EEVblog Electronics Community Forum

Electronics => Projects, Designs, and Technical Stuff => Topic started by: mrpackethead on May 05, 2017, 11:10:16 am

Title: Wanted..
Post by: mrpackethead on May 05, 2017, 11:10:16 am
I am looking for a Buck Converter to nomally convert 48V to 5V @ 200mA.    Cost is the big driver on this one,  i need to build 10's of thousand so them and i'm willing to compromise on effiency..

Any suggestion for a chipset.
Title: Re: Wanted..
Post by: CJay on May 05, 2017, 11:29:06 am
Specifics, I don't know chip part numbers but there are multitudes of PoE adapters that do just that (DWL-P50 being one I've used and torn apart) with the addition of the Ethernet 'stuff' so might it be useful to grab a couple and see what they're using?
Title: Re: Wanted..
Post by: BrianHG on May 05, 2017, 11:32:33 am
Just use this 1 TI chip.  All it needs is 1 cheap inductor 5 resistors, 4 caps, 1 diode and you can operate between 7v in and 95v in:
LM5008, or LM5008A.  $1.29 for 1000 at digikey.  Maybe less for 10k quantity.
Title: Re: Wanted..
Post by: BrianHG on May 05, 2017, 12:02:44 pm
TI's TPS54061 is $1.01 for 2500, but, it's exactly 200ma out.
Title: Re: Wanted..
Post by: fcb on May 05, 2017, 12:19:02 pm
If cost is the prime consideration, i'd roll a buck converter using TL494 as the core - very low cost (use a 12-15V zener diode to drop the supply for the TL494).
Title: Re: Wanted..
Post by: tszaboo on May 05, 2017, 12:31:25 pm
According to webench, LM5085 is the cheapest TI part for this, with a BOM of 2.28 dollar @ 1K. I would use this figure as a benchmark to make something cheaper. Its not hard.
Title: Re: Wanted..
Post by: MagicSmoker on May 05, 2017, 12:35:58 pm
The self-oscillating 2-transistor buck (most popular version is by Roman Black) is probably the cheapest approach.

http://www.romanblack.com/smps/smps.htm (http://www.romanblack.com/smps/smps.htm)

EDIT: I had a few spare minutes so I recreated the Roman Black circuit in LTSpice and tweaked the values for 48V in, 5V @ 0.25A max output. It behaves surprisingly well with a step change in load. One interesting aspect of its operation is that at very low output currents switching stops and it basically just acts like an amplified zener.

Title: Re: Wanted..
Post by: BrianHG on May 05, 2017, 12:48:12 pm
According to webench, LM5085 is the cheapest TI part for this, with a BOM of 2.28 dollar @ 1K. I would use this figure as a benchmark to make something cheaper. Its not hard.

The LM5085 is actually 5 cents cheaper than TPS54061,but it requires more $ in parts thanks to the external P-Channel fet you need + additional filtering caps and resistor + assembly and board space makes it more expensive.
Title: Re: Wanted..
Post by: tszaboo on May 05, 2017, 01:00:38 pm
According to webench, LM5085 is the cheapest TI part for this, with a BOM of 2.28 dollar @ 1K. I would use this figure as a benchmark to make something cheaper. Its not hard.

The LM5085 is actually 5 cents cheaper than TPS54061,but it requires more $ in parts thanks to the external P-Channel fet you need + additional filtering caps and resistor + assembly and board space makes it more expensive.
You are right, for some reason it completely ignored my input of 0.2A, and calculated it with 2A output power.
In any case, I wouldn't just do a full design online for someone, selecting the absolute cheapest parts for a project with 10K+ quantities. Because optimizing it can be a few days of work (since price is that relevant) and the OP has the budget to actually hire someone as a consultant, if he does not know how to do it.
Title: Re: Wanted..
Post by: MagicSmoker on May 05, 2017, 06:18:12 pm
I edited my previous reply to include an LTSpice .asc file of the Roman Black regulator. Careful selection of the transistors and buck inductor should get the total BOM down to less than $2. If the OP is feeling generous for me solving his problem...  :-DD

Title: Re: Wanted..
Post by: Zero999 on May 05, 2017, 11:36:55 pm
The self-oscillating 2-transistor buck (most popular version is by Roman Black) is probably the cheapest approach.

http://www.romanblack.com/smps/smps.htm (http://www.romanblack.com/smps/smps.htm)

EDIT: I had a few spare minutes so I recreated the Roman Black circuit in LTSpice and tweaked the values for 48V in, 5V @ 0.25A max output. It behaves surprisingly well with a step change in load. One interesting aspect of its operation is that at very low output currents switching stops and it basically just acts like an amplified zener.
Whilst I appreciate efficiency isn't the prime design goal here. I haven't done an extensive simulation, other than power in vs power out which suggests that circuit is only around 25% efficient (still much better than linear though) and most of the losses are in Q1, which should be checked to see it isn't getting too hot.

