DSE A2760 Stereo Speaker Amplifier

Introduction

I picked up this Dick Smith Electronics A2760 Speaker Amplifier many years ago for a bargain price since it was an ex-demo unit. These amps are a rebadged Class-AB amp made by Koda and sold under many different brand names worldwide. Good points of the construction of the amp include a ~300VA toroidal transformer with individual secondary windings for each audio channel, preamp and relay/protection circuit, a generous amount of filter capacitance on each channel and what appear to be genuine Toshiba output transistors. The rest of the unit is made with fairly generic components (good for serviceability) and the build quality isn't too bad considering the price. It does deliver on it's claimed 80W per channel into an 8ohm load.

Recently I wanted to use this amp as part of a speaker driver measurement setup, so I decided that it would be a good idea to measure it's distortion performance and see if anything could be done to improve it. I found some information and discussion of modifications for this amp on various forums, but no measurements to back up their claims. Shown here is a collection of modifications which deliver quantifiable improvements.

Looking at the schematic of the amplifier, the most obvious area for improvement is that the output transistors are not in the global feedback loop. Instead, the feedback is implemented after the driver transistors (before the main output transistors). I'm not quite sure of the reasoning for this - perhaps the aim was (near) unconditional stability, or they wanted to tout an impressive (read: unnecessary) frequency response / bandwidth. The downside of this feedback scheme is serious non-linear distortion from the output transistors. This also introduces a DC offset which is not compensated for by global feedback and is probably the reason why they have included a DC offset trimmer at the input stage.

Left: KD261 Schematic (some component values in the DSE (KD260) amp vary slightly). Right: DSE amp internals

Measurement - Before Mods

The amp was measured at 1watt (2.83Vrms) with an 8ohm resistive load. I chose to measure at 1watt as it is comparable to general listening levels. I used my PC soundcard (Asus Xonar DX) as the DAC/ADC. A 1khz test tone was played and the output of the amplifier was recorded. The recorded signal was processed with a MATLAB script to produce the following FFT figure:

1Khz sine wave spectrum:

Harmonic distortion performance isn't great, but probably not overly audible unless you have high end speakers (and a reference amp to compare to!). It is however unacceptable for speaker measurement purposes. The residual distortion plot shows that the non-linear distortion is dominated by typical Class-B crossover distortion, obviously caused by the output transistors only having local feedback.

Modifications

The mods involve reorganising the feedback loop so that the output transistors are within the loop, increasing the output transistor quiescent current to 200mA to bias it into Class A at lower volume levels and lowering the closed-loop gain. Additional compensation is required to prevent global oscillation when the output transistors are brought inside the feedback loop, so 2-pole compensation is implemented at the VAS (voltage amplification stage). A two-pole compensation scheme allows for higher open loop gain to be maintained within the audible frequency range (compared to single pole), lowering non-linear distortion significantly.

Method:

• Cut the trace between R231 and the junction of R224/R225. Connect R231 to input side of L201
• Cut the trace between R281 and the junction of R270/R272. Connect R281 to input side of L202
• Connect a 220pF C0G/NP0 ceramic cap in series with a 100pF C0G/NP0 ceramic cap. Connect 220pF to collector of V214, 100pF to base of V215. Connect a 2.2K resistor from the junction of the capacitors to the rail side of R232
• Connect a 220pF C0G/NP0 ceramic cap in series with a 100pF C0G/NP0 ceramic cap. Connect 220pF to collector of V232, 100pF to base of V235. Connect a 2.2K resistor from the junction of the capacitors to the rail side of R278
• Replace R231/R281 with 33K
• Replace R230/R253 with 1.5K
• Shunt R209/R257 with 1K multiturn trimpots (make sure they are set to 1K initially!)

Schematic of Mods (only left channel shown)

Calibration Procedure after modification:

• Power up the amp, adjust RP201/RP202 to set DC Offset (aim for less than 2mV). Let the amp get up to operating temperature and adjust again as required.
• Slowly lower the 1K trimpot for each channel until the voltage across R212/R262 is 56mV. This achieves 200mA quiescent current through the output transistors.
• Readjust RP201/RP202 if required, wait for the temperature of the amp stabilize and readjust all trimmers again as required. It is important to monitor the quiescent current as the amp warms up, as the negative temperature coefficient of the bipolar transistors can cause thermal runaway if the quiescent current is adjusted too high.

With the above mods, gain is roughly halved (to ~23 from ~48) and distortion is lowered significantly. The reduction in gain should not be a problem as even without a pre-amp the amplifier will clip at around 1Vrms input

Measurement - After Mods

1khz sine wave spectrum, 1W/8ohm

There may be further gains (particularly with regards to noise floor) to be had by modifying or removing/bypassing the preamplifier board, however performance is already good enough for my purposes so I left it alone.

As of posting this article, these amps can still be found in Australia second-hand (ebay, gumtree) for around $50-$100.