Notes on MPPT and buck converter
February 16, 2017
Simple buck converter with 2n7000 mosfet and gate drive from a signal-generator
I have been thinking about what would be required to create a switch-mode analog MPPT system to charge a small battery. Ideally, small and low-powered - something suitable for a solar-powered wireless 3.3V ESP8266.
- two differentiating op-amp circuits. One for change in power over time, and the other for change in regulation point over time.
- a four-quadrant analog multiplier to divide the two values to produce delta power/ delta regulation point - for the basic negative feedback response
- a pid loop for regulation
- perhaps another multiplier to generate the power signal from voltage and a current-sense resistor. Although a LiPO4 battery has very small voltage differences between charging and discharging - so may be able to rely just on current.
- Using the battery as sink, a design could be quite similar to a buck-converter. The feedback would control the sink current via PWM, to maximize power.
- LTC6992 pdf looks nice. Voltage control of frequency and PWM. Upto 1MHz. And 0-100% PWM. Imporantly it can run off 2.25V to 5.5V single supply.
Multipiliers are hard!
- A log/sum/anti-log approach with op-amps only works with unipolar inputs while we need a bipolar inputs and outputs - a four quadrant multipler.
- Sophisticated multiplier ICs like the AD633 pdf use an embedded zener as a reference. They all seem designed around higher voltage supplies and can’t be used for a lower voltage (eg. 6V) single rail design.
- A gilbert cell based on discrete transistors needs trimming. Use a matched transistor array?
- THe HFA3101 pdf is a Gilbert cell implemented with six matched bjts. But it’s designed for GHz RF and not that widely available.
- The MC1496 pdf is a RF modulator multiplier based on a Gilbert Cell, but suffers from non-linearity - else it looks pretty good. Fig 23 has an IC schematic. It pulls out all the useful pins - so can control the biasing.
- Another choice is an OTA such as the LM13700M pdf. There is a four-quadrant multiplier based on the OTA on Fig. 24.
Some good notes on multipliers can be found in Analog Devices tutorial pdf
Conventional buck converters and drivers often use two n-channel devices, and rely on a boost capactitor to drive the high-side mosfet gate in a source-follower configuration. Not sure if a p-channel on high-side might not be simpler - even with higher RDS(on) compared with an n-channel. In a high voltage application (> 10-14V) the gate wouldn’t need to be driven through as wide a voltage range. For low voltage - need to be logic level, or use voltage booster?.
mcp1416 non-inverting, and mcp1415 inverting - supposedly matched propgation timing. Up to 18V - could use for both low-side and high-side up to 18V. If shoot-through isn’t a problem then could control from single ltc6992. or use 74hc AND logic on gates.
- LiPO4 rather than Lithium for safety, and simplified charging management. See, float charging etc
with digital - 3.3V arduino pro mini - current is nice and under 1mA. But
but at 8MHz clock with 1MHz switching, we onlly get 8 clock counts PWM of phase resolution
An opinion that arduino is not appropriate for SMPS here
- Use two (eg. irf540) n-channel mosfets and Dickson charge-pump as bootstrap voltage for mosfet gates?
- There are p-channel mosfets with low RDS(on) 0.02ohms STP80PF55