aqua computer AMPINEL Exclusive Review – The end of all problems and fear with the 12V-2X6 or 12VHPWR connector

1 - Introduction and background 2 - Teardown with PCB and topolgy 3 - Security levels and active feedback 4 - Settings in the aquasuite software 5 - Transients and conlusion

Circuit analysis of the AMPINEL prototype

I read the circuit board from the high-current side towards the logic. The six 12 V lines of the 12VHPWR come in on the left via the device connector, each run through a defined measurement path on the front and go back to the output on the right. The large low-impedance shunts with R001 mark the current measuring points for each line; the line currents are recorded here for each channel. The power MOSFETs are located directly next to these shunts as active actuators for each channel; I am deliberately only naming them by function and not by type, as I will deliberately not identify the specific parts. The hurdle for imitators can be set a little higher at the beginning. These special, very low-impedance and low-loss MOSFETs are connected in series to the respective path and allow a controlled increase in the voltage drop per wire, whereby a disproportionately loaded line is specifically relieved and the current is shifted to the other lines. This is the actual balancing mechanism, which must be fast at the critical moment so that peaks do not end in heat at a contact. This then goes up to an internally defined maximum power loss per line before the whole thing starts to pulse so as not to overload individual channels.

I can also see a row of identical TSSOP-8 components below the shunt and MOSFET zone. The position of these double OpAmps in relation to the shunts and the symmetrical wiring with short, paired conductor tracks indicate the control (regulation) of the current per channel. These components – after the small, digitally controlled high-side regulators on the rear in the immediate vicinity of the input connector – form the analog front edge of the regulation. This is precisely where the speed advantage lies, as the first intervention is always primarily analog before the firmware monitors, evaluates and, if necessary, intervenes digitally. The gate signals to the MOSFETs are short and wide, which ensures low loop inductances and fast, clean gate transitions. Leakage inductances and ground references are kept local, which improves stability in partial load and linear operation and reduces oscillation. Not easy with such a tiny circuit board.

The central MCU is also located at the rear in a large QFP housing. It handles the higher logic, i.e. the aggregation of all sensor values, control fine-tuning, alarm logic, profile management, logging and USB communication. To bridge the gap between analog speed and digital decision logic, several discrete RC networks and level conditioning units are located close to the MCU. You can also see the internal connectors for USB, the two signal outputs and the temperature sensor, which are designed in the layout with clearly separated reference grounds and protective circuits. The signal paths from the analog front ends to the MCU are short and repeatable, which serves to ensure measurement consistency of the six channels. All in all, this results in a hybrid architecture in which fast analog paths provide short-term limitation, while the MCU controls the sustainable load conversion, escalation stages and logging.

The power loss is optimally distributed. Copper areas and via fields under and next to the MOSFETs and the shunts ensure low-resistance current conduction and heat dissipation. Thermal via arrays and wide busbars in the inner layer can be seen in several places, which dissipate the heat flow into the PCB and mechanical carriers. This is crucial for a balancer because the actuators operate briefly in the linear range during the control torque and have to dissipate voltage in order to divert current. A clean Kelvin connection of the shunts can also be seen, the measurement taps go separately to the analog ICs so that the load current path does not distort the measurement.

The OLED board is built separately. The flex cable of the display ends on a small intermediate board with fine conductors and a connector to the main board. Series resistors and small capacitors are located directly at the FPC input, which indicates signal conditioning and ESD protection. This decoupling reduces interference in the high-impedance analog parts and prevents display activity from increasing the measurement noise. The mechanical integration of the OLED via a carrier bridge shows that the electrical system should be voltage-free, which makes sense for a device with high path currents and pulsating loads.

I classify the overall concept in such a way that the load distribution takes place on two levels. Level one is the analog immediate reaction per wire, which works with shunt, amplifier or comparator and control MOSFET, where current peaks are taken away before software makes decisions. Level two is the digital coordination by the MCU, which handles limit values, profiles, timing, alarm cascades, power limitation via the GPU’s sense lines and the entire monitoring. This separation explains why the AMPINEL can smooth load peaks visibly and effectively without stalling the system, and why the interventions are reproducible.

I would like to expressly point out that the circuit board shown is a predominantly manually manufactured prototype and not a production version. Solder joints, test points, provisional fixings in the plug-in area and the mixed assembly indicate a development stage. Behind the analog front stage visible here, the gate control, the measurement path equality and the clean separation of high current and logic areas, a great deal of development time has been invested, because balancing with six parallel lines requires not only control technology, but also layout discipline, thermal fine-tuning and EMC care. I’m deliberately not specifying the concrete type of MOSFETs, that was a requirement; functionally, they are to be understood here as controllable series controllers which, together with the shunts and the fast analogue circuitry, form the basis for load redistribution.

To all tinkerers and optimizers: the installed 0.5 mm pads are good, even if I used putty when assembling them. Better safe than sorry, but it hardly brings any improvement.