Step Up Converter: N-channel choice

Other Hardware General Electronics
1

Hello,

I'm trying to build a step-up-converter for arduino, converting most everything < 5.0 V to 5.0 V. I'm trying to use http://nl.farnell.com/jsp/search/productdetail.jsp?sku=1460676. The datasheet advises a nice clean setup to use it:

749×477 41.4 KB

with a parts list in the same datasheet:

912×278 66.3 KB

However, part Q1, the N-channel mosfet (part http://nl.farnell.com/on-semiconductor/ntgs3446t1g/n-channel-mosfet-20v-5-1a-tsop/dp/1705289?ref=lookahead) is not readily available in Europe, so I thought I could settle for one with very comparable specifications http://nl.farnell.com/jsp/search/productdetail.jsp?sku=1864593.

Will this substitution work ?

2

You should make sure that:

  • The new Vgs is equal or lower than the original
  • The new Id is equal or higher than the original
  • The new Rdson is equal or lower to the original
  • The new Vds is equal or higher than the original
  • The new Ton and Toff are equal or lower than the original
3

Thank you Muddy,

that works out so I'm going to give them a try.

Cheers, Jack

4

The results are not too well, it seems instead of "stepping it up" it yields the same voltage at the output as the input and as input voltage increases suddenly a large current is flowing

I am wondering: I have made breakout boards for all components (since all are smd) and build the circuit on a breadboard. Could this be the reason it doesn't function ?

I made breakout boards for each component so I can test them seperately.

As far as I can tell all capacitors are working
The n channel mosfet can switch a load (test led) on and off
The coil makes contact, but I can't test its magnetic abilities
The diode is just a diode, I have no idea what to test about it
The step up converter is or is not working, I have no way to tell except for that the entire circuit is working

Any suggestions for tests on how to find errors in my circuit are appreciated.

5

Did you use the same model of inductor as specified in the parts list for the evaluation board?

Did you connect the CE pin to Vin?

Breadboard layouts don't work well for switching regulators. In particular, you need to star the ground connections out from a single point, as shown in the schematic.

6

I did some checking:

the required inductor is 10uH (20% tol), Rdc_max = 0.1 Ohms, Idc = 1.44 Amps, however, not available to me

the one I used is 10uH (20% tol), Rdc_max = 0.06 Ohms, Idc = 2.6 Amps, so my guess is this should work out

CE is connected to V_IN

However, your remark concerning ground lines might point to a flaw in my design: I connected all ground lines to the ground connector rail of my breadboard, I actually used both rails and connected them with wire. Is this wrong ? Can I use one connector rail, should I put all wires in adjacent holes in the breadboard ?

Much obliged to hear from you

Jack

7

Usually a switch mode power supply circuit will need tight lay-out on a ground plane. Breadboarding isn't
likely to give satisfactory results. Dead-bug-style construction on a copper ground plane is far more likely
to be workable.

8

CaptainJack:
Can I use one connector rail, should I put all wires in adjacent holes in the breadboard ?

That's more likely to work, but as MarkT says, you really need good ground area. Also, keep all the wires short, especially the input capacitor, output capacitor, diode, and source and drain of the mosfet. Did you use the recommended diode, or a similar fast Schottky one?

9

Gentlemen,

thank you for the kind replies. Building this dc booster is harder than I expected...

I used the prescribed schottky diode.

The reason I breadboarded was because my initial attempt didn't work: I integrated all components on a single etched board. The board is single sided but ground goes along all the edges, so I presumed this would do the trick, see the figure below.

469×512 66.8 KB

With annotations:

469×512 89.5 KB

Now: it didn't work. I had no way to check the functioning of all seperate components... is this possible while they are integrated in a circuit? I could have made an error in the design, I could have made an error while soldering, I could have destroyed a component without knowing. Can anyone offer an approach so I can test during the building process?

The datasheet of the NCP offers an evaluation board layout, but the quality was not good enough to use for etching. And what would be the fun of not building it yourself?

So, open to any suggestions before etching a new board.

Cheers,

Jack

10

I would have expected that layout to work, although the ground trace at the bottom of the board would ideally be wider. What mosfet did you use in the end? You've put pads for an SOT23 package on the pcb, but the one you said you were going to use is SOT26.

