Repairing a Little Mark III Bass Amp

Stock photo of amplifiers, from https://www.pexels.com/photo/guitar-stage-set-in-rock-studio-4413723/

Markbass amps are well-known (at least within the bass player community). Their light weight (as compared to amplifiers with hefty 50/60Hz mains transformers) makes them highly portable, while still sounding great. I received a Little Mark III amp that was not working, not even the power LED came on.

Front of the bass amplifier, with user controls
Back of the bass amplifier, with mains and speaker connections

What’s inside the case?

The top of the case comes off easily by removing 12 screws. The preamplifier stage (on the bottom) connects via a flat ribbon cable to the effect send/return and the XLR and tuner outputs (on the top left). The slightly tilted vertical PCB (top center) carries the Speakon and 1/4" output connectors. Next to it is the 12V 0.14A fan and the IEC connector for mains power. The main board in the center is home to the switch-mode power supply (right) and the power amplifier stage (left). The fan is installed such that it pulls air through the case and blows it out in the back. Thus, the heatsink attached to the power supply transistors receives an airflow that’s been preheated by the power amplifier stage.

Inside view of the amplifier, with removed top cover

Initial measurements showed no continuity from the mains inlet to the power switch. Because there was no fuse in the compartment of the IEC connector. Thanks to the sticker on the back (Fuse T 4 A 250V) this was easy to fix. Moreover, the cable from the IEC inlet had been disconnected from the mains switch. Pushing the blade connector back in was straightforward.

According to the writing on the transformer, the device had already been repaired once in June 2015. Luckily, a service manual can be found on the web to find out the original part numbers.

Fault diagnosis

Rather than switching the amplifier on right away (as the obvious faults had been cleared now), it was easy to take some preliminary measurements with the case being open. Here’s the results of continuity tests of the eight beefy transistors in this amp (note that I measured them in-place, rather than unsoldering any components yet):

So both the power amplifier stage and the switch-mode power supply are defective. Not a good idea to connect this amp to mains and blow another fuse for no reason!

Rather than this, I’ve removed the board from the case to take a look at the bottom of the PCB. There is considerable flux residue around the solder joints of the eight power transistors, hinting at the fact that they might have been replaced before. Besides this, flux residue is also visible around the pins of the DIP-8 power supply controller IC. The PCB traces under the flux partially revealed blank copper, which makes it appear like a corrosive flux had been used in the previous repair without washing the residue off the PCB afterwards. Not good.

Bottom view of the PCB, with flux residue around the solder joints of the amplifier MOSFETs

Upon closer inspection, it moreover turned out that all of the silicone blobs connecting parts on the board to the heatsink had been separated, as if the heatsink had been removed previously. Also, the manufacturer wouldn’t have shipped a device with thermal paste smeared all over the heatsink (I hope). Lastly, while using matched MOSFETs in the amplifier stage is honorable, adding labels and colored dots to each of them seems impractical for manufacturing at scale.

IRFP9240 transistors in the output stage, marked with handwritten labels
IRFP240 transistors in the output stage, marked with handwritten labels

The transistors themselves were mounted using mica pads with lots of thermal paste in the power amplifier, and slightly oversized ceramic spacers (2mm larger than standard TO247 types) for the power supply. One of the ceramic insulators was cracked when I unmounted the SMPS transistors from the heatsink.

Mica heat transfer pads with dried-up thermal compound
A cracked aluminum oxide isolator

Output stage transistor replacement

After unsoldering the six transistors in the output stage, it turned out that only two of them (one towards the positive and one towards the negative rail) had been shorted. Rather than ordering just a single replacement candidate, I placed orders for a total of seven (from two different sellers) in an attempt to find a triplet that would be sufficiently similar in terms of their electrical characteristics.

The easiest way to match transistors is to use a source meter, connect its positive terminal to the gate and drain pins of the MOSFET, and its negative terminal to the source. Then sweep through the range of drain currents (I went from 1mA to 500mA) and record the gate (=drain) voltage for the desired current. I clamped down the transistors to a heatsink, used a 4-wire measurement and recorded the values over the meter’s serial output for convenient plotting.

