Home> GEQ Shootout
 | GEQ Shootout
On December 17, 2003, I put together a GEQ Shootout with the intention of finding the best 1/3rd-octave equalizer for use with monitor wedges. The Noe Valley Music Series performance space was available for the day so we met there at the Noe Valley Ministry in San Francisco. A couple of local audio guys that I knew and some that I hadn't yet met from www.live-audio.com came with EQs to test. Mike Allen ran a set of precision bench test equipment and I ran a SmaartLive station. There was a monitor wedge setup that several guys used to compare real world use. All in all there were 32 EQ models that got tested in some fashion. Because of time constraints, not every EQ was given the full treatment.
I announced the event on LAB and there was some interesting discussion online. Here's the original announcement and people's comments. Here we have further suggestions regarding graphic equalizer models that would be interesting to test.  Photo by Carl Liss.Precision Bench Michael Allen came up with a series of exacting precision tests using his HP signal generator and balanced input oscilloscope. He had time only to test eight of the 32 1/3-octave EQs plus a vintage biamp EQ210 stereo 10-band he brought as a personal reference. A stowaway! He examined noise on the outputs both loaded and unloaded, line driving abilities, CMRR properties and silk-screening accuracy. Mike's data will be available shortly.  Acoustic Feedback Allen Craft, Gabe Nahshon and Luke Sheridan assembled a vocal mic and wedge system on the carpeted stage for testing acoustic feedback reduction properties of EQs. They used a pair of Tim O'Connor and Allen Craft's Harbingers and also a pair of Luke's Nexo wedges all sourced with a Shure Beta58 vocal mic split to several faders of an A&H GL-1. One fader would bring up one EQ. The mixer fed a Crown MA1200 bridged. Both wedges of a pair were tried and then just one. Radley's Crown CM-310A differoid vocal mic was auditioned as well. Each EQ was tweaked for maximum GBF then compared with the previous ones. Our brave crew tested seven of the 32 EQs, achieving the greatest gain with the Audient ASP231 and Rane DEQ 60L models. Every one of those seven sounded smooth except for the Behringer GEQ 3102 which Luke thought was 'choppy' and Allen reported as 'dark and small.' For some reason, the XTA required 8dB of makeup gain to get the Nexos going.  SmaartLive tests I put together for this event: Unless included as part of the test, EQ processing options such as HPF, limiting and notch filters are bypassed or dialed out as much as possible. Pink noise at 0.7vrms from a balanced CD player was the default source signal.  Flat Pink First thing we looked at in SmaartLive is frequency and phase response with all faders at zero. If an EQ doesn't do well at this stage you've got bad news! One of the vintage UREI 527-A units pulled from storage and the new Presonus DEQ624 displayed problems during Flat Pink. We're talking about full frequency problems for UREI #518 and two minor 1kHz and 5kHz dips on channel A of the new Presonus. A much better result is shown above: the KT DN360's freq line is nearly flat and the top phase line curves gently down in the HF region.  Death To 100 In this test we bottomed out the 100Hz fader and every other fader lower than it. I don't recommend this as a method for controlling lows -- I've always been told that it messes up the phase response, has weird frequency issues and often isn't necessary since so many EQs come with variable HPF. But, MAN, there's someone doing this at nearly every show I go to. What's up with that? Here we see the Rane DEQ 60L demonstrating 'Death To 100' although this model is perfectly capable of cutting lows by using the variable high-pass filter labeled 'Low Cut.'  Frown Next we returned the faders to zero and proceeded to carve a small frown going from 250 to 500Hz. Our accuracy in positioning faders depended in no small part on front face silk screening clarity. Specified settings are 250Hz down 10dB, 315Hz down 5dB, 400Hz down 5dB and 500H down 10dB. Here we see the 'Frown' demoed on one of the vintage UREI 527-A EQs.  Capital V This procedure moves three sliders to examine filter interplay. 315Hz goes up to +10dB, 500Hz stays down at -10dB from the last test and 800Hz is moved up to +10dB. It's plain as day if you look at the Audient ASP231 pictured here. The boosting you see is quite a bit more than one would typically use on monitor wedges but I figured that by examining the responses I would be able to separate unfamiliar GEQs into one of several filter classes. One thing to look for is to note the level of 500Hz on the computer screenshot. Is it higher than -10dB? Are the 315 and 800 filters pulling it upward?  Wavy Now we get a combination of boosts, cuts and a low cut filter placed at 200Hz, if possible. Some EQs had no HPF and some had a fixed setting but on any of the ones with variable low cut I began by flattening out the faders, noting the dB level at 200Hz and then increasing the frequency of low cut until 200Hz showed a reduction of -3dB. Fader settings for this test were 400Hz -3dB, 500 -6, 630 -6, 800 -3, 1.25k +3, 1.6k +6, 2k +6 and 2.5kHz +3dB. In the photo you can see the ART HQ-31 set for the wavy test; its High Pass filter knob is shown in the inset.  Dude Is Deaf Silly? Yes, this test might be called that. We've pumped the high frequencies by a huge amount to see if phase and frequency stay smooth. Monitors might need a little of this for an aging rock'n'roller who came up the hard way. Hopefully, none of you will ever get this particular guy in your club! With the Nady GEQ-231 early blackface version serving as the model you can see these fader settings: 2.5kHz +3dB, 3.15 +6, 4k +9, 5k +9, 6.3k +6 and 8kHz +3dB. Ouch! Where're my earplugs!  