Working with MCP1640 boost converter - high Iq at low load

[tom66]

I just spent about 4 hours trying to get this to work :/ until I noticed the GND pin of the SOT23-6 wasn't very well soldered. 

With poor ground:
 - Vadj > 1.2V (follows output voltage) which is WRONG...
 - Switches infrequently, probably due to bias current flowing through FB pin.
 - Draws about 100uA~200uA.

Anyway, that is now sorted...!   

Things I have learned:
The MCP1640 will not start into heavy load. It will start into about 5.1V/150mA. However, it will not start into 5.1V/300mA. It will do a few cycles then give up with low output voltage, this is a latching condition which will only reset once power is fully cycled (Vin < 0.6V.)  Once running, it will happily accept additional load with no problem. This isn't a problem for my application, but was a little confusing at first. I do need the 300mA, but that can be switched on or off, with a high side P-FET.

It will give 5.1V@300mA down to about 2.8V, which is pretty close to the minimum cut off for a li-ion cell. With no load, it will start from about 0.63V, and work down to 0.5V, giving 5.05V out.

The datasheet shows only DFN-10 package being used for 5V,300mA application; with a 10uH inductor. I'm using SOT23-6 and a 4.7uH inductor. Presumably this is a thermal thing or maybe peak current related. But, I have run it for over 10min with this load, and it only rose 15 deg above ambient. I can touch it with no risk of burnt fingers. It shows no signs of problems with this kind of load.

I intend to run a Noritake Itron GU112X16G-7002 graphical VFD from this, along with a TI ARM (Tiva MCU, TM4C123G) processor. In total, the load shouldn't exceed about 270mA with all running, less if the VFD is dimmed or displaying less bright pixels (though this only varies power usage by about 10%, or 20% if you use the dimming feature.)  It will all run from a single 18650 li-ion cell, 2000mAh known capacity.

I'm unable to get the low standby power the IC advertises :/. I can get down to about 0.3mA at no load at 4.2V in, increasing to around 0.6mA with 3V in. I figure this might be due to my poor layout. It's probably switching too frequently, perhaps due to feedback pin pick up.  This will only give about a year's standby usage. I was trying to target around 50uA, which should be about the self-discharge rate of the li-ion, which would put me closer to 5 or 6 years, instead of the year or so of usage I'm looking at right now.  Does anyone know why the Iq might be so high?

Now, it's only a project for my degree, but I'd still like to engineer it right .

Excuse the photos of the scope - I've misplaced my USB key. First shot is under full load, 5.1V @ 300mA. Second is with no load, aside from feedback resistors.

[Bored@Work]

Crappy soldering, lots of crap remaining on the PCB, and by your own admission a crappy layout. And you are surprised things don't work out as you intend them to be? 

Apply some care and pay some attention. E.g. what is wrong with the recommended layout from the datasheet? What's wrong with cleaning the board from splashed solder and splashed and burned flux?

And what is wrong with improving your soldering skills? Random picks from eBay:

eBay auction: #400674054832
eBay auction: #251449236934
eBay auction: #291092396373

[mikeselectricstuff]

Re. High Iq, there is a gotcha with most of these type of regulators -the current draw on the front page of the datasheet is drawn from the output, so the current at the input is multiplied by the step-up ratio, plus losses.

For single cell always-on type applications, the 1640 is not a good choice. A while ago I spent a lot of time looking for a more suitable device, and ended up with a TI TPS61221 - more expensive but much lower draw - about 15uA at 1.5V in/3.3v out
 

[tom66]

That was after a midnight stint getting this thing working. I've used IPA to clean up the flux now and it looks nice. There are bridges between the adjacent (parallel) capacitors because I took them off and removed them a few times trying to track down the fault and was getting low on wick.

Good news is it ran perfectly fine overnight with full load, so it's looking like the SOT23-6 is fine. Barely warm to the touch.

Re: crappy layout - I think it's OK but it's not as good as possible. Yellow: charging loop, orange: discharging loop, green is the feedback loop.

