A steady parade of digital tape machines marched through my shop last year, interspersed with a few pleasant diversions that included five condenser mics, four dynamics processors, three mastering projects, two guitar amp modifications and (drum roll please) a cartridge in a turntable. Oh yeah, and a gray-market quad-panner that had the ugliest spray-painted gold front panel I have ever seen!
As a user, reviewer and techno-geek, I can't learn enough about compressor/limiters. Two very different beasts piqued my curiosity last summer: an Altec 438 microphone preamp/compressor and an EMT 156 stereo compressor/limiter/expander (to be explored in the few-cha).
Note: While this column's spotlight is on the Altec 438 — a vacuum tube device — troubleshooting basics apply to any product, hence the occasional mention of transistors.
THE LOAD DOWN
In the studio, I often found Altec's variable-mu compressor complementary to bass guitar. On the bench, I've learned that “color” comes from many factors — response time, ratio, amplifier design and especially the transformers. “Color” is never due to just one tube, nor does it depend solely on an “optical” or other gain-reduction device; rather, it's produced by the sum of all the parts.
The Altec 438 has three transformers: mic, inter-stage and output. The first transformer is the most colorful, the second drives the variable-mu (dynamics) stage, while the output transformer is fairly robust. Because input and output transformers have tapped windings, adding switches improves the accessibility of the impedance options, making it easier to optimize the interface or explore alternate sonic effects.
A VU meter is one overlooked component that's common to all vintage compressor/limiters. It makes a sonic contribution by way of the “feedback” it supplies (or doesn't). The short story is that mechanical VU meters are perfect for voice but slow to respond to transients, whether monitoring level or gain reduction. Having relearned this valuable lesson a few years ago (when reviewing the Pendulum Audio 6386/6ES-8 variable-mu compressor), I can't emphasize it enough. Here are the key points:
- Listen first.
- Know the difference between peak limiting (fast attack and release) and compression (medium attack and release), as well as the appropriate threshold settings for each.
- Peak limiting probably does more to “wake up” a stereo mix, but you won't get the best results looking at a VU meter. In fact, I encourage retro manufacturers to incorporate additional peak display devices, such as LEDs, so that the metering reflects the work being done. (For more info about this and other “color” commentary, read the sidebar.)
In my experience, Altec products are consistently designed around a simple, no-frills philosophy (e.g., the industrial dark green or gray finish). Like the Teletronix LA-2A, and consistent with their other “packages,” the Altec 438 features a rackmounted chassis with a hinged front panel that facilitates service. The compressor uses a variable-mu tube, the 6BC8, with a built-in 12AY7 mic preamp. (The compressor-only version was called the 436. Both models have similar, but simpler, circuitry compared to the Universal Audio 175.)
There were at least three versions of the 436 and 438, which were designated A, B and C. The “A” version had only a volume control. The “C” version added threshold and release (variable from .3 to 1.3 seconds at 63% recovery). Attack was fixed at 50 ms. According to the Application Notes, “The 438 Amplifier is intended for use as an Automatic Level control in Recording, Broadcasting and Public Address systems. It can provide up to 30 dB of compression.” Like most dynamics processors from this era, the 438 is a feedback-style compressor, in which the Detector circuit is driven by the output amp. Surprisingly, the documentation goes on to say, “The 438 can be used without the compressor by removing the 6AL5 Detector tube from its socket.”
Regarding sonic color, the distortion spec for the 438 is fairly high: 2.5% at 30 dB of compression. The frequency response spec is 40 to 10k Hz, ±1.5 dB. Of course, the response goes well beyond 10 kHz, but in the late '50s, 40 to 10k was the critical window — everything above and below was icing on the cake. The 438 that I received from Customer X had been used as a mic preamp only — for voice-over work — and the customer had pulled the detector tube because the compressor was not functioning well. When the output amplifier got really funky, the unit was sent in for service along with a nice letter detailing normal use parameters, problems and — joy of joys — boxes of spare tubes. I had free reign to make the compressor usable.
Safety note: The capacitors in vacuum tube power supplies can retain their charge even after power has been removed. Please wear shoes and observe the “one-hand-in-pocket” rule. Do not lean on the unit with one hand while probing for the best shock-therapy test point with the other. Caps can be discharged through a resistor to chassis ground using insulated clip leads.
