Restoring and modifying a GR 1562-A SLM Calibrator

Most recent update 7-26-2012

Restoring and modifying a General Radio 1562-A Sound Level Calibrator
(with notes on the GR 1567)

Background
People engaged in acoustics work depend on some basic tools of the trade. One of the more important tools is a precision sound level meter (SLM). There are many on the market, but some of the best were made by General Radio (GR or GenRad) -- and some of these continue to be made by GR's successor, IET Labs.

But in order to use the SLM with confidence, it has to be calibrated, which requires a Sound Level Calibrator -- which is essentially a loudspeaker driving a known volume of air (acoustic coupler) into which the business end of a microphone is placed. The loudspeaker is driven by a relatively pure source of sine waves, usually at a single frequency, and it's highly stable output level is calibrated to deliver a precisely known sound pressure level (SPL) -- typically 114dB, 104dB, or 94dB, or some switchable combination of these.

So, to keep it's SLMs calibrated, GR also made some calibrators -- the ones I'm familiar with are the 1562-A and the 1567. The 1562-A offers five output frequencies from 125Hz to 2kHz in octave steps, all at an output in the microphone coupler of 114dB SPL. The simpler 1567 has only a 1kHz output, also at 114dB SPL.

1562-A
The 1562-A uses a Wien Bridge oscillator with frequency determined by two fixed capacitors and switched pairs of precision resistors. Its amplitude stabilization is via a very low-mass thermistor, housed in a vacuum bulb, and placed in the negative feedback loop. The vacuum housing of the thermistor means that it is not affected by ambient temperature variations which, for a thermistor mounted in air, will cause significant ambient-temperature-dependent level changes -- clearly a bad thing for a precision level calibrator.

The thermistor system works well, with a little extra thermal compensation thrown in, but the thermistor's large change in resistance from cold-off to hot-operating temperature means that it needs to be pre-heated in order to get the oscillator to start up reliably. The combined power and frequency switch has a position that both checks the battery level and pre-heats the thermistor, and also allows selection of the operating frequency -- switchable from 125Hz to 2000Hz in five one-octave (2:1) steps. Here's a simplified schematic for the original 1562-A:

1567
The 1567, having a only a single frequency, uses two LM301 opamps in a phase-shift oscillator which has three capacitors and three resistor to set frequency -- a topology that is OK for a single frequency, but which is less practical for switched frequencies than the Wien Bridge, which in this application would use two capacitors and two resistors. Amplitude stabilization is through a pair of paralleled, reversed-polarity silicon diodes in the feedback loop, with level adjusted via a variable resistor in the loop with the diodes.

This system provides fairly low distortion and very good stability, provided that the temperature-dependent forward voltage drop of the diodes is accounted for. GR used a feedback resistor + thermistor combination with a large, well-understood temperature coefficient of resistance to compensate for the diodes. This system works well. You can download the PDF of the 1567 manual, but be warned -- it's a download of 7MB. Here's a slightly simplified schematic of the 1567:

Getting a badly mistreated 1562-A back in working condition
I have a 1567, but I've always hankered after a 1562-A. I recently found a 1562 on eBay for a very low price. It was listed as "for parts or not working" -- its condition basically unknown. This is often a bad sign, but if the price is low enough, some good deals can be had -- but you have to know that work will be needed.

While I was waiting for the 1562 to arrive, I was able to download a service manual for it, but the schematic diagram was essentially unusable -- a very poor scan of the schematic, although the rest of the manual scan was fine. But I was able to buy an original manual from Ridge Instruments in Maryland, who have a large store of old manuals for test equipment. I then made a simplified schematic and am posting it here. Once I have some editing done on the scanned schematic, I will post it here as a PDF for download as well. Now I was ready to get the 1562 and see whether I made a good buy or not.

A very rough unit with a bad thermistor
When the 1562 arrived, I saw that the serial number is 170 -- talk about hoary with age! I determined that the internal oscillator was not working. I disassembled the unit and found that it was full of dust, gummy residue, and corrosion, and the crucial control element for the oscillator, the thermistor, was burned open. This thermistor is now unobtanium -- it has a cold resistance of around 40k ohms, and in operation in the 1562 it runs at about 440 ohms. This huge change in resistance means that the thermistor runs hot, and due to its vanishingly small size, is quite fragile. Generally the glass bulb provides good mechanical protection, but some people, not knowing the electrical fragility of the thermistor, check its resistance with a high-current ohmmeter and actually burn it out while testing it.

