Multiple Cuts

<<Previous  Next>>

Here's where the whole process of analyzing the GEQ Shootout results came to a head. I kept thinking about how to challenge an equalizer in a way that represented a tough monitor wedge situation and the idea of notching out multiple cuts started to take shape.

In the Multiple Cuts trial, I set overall SPL level with bandpassed pink noise, then I fed four sine tones simultaneously to the Device Under Test (DUT.) Using faders or other filters, I cut the four sine tones down by a specified amount. Then I returned to bandpassed pink and noted the resulting reduction of SPL.

Sound familiar? This procedure is the same as 915Hz Down 6dB but with four sine tones instead of one. In cutting the four, I went back and forth between filters, checking and rechecking levels. Filter interplay was such that a slight change in one fader meant going back to retweak the others. Much more time per DUT is required in this series!

By the time I formulated the Multiple Cuts trial, there were only three EQs left to test: the three that were on loan as demos from manufacturers. All others had been returned to their owners. Participating GEQs included the Audient ASP231, the Klark Teknik DN9340 Helix and the Rane DEQ 60L.

Five weeks later the IRP TEQ DG-4023 transversal EQ arrived and joined the Multiple Cut trial.

To select four sine tones in an unbiased manner I used a random number generator available free online, limiting the results to integers between 200 and 2000. Then I nudged the four numbers a few Hz nearer to specific SmaartLive cursor points -- the kind that show precisely as you mouse over the trace. These SmaartLive points change with sample rate and viewing window size -- your results will vary. Here are the Multiple Cuts sine tones and the target amount each one should be reduced in this test:

Signal level of the four-tone chord coming out of the CD player metered 1.47 volts between XLR pins 2 and 3 on my tiny Radio Shack voltmeter.

This version of bandpassed pink noise is one octave wider than in the last procedure, being crossed over at 100Hz and 4000Hz. It's also compressed somewhat to make overall SPL determination a little easier. I made a test CD with three tracks including bandpassed pink, the four-tone chord, then bandpassed pink again. Signal source was the same balanced output CD player used in all previous GEQ Shootout Smaart trials.

I tested each EQ in all relevant modes of operation. The Helix Proportional Q mode and its DN360 emulation mode proved to respond differently than each other though I thought they would be identical. The Rane included a Cut Only mode which gave finer adjustments but didn't change the end results in Perfect Q mode. However, the Cut Only mode prevented any of the gentle boosting of neighboring sliders I felt was needed in Proportional Q mode because of filter interaction. Finally, the Audient's two modes were (assymetrical) Narrow Cut and (symmetrical) Wide Boost/Cut. Its Narrow Cut mode was the one that did so well in cutting 915Hz.

The Helix changes level in 0.5dB increments on all faders. This kind of chunky stepwise increment prevented it from sticking as close to target sine tone reduction levels as the Rane and the Audient. While nudging filters this way and that, I noticed that an excellent response might be obtained at a point maybe 0.1dB short of the filter level I was aiming for. Even though in live sound show mode I would have settled for slightly lower reduction of one frequency in favor of retaining a better total response, on this test the hard numbers were king. At no time was a DUT allowed to post a performance if the target sine tone reduction was not precisely met or slightly exceeded.

A big problem was immediately apparent on channel A of the KT: while it passed pink noise quite well at a level showing as 9dB below max on the front panel LED ladder (Helix), the four-tone chord showed a large amount of distortion at the same input level. The Helix's channel B and the Audient had distortion at the same frequencies but at lower levels. I saw multiple distortion artifacts that were lost in the noise floor on the Rane. These frequency spikes appeared about 56dB below the four sine tones on channel A of the Helix and about 24dB above the noise. They appeared down about 75dB below signal on the Audient and on channel B of the Helix. Less pronounced, though visible. They appeared at frequencies related to the interplay of the four sine tones; for instance, one frequency came in at around 1.21kHz which is most likely the sum of 446.8 and 759. Ouch! Just to determine how much intermodulation ugliness was going on, I reduced input level to the Helix until the distortion artifacts disappeared into the noise floor. That input level was ~16dB lower, lighting up the bottom two of the DN9340's eight green input LEDs. The absence or presence of signal on the EQ's other channel did not change anything. Changing EQ level on the Helix didn't change things, either, which seems to point to D/A converter distortion. Barreling onward, I returned the faders and gains to unity, ignored the IMD spikes, and continued testing the Audient and the Helix at the same input level which the Rane handled without trouble. Fair's fair. Any Smaart user who would like to examine the full results from inside SmaartLive can download the RGP file with six snapshots I took including one of each channel on all three equalizers. Right-click on the link and save it to disk. Open SmaartLive and load the RGP file using the shortcut ALT-R and clicking on 'Load All' -- you'll need to browse to the place where you stored the RGP file. Get rid of these traces and back to your live sound duties by using the shortcut Ctrl-Shift-Delete. Have fun!

