To Preserve or Enhance?THAT IS THE UPGRADE QUESTION 11/01/2006 7:00 AM Eastern
Once upon a time, repair requests were “simple” and most customers were happy if the item in question passed audio. Ever since digital audio put analog gear under the microscope, questions about upgrades became just a bit harder to answer. No op amp is going to sound like a discrete Neve module. Op amps can and should be transparent. The past 40-plus years of IC evolution have left some room for improvement, although design engineers have managed to squeeze remarkable results from the tools of the day.
Op amps are not one-size-fits-all, but you can only learn by doing, even if a mistake is part of the equation. But knowing what's on the other side of the connector is helpful when you're doing an installation. It's taken years to acquire the discipline to preserve flaws that enhance sonic character — guitar amps being the most obvious. ICs that are wrapped around any flavor of gain-reduction circuit should be “colorless.” Vintage op amp flaws that get in the way of transparency can easily be treated. The minor battles are won at dynamic range extremes: just before headroom runs out and all the way down to the noise floor.
Awareness of, and striving for, 0dB full scale (kissing digital zeroes) has done a disservice to certain analog circuits. Pushing amplifiers toward clipping was not what the designers intended (unless you wanted a fuzz box). For example, it's not possible to overload the API 2520 (discrete op amp) if its 3-output transformer (xfrmr) windings are wired in series (most are).
The 2520 is slow by modern standards — 2 volts per microsecond — but you'd never know it. The transformer's 1:3 windings ratio provides more than 8 dB of headroom, and that increases the slew rate. You can apply this approach to any amplifier by taking advantage of 24-bit headroom, dropping the level 6 dB under the guise of maximizing resolution.
ON THE BENCH
I recently treated two stereo compressor/limiters, a UREI 1178 and a dbx 166. The 1178 preserves a portion of its 1176 heritage — the FET gain-reduction circuit — but to shoehorn two channels into a two-rackspace package, IC op amps replaced the discrete transistors. So much has been said about how the discrete circuitry and unique transformers contribute to the full-on 1176 experience, but for the optimist, here is a chance to hear the FET processing — solo.
The dbx 166 was an entry-level VCA processor with no-frills circuitry. The customer had several and wondered if an upgrade of ICs and/or passive components could improve the transparency. (There are less than a handful of ICs in the signal path.)
The UREI 1178 and the dbx 166 have transformerless inputs, a type of “line receiver” configuration also known as a differential amplifier because it subtracts one signal line from the other. “Balanced” signal transmission implies that there are two signals (pin 2 and pin 3 of an XLR; tip and ring of a ¼-inch connector), where one signal is of opposite polarity to the other on a twisted-pair cable. Sometimes, only one pin is modulated. In that case, the source/output impedance (of the previous device) should be the same for both of the signal lines so that noise will be equally radiated into the twisted pair.
Assuming the above, noise polarity is identical on both wires and subtraction literally removes the noise from the equation. The ability of the differential amplifier to amplify the signal and ignore the noise is known as the Common Mode Rejection Ratio (CMRR). Good CMRR is more about design and implementation than actual IC choices. You can see three of the common designs in Fig. 1. All but the last silver-faced version of the 1176 used an input transformer. Transformers are the most effective way to reject noise that is induced into the cabling, but they require more space, weight and cost.
The goal here is to heighten awareness of input circuit variations — they do affect CMRR — and of the fact that the variations exist partly due to a product's place in the audio timeline. Better designs contrast with the need to streamline production and minimize component costs.
How an op amp affects the audio signal depends on its demands. The post-processing makeup gain and output amps are separate, application-specific circuits. To best determine which op amps to choose, we need to determine the packaging options (Fig. 2).
To determine circuit performance, I apply a 1kHz square wave to the input and then poke a 'scope probe to the output of each amplifier. Figure 3 shows how the raw square wave should look on an oscilloscope; use a “x10” probe and calibrate according to the instructions. Note how the vertical component is nearly invisible on the upper trace (the raw oscillator output), as compared to the lower trace after traveling through a slow op amp at max output.
If you don't have a 'scope, capturing the wave on your workstation at the max sample rate will at least shed light on the one hazard in this game — oscillation outside the range of human hearing — although it won't be enough to determine performance.
Always monitor the oscillator's output to make sure that the signal is not loaded down by cabling and input circuitry. The capacitive loading effect will be similar to the lower trace of Fig. 3 and may be caused by capacitors such as C1 and C2 (across the 1178's input). Their purpose is to filter unwanted, out-of-the-audio-band interference, but they may also slow down the square wave enough to make it hard to see the effect of different op amps.
You may want to de-solder and lift one leg of each cap, as well as C3 (in the feedback loop of UREI's IC-1b), during IC testing, and then replace and see what happens. The wave should be square, with neither rounded edges nor spikes. Square waves are also good for checking capacitors.
Once you've determined the quality of the input amplifiers, the next test point is the amplifier following the gain-reduction circuit. An optical limiter may slow down the vertical portion of the square wave due the attenuator network's high impedance — it becomes very sensitive to capacitance. However, with 6 dB of gain reduction, the wave's response will improve. All transformer-coupled audio gear required a 600-ohm load, and without a transformer as destination, a load or termination resistor is required. SCSI, video and S/PDIF also require termination.
When upgrading a dbx 166's VCAs, the original parts required trimming to minimize distortion — a big difference. The new parts deliver the same low distortion at 1 kHz — about 0.05 percent — without adjustment and with plenty of room for improvement.
Are you still waiting for op amp recommendations? For dual op amps, I generally start with the Analog Devices OP275 and the Burr-Brown OPA2604. These parts are similar in terms of slew rate (22V/µS), but have much different current consumption. There are plenty of tweakheads looking for greater performance and recommending even better op amps. To them, I say, “Go for it!”
Eddie spent the summer teaching a class in guerrilla recording. Visit www.guerrillaremote.com for sounds and video.