If you like to experiment with electronics, sooner-or-later you’re going to build something you’d like to use in a portable way, perhaps running off batteries. While things that run off 5V are easy to cater for, as USB is ubiquitous, other voltages can be a little more tricky. For 3.3V devices, you could run them off 5V with a low-cost foolproof linear regulator, but you’re wasting 33% of the power as heat. If you wanted to avoid the regulator though, then you’ll run into the trouble of battery voltages not being a “neat” multiple of 3.3V. You could try two AA cells, but then you’d start off at 3V and as they run down, the voltage could fall too far. If not, then you could try three cells, but then it’ll be a bit too high at 4.5V and could cause damage. Likewise, a Li-Ion battery can see voltages from 3V up to 4.2V fresh-off-the-charger. What’s really needed is a buck-boost converter that can accept voltages below-and-above 3.3V and convert it to 3.3V.
As a result, I went to eBay to try and find a relatively cheap and small boost-buck converter that could run an ESP8266 or similar device requiring 3.3V from a range of input voltages. I settled on a generic module that was listed for AU$1.84 including postage.
The module is quite diminutive in size and came in a zip-lock bag with no other information. It appears to be as described – two inductors indicating two stages with a number of on-board ceramic surface mount capacitors for smoothing. The chip is marked with 2149F, so I suspect it is a MicrOne ME2149. While the chip seems capable of switching up to 4A, the SS34 Schottky diode is only rated at 3A. Looking further, the power dissipation limitations of 0.8W for the package suggest that maybe 1A is a more realistic expectation.
The inductors themselves are of the open type, so I suspect this is probably not going to be the most efficient design possible.
Only after looking at the back of the module did I find anything to identify it – it’s a Canton-Power DDO603SA which appears to be available in 3V, 3.3V, 3.7V and 5V variants. These variants may only be down to a change of the surface mount feedback setting resistors.
In order to most accurately test the module without making the set-up needlessly complicated, I decided to mount the unit onto a high-quality breadboard. Power to the module was supplied by a Keysight E36103A 20V/2A programmable power supply in 4-wire remote sense mode. The module was loaded using a B&K Precision Model 8600 DC Electronic Load also with external sense mode turned on. The unit was plugged into the breadboard with the adjacent pins used for load and supply to minimise breadboard resistance contributions. The pins next to this were used for sensing – as very little current flows through the sense connections, this would not affect their performance and allow for lead resistance to be compensated.
A Panasonic low-ESR 1000uF electrolytic capacitor was added (not in shot) to the adjacent pins at the input to the converter to ensure any current-transients were taken care of and avoid the length of the supply leads from affecting module performance.
Voltages and currents were read from the power supply and load respectively to avoid needing further external instrumentation and introducing further connection resistance. As a downside, the current readings were only to the milliamp when above 20mA, although the introduced error in efficiency appears to be ~1% at the most. For the most part, we are interested in the “overall” big-picture performance of the module rather than fractional-percents.
Because of the lack of proper specifications for the module, I had to work conservatively to get results without destroying the module. I found that it didn’t reliably work below 1.1V despite some claims that it works at 0.9V, although the claimed maximum of 6.5V didn’t seem to be too strenuous for my module.
The first thing I was interested in knowing was the quiescent no-load current consumption. This is important especially for projects where the powered equipment remains asleep or in very low power consumption states as the power consumed by the converter can be quite a bit higher and limit the run-time from a set of batteries.
From the graph, it seems the quiescent consumption is not too bad. It won’t break any records, but seeing as the consumption is below 1mA throughout the voltage range, getting down to about 180-300uA in the 3-5V range, you should still get quite a bit of life out of batteries.
Converter efficiency was not too exciting either, with most of the results in the 70-80% band depending on voltages and load. Due to the two-stage topology, such results are somewhat expected, especially in lower current ranges where the quiescent contribution to reduction of efficiency dominates. The module seemed pretty well-behaved up to 250mA, but above this, the module would not start-up under load, instead chirping quietly but pegging the current limiter on the power supply at 2A. This suggests maybe some type of latching-event under high-current output which may lead the module to wasting energy or even self-destruction depending on the current available. As a result, I don’t suggest using the module with any more than about 250mA of load on the output at start-up. However, if the module is running, it seems happy to have its load increased up to about 1A although at higher currents, the module cannot operate with the lower input voltages.
Unfortunately at this efficiency, if your voltages are strictly higher than 3.3V, then a linear converter wouldn’t be that much worse (especially if the quiescent current is lower). You’d really only choose the buck-boost when the voltage varies above and below 3.3V.
Looking at the output voltages, once the input reaches 1.5V, the output appears to be regulated. The voltages up to 100mA of load are relatively close to the intended 3.3V but with a notable reduction as current increases. By the time 250mA is reached, it has deviated quite noticeably, with higher currents further reducing output voltage.
The testing did not assess the output ripple and noise nor whether the unit has any overheat, over-current, short circuit protections.
For an AU$1.84, this boost-buck converter module is cheap and relatively “anonymous” in terms of proper specs. While I didn’t test the ripple or protection, it seems the unit was most happy in the 1.5V to 6.5V range with output currents of 250mA or less. If started below 250mA, then it can provide even 1A (as tested) but with increasing levels of voltage drop as a result. It might be just enough for an ESP8266 but probably not ideal, but what do you expect for the price?
The efficiency was an “expected” 70-80% as a dual-stage converter with the component choice obviously affected by price. You would only choose this over a single-stage buck or boost if your input voltage is known to vary both above and below the output voltage of 3.3V. The quiescent current was also decent, being 180-300uA between 3-5V, meaning that you’re not wasting too much energy just powering the converter but it is still quite high compared to an ESP8266 in deep sleep that consumes just 20uA or a linear regulator which often is around the same.
The biggest issue seems to be that above 250mA, my unit failed to start-up and instead shorted out the supply, suggesting that the unit might self-destruct in some cases (e.g. with a high current supply).