Boost converter with abnormally high current draw?

[GarthyD]

I am using a boost converter (MCP1640) to convert a regulated 3.3V up to a higher voltage (5V in my current testing, but 4V ultimately). I am seeing a current draw of 150mA-350mA (depending on conditions) *before* adding any load. Going from the datasheet, unless I am mistaken, this should be under 1mA.

Link for the MCP1640: http://www.microchip.com/wwwproducts/en/en547080

I am wondering why the current draw is so high. I am not particularly familiar with boost converters and inductors, so I don't have a lot of experience to draw on here.

If I add a small load (eg. a single LED), the current increases roughly as expected. The amount is generally a sensible ratio of the current you'd expect to flow through a LED given the voltage difference, eg. maybe 10mA or so depending on specifics.

This is a MCP1640B (note: "B"), which has an enable pin that when driven low should pass the input straight to the output. Even when I do this, the current remains around 150mA.

I don't have the recommended capacitors from the datasheet. I realise that this may be the cause of the problem. The datasheet recommends a 4.7uF ceramic for the input and 10uF ceramic cap for the output. Unfortunately I don't have any ceramic caps above 1uF on hand (plenty of electrolytics though). I've experimented with combinations of the ceramics and electrolytics that I have instead. The results are fairly consistent.

I have swapped the chip out with another (same batch), and it appears to behave the same way.

I've dropped the part of the circuit that fed into the converter and used an AA battery as input. I see similar results.

I've heard of problems with the MCP1640 before when connected with long wires. Whilst a breadboard-based design, I've still managed to pack all of the components into a 20mmx15mm area. The chip itself is on a small breakout board.

I have experimented with different inductors. When using a 4.7uH Murata 1100R, I generally saw around 300mA current draw. A 4.7uH Wurth WE-ZB caused about 180mA. A 10uH "ebay special" (bought cheap off of ebay, I don't know exactly what it is) used anywhere from 100mA-300mA. I lit a single LED with each, and the first two inductors did so reliably. Note that these inductors are *not* in the recommended list in the datasheet, they're just what I had available.

The generated voltage was close with each inductor. I used resistors that should generate 4.8V (330k bottom, 1M top), and saw voltages of 4.95V-4.99V being produced. So the voltage is roughly right.

I've checked the wiring repeatedly and cannot find a fault. The setup fairly closely follows the example given on page two of the datasheet, but uses 330k and 1M resistors instead on the feedback pin.

I'll probably be able to determine more once I have the recommended parts, but that may be some time away.

I am wondering if anyone familiar with boost converters and inductors is able to offer insight into what may be happening? Is this sort of current draw normal? Are there any critical elements to working with boost converters that I am not aware of? How suitable are the inductors I ended up trying? And so forth.

[amitchell]

Sounds like you have a short somewhere, what is getting hot?

You are going to have a hard time breadboarding a SMPS.

I would suggest that you play around with one of the evaluation modules found here under Development tools: http://www.microchip.com/wwwproducts/en/en547080

[GarthyD]

Sounds like you have a short somewhere, what is getting hot?

I'd wondered that earlier with the first IC, and tried it out, not noticing anything at the time.

However I gave it another shot just then for a bit longer with the second IC. The IC itself becomes uncomfortably hot (although you have to press it firmly to tell). The inductor also seemed a little warm, but it was hard to tell. Nothing else stood out, but the parts are packed in fairly closely so I might have missed something.

I've swapped back to the first one, and the problem is the same.

One thing I noticed is that when tested separately, the first IC appears to only have 100 ohm resistance between the inductor pin (SW) and GND, and seemed to vary based on time. The second did not. So something funny is going on here. Apart from that, I don't think there are any shorts, at least nothing that I noticed.

You are going to have a hard time breadboarding a SMPS.

I had wondered if that might be the case. Are they (switched mode power supplies, boost converters) generally very picky re layout?

I do have another angle I can approach things from if the boost converter path is likely to be difficult. It's just a different set of tradeoffs.

I would suggest that you play around with one of the evaluation modules found here under Development tools: http://www.microchip.com/wwwproducts/en/en547080

Cheers for the suggestion and link.

Thanks for having a look at things, much appreciated.

[rx8pilot]

SMPS's are definitely very layout sensitive. Getting them to work well with PCB's is no small task. Getting one to work reasonably well on a breadboard would be unusual.

Sent from my horrible mobile....

[bktemp]

MCP1640 works at low voltages and a rather high frequency. I wouldn't use it on any pcb with less than 2 layers. (On a 2 layer pcb is works fine if the layout is good).
The most critical path is the connection between OUT pin and the capacitor and its connection back to the GND pin. It should be very close (<5mm) to the ic. If there is a large distance between ic an capacitor there will be voltage spikes, possibly damaging the ic.
The input capacitor should also be as close as possible to the ic.
Since it is a synchronous boost converter, the switching node is less critical.

[tautech]

I've used the 1640 for a couple of projects and layout is critical, follow what it specifies in the datasheet as closely as possible.
There's NO WAY electrolytics can do the job here @ 500KHz, you must use MLCC.
Furthermore use of other than the inductors listed in the datasheet will give only shit results.

[GarthyD]

SMPS's are definitely very layout sensitive. Getting them to work well with PCB's is no small task. Getting one to work reasonably well on a breadboard would be unusual.

