Electronics > Projects, Designs, and Technical Stuff
Low power isolated DC/DC converter
jaromir:
For those not interested in reading wall of broken english text, TLDR version:
Please suggest isolated low power +-15V DC/DC converter from 6-10V battery supply. LT8300 or LT3001 may work, while it's being delivered I made experiments with DIY DC/DC converter with mixed results.
-----
I'm thinking of a project where I'd need isolated analog portions, each one with isolated, floating +-15V supply. Power consumption is say 10mA per portion, that is 2x10mAx15V=300mW, let's say under 0,5W. To make things complicated, it needs to be powered of batteries, probably 6xAAA or 6xAA cell, that implies working range 6-10V.
At first I thought like this can be done using off-the-shelf DC/DC converters, but somehow I can't find good candidate. First problem is input range. Majority of the "black bricks" is designed for stable input voltage, say 5V+-10%. I can add switching preregulator to bring 6-10V input down to 5V for DC/DC converters, but having two DC/DC converters in series looks like unwieldy solution. What is worse, most of the "black bricks" do have quite bit of current consumption. No load consumption is usually 100-200mW (often not specified) or thereabouts and full load (1W) efficiency 70% (getting worse at my 300mW conditions), together with switching 5V preregulator that would give horrible total efficiency.
There are types with wider input range, but with somehow worse efficiency, not much of improvement over solution with preregulator.
Fortunately, there are ICs designed for low power DC/DC converters, like LT8300 or LT3001. Though not exactly cheap, those seem to be designed for low power supplies, so I ordered a few pieces. While mail delivery does its job, I thought like "let's see if I can do it better", at least for education purposes. It turns out that... not yet and probably I need some consultation.
I opted for symmetrical push-pull converter like this https://imgur.com/a/lLrmKPx and thought of having LC low pass filter and perhaps voltage regulator after the rectifier.
First things first, I took two cores https://www.tme.eu/sk/en/details/efd15_8_5/ferrite-cores/feryster/ and bobbin https://www.tme.eu/sk/en/details/efd15-k-8p-smd/coil-formers-and-accessories/feryster/ and made two bifilar windings with 0,18mm diameter enameled wire. First (primary) is 2x40 turns, seconday is 2x80 turns. I know the transformation ratio is not enough for 6V->15V conversion, but the bobbin was already full and it's definitely good enough for tests. For primary, I connected end of one winding to start of another, creating longer winding with tap in center; the same for secondary. To check whether the polarity is OK, I grounded both primary and secondary taps and applied 80kHz signal to one end of primary, observing voltages on other winding ends. See attachment 1_check.png Yellow trace is excitation signal, blue is another end of primary - having opposite polarity, I guess that is correct. Purple and dark blue are ends of secondary, both being in opposite phase to each other and having correct transformation ratio.
I breadboarded the push-pull converter with two BC337-40 transistors, fed with 8V from lab power supply. I inserted 4R7 resistors into transistors emitters to observe current waveforms and experimented with diodes from transistor collectors to ground. Bases of transistors were fed from dual channel arbitrary waveform generator, set to 80kHz square wave, 5V, adjusting duty cycle.
The good news is that the converter basically works. When fed with 20% duty cycle, for 8V input I have roughly 15,4V after rectifier with 10uF smoothing cap and 4k7 load resistor. With 10V input, output is above 19V; with 6V, something around 11V. So far so good. That being said, there are aspect of this simple circuit that escape my understanding.
Let's look at waveforms, see attachment 2_transistor1.png Yellow trace is excitation of one transistor (blue is another transistor), purple is voltage on emitter resistor and dark blue is its collector voltage. The messy waveforms indicate suboptimal probing, but perhaps it's obvious that voltage on emitter resistor continued flowing after base voltage fell down. Is that expected? I though emitter current would rise only until excitation impulse goes off.
I varied the duty cycle, hoping to vary the output voltage, but the voltage remained more-less constant in range 5-30% of duty cycle, falling down quickly under 5%. I expected somehow stronger output voltage dependency on duty cycle. Above 30% of duty cycle I got increased power consumption, even with slight or no secondary load. I'm aware this kind of converter does need some dead-time, but this seems too high.