915Hz Cut In this procedure we fed a hot 915Hz sine tone (measuring 3.36v on my tiny Rat Shack auto-ranging digital voltmeter) from the CD player into the Device Under Test (DUT), noted the SPL of 915Hz showing on screen with faders flat then cut as much 915Hz as possible and jotted down the difference. The idea was that we would see how well an EQ performed on difficult feedback tones that fall between ISO centers. I picked 915Hz because it is pretty darn close to halfway between the 800 slider and the 1000 slider, making it tough to control. After applying this test two or three times it dawned on me that it didn't really tell us much about real world usage and that there were too many variables at play such as 1/3-octave filter depths being 10, 12, 15 or even 24dB down and whether or not there were notch filters included in the DUT. Still, we went ahead and kept this procedure in the sequence.  915Hz Down 6dB After doing some thinking about how to have the 915Hz test relate to the real world I came up with a better idea and applied it to as many EQs as I had time for. In this vastly improved version I fed some bandpassed pink noise into the DUT to get an overall SPL reading. To obtain bandpassed pink the original pink noise WAV file was high-passed at 200Hz and low-passed at 4kHz using 24dB/octave filters applied in the audio editing program and burned to CD. This allowed me to level the playing field somewhat by excluding extreme LF and HF responses and paying closer attention to the mid-freq power bandwidth. After noting the SPL, I stopped the bandpassed pink and started the 915Hz sine tone. The level at 915Hz is noted, then steps are taken to reduce it by 6dB. Most EQs got to this point using equal amounts of 800Hz and 1000Hz slider cuts but a few included assignable notch filters or anti-feedback circuitry. Finally, I stopped the sine tone and restarted the bandpassed pink noise, retaining the altered EQ settings. I noted the change in overall SPL. Each EQ was examined using as many methods of cutting 915Hz as seemed appropriate; parametric or semi-parametric notches, anti-feedback filters and plain ol' regular sliders. This test should be able to predict how much power and fullness are retained by the EQ when you have to cut a feedback mode that falls between sliders. Your musician clients will appreciate your ability to keep levels up and musicality high even if you have to chop a few room modes and feedback tones. A small change in overall SPL is better than a big drop. Extreme phase changes need to be assessed as well. A sharp filter that throws a phasey curve at your signal might be detrimental to the musicality. In terms of frequency alone, most EQs fell in the window of 1.0 to 2.0dB reduction in overall SPL after cutting 6dB of 915Hz. A few were better and any with extra-wide filters were worse. In the screenshot you can see the Audio Logic SC31 (out of production) demonstrating typical performance of a mid-pack front-runner -- only 1.1dB of overall SPL drop.  Multiple Cuts The concept of how to test for the ability of a GEQ to do well at monitor wedge duties kept evolving with each new round of information returning from the previous trial. In this late-stage procedure, I made a four-tone chord out of pure sine waves and sent this to the device under test. The EQ being tested was to attenuate a specified amount each of the four frequencies and the resulting overall loss in SPL was to be quantified. The idea was that EQs which had adjacent filters summing to the center might not be as good at cutting a grouping of several frequencies as they are at cutting one alone and that engaging a few filters in a row might result in unavoidable compromises in terms of retaining SPL. Monitor wedge feedback is usually such that there is usually more than one feedback frequency and yet wholesale cutting of modes makes for a weak, tinny or distant wedge mix. I also was curious about how the new various filter topologies compared to one another. Perfect Q, Proportional Q and Constant Q were among the contenders as well as Assymetric Q and parametric filtering. The KT DN9340's the one with parametric filters so naturally, it performed the best. But I also ran into intermodulation distortion (IMD) artifacts during this test. What a surprise!  Torture Pulse For this test we changed over from SmaartLive mode to recording mode. A series of non-frequency-specific square pulses were fed to the DUT and recorded onto my IBM ThinkPad T30's line level minijack inputs. A few cuts were introduced into the DUT in order to ensure that our EQ was doing some work and not idling with filters disabled at their detent positions. In the photo you can see the NEI 2711 (out of production) displaying the specified EQ settings of 315Hz -3dB, 400 -6, 500 -6 and 630Hz -3dB. One channel of the ThinkPad was fed EQ'ed pulses and the other was fed straight pulses. This allows for the subtraction of the ThinkPad's response from the EQ'ed DUT response, though some argument can be made that the EQ'ed signal will hit the ThinkPad a little differently making it respond in a non-linear fashion between its Left and Right channels. At least all DUTs were treated the exact same way, so intelligent comparisons are still possible.  At the top of this graph you can see the original pulses taken from the first of two groupings. They are perfectly square with every non-zero sample at the same value -- about 70% of maximum level. This what they looked like coming in to the EQ getting tested. Next level down shows what the computer thinks of as an ideal response given the EQ settings specified. At the bottom you can see the so-called ideal response subtracted from the original pulses. Analyzing the resulting recordings allows us to look at several things. We can check to see if any device takes too much time to ramp up to full level after a period of silence, we can compare ski-jump tops of EQ'ed pulses and we can see how a device responds to a sudden absence of signal. A further inspection of these audio files might yield additional information.
