boost-converter trouble

Ah yes, the sausage is leaking out...

(Ever more, these days, designs are -- I assume -- driven by digital engineers, following specs written by other digital engineers, few if any of whom have a nuanced, general understanding of control theory or DSP. It's all sausage inside. You don't want to see it getting made.

Such a device is presented in such-and-such way, following a block diagram, and seemingly emulating an analog chip (which would do the same job with fewer transistors, but higher development costs). But because of all these compromises, there are innumerable hidden states, which if the end user stays within the spec's expected operating envelope, never show up, but when you have situations like this, well... the ugliness spills out.)

It bothers me that, e.g., their "Eco-Boost" (some TPS54xxx's and such) controllers are reasonably nice, yet, when they go to burst mode, they don't slow down (the analog version), or pulse skip (the simple digital version), no, they have an internal oscillator and counter keeping state, and a mux to select a different tap from the [binary] divider, so the frequency only goes down in halves. But no one noticed that that's a horrible nonlinearity in an otherwise reasonable control loop, so, that's what made it into production.

Tim

-- Seven Transistor Labs, LLC Electrical Engineering Consultation and Design Website:

So? The soft start is only 512us. Cyclically disable the chip as originally planned but then delay the LED current pulse by 1ms after the 61040 is EN'ed.

Yeah, and the datasheets lie through their teeth, especially about forced-PWM vs stutter mode at low output current. I don't ever use new switcher parts without prototyping them.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Is there a uP? Make your own software controlled boost converter maybe. Mosfet, inductor, diode.

I've done converters using a schmitt-trigger gate as the control element. Sort of fun.

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

That would work if the TPS61040 had an oscillator. Instead it only makes cycles as needed, and without a load, 512 cycles aren't needed. One software solution is to run it at full load for 1 to 2ms before using it. But that wastes much more power, 16mW, than simply leaving the converter on all the time, 0.1 mW.

--
 Thanks, 
    - Win

This one passed multiple prototyping tests. Then a new preferred software scheme uncovered the issue.

--
 Thanks, 
    - Win

Someone has defined history as "one damned thing after another." ;)

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

I should hope so.

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Known as a "hysteretic-mode converter". They're fairly common these days, usually called something like "D-Cap" control.

You can make a schmitt fire in fixed-width pulses, which becomes a frequency control loop, not hysteretic. Each shot dumps a pretty much equal bump of energy into the load.

Hysteretic works fairly well in a buck switcher, less well (or not at all) in boost mode.

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

Okay, I see now. I don't like this TPS61040, they're using a really inferio r methodology for soft start just so they can include it in their highlight ed marketing bullets. The better engineered parts terminate soft start when the output reaches full voltage, along with output voltage status outputs, the TPS kluge does not, it just hangs in soft start mode until whenever.

As for better engineered parts with approximately the same meta-scale param etrics but better control, like internal oscillators, you have the Microchi p MCP1650/51/52/53,

pdf with cost about 25% more in quantities of 1000, but still purrretty che ap.

Then there is this premium performance Austria Microsystems part, AS1340 an d AS1344,

Price is not so bad. The only question is who are Austria Microsystems.

Most of these parts have soft start current limits way bigger than the TPS, even though the output current rating is about the same.

The way you're using the TPS now, which is effectively bypassing soft start , may run into problems with incomplete utilization of the full battery cap acity. The partially discharged battery always has issues with output imped ance and time response to surges, which may trip your UVLO and cause a rest art, all this happening just when you start powering your LEDs, and you're back to square one. Just something else to be tested...

I take it you're driving 4x white LEDs in series.

Yes, really bad.

Thanks for the suggested parts. Unfortunately I'm stuck with only a small amount of space, and therefore only considered sot-23 package parts. I had gathered a huge list of candidate datasheets in my computer, but it's the little unknown gotchas that lurk behind one's design choices that rise up and bite. There may be a great sot-23 alternate, but after 3 pcb passes, 15 boards assembled, and a season of live beehive experience, I had hoped to be done with the hardware optimization phase. It's working OK now, at the expense of 10 to 20% wasted power.

Actually, there are eight IR proximity-sensor LEDs in series. There are 48 proximity sensors for 24 bee-trip channels, and we scan all of the channels 20 times a second. Sensing is with 96 measured (on and off) analog ADC reflection levels. It works well even with sunlight shining into the channels.

--
 Thanks, 
    - Win

Uncovered.....or created?

Preferred because a lower power consumption was theoretically possible?

RL

What is the cheapest uP that has an ADC input? One of those and a three cent mosfet would make a nice boost switcher; just needs a little code. 30 cents for the uP and fet and a diode maybe. Serial interface is probably free.

--

John Larkin   Highland Technology, Inc   trk 

jlarkin att highlandtechnology dott com 
http://www.highlandtechnology.com

I really like the LPC802, at 50 cents in reels.

Cheers

Phil Hobbs

--
Dr Philip C D Hobbs 
Principal Consultant 
ElectroOptical Innovations LLC / Hobbs ElectroOptics 
Optics, Electro-optics, Photonics, Analog Electronics 
Briarcliff Manor NY 10510 

http://electrooptical.net 
http://hobbs-eo.com

Guessing, cheapest with >200kSps isn't going to be as competitive as you're thinking.

That's basically what the stupid phaux-analog controllers are anyways.

Tim

--
Seven Transistor Labs, LLC 
Electrical Engineering Consultation and Design 
Website: https://www.seventransistorlabs.com/

Silabs Busybee EFM8BB1, comes in at below 20 cents, and has 800kSa ADC, plus deadtime enabled timer, 2% voltage reference

Cheers

Klaus