SELECTING OP AMPSLast month, I wrote about upgrading analog gear performance: "Caps and Op Amps - Analog Maintenance and Upgrade Considerations." I wanted to present more details on the subject, so this column will continue that topic.
Bad caps can be weeded out using a square wave oscillator and an oscilloscope. This month, those same basic tools will be used to demonstrate a technique for selecting op amps to upgrade a UREI LA-4 optical limiter.
Unlike the LA-3A, which uses discrete transistors, the LA-4 is full of slow op amps - two or three RC4136 quad op amps, depending on the version. Early op amps left much to be desired: None could drive a 600-ohm load, many were noisy and, as I discovered, the RC4136 suffers from slewing-induced distortion or slew limiting. Translation: The op amp speed slows down as level increases.
OP AMP GEEK PARAMETERSThere are many uber-geek parameters that detail op amps' performance, and ultimately, their sonics. I chose two: power consumption and slew rate, which were detailed in Tables One and Two in Part One of this article. (Check out the September and October issues of Mix or visit www.mixonline.com.) One spec that I did not include was bandwidth, a close cousin to frequency response. Gain plays a role in the amount of usable bandwidth: High-gain circuits run out of usable high-frequency bandwidth, resulting in a visible roll-off as seen (via 'scope) by the slower vertical rise time of a square wave, shown here in Part Two. For more information about op amps and their many interactive parameters, both tables included links to the manufacturers to easily surf or download pages of data and application notes. What would normally be a simple upgrade is, in this case, hampered by the pin configuration - the RC4136 is not compatible with most standard quad op amps, such as the TL074. Finding a good op amp with the same low current consumption - 5 mA - is a challenge that requires this warning: When considering an op amp upgrade, it is important to know and monitor the current consumption of the original component and those you are auditioning. Never assume that the power supply - for a piece of outboard gear or a console channel strip - will have adequate reserve! In too many cases, power supplies operate too close to capacity.
As mentioned, the tables in Part One included just a few dual and quad op amps, both old and new. One of the most common older quads is the TL074, which consumes 6 mA of current and has a slew rate (speed) of 13 volts (vertical scope deflection) per microsecond (horizontal deflection), abbreviated as V/æs. If it were incompatible, I might have stopped there and investigated how it might have improved performance. The Motorola MC34084 is newer, about twice as fast and pin compatible (with the TL074); however, it requires nearly twice as much current. Ignore this warning and you'll be setting the house on fire in the most negative way.
PREPARE TO DUAL!Because the LA-4's RC4136 does not have a modern pin-compatible quad alternative, I decided to use two dual op amps, increasing the options. The Analog Devices OP275 (6 mA) was the obvious choice, but as I also tried the 10mA Burr-Brown OPA2604, I was witness to the potential negative side effects. Even though the OPA2604 seemed to perform slightly better than the OP275, it was prone to oscillation. (This is not the fault of the Burr-Brown part.)
Circuit design, circuit board layout, power and ground distribution issues (even the IC socket) all contribute to the excitement that prohibits you from simple plug-and-play. You've got to have a 'scope to see what's going on; otherwise, there's a damage risk to the IC and the parts around it, not to mention your monitors. But I am jumping the gun...
In order to experiment with any modern ICs, it is first necessary to create a temporary, nondestructive socket, re-pinned in this case for dual op amps. After creating a "map" of the here-to-there, a wire-wrap IC socket submitted to my will, allowing its long legs to be twisted and turned, as seen in Fig. 2. (A wire-wrap socket is designed to accept tools that cleanly wind special [and thin] solid wire around legs that are long enough to permit more than one connection. I hope to have a retrofit adapter before 2001.)
COMP SENSATIONWith two LA-4s on the bench, I could compare various "upgrades" with the "stock" version. The square wave oscillator fed both units, and both outputs were monitored with an oscilloscope. When no significant difference was noticed, I followed the signal path from stage to stage. The input stage of the LA-4 is transformerless, using two amplifiers followed by a third that can be switched for either normal or high (30dB) gain. Using the OP275 made a significant improvement to the first three stages as viewed at the output of U3.
U3 is then followed by a "build-out" resistor, R13, that, when combined with the photoresistor, creates a voltage divider/gain manipulator. R13 is a rather high value - 82k ohm - that is large enough so that any "stray" capacitance (from U4's input and/or the socket and the circuit board layout) slowed the rise time of the square wave.
Obstacle One: The value of R13 created a bottleneck that masked the speed gains achieved by the op amps in the first three stages. Because the output amp also needed improvement, I spent most of my time pointing the 'scope at the output of U4, just before the Output Level pot.
