Technics SE-9600: A New Power Supply Circuit Board

Most audio power amplifiers are supplied by an unregulated power supply: A transformer with a center tap, a bridge rectifier and then two large filter capacitors. All the ripple/noise rejection happens inside the amplifier circuit itself. The Technics SE-9600 is a fairly rare exception. Let’s have a look.

❗ 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.

The power supply circuit

The Technics SE-9600 uses a quasi-complementary symmetrical linearly regulated power supply that drops the unregulated voltage of nominally +/-67V down to regulated +/-50.5V. This explains why Technics easily got away with using two physically quite impressive but capacitance-wise rather anemic 10,000µF/80V capacitors (at least for an amplifier that is rated at 2x165W into 4 Ohms).

Each rail has its own Zener reference and uses two NPN series pass elements in parallel, each with its own emitter resistor for current balancing and current sensing. The voltage dividers for the voltage feedback are adjustable for both rails individually. Their output is fed to the base of transistors TR503 (positive rail) and TR507 (negative rail) respectively, which in turn modulate the control signal of the series pass elements.

The bandwidth of the control loop is very substantially limited by C502 (positive rail) and C505 (negative rail) - Technics specified a capacitance of 18nF. Only for the positive control loop they added another 18nF capacitance from the output to the base of TR503, for a more reasonable transient response. Notably missing on the board is any output capacitance that many voltage control loops absolutely need for stability. Instead, such output capacitance is found on the driver boards (one 100µF electrolytic capacitors per rail and per board).

A separate +24V Zener-based supply is fed off of the unregulated positive rail.

Technics implemented a couple of protection circuits. There is an over current protection circuit that measures the current through the series pass elements' emitter resistors and triggers an SCR (there is one per rail). The SCR in turn pulls down the control signal to the respective series pass elements, effectively disabling the rail until the power is turned and kept off until the SCR resets (i.e. the capacitors are somewhat discharged). To turn off the positive supply to the amplifiers in case the negative supply is interrupted, the positive SCR is triggered once again.

Both the DC protection and the over temperature protection circuits rely on a different mechanism: They actively open two normally closed contacts of the protection relay that feed the transformer outputs to the rectifier diodes. During normal operation this circuit is powered by the regulated +50.5V supply (using an or-ing diode); therefore it would cut itself off of power as soon as it triggers the protection relay. To avoid this (and implement proper latching), the normally open contacts of the relay - which will close in case of a protection event - are used to supply the unregulated power from the transformer to the protection circuit via a rectifier and filter through another or-ing diode. Once the protection relay has been triggered the power has to be turned of long enough for the relay to switch into its initial state.

I’d like to mention that I much prefer a normally open protection relay to be in the signal path between the amp and the speakers even if this increases the output impedance.

The redesign

As mentioned in the previous post I wanted to keep the circuit close to what Technics intended. Nevertheless, slight changes and especially adjustments in component values are required for proper operation with the more modern replacement transistors I selected.

Component selection

ℹ️ Note

The component selection is meant for the redesigned board. The values might or might not work as replacement for a defective component on the old board.

ℹ️ Note

The schematic in the service manual has a couple of typos or Technics changed something along the way. Always compare the components/values with what is actually installed in your unit.

Diodes

Silicon Controlled Rectifiers (SCR)

Bipolar Junction Transistors (BJT)

Those TTA004B/TTC004B and BD243C transistors are probably overkill, but generally match the originals reasonably well, I think.

Resistors

For the most part Technics used 5% carbon film resistors. Today, 1% metal film resistors are not that expensive at all, so I went with those instead. Does it really matter here? I don’t think so.

The 3W emitter resistors are metal film from the factory. I replicated this, but there is one “issue”. Modern 3W metal film have a significantly smaller body, meaning they have to have a significantly higher temperature to dissipate the same power. I paralleled two 3W resistors with (roughly) twice the resistance: 1 Ohm || 1 Ohm vs. 0.47 Ohm in the original design. Doing this has the potential disadvantage that the resistors might not fail as fast and protect other circuitry in case of catastrophic failure. I’d be interested to see actual data for the pratical impact this might have.

And of course, to make my life easier, I chose multiturn potentiometers for the adjustable voltage divider in the feedback loop.

Capacitors

General capacitor selection

Firstly the main filter capacitors: Since the capacitors (Technics branded) are exposed, any replacement should have a similar apperance, i.e. same physical size and color. Combine this with the required electrical specs and you know in advance that there won’t be too many options. Luckily, my main capacitors seems to be in fairly good condition, so those will stay in there. Still, I did a quick search: The closest capacitors I found had the same physical size, were probably black as well (if I trust the documentation; who knows what you end up with), but were rated at 15,000µF/100V instead of the 10,000µF/80V of the originals; and those were very expensive.