A source meter, connected to a heatsink-mounted MOSFET, configured to measure transistor I-V curves

The goal is now to find three devices whose I-V characteristics are as similar as possible. I numbered the transistors I had newly purchased from 1-7, and also took measurements of the two previous ones that weren’t shorted (labeled 810 and 818).

A diagram showing the I-V curves of nine IRFP240 transistors

Luckily, out of the 9 devices I had sampled, three (003, 006, and 007, which actually came from two different dealers) were very close in terms of their characteristics.

I only ordered three IRFP9240s to drive the output towards the negative rail, and all of them were very similar. Seems I lucked out.

Prior to mounting the transistors, I polished the heatsinks using steel wool and placed them on ceramic insulators using Keratherm KP99 thermal paste. I needed a total of about 1ml of the paste to mount all of the transistors (in case you’re considering how many syringes to buy).

IRFP240 and IRFP9240 transistors mounted on heatsinks
Three MOSFETs on a heatsink, mounted to the amplifier PCB

Unfortunately, on both sides of the board, a trace on the upper side of the PCB had lifted from the board when removing the middle transistor. I added little bodge wires to fix this.

A PCB trace that has separated from the top of the circuit board
A bodge wire to ensure connectivity of the separated trace

With all transistors replaced, the output amplifier section was ready.

Photograph of the output amplifier section with all transistors mounted, taken from the side

Switch-mode power supply repair

The manufacturer apparently also had a good time scraping off the chip marking on the power supply controller (hint: it’s an IR21531D). Larger-than-usual solder connections at the terminals of the MOSFET gate resistors (like R76 and R77) document more of the previous repair attempt. This is likely more of a cosmetic issue, though. Black charring is visible around the pins of the SMPS controller. This raised concerns about whether it was still working.

Photo of the SMPS driver IC on the PCB
PCB with SMPS driver IC removed

A quick check (hooking it up to a scope without any MOSFETs attached), however, nicely confirmed that it’s outputting non-overlapping square waveforms at the expected frequency. So no replacement of this “not recommended for new designs”-part is needed. That’s good.

Oscilloscope probes connected to the SMPS driver IC to test its function

Knowing the driver was still working and without any other visible damage, the two FETs to drive the transformer were swapped for brand new ones. This included polishing the aluminum rail they were mounted to, replacing their ceramic insulators, and adding new thermal compound.

Two IRFP27N60K MOSFETs mounted side-by-side on an aluminum heatsink

Overtemperature protection

According to the service manual, the amplifier contains two 10k NTCs that are used to trigger an overtemperature protection circuit. While the NTC mounted close to the output amplifier’s heatsink was easily accessible and properly soldered, it took a while to locate the second temperature probe on the PCB: It’s placed under the transformer. Through measurements, I could tell that it was disconnected, so it was inevitable to remove the transformer in order to access the solder pad. Apparently the previous repair attempt consisted of shoving the soldering iron underneath the transformer without confirming that the added large blob of solder actually connected the NTC to the pad. Spoiler alert: It didn’t.

Disconnected NTC under the SMPS transformer
Brown flux residue after removing the disconnected NTC

While I was at it, I was curious to learn how the fan speed depends on the temperature. Thus, I copied the relevant part from the service manual into KiCad and used its SPICE simulation feature. The results were initially far off due to an error in the service manual’s schematic (R53 is actually 1k5).

By having SPICE sweep through the resistance range of the NTC, the fan runs slowly up to a temperature 50 degrees Celsius at the NTC. Once the temperature rises further, i.e., the NTC resistance decreases, the fan speeds up. The fan’s maximum speed is reached at approximately 83 degrees Celsius. However, the overtemperature protection kicks in at an NTC voltage of 4V or less (which is the case at around 75 degrees Celsius), and disconnects the speaker in an attempt to bring the temperature down.

Simulation of the fan control circuit

Testing

Now that both the power supply and the output stage have been fixed, it’s time to take the amplifier for a test. It’s conveniently possible to disconnect the +-80V power amp rails by removing a solder jumper on the bottom of the board. An ingenious idea, as it allows for testing the preamp and power amp parts separate from one another.

Disconnected +-80V jumper connection on the PCB

In a first test, the output stage was supplied with a current-limited 32V from a lab power supply. While this did obviously not power the driver circuitry, any shorts would be detected right away. At a negligible current consumption, the problem of shorted output transistors seemed to be resolved.

