Is there really anything wrong with the MC34063?

[SingedFingers]

I am building a 12->5v 100mA buck converter to drive some mixed HCT/TTL logic from a SLA battery to drive a rather ancient alarm. The 78L05 that was in it before snuffed it because it was completely overdriven and just glued to a piece of metal (  ). Unfortunately the glue snapped off by the looks. The hot weather suddenly a couple of days ago finished it off by the looks as it blew it's arse out. This was not my design for ref - I inherited it from my father who has a dubious engineering background. I don't want to throw it away though as it works pretty well when it isn't exploding.

Soooo today, on the breadboard I am using an MC34063 (ST variety) and can get it to kick out that at around 82% efficiency and not even warm to the touch which is the main thing. It's also cheap, I've filtered the output so it's not too noisy with a Pi filter and I can't seem to destroy it whatever I try as it has a suitable current limit programmed via sense resistor. This also knocks the idle consumption of the device down by 75% which is nice so the battery will last a ton longer before I have to recharge it.

However I hear all sorts of horrible things about it when googling. Is there really anything fundamentally wrong with this IC at all or any edge cases that I don't know about before I go and sling this all back together to return it to its usual duty of stopping the pikeys nicking my stuff out of the garage

[bktemp]

It is old, it is not super efficient, it is not really a PWM controller more like a gated free running fixed duty cycle oscillator with a duty cycle reduction circuit on overcurrent. So the switching waveform often looks more random, therefore the frequency content changes depending on the load and input voltage.
If you don't care about the slightly higher ripple on the output voltage (a simple filter often does not help much for a 34063 for low loads), a 34063 is ok. It is cheap and it works.

[snarkysparky]

Also the output switch is a darlington configuration which means there will be 0.7V drop which causes the heating issues many report.

[SingedFingers]

Also the output switch is a darlington configuration which means there will be 0.7V drop which causes the heating issues many report.

Thanks that actually makes sense so you're losing 70mW @ 100mA on just that transistor. Not enough to make this warm on this load though which is fine.

If you don't care about the slightly higher ripple on the output voltage (a simple filter often does not help much for a 34063 for low loads), a 34063 is ok. It is cheap and it works.

I'm not seeing much ripple on it at all even when it's switching 100% of the time. I've got 2x AVX 47uF tant and a 100uH inductor in Chebyshev low pass arrangement and I can't see anything on the scope with a  50 ohm termination which corresponds to a 100mA load on 5v.

[coppice]

The MC34063 is only cheap to use if you get inductors for free. A more modern chip will run at a much higher switching frequency, and allow the use of cheaper inductors.

[bktemp]

I'm not seeing much ripple on it at all even when it's switching 100% of the time. I've got 2x AVX 47uF tant and a 100uH inductor in Chebyshev low pass arrangement and I can't see anything on the scope with a  50 ohm termination which corresponds to a 100mA load on 5v.

At 100mA it will operate in a continous switching mode. Try reducing the load to <10mA, then it will probably go into some erratic burst mode. But the behaviour also depends on the ESR of the output capacitors.

[SingedFingers]

It's only driving about 25mA at average. There is a peak of 85mA when the alarm is sounding. I'm replacing the output driver with a logic level MOSFET as it's currently a tip41c.

Output caps are around 1 ohm (AVX TAJ).

Inductor cost 3x more than the IC did!

I'm actually starting to think I should just redesign the whole thing and use a PIC I've got lying around. Or get an esp8266 and get it to send me a text message when some pikey scum breaks in.

[cyberfish]

For your requirements, you can actually just get a To-220 drop-in replacement.

Something like:
http://www.digikey.com/product-detail/en/murata-power-solutions-inc/OKI-78SR-5-1.5-W36-C/811-2196-5-ND/2259781

[Zero999]

For your requirements, you can actually just get a To-220 drop-in replacement.

Something like:
http://www.digikey.com/product-detail/en/murata-power-solutions-inc/OKI-78SR-5-1.5-W36-C/811-2196-5-ND/2259781

Note that pins 1 and 3 are the opposite of the LM78L05.

[SingedFingers]

I had no idea they existed. Looks like a handy product.

[T3sl4co1l]

As I'm (in)famously known to rant,

Yes.

Crap operating frequency, crap current limiting (try "accidentally" shorting the output!), crap output ripple, crappy huge inductors required, crappy high-ESR filter cap required.

Tim

[David Hess]

There is nothing wrong with the MC34063 which is a low cost version of the old 78S40.

It is a constant off-time switching regulator which has the advantage of not requiring frequency compensation and its current sensing protects it against many failures.

[SingedFingers]

Thanks for everyone's inputs. Much appreciated.

[fcb]

I design the occasional MC34063 into products - it's a bit shit, but it's really cheap.  And whilst inductors aren't free - they can be cheap.

Recently I used them in a commercial product that needed a -12V, and another commercial product that needed to take +5V to +8V.

The product that had the MC34063 generating -12V also had a TPS562200 in it to generate +3V3 from the main +12V for the microcontroller & GSM, as well as 7805 & LM2951-3.3 for low noise analog supplies.  So it's horses-for-courses.

[SingedFingers]

Yeah it's REALLY cheap. 50 pieces in SOP8 for £1.50 on ebay at the moment.

