ETF4; MIDILyser

ETF4 can display captured data in a huge variety of ways, so that you can analyse any problems with the existing acoustics of your room. Fortunately the mid‑range waterfall plot of my studio is fairly problem‑free.

This month, Martin Walker reviews PC software packages that help you check out your own studio acoustics and examine your MIDI data.

ETF4 (Energy Time Frequency) is a low‑cost PC measurement program designed to extract a wide variety of acoustic information from a specially created test signal. Unlike Acoustic X, which was reviewed in the December '98 issue of SOS, ETF4 is primarily designed to troubleshoot the acoustics of existing installations, rather than suggest acoustic treatment based on room dimensions alone. It provides graphical readouts of room‑related frequency response, room resonances, early reflections, late reflections, comb filtering, RT60 (reverberation time), and harmonic distortion. It also features a separate utility to aid design of Helmholtz resonators and QRD diffusers.

The effects of room acoustics are important to every musician, since they may well determine whether your mixes will sound good in rooms other than the one where they were recorded. Using nearfield monitors avoids many of the problems associated with boominess caused by low‑frequency room modes. However, you can still improve your monitoring system by making sure that you minimise the effects of early reflections from nearby walls, the ceiling, and studio equipment, which will otherwise 'smear' the stereo image, and stop you hearing the subtleties in your mix. This happens because each reflection arrives at your ears at a slightly later time than the original direct sound, since it travels along a longer path. This is exactly the same as setting up a multi‑tap delay in an effect unit, and with taps set at a few milliseconds we all know what you get — a comb filter response. As each reflection arrives at your ear, another set of notches and peaks will occur. Fortunately, the reflections will be at a significantly lower level than the direct sound, but they will still colour the frequency response. Later reflections (after the first 20mS or so) should be well diffused, so that they support the overall sound as reverberation, rather than adding discrete echoes to it.

Although you can use a frequency‑sweep tone to determine the effects of room modes, it is normally much more difficult to evaluate the effects of reflections, since they are constantly changing over time. The most practical advice is to place absorption on the ceiling and two side walls at the positions where primary reflections occur — the so‑called 'mirror' points (as you sit at your normal listening position, you can see your loudspeakers in a mirror temporarily placed at these positions). However, while having golden ears is helpful, any computer program that can provide results in the time domain, rather than the frequency one, will be extremely valuable.

Slow Fix

MIDILyser is a handy utility for fault‑finding in MIDI systems, especially when you need to see what data is being sent to each channel.

ETF4 allows you to take just such measurements using a PC with a soundcard (the author claims that even low‑cost soundcards will provide perfectly adequate results). It is supplied on a mixed‑mode CD‑ROM, and should work on any Pentium (or equivalent) machine running Windows 95 or NT 4.0, although calculation times will be significantly longer with slower processors. Track 1 contains the program data, along with the manual and a selection of presets, and Tracks 2 to 11 are identical audio test signals, separated by 15 seconds.

You will also need a reasonable quality omnidirectional mic. The extensive help file suggests that the Tandy (Radio Shack) SPL meter will provide a good low‑cost mic/preamp combination, and since I already had one of these to hand, this is what I used.

You can measure in any one of three ways using the supplied test signals. The most automated solution is to use ETF4 to play the WAV file test signal through your test loudspeaker using your soundcard, while simultaneously recording the room response using a mic placed at the normal listening position (full duplex). The other two methods both involve playing back the same test signal from the audio part of the CD; the room response can then either be recorded using ETF4 and your soundcard (half duplex), or simply recorded into any WAV editing package (such as Wavelab) or using a DAT recorder, and later opened and analysed using ETF4. This latter option is ideal if you want to measure rooms where there is no computer close to hand.

I had a few problems running the full‑duplex tests inside ETF4, due to the lack of an input level meter, and so opened Wavelab. I then used this to play back the supplied WAV test signal, while simultaneously recording the output of my Tandy level meter, placed where my head normally is at the listening position. The WAV test signal contains two identical sets of tone sweeps, and the program uses pattern recognition to locate the beginning of the test signal (within 0.5mS), so that it can display early reflection times correctly. Even using the CD audio test signal, as long as you capture several sets of sweeps, ETF4 will have enough information to accurately identify the start point of the signal.

The program uses Time Delay Spectrometry during the analysis, and on opening my recording in ETF4 the full analysis took nearly a minute, even with my Pentium II 300MHz processor. However, this only has to be done once for each file, and the analysed file can be saved separately for instant recall. This also saves disk space, since my 3Mb WAV files were reduced to TDS files with a size of 222Kb. You can also average several sets of results to achieve more accurate figures, although I found good consistency between my measurements, despite widely varying recording levels.

Final Analysis

Once the analysis is complete, you can view the results in a variety of ways. The Energy/Time Curves (ETC) for the first 50 milliseconds are shown by default, for four different frequency bands, with the initial sound shown at 10mS. The results for my small studio were encouraging, and had the desired response of not too much activity above about ‑20dB during the initial 15mS or so after the direct sound (displayed on the graph between the 10 and 25mS points). The ETC can also be viewed over a longer 300mS period. To examine more closely the changing frequency response as the room decays, more familiar 3D waterfall plots are available, and these will show room resonances that may cause booms and honks. These can also be viewed as individual time‑slices, which makes it easier to study how such resonances develop over the decay time of a room. By comparing two sets of results — with and without a duvet spread over my mixing console — I soon identified a notch at about 170Hz caused by reflections from its top surface.

RT60 (reverberation) measurements can also be made, as well as Gated Frequency Response (which allows you to see the effects of early reflections on frequency response). Three gate times are available. The 3mS one will exclude most of the room effects; the 10 and 20mS ones show how comb filtering enters the picture, with the spacing of any peaks and notches helping to determine the extra path length, and therefore the likely source of the reflection.

A Harmonic Distortion plot shows the measured levels of 2nd through to 5th harmonics, which will mainly show the linearity of your loudspeakers, but might also pick up sympathetic resonances from other objects (computer monitors and radiators can be culprits). Once you have identified any problem areas, the Device Designer utility will provide you with all the correct dimensions to build suitable DIY Helmholtz resonators and QRD diffusers, for more absorption or diffusion at the problem frequency.

I've deliberately not attempted to over‑simplify this application, since you will need some basic understanding of acoustics to use it effectively, and interpreting the results will still take some experience. ETF4 is perhaps initially a bit daunting, but once you have read the manual it is in fact extremely easy to use, and both this and the helpfile are full of useful information. The graphics are a bit basic in places, and it would also be easier to make recordings if a level meter was available before or during recording. As it is, the recorded waveform is simply displayed after the 30‑second test signal has finished. Despite this, ETF4 should prove extremely useful to anyone designing loudspeakers, improving their studio acoustics, or even evaluating car stereo systems. At £94 including VAT, it is excellent value for money, and you can download a demo version to try for yourself from the ETF web site (www.etfacoustic.com). Marton Music can supply the full version in the UK, and also have suitable mic/preamp combinations available, as well as more advanced PC measurement systems.

E 106505.251@compuserve.com