LM2674 vs MC34063: efficiency vs cost

[mainakae]

Hello again every body!

After my first post, in which I drifted awfully from what I wanted to ask, I'm trying to get more specific in the subject that has kept me struggling for the last month.

I started working on a better power supply for my Dromo project. At first I used a linear regulator for my prototipes, and worked great, but after modeling and designing the power distribution for the whole system, soon found that the linear regulator was an awful awful idea. Not only it's inefficient (which is something I'm actively addressing in Dromo, I plan to make my house more efficient: bad start if my domotic system is inefficient) but also nearly impossible to use in my current schema (see my next topic post ).

As if Dave read my mind in the distance, just when I was getting crazy trying to find the way to a swithced regulator, he designed a switching power supply in a magnificent video in youtuve! (hell, thanks again Dave!!!). But yes, I'm one of those guys... after building my own, and characterizing the supply, I got some less than amazing results. 60 to 70% efficient  sure: better than linear regulator, but... it also takes more space!. After noticing that the MC34063 Dave was using is quite an old IC (ttl logic for what I know) I started searching for more efficient regulators (MOS?) with higher frequencies (smaller coils) and soon found the LM3674. The datasheet promises really good efficiencies for my project, abobe 80% (my devices consume 5v 60mA to 200mA, but I'm planning for a max of 300mA just in case an external module is plugged to the Dromo device) but the step difference in parts costs are making me consider if that difference in efficiency is worth it (.34 € for the mc34063 to nearly 2€ for the LM2674).

What do you think?.

[c4757p]

What do you think?.

I think MC34063 can do better than 60-70%! I use it frequently and tend to get 80-90%. How are you using it? (And what manufacturer - I use the ON Semi or ST Micro ones.)

Also, you can do better than LM2674. It's only expensive because it's less common than some others. You can get 20 LM2575 for US$9 on eBay. (Yes, I've bought those exact ones from that seller, they're genuine.)

Edit: LM2674 appears to be a bit more efficient. Still, so what? You were already considering a linear regulator, the difference between 80% and 90% can't be that important.

[mariush]

LM3674 is a buck regulator, so I think you would need more than 5v at any time to get 5v out. All the graphs in the datasheet show maximum output at 3.3v and the fixed versions go only up to that.

An IC that I have my eyes on is ISL9111, a 1.2 Mhz boost regulator that also comes in fixed 5v version:

fixed 5v: http://www.digikey.com/product-detail/en/ISL9111AEH50Z-T7A/ISL9111AEH50Z-T7ACT-ND/2816339
adjustable: http://www.digikey.com/product-detail/en/ISL9111EHADJZ-T7A/ISL9111EHADJZ-T7ACT-ND/2816327

It's a bit expensive at around 3$ for one piece, but it can take 0.75v-5.25v and outputs 5v.



As you can see from the picture, it looks like it can just about do 5.25v @ 200mA with 1.8v input, and seems like it can do about 400mA with 3.6v input, so it would be great if you'd plan to use lithium batteries that are about 2.7v discharged and 4.2v charged - you'd still get about 100-150mA with 2.7v  .... and at least from the graph above you'd get over 90% efficiency when battery is full.

Also, as an idea, maybe get the adjustable version and set the output voltage to about 4.5-4.6v and see how everything behaves - most ICs running at 5v would be stable at 4.6v-4.75v and run just fine, they have some tolerance regarding input voltage, so this small drop in voltage could squeeze a few more mA at low voltage levels.
 

[Alana]

Be extra careful with board layouts for those 500kHz+ regulators. I had to make 2 layouts for it to work, and its not breadbordable at all.

[mainakae]

Woa, that whas fast!, thanks every one of you for your answers!

c4757p:
I thought that it could do better too. In fact I got it at about 78% efficient (at best) using a huge ferrite toroidal coil wound by myself (got the number of turns for the given inductance from somewhere in the internet). The thing gets better at higher currents, but my device is consuming 60mA minimum (most of the time) and up to 200mA when it's consuming to it's full potential. It could get even further if I connect some external module (like a bluetooth device, a PIR, temperature probes or things alike) but not much I suspect. The problem is to make it efficient at 60mA with an input voltage ranging from 24v up to 40. Perhaps I'm just asking too much. I could fix a little more the input voltage. In fact, perhaps to keep things in the safe side, I should choose 40v, but somehow I don't like that number. Yup I suck that much at engineering as to not to have a proper reason 


Mariush:
It looks amazing, I was considering using a boost regulator to get all the juice out of my drained alcaline batteries, but havn't started with that project yet now I have a good point to start with, thanks for that!
In my current project, I've decided to use a buck one in order to deliver power to several devices at higher voltages, thus minimizing loses in the power distribution line due to high currents.

Alana:
Thanks for pointing that out!, any suggestions on how should I outline it in order to minimize EMI and general troubles?. For a start, I've chosen shielded inductors. Also, I have to share board with CAN Bus circuitry (both driver and transducer), so I'm a bit concerned with that. Do I have to take special measures to avoid noise getting into the CAN and killing my communications?.

[mainakae]

c4754p, I forgot to say that I got that test running on a breadboard. I would not be surprised to be told that it kills efficiency totally, if that's the case.

[c4757p]

c4754p, I forgot to say that I got that test running on a breadboard. I would not be surprised to be told that it kills efficiency totally, if that's the case.

MC34063 isn't fast enough to be ruined by a breadboard, though it can be ruined by sloppy layout on a breadboard.

[mainakae]

That could deffinitely be the case: my breadboard layouts tend to be messy at best. I'll try again with better components, specially the coil. Mine made my wife look astonished, but I suspect it's less than optimal, as much as own kids are always good looking.

[c4757p]

Sometimes it can be useful to do a prototype circuit like this. Looks messy, but parasitics are low. That circuit wouldn't work at all in a breadboard, actually.

[c4757p]

Holy shit, that inductor's huge! It's bigger than my hand!

Oh wait, you just forgot to shrink the image. 

[Alana]

My layouts that worked were strongly based on "recommended layout" section in chip's datasheet. There is a good example in LMR62014 datasheet. General rule is to keep switching current loop as short and compact as possible. This include In and Out caps, inductor, diode and regulator chip - those parts should be as close together as possible and connected with thick traces.
If it comes to EMI I'm as green as you.

[mainakae]

Holy shit, that inductor's huge! It's bigger than my hand!

Oh wait, you just forgot to shrink the image. 

hahahaha

I forgot, true. Anyway it's too large for what I plan  .