Last point is efficiency. I measured 65% at most; that is not exactly convincing.
How can I achieve better regulation range? Is transformer such as this suitable for this job? Can I achieve better efficiency?
T3sl4co1l:
How isolated does it really need to be, if it's battery powered to begin with? (Are you using the battery power, directly or with a common-ground converter, for something else? Is it because the battery can be substituted with a possibly-common-ground supply?)
First off I'd suggest an ordinary flyback supply, and this can be made with off-the-shelf coupled inductors (1:1 or 1:2 ratio). Use two inductors, primaries wired in parallel, secondaries independent (a diode and cap, each), then the DC outputs wired in series to get the +/-15V.
Any flyback controller or regulator will do, just make sure it's current mode control, of some sort or another. Usually peak current with slope compensation. If they don't say this in the description, the block diagram will show a current sense signal and a comparator, and a ramp output from the oscillator section being added into the current sense signal.
The tricky part is feedback. You could do primary side feedback, but regulation stinks (limited by leakage inductance, which is not usually terrific; a lot of coupled inductors are kind of intentionally bad, it helps for some applications but hurts this one). Secondary side feedback (usually with a TL431 + optocoupler) is very accurate, but kind of wasteful (TL431 needs >= 1mA for accurate regulation, and optos are usually driven at 5-20mA for better... I don't really know, speed I guess?).
What's the lowest current consumption you will have? Should the supply have an enable signal, so it can be turned off when not needed? If it's always on, and the load can be light (like <1mA), the bias current drawn by the feedback circuit will be relatively substantial! So it would be nice to have a way around that.
Maybe just using a better error amplifier (AP, ZR, TLV and other -431 prefixes?), and if possible, a more compact opto that still performs well at low currents.
If you don't need low output ripple, you may also consider a hysteretic control; this could use very little idle consumption (a suitable comparator might draw 10s uA), and could also be wired so that the opto is only turned on when load current is demanded. (This makes a bit of a chicken-egg problem at startup (how can the secondary side command more current, if it doesn't yet have any supply voltage to work with?), but that can be worked around.)
Regarding push-pull: a self-oscillating circuit incurs full idle current due to the controller, drivers, and core loss. It could of course be triggered with hysteresis to give a mostly-stable output while consuming less power under light load. The circuit tends to be more complicated, I think, for the same feature and power level, so I would suggest a more conventional flyback.
BTW, what's wrong with off-the-shelf modules?
BTW, one thing that is wrong with them, or a lot of them in this range (<= 1W converters): they often do use a push-pull circuit, self oscillating with two transistors and some bias and signal coupling parts, and diodes for the output. These are not regulated, and typically have a tighter supply voltage range (say 8-16V, versus say 5-30V for a flyback). So as your batteries droop towards EOL, so will the +/-15, which I'm guessing is undesirable.
Note that such a circuit (like the one linked) is a charge pump, transformer coupled -- not a switched inductive converter. This is why the output voltage follows the input, and this is what you observed. What's missing? An output filter inductor, between the rectifier and capacitor, to make a forward converter.
Forward converters are perfectly fine as well, and controllers like UCC3808 come to mind, or some of LT's fanciest offerings (the slew-rate-controlled regulators come to mind), or integrated with some of AD's signal isolators (ADuMxxxx family). And probably a heck of a lot of other things, since it's been closer to a decade since I last looked over these product lines, hah.
You can also make a one- or two-transistor forward converter, but you still have the primary downside of the topology which is needing a transformer and a filter choke. Flyback combines them in one, basically, at the expense of higher input and output ripple; but that's easily tolerable at this power level, so it's a common choice.
Tim
jaromir:
Thank you. This is really detailed and valuable post, as is vast majority of your posts. :-+
Because of length and number of inputs in your post I snipped a few bits of it and tried to response to main points below:
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---How isolated does it really need to be, if it's battery powered to begin with? (Are you using the battery power, directly or with a common-ground converter, for something else? Is it because the battery can be substituted with a possibly-common-ground supply?)
--- End quote ---
I need to have two separate voltages isolated from each other. The battery referenced part will have exposed USB port and external power jack, so I concluded it may be good idea to have battery "earthed" and the two circuit portions "floating". Your deduction is correct, I should be more explicit in my post.
