were set. The only noticeable thing is, the OGG’s sounding is more transparent and upper middle frequencies’ reproduction is much better. Note that the difference is so minute and can only be heard when high quality equipment is used. Both coders scored high in this test with the OGG having a higher bitrate average, usually within the 340-380 Kbit range. There are no significant differences between the original and the coded samples when it comes to average AFCs despite the extremely high frequency handling.

process6_14s process6_17s process6_15s process6_16s
Moving on, the original recording and coded samples’ corresponding delta signals will now be calculated and compared.

The LAME samples have delta signals that sounded like a quiet wide band where the main sound can be heard weakly, with rough sound pattering and severe high frequency distortion. For the OGG samples, the sound is more complicated since it is reminiscent not only of noise but of a highly distorted original recording with phase distortion effects. In the OGG, better processing of different ranges is evident. They appear to be more thorough as compared to the LAME where parameters are indistinguishable with psycho acoustic models for most of the subranges. It is more apparent when the AFC of the delta-signals were analyzed. The lower the signal is, the better is the sound quality on corresponding frequencies.

process6_18s process6_19s
The OGG developers simplified the coding for low and middle frequencies that are lower than 2 KHz and improved the reduction of the upper middle and high frequencies up to 16 KHz.

process6_20s
It was shown that the LAME has better sound reproduction in the range of up to 2 KHz while the OGG has shown better results in the range of 2 to 16 KHz.

process6_21s process6_24s process6_22s process6_23s
The difference between Lame and OGG is that LAME allows manipulation in coding, psycho acoustic parameters and filters. Better sound quality can be reached when coding at 320 Kbit if cutting higher than 20 KHz frequencies.

process6_26s process6_25s

Results and Recommendations

The exceptionally high bitrate gave excellent sound quality but it is not favorable for high bitrates because a file streaming at 25 Mbytes/s is considered to be big. Many users prefer bitrates with 256 Kbit for a lower file size despite the lower sound quality. For this reason, we compared the sound quality of both coders in lower bitrates and estimated the losses relative to the highest bitrate. For the LAME to enhance the quality of the basic audible range, frequencies above 20 KHz were cut. We tested the LAME with a constant and a variable bitrate.

process6_27s process6_30s process6_28s process6_29s
Upon checking the samples’ frequency dynamics where AFC changes with time and averages in 20 to 100 millisecond intervals, the losses in the high frequencies seemed to be constant.

process6_32s process6_31s
The OGG cut frequencies above 18 KHz more frequently than the LAME.

The OGG’s frequency reproduction in the 15 to 18 KHz range was much better since the LAME cut the sound at lower frequencies and in pauses between high amplitude splashes. The LAME showed better reproduction of frequencies when in the average bitrate mode than when in constant bitrate mode.

process6_33s
The developers of the LAME have transformed the coder a lot. No difference is noted in the high frequency sounding anymore.

process6_34s
Coding samples in the average bitrate mode is better than doing it in the standard one. With this, average bitrate is more recommended than constant bitrate. OGG’s cutting of frequencies above 18 KHz has no significant effect on the sounding of the samples and neither does it differ much with that of the LAME. It is still advisable to use 256 Kbit as long as you use the latest LAME version.

The sounding for the delta signal and the 320 Kbit mode are similar. In high frequency range, OGG samples have hoarse sounds and increase in noise is apparently heard in LAME samples.