A good power amplifier should disconnect the speaker during power-up and power-down. The signal chain is likely in a chaotic state during this period and the chaos produced better does not reach the speaker in order to protect it from potential damage. Furthermore, the loudspeaker should be protected from potential catastrophic failure of a DC coupled power amplifier driving the loudspeaker. The typical failure mode is that one of the amplifier output stage power transistors forms a short circuit with the power supply, leaving the loudspeaker connected to the rail, resulting in destruction in most cases. I will discuss electromechanical relays and solid state relays and present my project based on a solid state relay.
Usually the loudspeaker relay is an electromechanical one. While this is a good solution for muting the loudspeaker, an electromechanical relay may fail to interrupt the current that flows during catastrophic output stage failure, especially for high power amplifiers. The advantage of electromechanical relays is that they may have low distortion in theory. This benefit may vary from relay to relay and also dependent on age and wear of the relay. AC operated relays may induce noise into the signal. The disadvantage of electromechanical relays is limited DC current breaking capability. Wiebe Cazemier conducted a very illuminating study how to properly connect electromechanical relays so that they are more likely to disconnect the speaker during a DC event. He has also demonstrated that relays are prone to failure interrupting large DC currents. Conventional relays have very limited capability to switch high DC currents at high voltages. Special relays become very expensive or even unobtainable. Power relays that are used in electric installations may be able to break high current at high voltages, but fail at low signal level since they are not designed for this application. The result is crackling sound and even signal interruptions at lower levels due to oxides on the contacts. Actually those relays are built for switching mains voltage only and high current operation. Summary is that electromechanical relays may work fine for low power amplifiers, but are not suitable for loudspeaker protection in high power amplifiers.
Solid state relays discussed here are made of two N-channel MOSFETs in series with their sources connected together. There are other types as well, which are unsuitable for audio applications. The advantages of such solid state relays are that they have low and predictable on resistance with generous current conducting capability, near zero "contact" wear, constant properties over lifetime and they can switch very fast, which allows to actually switch high currents. Disadvantage is sensitivity to excess voltage and non-linearity, which contributes to distortion. The main non-linearity of concern is that the RDSon is a function of the VDS across the MOSFET. Also, the CDS has impact on the ability to mute the speaker and is frequency dependent. Someone has performed a very interesting study how the non-linearity of solid state relays may contribute to distortion. Unfortunately I can't find the website anymore. Conclusion is that the contribution to distortion is negligible given that the ratio between relay resistance and loudspeaker impedance is reasonable.
There are some considerations regarding the DC detection and control circuitry. The DC detection shall not falsely trigger during normal operation. Music may contain some DC due to asymmetric wave forms and low frequency content that may also trigger the detection. When using low pass filters to distinguish between signal and DC, less sensitivity towards false triggering means slower response time. Control circuitry needs to be reliable and robust. It can be kept very basic or more advanced by adding just a few low cost parts. For this project, versatility was given priority and therefore, higher part count and larger PCB real estate were accepted.
- Delayed connect of loudspeaker after power on
- Instant disconnect of loudspeaker on power down (AC loss) triggered by power supply via SHUTDOWN interface
- Automatic restart after AC loss
- Fast disconnect of loudspeaker in case of DC on amplifier output
- Does not false trigger with 10Hz signal at amplifier output
- Latch off on DC fault
- Manual reset after DC fault possible if fault is cleared
- Galvanic isolation between channels to fit dual mono amplifiers with separate power supplies
- Wide supply voltage operation from 12VDC to 90VDC
- No extra power supply required
- Extra SHUTDOWN input for manual shutdown (used for shutdown on mains AC loss by transformer control module)
- Extra AUX_CONTROL output that switches together with relay to control auxiliary circuitry like input muting relays
- Status indicator LEDs: Relay status, DC status, latch status
- Functional as module inside amplifier and also as external unit
Dependent on transistors used
|AMP_L_SIG||Power amplifier output left||Connect to the left amplifier speaker output. Reference to AMP_L_GND|
|AMP_L_GND||Power amplifier ground left||Speaker return left|
|AMP_L_PWN||Amplifier negative supply rail||Optional connection. Reference to AMP_L_GND|