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---First off I'd suggest an ordinary flyback supply, and this can be made with off-the-shelf coupled inductors (1:1 or 1:2 ratio).
<SNIPPED POST>
--- End quote ---
Thanks for suggestion, I'll try flyback.
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---Secondary side feedback (usually with a TL431 + optocoupler) is very accurate, but kind of wasteful (TL431 needs >= 1mA for accurate regulation, and optos are usually driven at 5-20mA for better... I don't really know, speed I guess?).
--- End quote ---
I found relatively low power optoisolators, like SFH618 being specified at 1mA, so it may be good candidate for secondary side regulation.
Also, TLVH431 are 431-like devices with much lower specs.
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---What's the lowest current consumption you will have? Should the supply have an enable signal, so it can be turned off when not needed? If it's always on, and the load can be light (like <1mA), the bias current drawn by the feedback circuit will be relatively substantial! So it would be nice to have a way around that.
--- End quote ---
Enable signal is not needed. Device will have clicky-clacky mechanical ON-OFF switch.
Current consumption will change depending on what the analog floating portion will do, but will not get under something like 5mA and not above 10mA.
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---<SNIPPED POST>
If you don't need low output ripple, you may also consider a hysteretic control
--- End quote ---
Thanks. It will power analog circuitry, but LC filtering after DC/DC will be needed anyway, perhaps even linear regulator, so hysteretic control may do the job.
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---<SNIPPED POST>
These are not regulated, and typically have a tighter supply voltage range (say 8-16V, versus say 5-30V for a flyback). So as your batteries droop towards EOL, so will the +/-15, which I'm guessing is undesirable.
--- End quote ---
Yes, this is what's wrong with them. In my first post I thought of workarounding this with switched buck pre-regulator to step down voltage to 5V, for example, so that off-the-shelf DC/DC modules do have stable input voltage. This is somehow unwieldy, though. What more, efficiency will suffer - quiescent power consumption of typical DC/DC 1W bricks is in the same order of magnitude as my circuit will consume, preregulator will knock it down a bit more.
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---Note that such a circuit (like the one linked) is a charge pump, transformer coupled -- not a switched inductive converter. This is why the output voltage follows the input, and this is what you observed. What's missing? An output filter inductor, between the rectifier and capacitor, to make a forward converter.
--- End quote ---
Aha, now I see it!
When searching for forward converter I usually found flyback with reset winding+diode, like this https://www.maximintegrated.com/en/images/appnotes/3983/3983Fig01a.gif and I disregarded the secondary part :palm:
What I'm really missed is the choke, this is very good point.
--- Quote from: T3sl4co1l on September 16, 2019, 01:54:30 pm ---<SNIPPED POST>
You can also make a one- or two-transistor forward converter, but you still have the primary downside of the topology which is needing a transformer and a filter choke. Flyback combines them in one, basically, at the expense of higher input and output ripple; but that's easily tolerable at this power level, so it's a common choice.
--- End quote ---
OK, so my main takeaway from this is:
1, employ choke, as per picture https://www.maximintegrated.com/en/images/appnotes/3983/3983Fig01a.gif
2, try flyback
I'll try both and will let you know.
Zero999:
Those ready made isolated DC:DC converters are fine, as long as you get ones with a wide supply voltage range and don't cheap out. A filter might be needed to git rid of any high frequency ripple, but that about it. Here's a link to the datasheet for the range I've used before.
https://docs-emea.rs-online.com/webdocs/16ed/0900766b816ed87a.pdf
jaromir:
Thanks for the tip, Traco is my long time favorite when it comes to DC/DC bricks.
The TRN 1-0523 looks fine, but its no-load current consumption of 35mA means power consumption roughly 200mW, while I expect it to deliver something about 150-300mW of output power. Not great, not terrible. I have a few of TMA0515D in my junkbox, with no-load consumption of 30mA. This, with switched pre-regulator could do the job, too. TRN has better load variation regulation.
I consider both TRN and TMA+regulator as acceptable, but first I'd like to explore other ways.
Navigation
[0] Message Index
[#] Next page
Go to full version