I love 19" rack gear, and the little Sansui AU-317 is no exception. With 2 × 50 W into 8 Ω, it is a medium-power integrated amplifier, of course built in the late ’70s like many of the greatest “vintage” amplifiers. Its specs are not bad at all, but there are limitations too. Let’s check it out.
❗ Caution
All information is provided “AS IS”, without warranty of any kind, express or implied.
❗ Caution
The elevated voltages inside this amplifier can cause damage, severe injury, and death. If you are not authorized or qualified to work on such equipment, do not do it. If you follow along, you do so at your own risk.
Technical specifications
According to the service manual, the Sansui AU-317 delivers 50 W per channel into 8 Ω with no more than 0.03% total harmonic distortion and has a frequency response of DC to 200 kHz (+0 dB / −2.5 dB) at 1 W when the signal source is connected to the POWER AMP IN inputs. The damping factor is specified to be approximately 70 with an 8 Ω load. In a brochure, Sansui advertised a fairly high slew rate of 40 V/µs.
The trouble with 4 Ω loads
The service manual does not specify the power into 4 Ω loads, and I have seen people online advising against using this amp with 4 Ω loads, with remarks to the effect of “that’s how you blow up your amp.” Others say that it’ll be fine. The rear-panel screen print warns:
WHEN YOU CONNECT TWO PAIRS OF SPEAKER SYSTEMS, EACH OF THEM MUST HAVE IMPEDANCE OF 8 OHMS OR MORE.
To me, this implies 4 Ω capability as long as only one speaker system is active. Also: The aforementioned brochure specifies 72W into 4 Ω.
Adding to the confusion, and maybe one of the reasons some people might not want to connect 4 Ohm speakers, the somewhat different Sanusi AU-317 II has a more limiting warning printed on the rear panel:
SPEAKER IMPEDANCE MUST BE 8 OHMS OR MORE. WHEN USING TWO PAIRS OF SPEAKER SYSTEMS SIMULTANEOUSLY, TOTAL IMPEDANCE MUST BE 8 OHMS OR MORE.
What’s right then for the first generation of the Sansui AU-317?
The Sansui AU-317 uses complementary power BJTs 2SC1403 and 2SA745 in the now almost extinct TO-3 package. Luckily, a datasheet is easily available (from ISC, not Sanken), but of course there is no SOA diagram. The datasheet specifies a maximum collector-emitter voltage of 100 V, a maximum collector current of 8 A, and a maximum power dissipation of 70 W. The transition frequency is given as 10 MHz typ. (NPN) and 15 MHz typ. (PNP), respectively.
Let’s assume the Sansui AU-317 actually delivers 72 W into 4 Ω, then we’d have peak currents of about 6 A for a resistive 4 Ω load. This is relatively close to the maximum collector current of the transistors. The real issue though is the power dissipation in the transistors with inductive loads like speakers: it can be much, much higher due to the phase shift between voltage and current (i.e., voltage is present across the transistor while current flows through it), often coinciding with reduced power-handling capability at higher collector-emitter voltages due to secondary breakdown. So, despite the fact that BJTs can usually withstand a higher power dissipation for short periods of time, I would say that using 4 Ω speakers with this amp should be possible, at least to some degree, but isn’t ideal.
Initial condition
From the start, my unit was in fairly good physical condition. As far as I can tell, no one has ever serviced the unit, and I couldn’t detect any major issues either. Some of the controls, such as the speaker selector, were very slightly scratchy, but I have seen much worse. I noticed a volume change when switching between tone control “on” and “defeat.” To be honest, I wouldn’t be surprised if it came that way right from the factory.
Since I planned to do more than just a basic service, I performed a substantial number of measurements to identify potentially hidden issues and to be able to compare the results with measurements taken after the service and modifications.
In contrast to the second generation of the Sansui AU-317, the first generation supports splitting the preamplifier and power amplifier, making separate measurements really simple.
Pre-service Measurements
This is an excerpt of the measurements performed.
Preamplifier
Power amplifier
Service and modifications
If there isn’t a good reason not to replace almost 50-year-old electrolytic capacitors, I usually replace them on sight. Exceptions can be expensive main filter capacitors in good working condition. (This should not be confused with a repair strategy; it is just maintenance.)