For all other electrolytic capacitors I had to use Panasonic caps, after all it’s a Technics amplifier. And I generally like the FR-A series a lot: Low ESR and rated for 105°C over 5,000h to 10,000h.

Technics decided to use a polarized capacitor for filtering the output in the DC protection circuit. This is acceptable because the circuit limits any negative voltage across the capacitor to one base-emitter voltage drop of TR511 (TR510 does the same thing for positive offset voltages). Still, I’m not too fond of this approach. One reason being that the capacitance of a polarized electrolytic capacitor in reverse polarity typically is a lot smaller, so I’d expect the circuit to have a different time constant depending on the polarity of the DC offset. And it’s simple to avoid by just using a bipolar capacitor for C506 - which I did.

For film capacitors I chose WIMA FKP for smaller values, MKP for the larger C507 (330n). One could argue I should have used Panasonic for the film capacitors too…

Required changes for loop performance/stability

To come closer to the performance of the original board while using the new semiconductors I changed/added some capacitors; this might reduce stability. I came up with the values by performing limited simulations (LTSpice) and limited real world tests. It shall be made clear that the power supply might not be stable in all applications/operating conditions. See also section Tests. In other words: Use at your own risk.

Especially the control loop for the negative rail seemed to be more prone to oscillations in my bench testing: Whereas the positive regulator was stable without any output capacitance for my test currents up to 1.45A - admittedly with poor transient response -, the negative regulator absolutely required some output capacitance. With this experience and some simulations in LTspice I just increased the capacitance of C118, C120, C218 and C220 on the Driver Boards from 100µF to 220µF. For now I also added 470µF/63V capacitors per rail directly on the Power Source Circuit Board. Might not be necessary, but I think this still increases stability rather than to hurt it.

LED debugging indicators

I added a couple of debugging LEDs, which turned out to be a very good idea. To limit the power consumption I used red high efficiency LEDs that light up well with currents as low as 100µA.

• LEDs with a two-transistor constant current source connected to the unregulated supply that are lit up as long as the voltage is above around 2V. This indicates that there is still a significant charge in the main filter capacitors.
• LEDs on the +/-50.5V rails with a 470k resistor in series: These LEDs indicate that voltage is present on the regulated rails. They will turn off when SCR-based protection circuit has tripped.
• LED with a 470k series resistor in parallel with the relay coil: Indicates that a relay-based protection circuit has tripped

Protection relay

I might have found a new old stock replacement, but luckily, I was able to clean the contacts of the socketed 48V protection relay.

Connectors

Although binding posts/wire wrap connections surely have their advantages, I wanted the modularity of having connectors instead, which can be especially useful for testing. I opted for pluggable screw terminals that are rated for up to 300V/15A.

The PCB

The new PCB was designed in KiCAD.

Tests

Since this is just a hobby for me, my verfication and test efforts are limited in scope and extent. Also, the amount of test equipment required to properly test such a power supply design is substantial. For example, it makes a lot of sense to do the first tests with current limited power supplies. With a target voltage of +/-50.5V a +/-60V power supply would be the absolute minimum (ideally it should be capable of +/-70V which is more in line with the actual output of the transformer). My only power supplies capable of delivering these voltages are my 0-120V linear power supplies, but those offer only up to 1.5A, limiting the testing a bit.

Anyway, I did some basic transient testing using the aforementioned power supplies in conjunction with a DC Electronic load at (up to 1.45A) with some modern onsemi power BJT in the earlier stages of the development; later some simple tests with sine waves at fixed frequencies in the SE-9600, that either did not show instability or I did not recognize it as such.

One possible problem I saw was a “ripple” on the amplitude vs frequency plot created with the QA403 audio analyzer using “expochirp” at 2x110 W into 8 Ohms, that disappeared at a slightly lower output power of 2x100W. I do not know the exact cause (e.g. power supply, amplifier, measurement setup etc.) and I did not investigate.

I inadvertently tested the over current protection circuit on the positive rail while measuring something (I really shouldn’t short the output. Oops!), but did not check others.

Open question

Why exactly did Technics decide to implement a linear regulator? Yes, the regulator pretty much eliminates 100/120 Hz ripple; but a well designed amplifier shouldn’t have much trouble to reject the low frequency ripple by itself. But this regulator is so slow, it’s not even close to my HP linear bench power supplies. And the output capacitors of the regulator aren’t that large either (and will therefore have a decent amount of ESR as well). So what does it do for the high frequency response? Does it help? Does it hurt?