Here are the results in Excel spreadsheet format.

Best retention of SPL: KT DN9340 using parametric filters (is anybody surprised by this?)
Best retention of SPL using only graphic filters: IRP TEQ DG-4023
Worst SPL loss: KT DN9340 in Assymetrical mode (what were they thinking?)

After obtaining the Multiple Cut results I refined the search for IMD. I finally decided to implement the IMD ITU-R test using 19k and 20k tones at a 1:1 ratio as suggested by John Roberts (former Peavey engineer) and as laid out in Dennis Bohn's Audio Specifications on the Rane website. The 19kHz and 20kHz tones were burned to CD and played back into the laptop where it was immediately obvious that either the CDR, the CD player or the laptop was putting a 1000Hz IMD artifact into the signal when stimulated by 19k and 20k at a hot level. This was exactly what was being looked for, but not on the test equipment! Noting this, I moved on to testing EQs. Each EQ is shown at idle here, no filters are engaged. To determine how each EQ handled the ITU-R IMD tones, you will need to mentally subtract its trace from the CD+laptop trace at the bottom. Here are the hard IMD numbers, corrected for the laptop's own contribution:

Rane DEQ 60L: 0dB of 1kHz
Audient ASP231: 15.3dB of 1kHz
KT DN9340 Helix: 49.6B of 1kHz

However, as you can see, the Rane DEQ added other frequency artifacts in groupings clustered near 1.5khz, 3kHz and 4kHz as well as four more clusters between 8k and 12kHz. I don't know the nature of these groupings or what they would be called other than distortion. With just an eyeball comparison it seems the Audient added the least amount of artifact power.

All three EQs show 60, 180, 300 and higher AC line power harmonics bumping into the signal. I wouldn't rule out a bal/unbal interconnect problem or ground loop problem with the test setup when looking to explain this. No isolation transformers, pin 1 lifts or AC ground lifts were used in the test rig. Having AC line harmonics present doesn't obscure the IMD results.

All four of these traces were recorded into the same laptop input channel at the same high input level, one at a time, then they were copied into a graphics program with different colors and differing vertical offsets for visual clarity. The input level showed as full green plus one yellow LED on the Rane's LED ladder and as 12dB below max on the Helix but my inadequate Radio Shack digital voltmeter was only able to see 0.088 volts between signal hot and signal cold due to the high frequencies involved. I'm sure the Radio Shack engineering team never expected more than 50 or 60Hz as the AC frequency!

I'd like to add that none of the distortion artifacts displayed are audible to me with music as the signal source. I can't tell the difference between the straight source and one sent through one of these EQs at a level hot enough to pick up distortion artifacts. I think the main reason would be the much stronger musical signal masking the low-level distortion. But another reason why the IMD isn't a big problem is that musical peaks are brief and any associated intermodulation distortion artifacts would dip in for only a short moment. Finally, even though IMD is clearly non-harmonic when using test tones, with a musical input the typical source tones are musically related, creating IMD artifacts that become part of the harmonic content. For instance, a strong synthesizer tone at 300Hz that has strong harmonics of 600Hz and 900Hz could create IMD points at 300, 600, 900, 1200 and 1500. Clearly, these frequencies would tend to subtly blend in as part of the musical timbre of the synth. My ears aren't 'golden' so I have no trouble ignoring IMD problems.