Given the relative fragility of the 1562, I wonder how many of these have been broken and have not been repaired due to the unavailability of the needed thermistor? I contacted IET Labs (the successor to GR) about possibly purchasing the thermistor, and they responded promptly, letting me know that the part, due to high cost, is not separately available -- but that they can do a flat-rate repair of the 1562 and calibrate it for around $800.

If I were still employed in this field and dependent on the instrument for my work, I wouldn't hesitate to have them do this. But I'm retired and no longer working in electroacoustics professionally, and that $800 can now be more usefully spent on grandkids, travel, a rainy-day fund, or another interesting piece of old but usable test gear.

The Shure "Controlled Magnetic Transducer"
I also tested the transducer (loudspeaker) that actually makes the sound in the 1562 and found its acoustic output to be quite low. Everything metal in the 1562 had corrosion and the transducer was no exception. It looked like the 1562 had been laying in liquid for a while, or maybe had been dropped into water accidentally and then not properly dried out and cleaned up. The transducer fits between the front housing of the 1562 and the aluminum frame that holds the circuit boards and switch. The frame assembly in this unit unscrews from the front housing to access the transducer (in the 1567 I have, the frame is held in the front housing by set-screws). Here's a picture of the disassembled housing and frame:

The transducer was made by Shure Brothers, a famous maker of phono cartridges, microphones, and sound reenforcement products. This transducer was used primarily in a very popular series of desk-top microphones, mostly for use in PA systems, but it is still prized today for recording such things as kick drums and other high-level, low frequency sound sources. Most folks don't realize that transducers are reversible devices -- they can turn sound into electrical energy, or electrical energy into sound -- the earbud on an iPod can make a quite good microphone. Not being sure that this transducer was repairable, I searched eBay and found one of these "Controlled Magnetic Transducer" microphone elements for sale. I bought it.

It arrived and was in great working condition, as advertised. Now, having a working transducer in hand, I felt brave enough to take the original one in the 1562 almost completely apart. Its coil was good, but the corrosion extended into the area of the magnetic gap, in which is located the armature, a small tongue of metal with a wire attached from one end down to the diaphragm of the speaker/microphone. The other end of this flat metal armature is at the center of the coil which provides the magnetic driving force. It's a very simple and very effective design. I disassembled the magnet and cleaned all of the corrosion I could from the magnet's faces and from the soft-iron pole pieces, and was able to clean the pole piece face sitting under the armature without completely disassembling the transducer. I put it back together and it worked, and worked well at 1kHz. You just gotta love old-school stuff that can be repaired with simple tools and a little care. Sadly, though, there is some buzzing and distortion at lower frequencies. It will need to be completely disassembled and cleaned up to get full operation back.

Here are pictures of the old transducer on the left and the replacement on the right in each photo. Note the corrosion on the magnet assembly in the second picture (you can see the gray metal trapezoid of the armature between the yellow coil and the magnet assembly) and a segment of liquid stain running up the left side of the front housing of the transducer in the third picture:

I removed the glue patches from the front of the replacement transducer before installing it. These must have been from its original mounting in a microphone. The dark gray rubber gasket goes between the front of the transducer and the housing, while the wavy spring washer goes between the back of the transducer and the frame.

What to do about the thermistor?
But now I faced a second, more serious problem -- what to do about the oscillator and the bad thermistor? Not being able to find a replacement for the vacuum-bulb thermistor, I decided to look at another amplitude stabilization system, one that also uses a "thermistor" mounted in a vacuum bulb -- an incandescent lamp. I really don't know why GR didn't design the 1562 with a lamp in the first place. It's a very common form of amplitude leveling element used in tens of thousands of oscillators through the years, including the iconic Hewlett-Packard 200CD and the hugely popular Heath IG-18 and its successors. But lamps are low impedance devices that take a lot of power to warm up, so they are not usually well-suited for low-power gear.

One reason to use the thermistor is to get less loading of the oscillator's output -- with a battery powered device, using less power is a good thing. But small, low-current, low-voltage lamps were, and are still, available. With a lamp, there is no problem with ambient temperature sensitivity, and the distortion and level stability from the lamp system can be excellent.

I again went to eBay and bought some small lamps. Surely one (or two, or three) would serve in the circuit of the 1562, with suitable mods, but this oscillator puts out only about 1VRMS, which means that the amount of current available for a lamp to warm up to the knee in its operating point is tiny. At 1V out, the voltage across the lamp is only 333mVRMS -- one-tenth the voltage across the lamp in the Heath IG-18, for example. So the lamp needs to have low operating current -- the lower the better, and a relatively low voltage rarting, in order to get its operating point resistance high enough that the feedback network won't effectively short out the output of the oscillator.

