Improving the Sound Technology 1000A

The Sound Technology 1000A FM Alignment Generator is an excellent stereo signal source whose design dates from the late 1960s. While its performance was adequate at the time, later FM tuners were capable of lower distortion, often lower than that of the ST-1000A itself. Tuner distortion is limited by that of the signal used to align it. Distortion measures lowest when the tuner and signal distortions cancel. A meticulous alignment effectively transfers the residual distortion of the signal generator to the tuner.

ST-1000A THD is specified as less than 0.1% in mono and less than 0.2% in stereo for 1-kHz modulation deviated 75 kHz. It is possible to improve these figures by an order of magnitude with simple modification of the ST-1000A circuitry.

Additional modifications make the unit more accurate, more versatile, and easier to use.

Internal Oscillator

ST-1000A modulation linearity is limited mainly by the distortion of its internal audio oscillator, a Wien bridge sinewave generator. A P-channel JFET acts as a voltage-controlled resistor to automatically set the oscillator loop gain to one. Variation in JFET channel resistance with drain-source signal voltage distorts the sine wave.

Trimmer R105 is in series with the JFET. R105 sets the JFET operating point, effectively trading distortion for reliable oscillation. Oscillation will not occur if distortion is set too low. The ST-1000A alignment instructions say to adjust R105 for about -64 dB (0.06%) THD. You can't go much lower than this and maintain reliable oscillation. The internal oscillator also generates tones for dual sweep (10 kHz) and SCA (67 kHz) modes. Oscillation may fail at one of these frequencies first.

There is a trick to linearize JFET channel resistance that may not have been known when the ST-1000A was designed. Applying one-half of the drain signal to the gate makes any variation in channel resistance symmetrical, cancelling even-order distortion. This kills the dominant second-order JFET distortion product, greatly reducing THD.

Another limitation of the internal oscillator is residual audio on the gate control signal. Any variation at the oscillation rate will distort the waveform. Filtering the gate-control signal more heavily solves this problem.

You can modify the internal oscillator to incorporate these improvements as follows:

Cut the trace to the gate of Q8. Bridge the cut trace with a 100kΩ resistor. Solder another 100kΩ resistor between Q8's gate and drain (the ungrounded terminal). Finally, add a 10 湩 15V electrolytic capacitor across C49 (observe polarity).

Adjust R105 for minimum distortion consistent with reliable oscillation in all modes. If R105 must be considerably advanced to obtain oscillation in one mode, distortion may be unnecessarily high in the other modes. By altering the capacitor ratio for the sluggish mode (C43/C47 for 10-kHz dual sweep, for example), you can make oscillation occur more readily. If R105 is set too close to the onset of oscillation, oscillation may fail at some temperature. It is normal for the oscillator amplitude to take a few seconds to stabilize after starting or changing frequency.

This shows the spectrum of the 1-kHz internal oscillator after modification. Horizontal is 0 to 5 kHz, vertical is 10 dB/div, and the analysis-filter bandwidth is 30 Hz on an HP 3580A spectrum analyzer. Depending on the individual sweep and temperature, a tiny second-harmonic pip may appear somewhere near the -89 dB noise floor. The 400-Hz spectrum looks similar.

The signals in the images that follow are not deemphasized. A tuner's deemphasis network will attenuate the distortion products. The second harmonic will be about 2 dB lower than the levels shown, and the third harmonic about 4 dB lower.

Stereo Modulator

The ST-1000A stereo modulator uses a pair of 2N3819 JFETs as switches. Capacitors C111 and C112 (68 or 100 pF) slow the gate signals, which are 38-kHz square waves. This may cause the FETs to linger at a resistance higher than when fully on, increasing modulator distortion. Disconnecting C111 and C112 can lower it. I unsolder the nongrounded lead of each capacitor, leaving each part on the PCB supported by a single lead.

This is the composite stereo spectrum from 0 to 5 kHz. I adjusted the spectrum analyzer sensitivity to place the 1-kHz pip at the top of the screen in mono. It is at -6 dB for left-only stereo. The second harmonic in stereo is at -82 dB. Its level varies somewhat with temperature but is always less than -80 dB.

