|
This study was performed in Spring 2007 for 1 credit of FA3662 practicum at Michigan Technological University as part of the degree requirements for Audio Production & Technology, B.S., by Kevin C. Dixon. The procedure and results are documented herein.
Objective
To attempt to determine and quantify the effect different compressors have on the tonality and atmosphere of a sound, even when performing the same compression (as designated by the compressor's controls: attack, release, threshold, and ratio).
Design
Program Material
Program Material was chosen in an attempt to emulate the two standard applications of compressors: 1) Processing a newly recorded track and 2) Mastering an already mixed song or track.
It was further determined that the material should all be stereo, since some of the effects of compression are much better understood when heard in a stereo setting.
Four samples of music were selected in an attempt to fill these two roles (two for each).
- Pink Floyd - Pigs (Three Different Ones)
A portion from the end of the song was selected to emulate a mastering situation.
- Les Claypool and Highball With The Devil - Holy Mackerel
A verse from the song was selected to emulate a mastering situation, as well as demonstrate compressor operation on an electric bass guitar and vocals.
- Tool - Lateralis
A portion of the song was edited to expose the drums, which in the recording moved in a circular pattern.
This was accomplished by placing A-B mix in the left/A channel, and a B-A mix placed in the right/B channel.
- Momentum - Mambo Swing
A solo recording of a grand piano using a microphone near the hammers, and one near the top of the piano were used to generate a stereo recording. A section of the rhythm melody was extracted for use.
Equipment
Four compressors were selected based on two qualities: diversity and availability. In an attempt to achieve significant results, compressors were selected that typically fill different roles.
- Alesis 3630
This is an analog, two-independent-channel compressor, and is generally considered an entry level, general purpose compressor. It can be had for about $100.
The compressor used is owned by the Visual and Performing Arts department, and is normally installed in the McArdle Theater Booth.
- Focusrite d3
The Focusrite d3 is a 24-bit TDM plugin for Pro-Tools. It is a very well regarded compressor, and runs $600, bundled with an equalizer plugin of similar quality.
- Joe Meek SC2.2
The Joe Meek SC2.2 is a photo-optical stereo compressor, designed for professional studio use, and is sought after for its "mojo". It is valued at $800.
The compressor used is installed in the Rozsa Recording Studio.
- Symetrix 528E
The Symetrix 528E is a single channel voice processor and runs about $500. It is generally used in professional broadcast scenarios.
WMTU was kind enough to lend me their pair of 528Es.
Evaluation Procedure and Criteria
Designing the evaluation procedure was challenging, in that it needed to be approached in such a way that the listeners would not be influenced in any way, and would also be able to readily understand how they were supposed to fill out the sheet. The optimal solution to this sheet is found here: Compressor AB Evaluation Sheet [pdf]
Procedure
Since the listener had to evaluate four compressors for each piece of program material, it was determined that they should only evaluate the compressors in pairs, to cut down on confusion. Each compressor was designated a letter for its pair, either A or B. Each pair was then assigned a number, resulting in a terminology such as "Compressor 1A" or "2B". The listeners would then compare Compressor 1A to 1B, and 2A to 2B.
Criteria
Four core items were chosen to represent the main effects compressors impart on their input signal. The users were to rate them on a scale from -5 to 5, with a score of zero indicating no change from the original source. What each extreme indicates was different for each item. The ranking guide found on the Compressor AB Evaluation Sheet is shown here, to explain how the criteria should be interpreted by the listener.
Ranking Guide
Rank each item relative to the uncompressed signal. If no change for a given attribute is observed, rank zero.
- Intelligibility (-5, muddier to +5, clearer)
Ability to distinguish distinct sounds from one another, such as words, notes, or other separate sounds.
- Apparent Size (-5, smaller to +5, larger)
Apparent size of the image across the sound stage, which can vary from localized/small or encompassing/large
- Presence (-5, further to +5, closer)
Relative distance of the image from the listener, far/background or near/foreground
- Dynamics (-5, none to +5, fuller)
Observed dynamics (difference between loud and soft portions of the program material)
A fifth attribute the users were asked to examine was the spectral balance. If they felt there was any weighting to any of the three traditional bands (bass, middle, treble), they were asked to indicate it by circling the band they thought was different.
Procedure
Pre-Recording
In order to save time and money, it was determined that pre-recording the total 16 tracks would be the best approach. All of the compressors were brought to the Rozsa Recording Studio, and were then connected so that all compressors could be recorded simultaneously. (see Figure 1)
 |
|
Figure 1: Diagram of signal flow setup for simultainious recording of compressors.
|
First, the list of program material was randomly ordered using the RANDOM.ORG List Randomizer tool. Then, using the same tool, the list of compressors were randomly assigned to be the "reference" compressor for that material. The results of this selection process can be seen below, in Table I.