Also, before changing any components, make sure you have schematics. Electrolytic capacitors are often hard wired with multiple connections made to each terminal. Just removing them can be quite labor-intensive. No matter what part is being replaced, don't rely on your memory. Do make a drawing of the component layout, and check to see that it agrees with the schematic so that everything can be put back the way you found it.
Another point to remember is that, in terms of meeting their specifications, vacuum tubes and transistors are moving targets. “Testers” can be used to match devices and quantify certain aspects of performance, but the best way to scrutinize for noise is to have a known quiet circuit, at least one functioning ear and/or a spectrum analyzer. The NTI Minilyzer is one example of an affordable, all-in-one unit capable of measuring level, noise, frequency response and distortion.
With all vintage gear, check capacitors first. Last year, I showed how square waves can be used to find failing signal caps. (See tangible-technology.com for stunning pictures.) Checking power supply caps in tube gear is another matter. (Please re-read the safety note above.) Electrolytic capacitors that are physically leaking are suspect. All caps can be measured — out-of-circuit only — for “electrical” leakage and for tolerance using a capacitance meter. Vacuum tubes should be tested for emission and “gas,” the latter being the most likely culprit when other treatments do not relieve the symptoms. The stripes on overheated carbon resistors may no longer be reliable indicators of the component's value, so try to correlate them with the documentation and an ohmmeter.
POWER SUPPLY ISSUES
If the unit blows fuses, then the suspect parts include C12, C13, the rectifier diodes and the power transformer. Disconnect the diodes to check the transformer and hope it's not bad. The Altec power supply and detector circuits are shown in the figure. Note that the diodes comprise two half-wave rectifiers (in a voltage doubler configuration) that convert 117 VAC from the transformer secondary into 255 VDC across 100uF caps C12 and C13! C11a (not shown, but connected to “Va”), C11b and C11c are part of a multisection “filter” capacitor that isolates the preamp, gain reduction and output stages, respectively. (The “V” designations indicate power distribution to these stages.)
Let's assume, for the moment, that the device in question can be plugged in without doing further damage to itself or its surroundings. Connect a sound source, adjust the signal level and monitor via speaker. Kill the source, shorting the input if necessary, and listen to the output for noise. Lots of 120Hz hum indicates C12 and/or C13 are bad; some hum may be normal. It is not unusual for vintage gear to have a signal-to-noise ratio of only 80 dB unweighted, some of the noise consisting of hum.
If you are unsure about the other supply caps, then try this in-circuit test: Starting with a 'scope set to its least sensitive position, select AC coupling, connect an X10 oscilloscope probe to the top of C12, where a saw-tooth wave — also known as “ripple” — would indicate that some filtering is being done. Then probe Vc, Vb and Va; each should have progressively less ripple than C12, the saw-tooth being replaced by a near sine wave. Though they are not shown, the 6BC8 and 6CG7 are differential stages and, as such, will reject any ripple at C11b/Vb and C11c/Vc. The mic preamp is the one stage that needs the cleanest DC (C11a/Va).
I once received a piece of outboard gear that barely worked. Most of the tubes were beyond tired and when replaced, the unit oscillated like crazy, as if using good tubes was to blame. High-gain signals from various stages can cross-pollinate if the filter caps are well below tolerance. For a quick in-circuit cap check, connect an oscillator or sound source to the input and terminate the output with a 600-ohm resistor. Now, monitor all of the same capacitors as in the previous test to see if the audio signal is superimposed over the ripple. Any significant amount of audio would also indicate cap fatigue.
OTHER HOT NOISE ISSUES
Although all “antique” carbon resistors are potential noise sources, start by scrutinizing values of 100 kilohms and above that are connected to a vacuum tube plate or a transistor collector — essentially, all voltage gain stages. Be sure to first change the tube or transistor in the stage suspected of generating the noise. Short the input and monitor the output. If “popcorn” still appears, then switching to metal film resistors should reduce the noise, shifting its spectral content to a less disturbing region. Next, tap each tube and listen for microphonics. Hand-select, if necessary. Tubes in high gain circuits are sometimes shock-mounted to isolate them from chassis noise. Replace dried rubber grommets (available at Radio Shack) to improve mechanical isolation.