Power supply
The main filter capacitors measured about 82% and 85% of their nominal capacitance, respectively, so in this case they had to go. Sansui used capacitors with two additional mounting pins. Although less common today, KEMET still offers replacements with the exact footprint (ALF70C103DD063)—no need to modify the PCB or use an adapter. Note that the new capacitors are about half as tall.
The AU-317 has a Zener-regulated series-pass power supply for the preamp, phono preamp, tone circuits, and microphone preamp. It lacks the closed-loop control of the higher-end models in the series (e.g., AU-717) and therefore exhibits more ripple. If one is inclined to reduce the power-supply ripple, an obvious method would be to increase the capacitance of the capacitors in the power-supply circuit (C605, C606, C607, C608, C609, C610). The generally smaller size of modern capacitors helps here.
That being said, I would strongly advise against doing so without understanding the roles of the different capacitors in the circuit and the consequences of changing their capacitance, including but not limited to:
- Larger filter capacitors will generally put more stress on certain components such as the transformer, fuses, rectifier diodes, and resistors. This may lead to damage to individual components, the power supply, or connected circuits.
- Increasing the capacitance might prolong the time until the voltage rails have fully stabilized. The protection circuit might switch the protection relay on before the rails have stabilized, resulting in a certain DC offset at the speaker terminals during power-on (and therefore an audible pop from the speakers).
It might be necessary or advantageous to mitigate these side effects, especially if the capacitance change is significant.
The same idea can be applied to capacitors C11, C12, C23, and C24, which together with R47, R48, R53, and R54 form a low-pass filter for the supply of the VAS stage. Be aware that the pulse energy dissipated in the resistors during power-on is substantial; one has to ensure that the resistors used can handle this.
I did increase certain capacitance values and made sure to mitigate potential issues.
In my opinion, it is recommendable to check and potentially replace heat-stressed transistors such as the series-pass elements (T601: 2SD356, T602: 2SB526). In this instance, one of the series-pass transistors was at the low end of its H_FE group. This could have been the case from the factory, or the transistor may have degraded over time. I chose to replace these with modern MJE15032/MJE15033 transistors.
Preamplifier
The amplifier had a gain mismatch of about 0.1 to 0.3 dB, explainable by variations in the stepped volume potentiometer and potentially tolerances in other components. This was reduced with a tiny, inconspicuous parallel SMD resistor that changes the gain of one channel just slightly.
Tone amplifier
The slightly audible deviations between “tone on” and “tone defeat” are likely predominantly caused by tolerances in the stepped bass and treble potentiometers and, as such, cannot be fully compensated without replacing the potentiometers.
By decreasing the effective values of R37, R38, R41, and R42, I was able to flatten the frequency response with the tone circuits engaged. The optimal values were determined with the help of my programmable precision resistor. Again, SMD chip resistors (parallel to the THT components) were used to make the changes permanent. This leads to a nicely matching frequency response for the ±2 and ±4 bass and treble settings in my particular case, but this can lead to worse results in other cases (see also post-service measurements). Again, this is due to the imperfections of the potentiometers.
Driver circuit
The four still-working transistors in the VAS stages (T07, T08, T11, T12) were replaced due to visual signs of heat stress accumulated over almost five decades with widely available TTA004B and TTC004B transistors. The TTx004B transistors have a lot of gain, usually even more than the E grade of the obsolete KSC3503/KSA1381. Unfortunately, the output capacitance (C_ob) is quite a bit higher. Not ideal… Still, I’ve found the TTx004B to work pretty well (see measurements): I think it’s a decent choice of a transistor that you can actually order from reputable sources in large quantities at a reasonable price.
Although the original driver transistors (T13, T14: 2SD357; T15, T16: 2SB527) were quite badly tarnished, the corrosion did not impact their electrical performance (yet?). By now, you might correctly guess that I replaced those as well. TTC004B/TTA004B transistors were chosen again, here for their comparatively high current gain and transition frequency, which should be helpful in this EF2 design. I want to make clear, though, that I am not sure these are the best choice. The reason is that they are far less robust than other common driver options such as the aforementioned MJE15032/MJE15033, even if the TTx004B’s current-handling capability is almost double compared to the original drivers (1.5 A vs. 0.8 A continuous collector current). TTx004B transistors are still lower-power devices, and from my simulations I would guess that they operate close to their SOA limits at very high loads with low-impedance speakers—especially without proper heat sinking as in this application. On the other hand, I am pretty confident that for my use case they should do just fine.