Testing the power supply is a little more tricky. Rather than connecting the amplifier to mains, I used a lab power supply first, set to 180V AC at 50 Hz. I had tried lower voltages first, but they weren’t sufficient to start the SMPS circuitry. The current intake of the amp with the output stage disconnected is 144mA. While that’s quite considerable for an amp with a disconnected output stage, at least the consumption did not increase over time and no part of the circuit heated up excessively.

A lab power supply set to 180V AC shows a current consumption of 144.1mA for the amp

With the power supply working, and no short circuit in the output stage, it was time to re-connect the 80V rails.

Solder jumper added to connect the 80V rails to the output amplifier.

With the amplifier connected, the current intake rises to 235mA at 230V AC. There’s no DC offset on the output voltage, as confirmed by measurements – and the output relay clicking and connecting the speaker.

The lab power supply set to 230V AC shows a 235.5mA current draw with no speaker attached

In a final test, I provided the bass amplifier with a 500mV RMS sine wave from the signal generator built into my scope, and connected it to a 23.15 Ohms load. The output shows a proper and undistorted amplification of 44V RMS (134V peak-to-peak) at an average current of 1.85A.

Oscilloscope showing its generated sinusoidal wave (captured from the amplifier's tuner output) and the voltage at the speaker terminal

Hot or cold bias

There’s one final part to the setup: The bias voltage. Keeping in mind that the power transistors were exchanged by ones that require an almost 0.25V increase in gate voltage in order to drive the same current, tuning is likely needed. Conveniently, there’s a trimmer easily accessible in the center of the power amplifier to aid in setting the bias current to the right value. Clockwise turning reduces the gate bias (and thus the bias current), while it is increased when turning the trimmer counter-clockwise.

Multimeter showing a bias voltage of 1.6224mV

Given that the output transistors have been matched, they can be expected to share one third of the total bias current each. Thus, measuring the voltage across the source resistor, computing the corresponding current using Ohm’s law, and multiplying it by the number of output transistors yields the bias current. For example, a bias voltage of 1.6224mV across the 0.33 Ohms source resistor equals a current of 4.916mA. Multiplying this value by three (because there are three FETs connected in parallel) gets us the total bias current of 14.75mA that’s flowing between the rails without any signal at the input. When I received the (repaired) amp, its bias current was set to 40mA.

Setting the current too high will lead to the power transistors running hot, setting it too low will lead to cross-over distortion around the zero-crossing of the output signal. I used the scope’s signal generator to supply the amp with a clean sine wave and used its FFT feature to measure distortion at the output. I started with a source resistor voltage drop of just 0.5mV and gradually increased it until no further improvement of the distortion was visible.

I ended up with a setting of 2.0mV across the source resistor at room temperature, i.e., a total bias current of slightly less than 20mA between the rails in idle mode (i.e. no input signal, gain and master volume turned down all the way). Lower bias currents led to visible distortion around the zero-crossings when amplifying signals above approximately 10kHz into a 4 Ohms resistor (not visible when using the aforementioned 23.15 Ohms load). Keeping in mind that it’s a bass amp (10kHz being somewhat above the signal of interest), I think this can be tolerated. Thanks to the integrated Vbe multiplier (T9 being thermally close to the heatsink), this value also drops with increased temperature.

I noticed that it’s quite tricky to reach the thermal equilibrium because opening the case leads to different temperatures on the heatsink and the Vbe multiplier transistor, which again changes the bias current. Repeated testing on consecutive days (with plenty of time for the heatsinks to cool down again), however, ultimately allowed me to determine the heatsink temperature when idle (25.1 degrees Celsius) and the corresponding bias current (2.0mV).

Is all well that ends well?

I eventually took the risk to connect the amp to a 4 Ohms speaker cabinet and use an actual electric bass guitar as the signal source. It worked flawlessly, even though it needs to be admitted the I have only played the amp using moderate volumes so far. The heatsinks didn’t get hot to the touch. Fingers crossed it had only overheated previously (because it was trapped in a rack without proper ventilation, or connected to a speaker of too low impedance).