The worst bit is the feedback, I couldn't figure out a good way to get the ground for the Vfb resistor near the IC. Note, the ground path for the feedback and switching loop is the same, though I do make an attempt to punch through to the ground plane as early as possible.  I could have ripped up the board completely, but I was in a hunger to get prototypes made and test.... maybe a bad idea in hindsight but they say that's 20/20.

That being said, it seems stable under load: output ripple is under 30mVp-p, with just 44uF output capacitance. Switching full pelt at 510kHz with about 50% duty cycle, decreasing as input is increased.

This is also my first time using EAGLE as I use Cadence Allegro at work/internship, and gEDA at home. Since it's a university project I was required to submit *.brd files. Now, I have many swear words for Eagle. I do not like using Eagle one bit...!

[tom66]

Re. High Iq, there is a gotcha with most of these type of regulators -the current draw on the front page of the datasheet is drawn from the output, so the current at the input is multiplied by the step-up ratio, plus losses.

For single cell always-on type applications, the 1640 is not a good choice. A while ago I spent a lot of time looking for a more suitable device, and ended up with a TI TPS61221 - more expensive but much lower draw - about 15uA at 1.5V in/3.3v out

Huh, that's interesting. I see that in the datasheet.  So that probably matches what I'm seeing. I know the IC derives its bias from the output once started, so I guess this explains the high Iq. I guess the converter is not very efficient under light load. The inductor current is completely discontinuous.

It does seem to be operating normally when in light load, pulsing the output. PFM mode. This is expected. I just wish it would PFM... less.

Single li-ion cell should last about 6.7khours, or less than a year in standby. at least the cell is protected.

Maybe if I do it a second time, I would consider allowing the MCU to put the converter into standby. The output would begin ramping down, MCU would reset, then it would put it back into standby. I think I saw a PIC application note on this. If the load current is small, the quiescent current can be reduced massively. The 5V gets regulated immediately to 3.3V for the MCU, by an MCP1703 which was selected for low quiescent current (about 2uA.) I know if I turn off the input supply the 3.3V stays up for a couple of seconds so the pulse rate could be every few seconds if the MCU only interrupts on PBOR.

Unfortunately I couldn't use a TPS61221 as my load current is 270mA. As the input switch limit is around 200mA it's probably going to hit current limit once any significant load starts up. It's actually really hard to find an IC with low Iq -and- high switch current limit. I think I found this part and one from LT. The LT one was about 5x the price and was not stocked by Farnell so that was the deciding factor.

[andyturk]

I've played around with those too. I needed to boost two AA cells up to a 3.3V and 5.0V (using two separate converters) for a project and tried a number of different boost converters. The 1640 works OK and is pretty cheap. As I remember, it couldn't supply a huge amount of current though. The LTC3527 was much nicer to work with, but a lot more expensive.

I just hooked up my 1640 test board to a power supply and measured about 117uA of current when the two 1640's are disabled. Divide by two and call it 60uA for each?

I used the DFN package, which (given what I read on the internet) allows you to get a tighter layout for the "hot loop". I think that mostly helps with Imax. I'm not sure what would cause your Iq to be higher. Maybe crud on the board is throwing off the feedback?

EDIT: OK, so what I measured was Iqshdn, not Iq. With no load and both converters enabled, my test board draws about 140uA. That's about 20 uA more than when the chips are disabled. The interesting thing is that the datasheet says Iqshdn should be 1uA or less, and my board is drawing way more than that. Not sure why. I also noticed that Iq jumps to ~500uA as the input voltage approaches 5V. I haven't put a scope on it, but it's probably oscillating.

[tom66]

That is odd. The MCP1640 should really only draw about 1uA in standby (i.e. EN is low.) Could it be the feedback resistors drawing that quiescent current?  If that is not the feedback circuit, kind of disappointing to be honest. Seems like the specifications were written by marketing, instead of engineering.

Still, it's only a silly second year project, so I'll probably leave it as is - will at least give me something to write about in the report!

[mikeselectricstuff]

That is odd. The MCP1640 should really only draw about 1uA in standby (i.e. EN is low.)

There are a few flavours of the 1640, one option being that it bypasses & connects the output to the input when in shutdown instead of disconnecting the output.