11

I'd try replacing the NCP with another one. Recheck wiring/schematic/etc.

I've been building simple DC/DC converters (12V to 48V, 12V to +15V, 12V to -15V). You can make them work on a breadboard using through-hole parts equivalent or breakout boards (it will still function), though probably efficiency won't be very high due to layout, etc. But it should still work as intended.

If you do the final PCB layout right, you'll see efficiency increase (or max power capability increase sometimes by two-fold) vs. the breadboard version.

You need to use a schottky diode, ordinary diode wouldn't do.

Exact brand/model# of inductor not needed to make it work, but DCR and inductance is very important.

C1,C2 should be low ESR capacitors.

Check temperature on your NCP. If it's getting too hot, it could be current limiting or shutting down, or worse-- dead. Need adequate heatsinking.

Usually a switch mode power supply circuit will need tight lay-out on a ground plane.

I designed (2) boards, one with ground planes on both top and bottom, and my original board design - ground plane only on bottom side, top plane just traces. The placement of components didn't changes, same everything.

The PCB design with the ground plane on the bottom, and just simple traces on top layer performed better (stability, lower ripple, etc) vs. the full ground plane on both top and bottom layer.

(This is just a 2 layer board, so maybe a 4 layer design will also behave differently.)

My entire DC/DC converter is on a PCB just 1"x1" big.

I thought that was pretty cool, until I saw this... 1Amp on 3mm x 3mm

dcdc.png

12

Ah, it was SOT23, that is the one I tried, with this N channel mosfet

Specs:

MOSFET, N CH, 30V, 5A, SOT23
Transistor Polarity: N Channel
Continuous Drain Current Id: 5A
Drain Source Voltage Vds: 30V
On Resistance Rds(on): 0.022ohm
Rds(on) Test Voltage Vgs: 4.5V
Threshold Voltage Vgs Typ: 800mV
Power Dissipation Pd: 1.3W
Operating Temperature Min: -55°C
Operating Temperature Max: 150°C
Transistor Case Style: SOT-23
No. of Pins: 3
MSL: MSL 1 - Unlimited
SVHC: No SVHC (19-Dec-2012)
Current Id Max: 5A
Operating Temperature Range: -55°C to +150°C
Voltage Vgs Max: 12V

However, when placing a new order they were out of stock, so I found this replacement, which is SOT26

Specs:

MOSFET,N CH,20V,5.3A,SOT26
Transistor Polarity: N Channel
Continuous Drain Current Id: 5.3A
Drain Source Voltage Vds: 20V
On Resistance Rds(on): 0.027ohm
Rds(on) Test Voltage Vgs: 4.5V
Threshold Voltage Vgs Typ: 700mV
Power Dissipation Pd: 2W
Operating Temperature Min: -55°C
Operating Temperature Max: 150°C
Transistor Case Style: SOT-26
No. of Pins: 6
SVHC: No SVHC (19-Dec-2012)
Operating Temperature Range: -55°C to +150°C
Voltage Vgs Max: 12V

However: I see now that the SOT23 are in stock again (they were restocked much sooner than expected) so I will get those. With these the first etched board did not work. For the breadboard I went with the second.

I'll widen the lower ground trace, recheck the circuit, etch a new board and place new components on it.

For C1 and C2 I used low ESR tantalum capacitors.

Would a ground plane on the back side help ? As far as I can tell from your experience it would make it better, but it should work single-sided.

If have no heatsink on it for the moment, would I need it? Usage will be 0.05 Amps regular, with peaks of 0.5 Amp for very short durations (~1 ms); I have some capacitors to help with that.

It's good to hear you have some working boosters, 1x1 inch would be very good for me indeed. I'll recheck everything and post my new layout.

13

I tinkered with this circuit in Linear's SPICE. I replaced the switch module with a gated pulse train at the appropriate frequency. I used a comparator to sense the output voltage and to control the gating of the PWM pulse train. At any rate, it was sensitive to the MOSFET I used. The default model, a generic FET wouldn't boost the signal. I guess it has to do with on/off switching speed of the transistor or some other characteristic like maybe RDS(on)? Maybe this is why the OP isn't getting any boost?

