I recently compared two similar 3V -> 5V boost circuits, the Sparkfun #08999 5V out @ 100 mA (based on NCP1400) and the Pololu #798 5V out @ 200 mA (based on NCP1402).
The details are below, but the executive summary is: the Pololu 5V boost circuit module has 2x max current and it is cheaper, smaller, has lower noise, and is more efficient at every load current, as compared with the Sparkfun module.
(note: I regularly buy from both Sparkfun and Pololu, but have no other relationship to either company.)
Sparkfun #8999 NCP1400 5V out 100 mA circuit. Measured at 3.2V input, resistive load:
Efficiency vs current out: 81% @ 25 mA, 67% @ 50 mA.
Noise on 5V, 50 mA out: 34mVrms, 300 mVpp (20 MHz BW) 40 mVrms, 600 mVpp (100 MHz BW).
With 5 V, 25 mA out: 18mVrms, 230 mVpp (20 MHz BW), 28 mVrms, 400 mVpp (100 MHz BW).
Current draw from 3.3V with no load: 1.2 mA
Pololu #798 NCP1402, 5V out, up to 200 mA with >2.7V in. Measured at 3.2V input, resistive load:
Efficiency vs. current out: 81% @4 mA, 84% @25 mA, 86% @50 mA.
Noise on 5V out: 10 mVrms, 50 mVpp in 20 MHz BW and 14 mVrms, 100 mVpp in 100 MHz BW
(independent of load current, over tested load range)
Ripple frequency variable, increases with load current:
5 kHz @ 4ma, 50 kHz @ 50 mA output. Vout: 5.004 V to 5.010 V (Vout increases with higher load).
(2nd unit tested, Vout= 4.98 V.)
Current draw from 3.3V with no load: 200 uA
i own a Pololu and i like it.
i only used it in a breadboard prototype context with 2 AA has input and less than 40mA out
so i didn't really push it but it works well :-)
Sparkfun's PCB has a veery poor layout. Pololu's is WAY better. Most DC-DC products from SparkFun have really really poor layouts, with thin traces (prone to parasitic inductance), messy grounding (seriously WTF? the grounding of that PCB is crap) and looks like they also can't properly choose the passives.
Seriously, 67% efficiency is nothing to be proud of.
When you design a DC-DC converter board, you must use as thick as possible traces and you must be very very careful with the grounding, to avoid parasitic inductances between ground zones, which could shift the ground potential between them causing extra ripple. A common good practise is using a star-ground topology or an isolated ground plane. They are lucky because the NCP1400 switches at only 200Khz....
Thanks for sharing. Indeed, the Pololu board looks like the typical switcher layout from a datasheet. It's not so much traces, rather the parts sit on islands of copper.
The TMS 828 (?) -5volt generator for the Dangerous DSO has similar layout recommendations in the datasheet. I didn't dig up the datasheet for these parts, but it would not surprise me if they also had a recommended layout. I see that in most switcher datasheets, even simple stuff like MAX232s usually have it.
[quote author="ian"]Thanks for sharing. Indeed, the Pololu board looks like the typical switcher layout from a datasheet. It's not so much traces, rather the parts sit on islands of copper.
The TMS 828 (?) -5volt generator for the Dangerous DSO has similar layout recommendations in the datasheet. I didn't dig up the datasheet for these parts, but it would not surprise me if they also had a recommended layout. I see that in most switcher datasheets, even simple stuff like MAX232s usually have it.[/quote]
Yeah, most times large traces translate into islands of copper. It's a very common practise and most designs are done like that.
But the problem with sparkfun's layout is mostly the ground. Just look at it. It has thin paths of copper between pads, it's not solid and has an easily avoidable ground loop. So well, you have parasitic L and R and a ground loop. That's an explosive combination for such an analog circuit. The low switching speed of that particular regulator saves them from epic failure, but the efficiency is significantly lower than what the datasheet says it could obtain with the same input and load.
Seriously, there could be infinite ways to route that particular PCB, and sparkfun chose one of the worse... and I even bet that they used stock capacitors without any consideration about ESR....
and along with the efficiency problem is also the switching noise on the output, which is about 5x worse than on the Pololu circuit. That's what annoyed me the most about the Sparkfun part. It looks like the Pololu device has a large-ish ceramic cap, probably low-ESR and the other one has something else, doesn't look ceramic.