This is the composite subcarrier spectrum from 33 to 43 kHz. The 37- and 39-kHz fundamental sidebands are 12 dB down. Third-harmonic sidebands at 35 and 41 kHz are less than -81 dB. These harmonic levels also vary somewhat with temperature.

Disconnecting C111 and C112 normally has no adverse effect, but in one ST-1000A low-level switching glitches appeared in the composite output. The resulting spectral spurs were low enough, but after modifying the stereo modulator it's a good idea to check the composite waveform for glitches and the spectrum for spurs.

An ST-1000A with 68-pF capacitors had no visible baseband distortion products, but subcarrier products were greater than -80 dB. Removing the capacitors reversed the distortion levels. 30-pF capacitors yielded the best balance between baseband and subchannel distortion, with everything below -80 dB.

In another ST-1000A the baseband second harmonic was well above -80 dB. Removing C111 and C112 mainly improves the gate signal fall time. To quicken the rise time, I added 22kΩ across R9 and R15. That brought the second harmonic down to -82 dB. It also reduced the gate drive, which is 8 volts worst case for 2N3819 cutoff. If you modify R9 and R15, check distortion with the modulation level raised to 150% to verify that the FETs turn completely off for large signals.

Check the pilot phase alignment after any change to the stereo modulator. Occasionally I've been able to adjust R93 to bring the phase within spec, but the trimpot was all the way to one end. To permit full adjustment, I added 3.9kΩ across R92.

Be sure to check distortion for both left and right inputs when modifying the stereo generator.

RF Distortion

I used a Marconi 2305 modulation meter to evaluate the overall RF distortion, which includes the distortion of the internal oscillator and stereo modulator plus that of the varactor frequency modulator. The Marconi contains a wideband, low-distortion, charge-pump FM detector. To check its residual distortion, I built a pulse-count FM detector and attached it to the Marconi's 1.5-MHz IF output. I looked at an FM signal with multiple harmonics below -60 dB. The harmonic spectrum was identical for the pulse counter and the Marconi detector, except for some difference in second-harmonic levels below -70 dB.

To characterize the second-harmonic distortion, I tuned the Marconi local oscillator above the ST-1000A signal frequency and measured the distortion level. Then I offset the LO below the signal and repeated the measurement. These LO settings invert the phase of the ST-1000A second harmonic. It will add to any Marconi second harmonic on one side and subtract on the other. I adjusted R52, the ST-1000A predistortion compensation for the varactor modulator, until the residual second-harmonic distortion in mono was the same at both LO frequencies. It was -68 dB, which is the residual Marconi distortion level. This procedure nulls the ST-1000A second harmonic and is the best way to adjust R52.

When the level of a fundamental drops, the second harmonic generally drops further and the third even more so. Stereo with a 9% pilot lowers the baseband signal 6.8 dB. The Marconi second harmonic originally at -68 dB drops to about -82 dB. The lower distortion levels for stereo in the images that follow are a consequence of the smaller baseband signal level.

This shows the demodulated spectrum of a 97.5-MHz monophonic signal deviated 75 kHz. In some units I've seen the second harmonic vary anywhere between the noise floor and -75 dB, depending on temperature. I obtained these results before I knew about the passband inversion trick for nulling the ST-1000A second harmonic. Its level here probably is a combination of ST-1000A and Marconi distortion. The third harmonic is stable, here at -75 dB (-70 to -80 dB in other units). Based on my tests of the Marconi detector, I believe this distortion product is due to the ST-1000A. A third harmonic at -75 dB in mono will drop below the noise floor in stereo.

This is the demodulated baseband stereo spectrum (9% pilot, 75-kHz deviation). The spectrum analyzer sensitivity has not been changed from mono. It is hard to identify distortion products in the residual noise. They may barely appear at some temperatures, but they are always close to the noise floor.

This shows the the demodulated distortion products around 38 kHz. All are less than -82 dB.

Intermodulation distortion products may appear on the demodulated 19-kHz pilot signal. Here all are less than -79 dB. In some units these products vanish.