[andyturk]

I used 976k/309k and 976k/562k for feedback resistors. That's maybe 6uA combined when the output is enabled. It should be basically zero if the output is off. The board has a FET inverting the enable signal, and the pull up/down resistors on the FET will account for some of the current, but not all of it.

[tom66]

Hmm, so I wonder where the excessive Iq comes from? Can you see if warming the board up, or cooling it down significantly changes it?

Mike, I'm using the standard MCP1640, no option code. On my board EN is tied permanently to VBAT.

[NiHaoMike]

Why are you running the microcontroller from the output?

[tom66]

The microcontroller is run from a 3.3V regulator, which hangs off the 5V rail. I need the 5V rail for the Noritake VFD and ultrasonic sensor.  The rated input voltage range is 2.8V to 4.25V. (Below 3V, it will lock out with "empty battery" display.) So I need the boost converter to run the MCU for when Vin is less than about 3.4V.

For quiescent tests, the only load was the LDO (rated 2uA, MCP1703.) I haven't yet tried isolating it,  I suppose, but other's results on here to see to confirm issues with the MCP1640 rather than a humble LDO. However, I should try to see if this is a potential cause.

[andyturk]

The microcontroller is run from a 3.3V regulator, which hangs off the 5V rail...

Consider running the mcu from the battery (maybe a 3.0V instead of 3.3V). That'll save a lot of current through your boost controller. If your mcu needs 100mA and your boost is only 75% efficient, then you'll reduce the input current to the boost controller by 133mA.

[tom66]

Run mode current isn't particularly critical - it's the standby mode that's important.

Over 85% of power is used by VFD. It has resistive heating elements.

With the clock I am using, MCU will need about 30mA.

I also have the benefit of a high accuracy 3.3V rail for use as ADC reference. It's rated within 1%, fine for measuring Vbatt.

[mikeselectricstuff]

Run mode current isn't particularly critical - it's the standby mode that's important.

Over 85% of power is used by VFD. It has resistive heating elements.

With the clock I am using, MCU will need about 30mA.

I also have the benefit of a high accuracy 3.3V rail for use as ADC reference. It's rated within 1%, fine for measuring Vbatt.

In which case run the MCU from the battery so you can shut down the boost reg when off.
For measuring Vbat, use Vbat as the reference and , measure the regulated 3.3v

[tom66]

So how do you run the 3.3V MCU off a 3~4.2V battery and have a guaranteed ref voltage? A second regulator? That seems excessively complex. I could use, e.g. 2.7V. But I'm already pushing it with my VFD, which is rated for 4V (or more) logic inputs. I've had no trouble giving it 3.3V, wouldn't be sure if it would work much lower...

I could use the MCP1640 with input bypass and turn it off, but ultimately my initial hope was that the MCP1640 would have a low enough quiescent current to run continuously. This doesn't appear to be the case.  It also doesn't seem to matter as andyturk has shown it draws huge Iq in standby anyway :/

[mikeselectricstuff]

So how do you run the 3.3V MCU off a 3~4.2V battery and have a guaranteed ref voltage? A second regulator? That seems excessively complex. I could use, e.g. 2.7V. But I'm already pushing it with my VFD, which is rated for 4V (or more) logic inputs. I've had no trouble giving it 3.3V, wouldn't be sure if it would work much lower...

I could use the MCP1640 with input bypass and turn it off, but ultimately my initial hope was that the MCP1640 would have a low enough quiescent current to run continuously. This doesn't appear to be the case.  It also doesn't seem to matter as andyturk has shown it draws huge Iq in standby anyway :/

There are a few options - one would be to use a MCU that can run over the whole battery voltage range. Maybe from the 1640 output which is put in bypass mdoe when in standby.

[tom66]

There are very few ARM processors that can work from 5V and the only moderately powerful alternative would be an old school dsPIC30F. I need about 4K RAM for display buffer and datalogging, plus fast processing to update the VFD and make it silky smooth.

I'm using the TIVA because that's what my lecturer wanted - normally I'd use an STM32 as I have a lot more experience.

Once it's all working,  I'll post a demo video.