I didn't get to test IMD distortion until two weeks later when the number of participating EQs was down to three. Too bad! John Roberts and I had discussed on LAB how this test might quickly determine problems in sloppy circuit design so I had a 19kHz plus 20kHz dual sine tone WAV file ready to go. Unfortunately, I was worried about my laptop or a CD player only doing 16 bits at 44.1kHz sample rate (standard CD quality) so I felt sure that the sources available didn't have enough resolution to make this test meaningful. I was hoping that a participant would arrive with a laptop-controlled USB device capable of putting out a stream of +4 line level audio made up of 20 or 24 bits at 96kHz. Better than that would have been two lab-quality analog signal generators buffered into one balanced output. It might have been Don Boomer or T.Alan Kraus who suggested an alternative of using two lower HF tones such as 15kHz and 16kHz -- I almost took them up on that. Time constraints kept this test out of the picture on Shootout day. It was only until later in Multiple Cuts that I ran across IMD artifacts, showing me just how easy they are to test for.
I also didn't get to test how well each EQ was able to withstand the RFI generated by a cell phone call. Steve Snyder and I both had Nextel phones in our possession at the venue -- we are all too aware of how much the Nextel ring sequence interferes with audio gear. The problem I had on Shootout day was that we needed to record the output of the DUT for subsequent analysis. My ThinkPad could have done the recordings but it was too busy with other tests -- we didn't pull aside another test device. Also, Mike Allen's bench test procedure measuring CMRR promised to predict which EQs were going to fail the Nextel phone test so it seemed redundant at the time. In the event, Mike's careful and exacting procedure proved time-consuming enough that only a quarter of the EQs were tested for CMRR.
It would have been interesting to see whether a tiny latency delay could have improved any of the analog EQs in terms of GBF during stage monitor duties. Don Boomer was wondering if the digital EQs all benefit from their inherent delay. Quite likely... In the end, we didn't go to the extra trouble of comparing an analog EQ at time zero with a slightly delayed version of itself. Doing so would have taken more time and would have introduced all the quirks and characteristics of the digital device assigned to delay duties. A complicated test, to be sure.
Another test I was asked to perform was to check on headroom differences between an EQ's �dB mode and its �B mode. Lew Veldas on LAB mentioned that one of the Yamaha EQs he's seen seems to clip sooner when in �B mode. I was intending to check this on every EQ but in the event I felt that it was more important to capture all the SmaartLive traces that I could. The single EQ model I chose to investigate this possibility was the one Yamaha Q2031B taking part in the Shootout; a piece of gear which defined the lower limits of anti-feedback performance due to its wide filters. I found that the clip light didn't really indicate actual clipping at �B but it came on nonetheless when the signal was approaching maximum. This is typical -- many EQs have a clip light that begins to shine when you are within 3dB or so of clipping. But in �B mode the Yamaha's clip light would come on 6dB in advance of clipping. The really odd thing was that in �dB mode the clip light would come on several dB AFTER the onset of clipping. Not a very good indication for the user! Perhaps the unit I was testing was faulty but both channels tested the same.