[Zero999]

Another option is to use a discrete switching regulator, which can be made with a couple of transistors and achieve efficiencies over 90%

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

[T3sl4co1l]

Although with less protection and only comparable performance (still needs huge inductors, makes lots of ripple).

A nicer one doesn't take many more transistors, which is sad because Black can't be bothered with it (I actually asked, years ago).  I did, though:



Flyback rather than buck, but a similar approach can be taken.

Tim

[SingedFingers]

Interesting designs. Thanks for the info. I will analyse them when I get a few minutes. I'm always interesting in switching supply designs.

I've built a couple of flyback ones in the past, one crudely breadboarded here. This little bugger, powered by a 555 and LM311 and IRF510 (does the job!) managed to kick out 75v @ 300mA quite happily which is to be quite frank, scary, especially as the max Vf on that diode is 40v and the output cap is 50v rated! This could be replaced with a 34063 and a slightly wimpier MOSFET I reckon. Operation is the LM311 compares the output voltage ratio against a dumb zener reference and turns the 555 off with the reset pin if it is on the mark. Similar operation to the 34063 but larger, more expensive and more complicated.



You can see the burn marks on the board at the bottom right when it went overvoltage when there was a bad solder joint and the divider fell off ( )and was discharged with a screwdriver and some croc clips.

I have another design in flight which was a dumb experiment to generate typical electrostatic CRT drive voltages which uses an E42 core with two TIP41C's in a multivibrator configuration with two primaries of opposite phase. That is already kicking out a selection of nice voltages but the efficiency is completely boiled shit. Sucks up 600mA idling and sounds like someone is murdering a dolphin when under load. Did some maths wrong when calculating the reactance of the primaries me thinks and the frequency is way too low. Needs to be >20KHz.

For ref, I built the 34063 into it now on a little daughterboard (stripboard - kill me) that slides into the steaming holes left by the 78L05 and it's working nicely. Total efficiency is around 82% on the final device which is good!  12mW versus 180mW when idling away now. BIG improvement. Means that SLA will last 10-15x as long before I have to charge it (currently monthly).

This is far more fun than the day job

[Mr. Scram]

What would be a modern day alternative for the MC34063 with similar capabilities? Or are modern chips inherently more specialized?

[T3sl4co1l]

Families, perhaps; I'm partial to a number of TPS54xxx parts.  Shop for types without voltage mode control, with whatever features you need (voltage, current, efficiency), and in whatever type (sync/buck/boost/etc.) you need.

Tim

[MagicSmoker]

The only real advantage of the '34063 - and I suspect the primary reason it has been in continuous production for decades now - is that it is hysteretic mode controller so does not need frequency compensation. That is to say, you get the optimum transient response (basically as fast as the switching frequency) with unconditional stability. You can't touch the transient response with a conventional PWM controller (current or voltage mode) without resorting to at least 5x higher switching frequency and/or without sacrificing some stability.

I don't have quite the same dislike of the '34063 that T3sl4co1l does, but these days there is only one very specific application I use it for: when I need to supply an adjustable or software-selectable output voltage on a product's IO connector. For example, accommodating encoders and resolvers for motor speed feedback that might run on 5V, 12V or a range of voltages. Attached is the schematic for such a circuit, as well as the part of the board implementing it. NB - NCP3063 is an "improved" version of the '34063.

[free_electron]

the only thing that can be done with a 34063 is  :
-first widlarize it.
-douse the remains in gasoline and set it on fire.
-take what is left over now and put in in one cubic metre of concrete.
-bury that block 6 foot deep.
-put a sign there that says not to eat whatever grows on that spot...


yeah they work, sometimes , under some conditions . but just as often they just freak out and explode.

these chips are old. have an erratic feedback mechanism , no fault detection . so the slightest thing that drifts out of spec and the thing goes crazy.

there is much better these days.

[Mr. Scram]

there is much better these days.

Suggestions and alternatives are more than welcome

[SingedFingers]

Ok I've breadboarded 4 converters now and put two into production use and I'm actually quite happy. Sure it's not Jesus' sandals compared to some but I'm really not having any problems.

I cannot make one explode that is set up properly. I have tried. If you kill the sense resistor, it will explode, that is it. Other than that I have overloaded, shorted, dumped large caps across the output, built them on shitty breadboards (see below) and they just go "meh" and carry on. I've pulled up to 500mA though them fine with no significant heating. These are ST ones marked "063 CHN".

I can filter the noise out adequately with a couple of 22uF tants and a 470uH inductor on each line out of the supply.

I knocked up a 5v -> +/-15v converter with a couple of them last night to have a play and that works nicely.

Excuse the potato camera... only had work handset handy.



Efficiency is between 76 and 85% depending on the load and inductor size which isn't terrible.

[langwadt]

It's only driving about 25mA at average. There is a peak of 85mA when the alarm is sounding. I'm replacing the output driver with a logic level MOSFET as it's currently a tip41c.

Output caps are around 1 ohm (AVX TAJ).

Inductor cost 3x more than the IC did!

I'm actually starting to think I should just redesign the whole thing and use a PIC I've got lying around. Or get an esp8266 and get it to send me a text message when some pikey scum breaks in.

if price matter, just get something like this http://www.ebay.co.uk/itm/162402905270