Finally, Each of the program materials were then sequentially recorded using the following process.
- Reference compressor (as determined in the above procedure) was tuned to the program material as described in Mixing With Your Mind, chapter 8 Cracking Compressors.
- These settings were then transferred to all of the other compressors.
- The program material was played through all equipment, and recorded simultaneously.
| Table I: Compressor Settings |
| Song |
Setting |
|
Reference |
| |
Attack (ms) |
Release (s) |
Threshold (dB) |
Ratio |
| |
| Pigs |
6 |
3.3 |
-4 |
4:1 |
|
Joe Meek SC2.2 |
| Mackerel |
2 |
2 |
-10 |
2:1 |
|
Alesis 3630 |
| Mambo |
125 |
2.3 |
-10 |
2:1 |
|
Symetrix 528E / Alesis 3630 |
| Tool |
4.5 |
42 ms |
-19.2 |
4.5:1 |
|
Focusrite d3 |
Notes:
- Input and Output Gains were set to 0dB
Concerning the Symetrix 528E:
- The Symetrix 528E does not have controls to adjust the Attack and Release times, it only has Threshold and Ratio. For the song Mambo, the Symetrix 528E was used to determine the Threshold and Ratio settings, then the Alesis 3630 was used to determine appropriate Attack and Release times.
- The Symetrix 528E was wired to pass audio through the entire module, but all other options, such as De-Essing and Equalization were bypassed.
Concerning the Joe Meek SC2.2:
- All recordings were performed with the Dark Mode switch in the "off" position (e.g. in Normal Mode). Use of the Dark Mode results in much less linear performance.
- The Slope control was set at "2" for all songs, which corresponds to a compression ratio in the 2:1 - 4:1 range
- The numerical values for Attack and Release were calculated using straight line interpolation between the values given as the maximum and minimum values found in the SC2.2's manual.
Samples & Preparation
Using the List Randomizer again, four different orders of compressors were determined. These randomized lists were then used to determine the identifiers for each compressor for each song. The resulting assignments are noted in Table II below. A CD was burned which included the uncompressed version of a song, followed by the four compressors.
You can listen to all of the different tracks and compressors by following the links in the table below.
| Table II: Master Track List |
|
|
|
|
Directly preceding the scheduled listening session (held in Walker 210), the room was cleared of extraneous equipment and furniture. Flats were deployed in an angular fashion to scatter reflections on the back wall, rather than cause slap-back echo.
Listener Evaluation
Upon arriving, each listener received four copies of the Compressor AB Evaluation Sheet , one for each of the tracks to be evaluated.
The name of the program material to be played was announced. Originally, the Uncompressed version followed by the A and B versions was to be played. This proved to be a difficulty for the listeners. Because they were to compare each compressor's effect on the uncompressed signal, the listeners requested to hear the uncompressed version between each example. The procedure was altered, and used for for all material throughout the listening session. This altered procedure is summarized in the following steps.
- Play the three examples all the way through in the following order
- Uncompressed
- A
- Uncompressed
- B
- Cycle through the tracks, listening to about 10 seconds of the beginning of each, using the same order as described above.
- Any further listening was determined by requests from the listeners
After all of the material had been thoroughly examined by the listeners, the compressors used for each song were revealed to the listeners.
Results
The initial results obtained from the experiment were somewhat inconclusive. Further analysis of the raw data showed that where one listener might have rated the change in Apparent Size a +3, another user may have rated it a -3. Throughout the data, such "opposites" can be found. It seems the listeners agreed that some change had occurred, but weren't sure how it exactly corresponded with the ranking sheet. Reasons for this could be inexperience in listening to compressors, different listeners focusing on different aspects of the sounds (e.g. focusing on vocals versus a certain instrument), or difficulty in using the evaluation sheet.
In an attempt to draw more information out of the data, the data was analyzed in two ways. First was a normal averaging of data. This produced semi-insignificant data. Second was to calculate the perceived differences. To accomplish this, the absolute values of all of the rankings were averaged. This produced much more meaningful data.
Shown in Table III is a summary of the results, both Averages and Perceived Differences. Recall that these values are on a scale of -5 to +5. Notice how many of the Rankings are nearly zero, but when compared to the Perceived Difference for the same category, there is a marked difference.
