A DC protection circuit is typically included at the output of most audio amplifier, and is meant to disconnect the loudspeakers if a significant DC component is present at the amplifiers output. This is important as excess DC current through the loudspeaker (or headphones) will generate significant heat and can damage it. Unlike commercial products, most DIY builds I’ve seen over the years, don’t include a DC protection at the output. This of course leaves the loudspeaker/headphones connected to it vulnerable in case of a problem in the amplifier. Therefore, when I was planning one of my previous amplifier builds, I’ve decided I should first design, a DC protection circuit to add the output of the amplifier. I’ve decided to slightly enhance the circuit to include a few extra features other than just DC protection, and make it as versatile as possible:
- Supply voltage of +/-12V to +/-75V (or single 24V-150V supply)
- Wide input swing of +/-55V
- Support single-ended/balanced/active-ground amplifiers
- 2 channels input per board
- Adjustable sensitivity
- Independent detection per channel for a robust design
- Support 2 outputs (A/B/A+B) with relay switching
- Visual notification(LED) of active output, and fault
- Delayed start-up
- Accelerated shut-down for reduced “popping” noise
- 40mA supply current with single relay energized
- Up to 8A load current with default relay
The circuit can obviously be modified if needed for a simpler build (not output selection for instance), and can be extended as far as voltage range is concerned. For instance, I have since used the same board with a few less parts with a single 12V supply for the output of a small headphone amplifier.
Circuit Schematic and Operation:
The circuit can be broken down into 3 main parts: power supply, signal filtering and detection, and switching circuit.
The power supply section is implemented as a simple Zener follower circuit meant to generate a nominal supply of roughly +/-12V:
For these not familiar with this topology, it is fairly simple to understand. Zerner diodes D4/D5 have current flowing through them, biased by R15/R16, creating a +12V/-12V accordingly. These voltage are then follower by BJT’s Q6/Q7, which drop a 0.7V across them, but offer much lower output impedance to supply additional current to the load. The diode drops will make the output voltage closer to +/-11.3V nominal value.
Signal Filtering and Detection
This part of the circuit is based on a differential amplifier structure, implemented with operational amplifiers (Op-Amps). Overall, a 3rd order low pass filter (LPF) is implemented inside this circuit. This is done to enable high selectivity between DC voltage and low frequency music content, and offer minimal probability of a false trigger.
The circuit starts with a differential amplifier built around ULA, with a 100Kohm input impedance. It is an attenuating stage (-13dB) to extend the input swing beyond the local supply rails, and includes bypass capacitors (C1/C2) for the first LPF. The output of ULA is fed into an additional amplification stage around ULB, with one more LPF (C3) and adjustable gain (via trimmer TRIML) to allow adjustable selectivity. This can be replaced for a fixed gain structure by using resistor BYPSL. The output of this stage goes through one last LPF (R7/C4) before reaching the detection circuit.
The detection of fault is accomplished by the 2 PNP devices Q1/Q2. If the voltage at the output of the last LPF is below roughly (-)0.6, Q1 will conduct through R8 to the next portion of the circuit where the switching will be taken care of. If that same voltage is above (+)0.6V, Q2L will conduct down the same path. As long as the voltage stays within (-)0.6V to (+)0.6V, not fault is detected.
To quantify the input detection threshold, all with have to do is divide this 0.6V figure by the circuit DC gain:
Punching in the numbers, the desired value for resistor TRIM/BYPS can be chosen. Using the default value of 250K for TRIM will allow the threshold to be adjusted in the range of 75mV to 270mV.
The switching circuit is where the control currents from both channels are combined and used to disengage the output relay in case of a fault. It is also where the delayed start-up, accelerated shut-down, and output selection are implemented:
To make it simpler to understand, lets follow the circuit from the end back to the beginning. K1/K2 are the relays used for outputs A/B, with SW1 used to select between them (or both of them). LEDA/LEDB and their series resistors are simply there for visual notification of the active output. Q5 is the switching transistor that controls the current (or lack of it) through the relays. The voltage divider and LPF made from R12/R13/C6 is there to offer a delay at start-up and protect Q5 from excess gate voltage.
Transistors Q4A/B are controlled by the current from the detection circuits discussed earlier. When they are activated, Q4A will discharge the gate node of Q5, disengaging the relays. Q4B will turn on LED1 for visual notification of the fault detection.
Finally, Q3 and its surrounding passives are used for accelerated shut-down. As the supply voltage starts dropping, the base of Q3 will be pulled lower, while its emitter will remain high due to capacitor C5. When this happens, Q3 will conduct, and will feed current into the same node that is controlled by the fault detection circuit, disengaging the outputs.
The board layout uses 2 layers, with wide traces for the audio signal, which are placed further away from the active components to minimize any change of noise coupling. As always, the ridiculously low price of PCB fabrication nowadays makes it much faster and more efficient to have the boards professionally made instead of trying to etch them at home.
To verify the operation of the circuit I have first only assembled one half of it, as if working in a single channel topology. I was interested in verifying mainly 2 things. The first, is that a fault is indeed properly detected at the expected threshold. The second, is that the circuit is sufficiently selective to allow large amplitude at low frequency without tripping. Don’t mind the flux, it was obviously cleaned before the circuit was put to actual use 🙂
As can be seen above, I’ve assembled mine with BYPS resistor instead of a trimmer, as for the application where I will use this board I have no need for the adjustment. I have used a 0R jumper for a (theoretical) threshold of ~270mV. Measuring the actual threshold yielded very similar numbers, with 280mV in one polarity, and 260mV in the other.
To quantify the selectivity I’ve decided to use a 10Hz input sine, which half the minimum frequency that we typically concern ourselves when it comes to audio. Measurement showed that at this frequency the circuit will only trip at an amplitude of ~28Vrms (~100W into 8ohm load), and this number rises sharply with frequency as all 3 LPF’s are located below 10Hz.
The circuit in its default form is meant fora stereo amplifier. However, there are people who build mono-blocks or dual-mono amplifiers, and want to use completely separated circuits for each channel. This can easily be done with the PCB by simply leaving out (not populating) one of the filtering/detection circuits. However, I recommend you use the remaining relay contact from the unused channel, and parallel them. This will halve the relay contact resistance, and double the current handling capacity.
Another concern is for amplifiers which incorporate both single-ended, and balanced outputs. This is more common for headphone amplifiers. Remember that the DC protection circuit can only detect an error between the signals connected to its input. Therefore if you only monitor one of the outputs, it won’t offer complete protection to the other. To overcome this limitation, I recommend you build a single board for each channel (as if it is a dual-mono build), and use both detection circuits on each of the boards. All you have to do is connect one input to monitor the balanced output, and another to monitor the single-ended output. I have done just that in one of my builds, and I have verified it reliably protects both outputs.
At the time of the build I’ve organized a basic BOM for the default build of the detection circuit, and I’m including it here. Most parts can be safely replaced with other equivalent parts, but I’m including the parts I’ve used so that you can verify the physical dimension and choose whatever parts you prefer.
Most of the information above, as well as a few other notes, as were posted in another internet forum are included in the following PDF file.