Definite progress was made at each step of the Altec 438 restoration, but I made the initial tests with a steady-state tone only. Various program sources showed that the dynamic response of the variable-mu circuit became unstable under changing loads. Returning to a test tone, the instability occurred right at the threshold, taking the form of a low-frequency modulation as if the sidechain was oscillating. (The 438 in question was the “A” version with fixed threshold.) I reversed a non-factory modification in the detector circuit to see what normal was.
From the 6CG7 output tube, a differential signal feeds the cathodes of the 6AL5 Detector via points A1 and A2. Each half of the wave is rectified, combined by tying plate pins 2 and 7. The raw DC signal is then smoothed by an RC network that sets attack and release times. Because I wanted the unit to be more versatile, I tweaked and listened and tweaked again until I selected the components to the left of the 6AL5, as detailed in the figure.
To achieve near-original performance, set P2 (a 2-Meg pot) to the max CCW setting and the Speed switch to S (slow). Medium and fast settings are achieved by choosing two smaller-value caps; turning P2 CW reduces attack time and increases release time. I liked the idea of having a master Speed switch plus a single pot to inversely affect these two parameters. The threshold control — P3, added in the “C” version — is shown to the far right. Switches labeled “A” and “C” show the before and after of the factory mod. I would have liked to gang P2 and P3 together in order to desensitize threshold when dialing P2 toward Peak Limit mode, and vice versa when P2 is in Compress mode.
CIAO FOR NOW
Signal path show and tell will have to wait for now, but I didn't want to leave without mentioning the EMT 156. It passes audio through a diode bridge using pulse-width modulation to control the gain. Unlike the Altec 438, there are multiple detector circuits for peak limiting, compression and expansion. Three DC signals are mixed together to drive the pulse-width modulator. What a bizarre beast that was, from the power supply on up. Maybe next time. Till then, keep your soldering iron tip clean.
Visit the Eddie archive atwww.tangible-technology.com.
Three issues come to mind regarding the desirable idiosyncrasies of old-fashioned analog: ease of use, the “color” source and user feedback via metering (if applicable). Electronic part values vary with temperature and manufacturing tolerances. When used in a mono signal chain, most differences are insignificant. When used in stereo, the tolerances are more critical. As with all vintage gear, component aging is a contributing factor to sonic character. The sound of some “magic box” might be the result of the design or the characteristics of marginal components. It is important to know which, especially if the plan is to clone the device and cash in on the magic.
Such was the case with the “new” Universal Audio 1176, one example of sonic time travel that propels the hardware side of our biz. Bill Putnam Jr. and Co. extensively interviewed passionate users, learned what version to replicate, agonized over the cost of finding “original”-style parts and then had their work subjected to further scrutiny. A handful of cherished 30-year-old originals (and their owners) set the high watermark. Golden ears celebrated the resurrection, attributing the subtlest of sonic nuances to component aging in the originals. This is a compliment to Universal Audio, but it must have made a few people sweat!
VROOM WITH A VU
I always knew to ignore traditional VU meters when recording snare, tambourine, handclasp, glockenspiel and steel drums (to name a few). Mechanical metering is, by nature, slow to react to transients but perfect for voice, bass and electric guitars. This is true for level as well as gain reduction readings. With medium-to-slow attack and release settings, VU metering will accurately reflect the job being done. But on program, using the fastest settings, the “magic” didn't happen until I stopped using the meter and just listened.
For pop music, close-miked drums create lots of peaks well above vocals and other “soft” instruments. I generally start out with Slow Attack and Fast Release settings, then gradually increase attack until it starts “biting” into the track, bringing the peaks into line with the “meat.” Attack and release response must be very fast — too fast for the meter to respond and too aggressive to do any more than take “a little off the top.” As I gradually increased threshold, the mix came alive; peaks of about 6 dB were tamed, yet the meters were barely moving.
The limitations of VU metering are an acceptable shortcoming of vintage technology, but one that requires a little intuition from the user. That was then. Now, precise metering can serve double duty as both a teaching tool and as a value-added product enhancement. By illuminating the differences between peak and average program levels — both in Level and Gain Reduction modes — ears-in-training will see (and hear) what the controls are doing, ultimately getting more from the product. Embracing the simplicity of the past is a great discipline, but I also encourage retro manufacturers to integrate the simplest bits of modern technology — a few LEDs — to deliver the feedback users need so that they can get more from their gear.
— Eddie Ciletti