In the Wien bridge oscillator, the ratio of the two negative feedback resistors is 2:1, so the feedback leg resistance will end up about twice the operating resistance of the lamp in the common leg. The oscillator sees the feedback network as its load, and low-power amplifiers don't much like little resistors hung on the output. As a floor for a minimum value, I was hoping to find a lamp with a cold resistance of at least 50 ohms, with higher being better up to a point -- there still has to be enough signal current flowing through the lamp to start getting it warm enough to raise its resistance off of the floor of its cold value.

Past experience had shown me that increasing the cold resistance by even a small factor was enough to get the lamp to provide stabilizing resistance changes. Enough current to raise the cold resistance by a factor of two is about ideal; but I knew this would be a struggle in this circuit due to the low signal current. For more on the subject of oscillators with lamp stabilization, see my web page here.

I ended up with a small selection of not very suitable lamps:
2187         28V     40mA     cold R = 65 ohms     wire-lead version of the 327 -- a Jim Williams fave
unk           12V     25mA     cold R = 47 ohms     RadioShack lamp with color-coded leads
unk          1.5V     15mA     cold R = 13 ohms     "grain-of-rice" lamp from the UK
90V           90V     30mA     cold R = 320 ohms     left over from IG-18 mod
120MB     120V     25mA     cold R = 350 ohms     works well in IG-18 as sub for 90V

I couldn't find any, but a good candidate might be the type 2158, a wire-terminal lamp rated at 3V and 15mA. One of those could be just the ticket.

What wasn't clear was whether the amplifier in the 1562 would be lamp-friendly with the high loading. The fallback would be a plan to use an IC opamp, but this would either mean piggy-backing the opamp onto the 1562's circuit board or require making a small circuit board to replace the existing one. I only wanted to use that plan if I had to.

Changing the amplifier
Some changes needed to be made. The lamp is a positive temperature coefficient device (PTC), whereas the thermistor is a negative temperature coefficient device (NTC). This means that the lamp(s) needs to go in the common leg of the feedback loop instead of in the feedback leg as the NTC thermistor does. The original amp used all germanium transistors -- 2N1304 and 2N1305. Some earlier owner/user had changed the two complementary output devices to silicon transistors, 2N3904 and 2N3906, with two 1N914 diodes as base-spreaders for lower crossover distortion. I replaced the two complementary 2N1304 and 2N1305 Ge voltage amp transistors with higher-Beta Si PN2484 and 2N5087. I breadboarded the circuit and it immediately became apparent that there were problems -- a lot of crossover distortion, insufficient output swing at the 1562's minimum supply voltage of 6V, and an inability to drive low-Z loads -- hooking up the transducer killed the output.

Clearly, given the battery operation and its range of supply voltage from 9.3V to 6V (the original was designed to use a 9V "C" size mercury battery with snaps on the ends, but running off a regular 9V snap-top battery works great -- it just squeezes into the space for the original battery), the germanium transistors were a good choice, especially in the output stage, so replacing those had been a step backward. Even unloaded, there were problems, thanks to the low resistance of the lamps.

I tried the various lamps without a load on the output. I thought the 2187 lamp would work, but there just isn't enough signal current to get it warm enough to move up the I-V curve and get some voltage variable resistance.effects -- the PTC in action. Oddly, the 90V worked, but not well enough -- again, not quite enough current for it to be usable. The grain-of-rice lamps showed promise -- they have the lowest current rating. I wired up a small piece of stripboard and soldered five of them in series, with pads so I could select any number from one to five lamps. Three lamps worked pretty well -- clean, fast startup and settling to the final output level, and OK in general. The 12V Radio Shack lamp was almost as good as the three smaller lamps. But nothing could solve the inability of the amp to supply output current to the low-Z load of the feedback network and transducer combination.

I breadboarded the opamp circuit and found that an NE5534 worked the best, better than an OPA134, OPA604, uA741, LM356, or LME49710. An LM386 low-voltage power amp designed for transistor radios would seem to be an obvious choice, but I couldn't get the one I have to work without serious stability issues -- ugly to work with and it used a lot of parts.

The 5534, certainly one of the world's best opamps by any standard, worked well with the three-lamp grain-of-rice combo and also with two of them and five of them, and worked with the RS 12V lamp, too. The 5534 worked well even with a 6V single-ended supply, although a single-supply, rail-to-rail opamp might be better -- but the one I have, an LT1006, wasn't good at all. The 5534 drives the combined feedback and transducer load without complaining too much at 1VRMS output. By the way, I didn't need any compensation around the 5534 -- if you see HF oscillation, add 39pF from pin 5 to pin 8.