Equalizing Deviation Levels

When switching from mono to stereo, the pilot increases total deviation. To keep the deviation constant, you must reduce the internal oscillator level. The ST-1000A will do this automatically if you install an optional resistor in the pads provided. One end of the resistor is at the junction of R120 and C54. Function switch section S1-6 grounds the other end in stereo mode, forming a voltage divider that lowers the internal oscillator level.

The yellow rectangles outline the pads for the optional resistor. In my ST-1000A, 160kΩ equalized mono and stereo deviation for a 9% pilot. This resistor is not mentioned in the ST-1000A manual nor shown in the regular schematic, but it is indicated as R206 on the option M3 schematic with a value of 47kΩ.

You may also see a deviation change when switching to L+R or L-R. In my ST-1000A the deviation rose in these modes. Adding 62kΩ across R113 equalized the levels. The pads for this 10kΩ resistor coincide with the lower line of the yellow rectangles. Its value is 8.2kΩ on the M3 schematic.

Pilot Injection

To help set the pilot injection level, the TEST pushbutton increases the meter sensitivity by a factor of ten. To check the times-ten accuracy, apply an external 19 kHz in mono, adjust the input level until the meter reads exactly 100%, and then measure the AC voltage at the COMP jack. Remove the 19 kHz, switch to stereo, press TEST, and adjust the pilot level until the meter again reads 100%. Remeasure the composite signal level, which should be 10% of the previous reading. Adjust the value of R142, a factory selected part, to correct any error. The green rectangle in the righthand image above outlines the pads for R142.

Dual Sweep

I found the PHASE control a bit tricky to set when trying to overlap the forward and reverse CRT traces in dual sweep mode. The control was too sensitive. Adding a 4.7kΩ resistor across it reduces its range and allows easier adjustment.

SCA Frequency

If your ST-1000A has option M2, which adds a switch to select an internal oscillator frequency of 400 or 1000 Hz, you'll find that the SCA frequency varies with the switch setting. This stray-capacitance effect is not mentioned in the ST-1000A manual. Pick one switch position as standard and adjust C44A for an SCA frequency of 67 kHz. The switch's effect on the 10-kHz dual sweep frequency is negligible.

Auxiliary Input

Option M1 provides a wideband auxiliary input on the rear panel that bypasses the stereo modulator and lowpass filter. It's mainly intended for injecting SCA signals at frequencies other than 67 kHz. I use the aux input to precisely check a tuner's deemphasis accuracy. This eliminates any frequency response variation due to the lowpass filter. The aux input also is the place to connect a composite stereo generator to test stereo separation at high frequencies. (Separation for the internal stereo generator is specified above 1 kHz only for option M3, although it still seems quite good at high frequencies without the option.)

To add option M1, install a BNC socket in the prepunched hole on the rear panel, install R34 and C150 in the pads provided at the right side of the PCB, and run a shielded cable from the BNC to the PCB. Adjust C150 for flattest response at high frequencies. About 2.5 VRMS is required for 75 kHz deviation. Although this sensitivity is adequate for normal SCA injection levels below 20%, it may not be enough for full-deviation tests. I added 9.1kΩ across R34 and 39 pF across C150 to increase sensitivity. The meter drops off above 100 kHz but the composite output goes to 200 kHz at full deviation before it slew-rate limits.

dBf Scale

The RF LEVEL dial on some ST-1000As has a second scale calibrated in dBf. This scale was originally intended for use with the Model 100 50Ω:300Ω matching transformer, which has a 1:1 voltage ratio (7.78 dB loss). If you loosen the two set screws you can rotate the knob and dBf scale with respect to the 猩 scale, which is fixed on the shaft. This lets you recalibrate the dBf scale for any amount of external loss. For no loss, 73 dBf should line up with 1 mV. If you alter this alignment to account for the loss of your 50Ω:75Ω matching network and test cable, the dBf dial will indicate the power delivered at the end of the cable.

Circuit

A schematic is here and part locations are here. A schematic for option M2 is here and part locations are here.

More is here.

Updated May 11, 2008