Finally, I would have LOVED to get hard numbers in the live acoustic testing part of the Shootout. My intent was to have a SmaartLive testing station with a measurement mic at musician ear height and to have a series of expert sound folks tweak a set of EQs for maximum gain-before-feedback. Then pink noise (or some appropriate version of bandpassed pink) would be fed to the monitor wedges through the DUT and our SmaartLive mic would tell us the overall SPL reading. Higher SPL would indicate that the EQ allowed more GBF and/or that the person doing the EQ tweaking was better able to complete the task when using that device. Shifting EQ tweak duties around to various participants would have helped drop the human out of the equation, letting all the best EQs shine through. Having two separate live acoustic setups would have been even better. At the Shootout, there were too few SmaartLive stations and too little time allowed for high-SPL tests on stage for us to arrive at numbers like this. If I had been able to set up a couple of testing areas a day in advance along with laying out a hard-and-fast testing regimen and if I had been able to rope in a willing Smaartie or two who could wrangle attendees and EQs for tweaking then such a real-world measurement would have taken on a much greater significance. As it turned out, the best we could do with the time we had was to jot down some subjective findings on the order of Sucks/Doesn't Suck!
EQs that got tested in some manner:
AB Intl 231 (donated [!] by Paul Bell [NOW what am I going to do with this thing!]) ART HQ-31 (courtesy of Dave Lowum of Tantra Tour Sound) Ashly GQX 3102 (courtesy of Rod Thompson of Suntrack Sound via Luke Sheridan) Ashly Protea 4.24G (courtesy of Rod Thompson of Suntrack Sound via Luke Sheridan) Audient ASP231 (thanks to Jeff Harmon at Audient USA for the demo) Audio Logic SC31 (courtesy of Michael Beyl of Swanson Sound) Behringer DSP8024 (courtesy of Allen Craft) [WHAT? This thing has no sliders! Get it outta here...] Behringer GEQ3102 (one from Bananas At Large and one from NY courtesy of Paul Bell BSS FCS-960 (courtesy of Don Lind of Golden Gate Sound) dbx 3231L (courtesy of Sarah Roos of BBI) dbx 480 (courtesy of Luke Sheridan) dbx iEQ31 (courtesy of Lee Jacobson) KT DN27 (courtesy of Jeff Cleland at Hyde Street Studios) KT DN360 (courtesy of Sam Lunetta of EventCom at the San Francisco Marriott) KT DN9340 Helix (thanks to Jay Easley of EV/KT/Midas/Telex for the demo) Lake Contour (courtesy of Miguel Hadelich of Lake) Nady GEQ-231 [early blackface] (courtesy of Russ Davis) NEI 2711 (courtesy of Dan Pockrus) Peavey EQ31FX (courtesy of Bananas At Large via Allen Craft) Presonus DEQ624 (courtesy of Paul Bell) Rane DEQ 60L (thanks to Stephen Macatee at Rane for the demo) Rane GE 27 (courtesy of Sarah Roos of BBI) Rane GE 30 (courtesy of Sarah Roos of BBI) Rane ME 30 [older model without XLR] (courtesy of Allen Craft) Rane ME 60 (courtesy of Rick L'Heureux of AVT) Rane MQ 302S 'Mojo' (courtesy of Rick L'Heureux of AVT) Sabine GRQ 3101 Graphi-Q (courtesy of Philip Brady) Sabine GRQ 3102 Graphi-Q (courtesy of Rick L'Heureux of AVT) TDM 30GE-2 (courtesy of Don Boomer of ICBM Live) UREI 527-A (a pair of them courtesy of Tom Hall) XTA GQ600 (courtesy of Greg Knight of Ultramagnetic via Luke Sheridan) Yamaha Q2031B (courtesy of Brett Bailey of American AV at the Palace Hotel)
Present but not tested:
Altec 1650 [vintage green-face with security cover] (courtesy of Radley Hirsch of San Francisco Audio) Symetrix 533E (courtesy of Radley Hirsch of San Francisco Audio)
Mentioned in pre-Shootout discussions but didn't show:
Alesis MEQ230 Behringer DEQ 2496 biamp EQ290 dbx 1530 dbx 1231 IRP TEQ DG-4023 transversal, although it was included six weeks later when it did arrive! White 4400
Mike Allen's stereo 10-band Stowaway:
biamp EQ210
Participants
Michael 'Bink' Knowles Michael Allen Steve Snyder Luke Sheridan -- here's Luke's Take R. Allen 'Alien' Craft, Jr. Gabe Nahshon Don Boomer Radley Hirsch Don Lind T.Alan Kraus Miguel Hadelich Ervin Grinberg Carl Liss Robert Botsford Philip Brady
Note: uncredited photos, screenshots and graphic renderings are by Michael 'Bink' Knowles
|
