Trust, But Verify – Frequency Response and Distortion Using Room EQ Wizard Mike Rivers © 2019 Prologue “Trust, but verify” is from a Russian proverb, a phrase that former US President Ronald Reagan frequently used when speaking about the Soviet Union’s compliance with a nuclear disarmament treaty. If you’ve followed my articles about specifications here - what’s important, what’s not, and what’s missing from the published spec sheet - I think you’ll agree that the proverb fits well. While manufacturers rarely publish a specification that a product can’t meet by hook or crook, measurement methods or missing tolerances can make something look better on paper than it actually is. In this occasional series, I’ll explain how to make measurements on your own gear using free or inexpensive software and tools you might already own. This is more “proofing” than “testing” in that you won’t get lab accuracy with the methods I’m describing here, but you’ll be able to compare its performance against its published specs, fill in the missing pieces, and have a better understanding of what it can do, and how well. We’ll be measuring voltage, current, and resistance or impedance, distortion, frequency response, delay (latency), dynamic response, and maybe some other characteristics. There’s plenty to learn, and much fun to be had in the process. Some Basic Concepts of Test and Measurement Most testing involves a stimulus-response setup. You connect a signal source (the stimulus) to the input of the device under test (DUT) and measure the output (the response). There are absolute and relative measurements. A preamp’s maximum output level is an absolute measurement, one that needs to be measured with a calibrated voltmeter. The characteristics of the preamp’s low-cut filter, however, involve relative measurements. We’re not interested in the actual output voltage of the preamp, but rather, the change in output voltage over the frequency range where the filter is doing its job. It’s important to understand the difference between accuracy and resolution. A digital voltmeter will measure a fresh 1.5v alkaline battery at around1.566 volts. The meter’s resolution is to 3 decimal places or 1/1000th of a volt (1 mV). However, if the meter’s specified accuracy is ±2%, the actual voltage could be anywhere between about 1.535 and 1.597 volts. Don’t be bamboozled by a measurement with lots of decimal places (resolution) without knowing the measuring accuracy. Digital voltmeters, from where these “to three decimal places” numbers come, have their own set of potholes, but I’ll address them in another installment. Before checking out a piece of gear, decide what to measure and have a realistic expectation of the results. A mic preamp’s frequency response will be very flat, but if it has a low-cut filter, you’ll want to measure its frequency response with the filter switched both in and out. If you’re investigating a preamp’s “character,” you’ll want to look at distortion both with a test signal that’s comfortably below the maximum input level and with it driven into distortion. And you’ll want to look both at distortion with the input overdriven, and with the output overdriven. There’s a difference. If your measurements differ substantially from your expectations, you might be measuring incorrectly, measuring the wrong thing, the device might be broken, or you might just be surprised! Lastly, document your testing so you can understand your results. Now, let’s measure some stuff. We’re Off To See The Wizard Room EQ Wizard, that is, REW for short. It’s a free (donations cheerfully accepted) program with versions that support Windows, Mac OS-x, and Linux. As its name suggests, REW is designed for measuring room acoustics, but we’re not going use it in that manner (though you certainly can, for other measurements). For its intended application, REW generates a test signal that’s sent out to the room’s speakers via your computer’s audio interface. A calibrated microphone connected to the interface’s input picks up the room sound, and the difference between the test signal and the microphone’s output is analyzed and displayed. For the measurements discussed here, we’ll replace the speakers, room, and microphone with the DUT, with REW both providing the stimulus (test signal) and measuring the device’s response. It’s a cool tool, but there are some caveats. While the accuracy of your measurements is limited by the linearity of your interface, most contemporary recording interfaces are sufficiently accurate and linear for making most audio measurements. Connections between the DUT and interface are mostly straightforward, but be aware that there’s no single best way to mix balanced and unbalanced interconnections – grounding or floating the wrong lead might degrade your test results. Remember, too, that every connection is a potential noise source, so avoid stringing adapters together and take the time to make or modify cables for the most direct connections. It’s also important to understand headroom and manage gain and signal levels appropriately. REW, by default, expects an input level from the DUT that allows about 12 dB of digital headroom, and will complain if it detects clipping (0 dBFS). You probably allow more headroom than this when recording with your DAW, so treat your REW setup with the same care. Driving a mic preamp into the “warm- and-phat” range from the interface’s output is easy whether the preamp’s input has a transformer or is transformerless, but your interface’s maximum output level may be incapable of driving a line level input transformer into saturation. Many of today’s interfaces offer “zero latency” input monitoring that sends the input signal directly to the output, avoiding a trip through the computer. For our use here, an interface’s direct input monitoring must be turned off. REW needs to see only the test signal coming from the DUT, and engaging direct monitoring can cause feedback. Lastly, recognize that you’re in Computer Wonderland here, a place where no two computers are alike, and sometimes differences can cause unplanned behavior. I’ve only used REW on Windows XP and Windows 7. It’s fair to assume that the other OS versions work similarly, but one never knows. I have some tips here for setting up and using REQ for measuring your audio gear, but this is intended neither as a tutorial nor a review of the program. REW Setup and Calibration Setting up REW to work with your interface is essentially the same as setting up a DAW. Open the Preferences window, click the Soundcard tab, and choose the input and output to which you’ll use for connecting the DUT. REW is a single channel processor, so you’ll need two passes to measure a stereo device. Select the ASIO driver if your interface has one, otherwise, for Windows, select Java, which uses the standard Microsoft sound card driver. The Mac version uses the OS-X Core Audio device selection. 44.1 kHz sample rate is adequate for measuring across the conventional audio bandwidth, but with a higher sample rate, you can measure at frequencies above 20 kHz. The rest of the settings can be left at their defaults for now. The next step is to calibrate the interface. First, connect a cable between the interface’s output and input. If your interface has input gain and output level controls, set them to their normal working positions to approximate unity gain. Click the Check Levels button and, if necessary, fiddle with the interface’s input and/or output controls to make REW’s meters match within a couple of dB. If the interface has a Clip indicator, be sure it remains off. Your goal is for both the Output and Input meters on the Calibrate screen to indicate close to –12 dB. When you get the levels set properly, click the Calibrate button, and instructions at the bottom of the page will walk you through the procedure. This measures the frequency response of your interface and creates a correction table so you’re your measurements will be corrected for any irregularities in your interface’s frequency response. Since the calibration sweep starts at 0 Hz, low frequency roll-off is normal, as is the sharp drop-off above the Nyquist frequency (half the sample rate). If all is well, the frequency response will be flat within a few tenths of a dB over the normal audio range. Finally, click the Make Cal button (that’s short for “use this as the calibrated interface”), name and save it, and your test system is ready to go. Measurements during a session are stacked at the left of the screen. REQ’s test signal provides data for several different measurements in one shot. Tabs above the plot area of the measurement screen select what’s displayed on the graph. “SPL & Phase” shows the most recent measurement or one you highlight in the stack. “All SPL” superimposes the frequency response of all of the measurements on a single graph. Check boxes at the bottom the screen allow you to choose which measurements are included in the plot. The graph here shows six frequency response measurements of a graphic equalizer with a different setting of the sliders on each test run. The camera icon at the left takes a screenshot to go into your test report. You can save your measurement data in a file so you can view measurements later or look at other data from the same test run.