So I built a little PC board for the opamp, the same size as the GR original. I put a socket in for the amp, just in case. The circuit with the 5534 really isn't much different from GR's discrete circuit, except that it delivers the 1VRMS output into 150 ohms with less than 0.1% THD at 1kHz -- a fair improvement. Here's the circuit and a picture of the new unfinished board and the original:

Nothing in this circuit is unusually sensitive to changes in ambient temperature, although I wouldn't try to use it in Death Valley or at one of the Poles and expect great accuracy. Note that I added a bias pot to optimize the output swing for a 6VDC supply voltage. The 1562's battery test lamp is driven by a circuit that extinguishes the lamp if the battery voltage is under 6V. With the transducer working, THD does rise at 125Hz to around 0.9% -- mostly 3rd harmonic, possibly because the thermal mass of the lamps is so low that there's less integration from peak to peak of the wave. As long at the amplitude is stable, that's acceptable -- even though it doesn't match the spec for the original, which is less than 0.5%. However, the distortion does meet spec from 250Hz on up, dropping below 0.1% at 1kHz and 2kHz. Here's a picture of the lamps on the osc. board -- you'll have to look close!

I put the 200 ohm feedback gain pot on the oscillator board to eliminate some switching -- the thermistor warm-up switching is not needed -- and to make adjustment easier. This pot will be around 100 ohms with three of the grain-of-rice lamps, which at 1VRMS output are at about 50 ohms total -- just barely up the knee of the I-V curve. The feedback pot got adjusted to give a 1kHz transducer output of 114dB with my calibrated SLM. All of the frequencies were within plus/minus 0.5dB with that SLM, which has a quite flat electret condenser mic. With the transducer connected, and with measurement at the phone jack test signal output, THD ran from about 0.07% at 2kHz to 0.9% at 125Hz. Here's a picture of the oscillator mounted in the 1562:

Update, 9-18-2011
After full assembly with the transducer, I discovered some issues --
1) that the fixed voltage divider biasing wasn't working well at the two ends of the battery voltage range -- I could optimze for one end or the other, but not both. I discovered that with increasing supply voltage, the biasing needed to increase too, but not at as fast a rate as the supply increase. My solution was to put a blue LED under the bottom leg bias resistor to act as a zener diode. LEDs have a quite sharp knee and work very well for this. The blue LED at this low current has a forward drop of about 2.6V. That, in combination with a change in resistor values, fixed the issue completely -- see the revised schematic below.

2) With three lamps, the resistance increase of the lamps was not enough to give good stable regulation. So I shorted out one of the lamps. Using two lamps instead of three lowered the feedback resistance somewhat but gave better operation.

3) The feedback pot was just too sensitive to adjust precisely and was also causing some amplitude drift issues. The solution was to restrict the range of the pot by adding a series resistor and a shunt resistor, limiting the adjustment to a range from 82.5 ohms to 105 ohms, with the optimum point at about 93 ohms for the calibrated 114dB level. This range gives an amplitude adjustment range of +/- 1dB.

4) The battery test function was pretty anemic, not having a sharp cut-off from lamp-lit-voltage to lamp-not-lit-voltage. The original design uses a #345 lamp, rated at 6V and 40mA -- this tests the battery under load, which is good. The two-transistor switch is supposed to extinguish the lamp if the supply is less than 6V, but lets the lamp illuminate weakly over a range of voltages so that you're not really sure if the battery is good or not. In the revised oscillator circuit, the minimum voltage for low distortion is 6.5V and since I had to change the circuit, I made a few changes to get it to work right at 6.5V.

The first change was to add 3.3k from the collector of Q106, the NPN, to the supply rail, to add some hold-off current to assure that the transistor actually switches fairly sharply from on to off. The second change was to shunt the 1k resistor from the base of Q106 to ground with 15k, to raise the switching voltage slightly. All is well. Here's the circuit with revisions:

###

Update, 7-26-2012
I recently acquired a second 1562-A, this one working correctly, all original parts (except for the transistor-radio battery mod), with thermistor intact. It's distortion from the phone jack output is fairly uniform across the frequencies, running at around 0.2%, well within spec. It reads 1kHz at exactly the same dB level on the SLM as my modded 1562-A, about 0.1dB less than 114dB. I'd like to figure out a way to get the distortion down on the modded unit -- I've stretched my brain, but haven't come up with anything yet. Extremely low-current lamps are apparently as rare as tiny little glass-bulb thermistors...