Electronics > Projects, Designs, and Technical Stuff
Choosing an inductor for a boost converter
aiq25:
--- Quote from: MagicSmoker on September 16, 2019, 09:58:03 pm ---
--- Quote from: splin on September 16, 2019, 08:16:44 pm ---Assuming you mean the TPS61236P (google can't find a TPS1236P)...
--- End quote ---
Hmm... OP definitely should have mentioned the part number and, ideally, that it wasn't a bog-standard switcher IC to give us some forewarning to check the datasheet ourselves; mentioning that sort of critical detail on post 8 is poor form, but sadly all too common around here.
--- End quote ---
I should’ve mentioned it, next time I’ll give more detailed. It’s just that I was asking this as a general question, applicable to all boost applications. I’m really not locked into a particular IC yet. I only found boost regulators from TI and Monolithic Power Systems that should work for my application, this is searching through Digi-Key. I’ll search again, thinking about Analog or Linear should have something.
MagicSmoker:
Right, well the trend among switcher ICs is definitely towards higher and higher switching frequencies and while that can minimize size, it often doesn't because AC losses in windings and core materials increase exponentially with frequency. For example, skin depth in copper at 1MHz is 65um, which is the same as the thickness of 1oz 2oz Cu plating on a PCB and limits the wire diameter to 35 AWG or smaller (if a significant fraction of the current is AC, as in a DCM choke or a transformer). The specific core material used is a lot more important at 1MHz, too - it has to be one of the advanced ferrites such as Ferroxcube 3F4 or TDK/EPCOS PC200, and even then you'll need to limit AC flux swing to 50mT or so to keep losses under control.
And I'm not even mentioning all the knock-on effects you'll have to deal with as a result of even tiny stray Ls and Cs being meaningful at this frequency.
So if you have a choice and you aren't too experienced at smps design - the latter seems to be the case based on the questions you are asking (hey, we all started off at zero knowledge) - then I'd aim for a switching frequency in the range of 100kHz to 400kHz, especially if you want to stick to using off-the-shelf inductors where you might not know the specific ferrite used, and which are invariably wound with a single wire or flat ribbon so there is no accommodation for skin effect.
That said, the IC you had tentatively chosen doesn't require you to design the frequency compensation and feedback network, which I know can appear to be an overwhelming advantage to the neophyte smps designer; resist the temptation! I'm not a huge fan of TI's switcher ICs these days - they seem to be shitting out new TPS part numbers like a refugee with cholera - but an alternate part to consider that is altogether more flexible is TPS61089/610891. It's about $1 more but has adjustable frequency of 200kHz-2.2MHz, an adjustable switch current limit with a guaranteed minimum value of 7.3A, and a minimum input voltage of 2.7V.
Of course, you could also use a much more common controller IC and an external switch, but that might not cost any less overall and be harder to get working so perhaps not the best choice, especially for those new to smps design.
Oh, and a handy site I always recommend is: http://schmidt-walter-schaltnetzteile.de/smps_e/smps_e.html
EDIT - removed unused asterisk; fixed plating weight vs. skin depth
aiq25:
--- Quote from: MagicSmoker on September 17, 2019, 11:20:27 am ---Right, well the trend among switcher ICs is definitely towards higher and higher switching frequencies and while that can minimize size, it often doesn't because AC losses in windings and core materials increase exponentially with frequency. For example, skin depth in copper at 1MHz is 65um, which is the same as the thickness of 1oz Cu plating on a PCB and limits the wire diameter to 35 AWG or smaller (if a significant fraction of the current is AC, as in a DCM choke or a transformer). The specific core material used is a lot more important at 1MHz, too - it has to be one of the advanced ferrites such as Ferroxcube 3F4 or TDK/EPCOS PC200, and even then you'll need to limit AC flux swing to 50mT or so to keep losses under control.
And I'm not even mentioning all the knock-on effects you'll have to deal with as a result of even tiny stray Ls and Cs being meaningful at this frequency.
So if you have a choice and you aren't too experienced at smps design - the latter seems to be the case based on the questions you are asking (hey, we all started off at zero knowledge) - then I'd aim for a switching frequency in the range of 100kHz to 400kHz, especially if you want to stick to using off-the-shelf inductors where you might not know the specific ferrite used, and which are invariably wound with a single wire or flat ribbon so there is no accommodation for skin effect.
--- End quote ---
I have done one SMPS design before (using Push-Pull topology) but all of the magnetic components were designed by someone else, so this is a good learning experience for me. I was aware of the issues with AC losses at the higher switching frequencies but you're right, I don't have all the knowledge yet.
--- Quote ---That said, the IC you had tentatively chosen doesn't require you to design the frequency compensation and feedback network, which I know can appear to be an overwhelming advantage to the neophyte smps designer; resist the temptation! I'm not a huge fan of TI's switcher ICs these days - they seem to be shitting out new TPS part numbers like a refugee with cholera - but an alternate part to consider that is altogether more flexible is TPS61089/610891. It's about $1 more but has adjustable frequency of 200kHz-2.2MHz, an adjustable switch current limit with a guaranteed minimum value of 7.3A, and a minimum input voltage of 2.7V.
--- End quote ---
I have looked at this IC and was going to use it for another application. This design is my first where I'm trying to design everything myself, so I was looking to use something easy but that doesn't seem to be the case anymore. lol
--- Quote ---Of course, you could also use a much more common controller IC and an external switch, but that might not cost any less overall and be harder to get working so perhaps not the best choice, especially for those new to smps design.
--- End quote ---
Cost is not a major factor for me, at this point. Efficiency is more of a concern and the TPS61236P datasheet shows very high efficiencies, one of the reason I chose it.
--- Quote ---Oh, and a handy site I always recommend is: http://schmidt-walter-schaltnetzteile.de/smps_e/smps_e.html
--- End quote ---
Thanks, this is a great website!
aiq25:
--- Quote from: splin on September 16, 2019, 08:16:44 pm ---
Assuming you mean the TPS61236P (google can't find a TPS1236P), you could take a radical approach to the problem - such as reading the datasheet:
--- Quote ---9.2.1.2.3.1 Inductor Selection
Because a 1-μH inductor normally has a higher current rating and smaller form factor than inductors of higher values, the TPS6123x is optimized for 1-μH inductor operation. Inductors of other values may cause control loop instability and so are not recommended.
--- End quote ---
--- End quote ---
Thanks, I did not catch this.
--- Quote ---Looks like a good reason not to use a 3.3uH inductor. I used the coilcraft DC-DC inductor selection tool I mentioned earlier:
https://www.coilcraft.com/apps/power_tools/dc-dc
The TPS61236P switches at 1MHz so I entered that and other parameters as shown in the attached. I limited the ripple current to 30% to get the tool to find inductors around 1uH, and used the filter to limit the size to 11mm x 11mm. The attached pdf contains the results.
The cheapest inductor @0.32 is the SD43-102, 29.7mohms, 4.7 x 4.2mm, 549mW total losses of which the AC losses are 34mW. 549mW is 4.5% of the O/P.
The MSS1038-102 @$0.62 is 5.4mohms, 10.5 x 10.2mm, 137mW losses of which 49mW are AC.
All the inductors shown in the coilcraft results have DC losses at least double the DC losses; I guess most similar ferrite inductors won't be very different so I would use DC resistance as the primary selection parameter. Obviously you will have to measure the actual losses of your short listed parts.
The TPS61236P switching frequency varies between 750kHz and 1250kHz so you would need to check the chosen inductor is suitable over the whole frequency range.
--- End quote ---
Thanks. I will definitely check out this tool.
Navigation
[0] Message Index
[*] Previous page
Go to full version