|AMP_L_PWP||Amplifier positive supply rail||Optional connection. Reference to AMP_L_GND|
|Control circuitry ground||Has no galvanic connection to amplifier ground or right channel ground|
|CTL_SUP_L||Control circuitry power supply||Wide voltage range input from roughly 12V to 90V. Reference to CTL_GN1_L / CTL_GN2_L|
|CTL_SDN_L||Control circuitry shutdown input||Leave floating for normal operation, pull to CTL_GN1_L / CTL_GN2_L for instant speaker disconnect|
|CTL_AUX_L||Control circuitry auxiliary output||Open collector output which switches together with speaker relays. Reference to CTL_GN1_L / CTL_GN2_L|
|AMP_R_SIG||Power amplifier output right||Connect to the right amplifier speaker output. Reference to AMP_R_GND|
|AMP_R_GND||Power amplifier ground right||Speaker return right|
|AMP_R_PWN||Amplifier negative supply rail||Optional connection. Reference to AMP_R_GND|
|AMP_R_PWP||Amplifier positive supply rail||Optional connection. Reference to AMP_R_GND|
|Control circuitry ground||Has no galvanic connection to amplifier ground or left channel ground|
|CTL_SUP_R||Control circuitry power supply||Wide voltage range input from roughly 12V to 90V. Reference to CTL_GN1_R / CTL_GN2_R|
|CTL_SDN_R||Control circuitry shutdown input||Leave floating for normal operation, pull to CTL_GN1_R / CTL_GN2_R for instant speaker disconnect|
|CTL_AUX_R||Control circuitry auxiliary output||Open collector output which switches together with speaker relays. Reference to CTL_GN1_R / CTL_GN2_R|
|off||off||off||Module is offline, Loudspeaker disconnected|
|off||on||on||DC present, Loudspeaker disconnected|
|off||on||off||DC transient stored, Loudspeaker disconnected|
PCB assembly in CAD
Schematic left channel
Schematic right channel
PCB and assembly options
In the schematic I used two transistor pairs: The SC1845 / SA992 and MJE340 / MJE350. The SC1845 / SA992 are pin compatible with the MJE340 / MJE350 and have high VCE breakdown and this is why I chose them. They are obsolete by now, but can be replaced by any other transistors. The circuit will work with any transistors. The only limiting factor is VCE and SOA and both need to be chosen to fit the power supply voltage used. There are two PCB revisions: One that is designed for small signal transistors only and one that has the larger power transistors designed in. The one with footprints for larger transistors allows installation of low power ones as well of course. The schematic presented here is for the high power variant. Usually one would run the control circuitry from any low voltage supply like 12V or 15V or whatever is present in the amplifier. In this case, literally any transistors can be used for the control circuitry. Only the DC detection needs higher power transistors since they are confronted with the audio output signal that may be high voltage.
The attentive reader might have spotted that the design is actually two completely separate circuits with one channels circuit being an exact copy of the other one and so is the PCB design. This is because copy and paste in the EDA environment seemed more appealing than spending more time at this stage to optimize the design. The modules form factor is determined by other modules that have the same form factor so PCB size reduction was not an obtainable improvement anyway and built for low supply voltage operation, the extra component cost is negligible. There is a lot of redundant control circuitry that can be shared between channels while maintaining almost all features listed previously. Only the DC detection and solid state relay circuitry needs to be separate for each channel. Meanwhile I have drafted a variant with optimized topology and added some additional circuitry that makes the module even more useful in some applications.
Status and outlook
The circuit works well both in simulation and reality. PCBs for the low power variant using TO-92 transistors are assembled and tested. I improvised the high power variant using TO-126 transistors for the CCS and found performance as good as the low power variant. I will use the protection as low power variant since I have +12V for such control circuitry available in my amplifier.
I found an issue with using the DC protection in my amplifier: The amplifier generates some DC voltage at the output during power on and this event is long enough to trigger the protection circuit. This means that the DC protection circuit needs to be reset after every power up. I would need to redesign the circuit so that DC events at the loudspeaker output only trigger the protection after the turn-on delay is over and the loudspeaker is connected. Prior to that, DC events should be ignored.
There are two possible solutions to deal with DC events during power on:
Disable the latch by removing R14 and R42. This will let the circuit reset itself a few seconds after end of a DC event. This is within the intended mode of operation. Manal user reset is an option and automatic reset is an option as well.
Better would be to change the circuitry in a way that DC events during startup are ignored. U1/U2 and U4/U5 could be connected to the auxiliary control output instead of ground. It seems feasible to improvise this on the PCBs without creating a mess.