The Burr-Brown OPA2604 seemed better than the OP275 in the U4 position, but not by much. Increased gain reduction revealed two phenomena - improved speed in the modified circuit (good) and slew limiting (bad) in the reference unit. In the modified LA-4, the "activated" opto lowered the impedance of the circuit, making the signal impervious to any "stray" capacitance that might follow.
Obstacle Two: Increased signal levels reduced the slew rate of the RC4136 - further convincing me to continue along the upgrade path. So far, at least one problem was solved by the upgrade. The left side of Fig. 4 shows two before-and-after square wave examples.
As a temporary fix (that stuck), I cut the value of R13 nearly in half, from 82k to 47k ohm. (A 68k ohm resistor was used in the LA-2A. In the LA-3A, the 68k ohm resistor is paralleled with a 100pF cap to improve high-frequency response.) I knew the gain reduction slope would be affected by the change to 47k ohm, but it solved the slew-rate problem while the output amplifier was being tweaked.
THE MAKE-UP GAMEThe output (or line) amplifier of any compressor/limiter is designed to recover "the lost gain of signal processing," as well as make the signal robust enough to re-enter the outside world. The newer LA-4 line amp combines one section from the RC4136 with discrete transistors. The earlier version used a single op amp (which I could not easily cross-reference) driving an unusual output-pair configuration.
The first thing I did was bypass the output amp, installing a temporary socket that would accept "large" op amps, like the John Hardy 990, an API 2520, an RS 30-533 (a British FET op amp) and a "bootleg" API using an NE5534 followed by a pair of discrete transistors. (Yes, I am guilty of temporarily ignoring my own rules about power consumption.) This brief detour defined the three challenges of Obstacle Three: The output transformer partially masked the improvements of the previous stages, the mystery IC was, uh, a mystery, plus the almost-necessary high recovery gain. All conspired to slow the speed of the output amp and mask the improvements that survived through U4. Are you ready to play?
I wasn't ready to remove the output transformer, even though doing so would have improved signal transparency. Like the Class A Neve line amp, the output transformer can be used for gain - 4 dB in this case - which also translates into additional headroom. Once I realized that the mystery IC was simply a standard, single-channel op amp, I chose the closest physically available device, an NE5534, because it was handy and far better than the mystery IC. (Rebiasing the unusual output amp was required.) I then reduced the gain of the output amp to improve bandwidth enough to reflect the modifications.
Before modifying the stock unit, I compared its gain-reduction characteristics with that of the modified unit, accepted the differences and decided to keep all changes. The stock unit was then modified, and both units were calibrated so that gain - and gain reduction - were identical. The customer was quite pleased with the improved performance, especially on acoustic guitar.
After improving the speed of the LA-4, it became necessary to deal with stability issues. Remember that the Burr-Brown OPA2604, through no fault of its own, had a tendency to oscillate. This was most likely due to resistance in the printed circuit board (PCB) traces that bring power to the various ICs. PCB trace-resistance increases with heat; so again, current consumption becomes an issue. (More heat equals more resistance.) The most common solution is to add "bypass" capacitors (0.1æF) from each power input (both "plus" and "minus" pins) to ground. This simple fix isolates the amplifiers from each other, minimizing interaction that can lead to oscillation. C18 and C19 in Fig. 3 were not in the units I modified, but were added in later production versions.
SPEED DEMONYou might think that an obsession with amplifier speed equates to a search for the Holy Grail of Audio, but that's not quite the case. Years ago, I found that square waves provide quick sonic snapshots, useful for finding marginal capacitors, the effects of transformer termination (and lack of same), as well as transient performance. Both the stock and the modified units probably had the same small-signal frequency response when measured from 20 to 20k Hz (using a sine wave), but the most significant improvement eliminated most slew-rate distortion at high signal levels.
This is an article, not a thesis, so please forgive me if "important issues" were glossed over or ignored. Based on frequent e-mail requests, the point was to demonstrate to those who have "the upgrade bug" that it is possible but not always easy to simply plug and play. Yes, the ear is the ultimate piece of test equipment, but the random oscillation is like an unexploded land mine.
This now leads us to next month: power and ground distribution. We'll look at a Trident Series 65 upgrade as modified by local uber-geek Dan Kennedy of Great River Electronics. As always, my thanks to Dan, as well as Dave Hill at Crane Song and Greg Gualtieri at Pendulum Audio for their insights and expertise.