I do like the Black regulator but I think the old LM7805, plus a zener, BJT VBE multiplier or emitter follower and potential divider before it, to keep the input voltage below its maximum rating.
Title: Re: Wanted..
Post by: mrpackethead on May 06, 2017, 04:08:04 am
Thanks for input folks.. interesting everyone goes to WorkBench.. :-)  They have made it easy for people..

The pair of transistors is an interesting idea, id not contemplated doing things discretly, but its well worth considering..

Title: Re: Wanted..
Post by: westfw on May 06, 2017, 09:44:49 am
48v nominal? (36-75v, usually?). Or closer?
Isolated, or doesn't matter?

Title: Re: Wanted..
Post by: MagicSmoker on May 06, 2017, 10:07:46 am
Whilst I appreciate efficiency isn't the prime design goal here. I haven't done an extensive simulation, other than power in vs power out which suggests that circuit is only around 25% efficient (still much better than linear though) and most of the losses are in Q1, which should be checked to see it isn't getting too hot.

I do like the Black regulator but I think the old LM7805, plus a zener, BJT VBE multiplier or emitter follower and potential divider before it, to keep the input voltage below its maximum rating.

Yes, my hasty tweaks to get the circuit to work at 48V resulted in appalling efficiency. The buck choke inductance is too high and the feedback capacitance too low resulting in too high a switching frequency and deep CCM operation.

Attached is a revised version of a 48V to 5V/0.25A Roman Black regulator with a bit more thought put into the components and efficiency is now 51%, which, in my experience, is about as good as you should expect from a buck at such a large step-down ratio and low output power. A self-oscillating current-fed push-pull can probably be designed for closer to 70% efficiency, but it would require a custom transformer in addition to a choke so it will inevitably cost more (albeit this is still a very simple circuit, and the topology of choice for those cute little on-board SIP/DIP 1-2W dc-dc converter modules).

Title: Re: Wanted..
Post by: Zero999 on May 06, 2017, 01:32:10 pm
Whilst I appreciate efficiency isn't the prime design goal here. I haven't done an extensive simulation, other than power in vs power out which suggests that circuit is only around 25% efficient (still much better than linear though) and most of the losses are in Q1, which should be checked to see it isn't getting too hot.

I do like the Black regulator but I think the old LM7805, plus a zener, BJT VBE multiplier or emitter follower and potential divider before it, to keep the input voltage below its maximum rating.

Yes, my hasty tweaks to get the circuit to work at 48V resulted in appalling efficiency. The buck choke inductance is too high and the feedback capacitance too low resulting in too high a switching frequency and deep CCM operation.

Attached is a revised version of a 48V to 5V/0.25A Roman Black regulator with a bit more thought put into the components and efficiency is now 51%, which, in my experience, is about as good as you should expect from a buck at such a large step-down ratio and low output power. A self-oscillating current-fed push-pull can probably be designed for closer to 70% efficiency, but it would require a custom transformer in addition to a choke so it will inevitably cost more (albeit this is still a very simple circuit, and the topology of choice for those cute little on-board SIP/DIP 1-2W dc-dc converter modules).
That's much better. The lower value inductor will also be cheaper too.

As I'm sure you know, a lower resistance inductor and larger diode (lower VF) will also help efficiency but it would cost more.
Title: Re: Wanted..
Post by: mrpackethead on May 07, 2017, 08:58:42 am
Thanks for all the awesome input.   I've had a good look at this problem for a few days, and i've decided that i can change the specificaiton a wee bit. If i can keep Vin to less than 40V, it really opens up the number of possiblitys

The MC33063A is a bit of an old school IC, but a pretty useful one to have in your junk box.. you can use it for a variety of topologoes and they are cheap! ( $0.15 or so )

Title: Re: Wanted..
Post by: BrianHG on May 07, 2017, 10:12:20 am
Just drop your 48v down to 30v with a 18v, 5w zener diode and use a linear regulator.  Dirt bottom cheap of around 35-45 cents for everything.  ZD=10 cents, 36v-5v LDO = 15 cents, + 2 caps, 15 cents...