NCP3170 switching regulator

[Clear as mud]

This is one of the cheapest switching power supply components on Mouser.  The 500 kHz version is about the same price as a standard linear regulator, although the switching regulator requires more external components.

Datasheet: http://www.onsemi.com/pub_link/Collateral/NCP3170-D.PDF

Does this seem like a good chip to use, to build a general-purpose power supply for prototyping?  I am thinking of using two of them, configuring one for a 3.3 volt output and the other for 5 volt, and putting right-angle male headers on the board that would be correctly spaced to plug into the power rails on a breadboard.  The input would be 12 volts via a barrel jack.  I probably would also put enable/disable switches on to turn off one or the other if I am only using one output voltage.

Suggestions are appreciated, especially if you don't like that idea and you would do it differently.

[Hideki]

Sure, why not. I like the NCP parts. You won't get proper adjustable current limiting, but if that doesn't bother you when breadboarding, then go ahead.  The price is right and the documentation is good as well.

I'm using two of the NCP1595 (1 MHz, DFN-6 package) on a DSP board without problems and have designed in the NCP3170 to generate the 5V (from 12V) rail on another board.

Board layout is important, so don't place the parts far apart and then connect them up with thin traces. Polygon fills is the way to go.

[peter.mitchell]

I've used them before, top notch. As mentioned, the documentation is good, pretty much all you need to get started.

[Clear as mud]

I spent a day on the design, but I started running into problems.  I wanted it to be useable up to a maximum current of about 2 amps, but it seems to require very large capacitors and inductors for such a large current.  I downloaded the Excel file from On Semiconductor's site, and entered the basic specifications as follows:
NCP3170A (500 kHz)
Input voltage: min 8; typ 12; max 15.5
Efficiency Target: min 89; typ 90; max 94
Output Voltage: min 4.95; typ 5; max 5.05
Average Output Current: min 0.1; typ 1.5; max 2.2
Output Ripple Voltage: min 10; typ 20; max 40 mV
Transient Regulation: 5%
Transient Current: 1 A
Input Ripple Voltage: min 30; typ 50; max 110 mV
Ambient Temperature: min 15; typ 25; max 40
fcross: 50 kHz (I'm not sure what this is for, I just left it as it was)

With those inputs, the worksheet was recommending:
Output Inductance: min 4.3; typ 10.6; max 30.8 (uH)
Inductor DCR: min 1.0; typ 58.7; max 1.0 (milli-ohm)
Output Capacitance: 223 uF, 8 to 32 mOhm ESR
Input Capacitance: 493 uF, 23 to 86 mOhm ESR

I spent a few hours searching for capacitors and inductors that would work and are not too expensive.  I decided on:
Output Inductor: 2102-V-RC, a vertical toroid with 12 uH and 7 mOhm resistance
Output Capacitors: 2 in parallel, C3216X5R1A107M160AC, they are 1206 mil surface mount size, and the parallel combination has about 200 uF capacitance and only .001 mOhm ESR.
Input Capacitors: 4 in parallel, 35SEPF120M, these are 10 x 13 mm SMD polymer electrolytic and the parallel combination has about 480 uF capacitance and 4.5 mOhm ESR.

So, I have two problems:

• All these inductors and capacitors cost about $15 I think (I didn't add it up exactly, but just got a general idea of prices) and that is about 3 times the amount I was hoping to spend, since I am building these with the idea of making a useful product to sell and I have to make some profit.
• The resulting circuit seems to be unstable.  I downloaded the compensation tool (based on LabView) from OnSemi, and it is a bit over my head since I don't remember all of that math, but on the transient response tab, I could not get it to behave in a stable manner by adjusting the compensation resistor and capacitor.  The oscillation increases with time instead of decreasing.

I'm taking a break from this and working on another project for a few days, but eventually I'll come back and try to finish this.  I'll spend some time reviewing and trying to remember the mathematical background for the gain/phase analysis, and reading the data sheet.  But first I wanted to ask here and make sure I am not overlooking something obvious.  I have a feeling maybe the high-quality inductors and capacitors I specified cause too much ringing because they don't have enough resistance.  In the compensation analysis tool, I increased the ESR by 1 or 2 mOhm for the resistance of the traces and connections, but it didn't seem to help.

-Daniel

[Clear as mud]

With the values I chose, the Excel spreadsheet is predicting a loss of .777 watts at full load with the maximum input voltage.  90% of that is in the NCP3170 itself, since I have little loss in the low-ESR components.  What is a general rule of thumb for power dissipation in a SOIC-8?  I thought 0.7 watts would be fine if I attach the pins to large copper areas as shown in the thermal management section of the data sheet.