14

CaptainJack:
Ah, it was SOT23, that is the one I tried, with this N channel mosfet...

As well as the other specifications, check that the total gate charge of the mosfet you used wasn't much higher (preferably lower) than the recommended one. The total gate charge affects how much drive the mosfet needs to switch quickly.

Also check the the self-resonant frequency of the inductor you used is well above the switching frequency (preferably no lower than for the recommended inductor).

15

You guys are introducing me to a new world. All right, comparing the data sheets I found that gate charge for the replacing mosfet is smaller and switching times are lower as well, so that should work out. Rds_on is also smaller for the replacement.

The NCP1450 switches at 180 kHz, my inductor has been tested @100kHz, but nowhere in the datasheet is a self-resonant frequency.

Can I check anything else ?

16

Well measure all the voltages and supply current both before and after activating the chip-enable line.

Your PCB layout seems poor, you have long thin traces where there needs to be short wide traces (to both capacitors).
Short = a couple of mm if possible. There's no point using low ESR and low ESL capacitors and then putting them at
the end of 10mm of thin PCB trace, that defeats the object. A 2-layer board is going to be a lot easier to get working
as the ground return path can be directly underneath, keeping stray inductance right down.

17

Why not try to use an NCP3063 instead? It would be easier. (it's a jelly bean part, very simple, easy to get going)
I just whipped a quick 3.3V to 15V converter using parts in my junk bin... on a breadboard.
3.3V to 5V should be even easier, just change values of 2 resistors.

If your design/wiring/components are right, it should work even though it's built on a breadboard.

When you put it on a proper PCB and layout is optimal, you'll see increases in efficiency. Power Out / Power In
But it should still work even when built on a breadboard.

18

Regarding the layout: I have been going at this from the wrong way. I mentioned earlier I couldn't use the layout in the datasheet, it appears I just didn't understand the simplicity of it: I only need to make a small modification for the different N-channel mosfet to make it meet my demands, it fits most all your recommendations, apart from the fact that it is single sided. But since they designed it, I suppose it will work. I'll post the design as soon as I am done battling the Eagle.

NCP3063... why on earth did I not discover this chip while searching for a good dc booster ? I searched Farnell for quite some time and apparently did not find it, though they DO have it. [time passes while browing farnell and reading data sheets...] It is because the output voltage is rated as 40 Volts, while I was searching for 5 Volts output. With external components you seem to be able to regulate voltage. I do not understand enough of dc boosters to have seen this one coming. Is this normal for booster chips ? Because then I will have to review the entire bunch of them.

What I need for my project is a booster that delivers 5.0 Volts from 3 AA batteries with peak rating for current of max 1 Amps. I prefer it to be (1) easy to build, (2) as cheap as possible, (3) with a preference for DIP parts. The NCP3063 seems to fit this nicely. Is the NCP3063 my best bet then ? If so I will read into the tytpical application circuit, check my supplies and order what is needed, so I can whip one up.

On another note: I already have all the components for the NCP, so I will try to build one, even if the NCP3063 will be the one I would like to use in larger quantities for projects, just for the sake of building it (what else would I do with these components).

19

DC/DC converters are not magic boxes. They're also not 100% efficient. On a good day, expect 80+%, and in some configuration 50% or so.

5V at 1Amp from a AA batteries may be expecting too much. The converter can only produce output power from some percentage of it's input power source. In your case (3) AA batteries (which is how many mAmps?). Don't forget that.

The NCP is very forgiving. Even if your short the output, output voltage clamps down to 1.25Volts or so. Just reset the power (unplug/switch on/off the input voltage source) and your DC/DC converter is back up and running again.

It is because the output voltage is rated as 40 Volts, while I was searching for 5 Volts output.

No, 40V is the max. input voltage for the NCP chip. From there, you can adjust it down or boost it up, or boost/invert it.

20

To maybe quantify and clarify what vasquo said at the beginning. Lets say you want 5V/1A output and your cells are at 4V. That means that your cells will have to supply 5/4 = 1.25A and that would be at 100% efficiency. Well that isn't happening, so at say 80% your cells would have to supply 1.25/.80 = 1.56A to get 1A output. That's going to be tough on battery life so you might consider more cells and a buck converter.