For instance, the Average Ranking for Presence for the Alesis 3630 is a 0.05, while the Perceived Difference for the same compressor is a 1.75.
| Table III: Summary of Results |
| Average Rankings by Compressor |
| Compressor |
Category |
|
Intelligibility |
Apparent Size |
Presence |
Dynamics |
| Alesis 3630 |
-0.45 |
-0.5 |
0.05 |
-0.95 |
| Focusrite d3 |
0.05 |
0.33 |
0.83 |
-0.35 |
| Joe Meek SC2.2 |
-0.25 |
-0.38 |
-0.02 |
-0.6 |
| Symetrix 528E |
-0.55 |
0.75 |
-0.05 |
-0.95 |
|
|
|
|
|
| Average Rankings by Song |
| Song |
Category |
|
Intelligibility |
Apparent Size |
Presence |
Dynamics |
| Pigs |
0.4 |
0.25 |
0.48 |
-0.75 |
| Mackerel |
-0.2 |
0.45 |
0.08 |
-0.75 |
| Mambo |
0.1 |
0.35 |
1.05 |
-0.3 |
| Tool |
-1.5 |
-0.85 |
-0.8 |
-1.05 |
|
| Perceived Difference by Compressor |
| Compressor |
Category |
Overall |
|
Intelligibility |
Apparent Size |
Presence |
Dynamics |
|
| Alesis 3630 |
1.45 |
2 |
1.75 |
1.65 |
1.71 |
| Focusrite d3 |
0.75 |
1.28 |
1.08 |
1.05 |
1.04 |
| Joe Meek SC2.2 |
1.15 |
1.83 |
1.23 |
1.3 |
1.38 |
| Symetrix 528E |
1.45 |
2.25 |
1.85 |
1.65 |
1.8 |
|
| Perceived Difference by Song |
| Song |
Category |
Overall |
|
Intelligibility |
Apparent Size |
Presence |
Dynamics |
|
| Pigs |
0.8 |
1.3 |
1.23 |
1.45 |
1.19 |
| Mackerel |
0.8 |
1.35 |
0.93 |
1.05 |
1.03 |
| Mambo |
1.2 |
1.65 |
1.55 |
1.2 |
1.4 |
| Tool |
2 |
3.05 |
2.2 |
1.95 |
2.3 |
The data presented in Table III can be found, along with a complete analysis of the results in Appendix A.
Conclusions
As a result of this study, it was determined that different compressors, while set to the same settings, all behave significantly differently, imparting their own qualities on a sound. It is hoped that this study can provide some starting point for other recording engineers as they mix their projects.
Using the Perceived Difference results, and some of the information from the Average rankings, some generalized conclusions about each compressor can be made.
- The Symetrix 528E scored the highest in overall perceived difference (1.8). This compressor is designed for broadcast, a business where if you don't have a distinctive sound, it can cost you customers.
- The Alesis 3630 was second highest (1.71), and consistently had the highest or 2nd highest scores in all categories. Listeners (as determined by their feedback at the listening session) could pick out the Alesis after they were acquainted with it. The Alesis performed well as a general purpose compressor.
- The Joe Meek SC2.2 received a 1.38 in overall perceived difference. It scored a good 0.6 higher in Apparent Size than in any other category. Correlated with an Average Apparent Size ranking of -0.38, this compressor should be considered if the size of a source needs to be reduced, without affecting its other qualities significantly.
- The Focusrite d3 received the lowest perceived difference ranking of 1.04. On average, it was considered to reduce the Dynamics of most signals while increasing Presence and largely ignoring Apparent Size and Intelligability. Given these traits, the d3 would perform very well in a mastering situation.
In terms of the program material, the Average Rankings were of little use, but the Perceived Differences were quite useful.
- The Tool - Lateralis sample was impacted the most by the compressors, coming out with an average perceived difference of 2.3.
- Momentum - Mambo Swing was also rated fairly high, with an average perceived difference of 1.4.
The fact that these two samples were just instrumental, as opposed to an entire mixed song, contributed highly to their malleability by compression. Either of these samples can be considered representative of how compressors affect drums and pianos, respectively.
- Pink Floyd's Pigs (Three Different Ones) received a score of 1.19.
- Les Claypool's Holy Mackerel received a score of 1.03.
These low rankings can be attributed to the fact that they were already mastered recordings, having been previously subjected to compression. In particular, Holy Mackerel is a modern recording, and so suffers from "soundloaf" syndrome, meaning it has already been compressed to such a degree that it looks like a large square wave.
Appendix A: Results
Appendix B: Raw Data
|