The fusible resistors were still within spec, the worst one was about 4% high; all were replaced with flameproof resistors with a specified fusing characteristic. My initial measurements revealed a minor gain mismatch between both channels of approximately 0.2 dB. This can be fixed by replacing the feedback resistors (R13, R14, R31, R32) with low-tolerance or matched resistors. I chose matched 1% 50 ppm metal film resistors, but want to make clear: I did this for the fun of it; such a tiny difference can usually be considered inaudible.
New output transistors
This is my own amplifier, and I am absolutely willing to replace perfectly well-working original output transistors. Of course, I will keep the retired ones. Requirements for the new output transistors include a compatible case (TO-3, TO-3P), a wide SOA, and a high(ish) transition frequency. Therefore, MJ21193/MJ21194 complementary power transistors are out, and there are not many other new TO-3 devices available anymore.
To me, the obvious choice is the complementary power transistor pair NJW3281G/NJW1302G (TO-3P), with a maximum collector-emitter voltage of 250 V, a maximum continuous collector current of 15 A, a maximum power dissipation of 200 W, and a wide SOA. With a transition frequency of about 40 MHz (PNP) and 50 MHz (NPN), respectively, at similar test conditions as for the original transistors, the new outputs can be considered much “faster,” although the output capacitance is likely much higher as well.
An alternative would be the slightly less robust 150 W variants NJW0281G/NJW0302G, which have a lower output capacitance. Luckily, the new transistors have a somewhat higher current gain than the original ones, reducing the load on the drivers.
Sansui specified to set the bias so that the voltage at the test point (i.e. across both emitter resistors) is 10 mV. For the new output transistors, I almost doubled the bias current to just below 20 mV in favor of lower distortion figures.
Heat sinking
New driver and output transistors are all well and good, but the heatsink solution is what it is. So, for short periods of time—where the thermal mass of the heatsink is large enough—the amplifier should be much more robust. Still, thermal limitations remain, both for the drivers and the outputs. I disassembled the two-part heatsink to reapply heatsink compound—the tiny amount of compound that was applied at the factory was completely dried out. For the output transistors I used Bergquist Sil-Pads (TSP K1300/K-10).
Protection relay
Of course, I had to use my new solid-state relay, which I presented in one of the previous posts. The replacement itself was fairly uneventful; it works as intended. Although the solid state relay does not require a flyback diode I left one in circuit, just in case someone wants to convert the amplifier back to a machanical relay.
Protection indicator
In the Sansui AU-717, the power LED doubles as a protection indicator by blinking as long as the amplifier is in protection. In contrast, Sansui saved a couple of bucks (or cents) by omitting a protection indicator on the AU-317’s front panel—the power LED is on as long as the positive rail voltage is present.
Since I use my solid-state relay, which does not produce an audible click when (dis)engaging, I opted to replicate the AU-717 circuit. The result is a small PCB that implements a multivibrator circuit like the one in the AU-717, made of two transistors, two capacitors, and a couple of resistors. To make the implementation simple, an optocoupler galvanically isolates the “not-in-protection” input (driven by the voltage across the relay coil/control input). The PCB is mounted vertically on the left-hand side of the chassis using holes predrilled by the factory.
Mains cable
Usually, I leave well enough alone and do not change the mains cable. That being said, I generally prefer power inlet sockets over directly attached mains cables, and this time the mains cable was in undesirable shape. So I removed one of the unswitched outlets and replaced it with an IEC type C8 socket with almost identical dimensions. Only a tiny bit of filing was required. Note that it is absolutely necessary to ensure that the connector can handle the applicable electrical and safety requirements.
Post-service/modification measurements
Preamplifier
Power amplifier
Conclusion
The service and modifications were successful: the amplifier should be more robust and less prone to damage with 4 Ω loads (although I would still be careful with sustained high loads and difficult-to-drive speakers due to the thermal limitations of this rather compact amplifier). Also, most measurements show similar or better performance for both the preamplifier and power amplifier:
- Less hum from the preamplifier
- Less or similar distortion for the most part
- A flatter frequency response from the tone amp in the neutral position
- Less difference between the amplitudes of both channels for both the preamplifier and the power amplifier
- Similar bandwidth and slew rate (the gain/phase margin might be slightly reduced, see comment on step response of the power amplifier in the image description)
- A relay that does not really care whether it turns on or off with a signal present
- A protection indicator (blinking power LED when in protection)
It’s a great little amplifier that performs surprisingly well. I enjoy using it.