Thanks. I've since been digging around for further information and everything I've found concurs with what you and amitchell say here. After the simplicity of LDOs I had assumed boost converters would be easy enough to work with. I think this assumption was deeply flawed!

MCP1640 works at low voltages and a rather high frequency. I wouldn't use it on any pcb with less than 2 layers. (On a 2 layer pcb is works fine if the layout is good).
The most critical path is the connection between OUT pin and the capacitor and its connection back to the GND pin. It should be very close (<5mm) to the ic. If there is a large distance between ic an capacitor there will be voltage spikes, possibly damaging the ic.
The input capacitor should also be as close as possible to the ic.
Since it is a synchronous boost converter, the switching node is less critical.

Thankyou for all of that detail, it really helps me put things into context.

When breadboarding the caps are probably just on the wrong side of that 5mm limit, even when using the closest slot to the IC.

...

At this point I think I'd best fall back to plan B. :} Or at the very least, if I continue to experiment with the 1640, use a fresh one and solder everything right onto the pins via very short wires.

Many thanks for all of the input on this one. I had not expected to run into the difficulties that I did, and hadn't realised how little leeway there is when working with boost converters.

[GarthyD]

I've used the 1640 for a couple of projects and layout is critical, follow what it specifies in the datasheet as closely as possible.

Thanks, this is really good to know. I'm not only new to 1640s, but boost converters in general.

There's NO WAY electrolytics can do the job here @ 500KHz, you must use MLCC.

I'd suspected as much. The best I could manage with what I had was parallel 1uF ceramic caps plus electrolytics for the rest.

Furthermore use of other than the inductors listed in the datasheet will give only shit results.

Thanks. It's good to know that things are that sensitive- I was unaware. I am used to there being a bit more flexibility- as in I can sometimes get things going in a prototype with less-than-ideal parts, put in an order for the proper parts, and add them in later. It looks like I'm not getting away with this approach this time.

[tautech]

Furthermore use of other than the inductors listed in the datasheet will give only shit results.

Thanks. It's good to know that things are that sensitive- I was unaware. I am used to there being a bit more flexibility- as in I can sometimes get things going in a prototype with less-than-ideal parts, put in an order for the proper parts, and add them in later. It looks like I'm not getting away with this approach this time.

I asked Coilcraft for samples and they kindly obliged. 
Looking at one, it's marked as 472D

[tautech]

Furthermore use of other than the inductors listed in the datasheet will give only shit results.

Thanks. It's good to know that things are that sensitive- I was unaware. I am used to there being a bit more flexibility- as in I can sometimes get things going in a prototype with less-than-ideal parts, put in an order for the proper parts, and add them in later. It looks like I'm not getting away with this approach this time.

I asked Coilcraft for samples and they kindly obliged. 
Looking at one, it's marked as 472D

I should add, even though I followed the datasheet as close a possible (SS PCB) I struggled getting ripple down to acceptable levels even with light loads and when a 70ms 20mA draw (part of this design) was engaged things got a little crazy. In the end I settled for some additional SMD Tantalum bulk capacitance on the output rail and then things behaved as it should.

[bktemp]

Other inductors than those listed in the datasheet also work, but it is often very difficult to estimate the losses in the inductor at the operating frequency + waveform, because most manufacturers don't give you all the necessary data for calculating the losses.Therefore you often have to try several inductors to find the best ones that fit your application.
Using the suggested inductors is a good starting point, but they are often rather expensive (for example the WE ones are good but expensive compared to other inductors).
My preferred inductor for MCP1640 is ELL-VGG4R7N, because it is small and cheap and works reasonably well.

The suggested layout in the datasheet works well, it can be done even using 0805 components for all resistors and capacitors.
Depending on the cable length to the input power source, you maybe need additional capacitance to the suggested values in the datasheet.
As tautech suggested, adding a couple of 10uF of tantalum or other low ESR electrolytic capacitor to the output often helps, but don't go too high (stay at <100uF), because too much capacitance at the output makes start up for the boost converter difficult.

[GarthyD]

Thankyou yet again for all of this information. When I have the right parts and the opportunity this gives me a very good chance of successfully putting something together. This advice has saved me a lot of wasted time trying to get something working with inadequate parts and layout. Very much appreciated. Thankyou.

[amitchell]

Just a few tips if you struggle with a boost converters output ripple.

Populate the output with a CLC filter, between 500Khz and 2Mhz I have used a 10uF 10uH 10uF combo with a high frequency bypass cap around .1uF 100V right at the output diode. Notice the tight area(switch node) incompassing the inductor L402, switch pin, and diode D1. Also the loop with the diode and HF bypass cap C406 (0402 for scale).

A few small cap footprints at the other side of the output filter can be used to remove high frequency ringing/noise if needed, its free to add footprints for them  Sometimes an input CLC filter helps cut the noise coming back into the input side of the circuit.

Here is some good reading.

https://e2e.ti.com/support/power_management/simple_switcher/w/simple_switcher_wiki/2243.understanding-measuring-and-reducing-output-voltage-ripple

http://www.kemet.com/Lists/TechnicalArticles/Attachments/5/Avnet2012PowerForum_CapacitorsSelection.pdf


Screen grab of my latest boost led driver showing the top and third layer. Second and Fourth layers are full ground planes.