So how about a Tablet power supply?

[Mr RB]

...
Well, if you do 240vac to 12vdc off line and something goes wrong it could possibly send the peak of a 240vac rms sine out where 12v is supposed to be. Can you spell "Fry Daddy"

I was going to use a SMPS transformer based design (ie forward converter), I would not try 240->12 at a couple of amps in something like a buck converter!

...
BTW have you ever tried to run a regulated switching wall wart off of a 12v to 120vac power inverter without it being a pure sine (like the 'modified' sine type which is really just a plus and minus 145v pulse) ?

I have a lot of experience with SMPS supplies, after a period of 25 years or so being a repairer of TV/VCR/consumer goods a lot of the repairs involve failed SMPS supplies in the appliances, and I've repaired and seen inside literally thousands of designs and schematics etc.

Those SMPS wall warts will use a rectifier to make the AC mains into DC, and into a large filter cap which will charge to about 170vDC assuming you're on 115v AC mains. So the SMPS part of the wall wart is actually a 170v DC to 5v DC converter, which will use some type of PWM to regulate the output 5v.

It really does not matter what quality of AC you are providing, it is not that sensitive to the freq of the AC (ie generator use) or even if you feed it DC and not AC.

Depending on the amount of overhead built into the PWM system it may even run ok from a much reduced DC input like 100vDC. Many of the offline SMPS supplies will work just as well from 115v mains as they do from 240v mains, and still have the overhead to work ok down to 70 or 80v depending on load.

 

[MrAl]

Hello again Mr RB,

Oh yes i didnt think you would try a transformerless offline power supply :=)

Thanks for the info on the switching wall warts. It's important to me because i cant risk blowing out my wall wart.

So you were saying that the input to the SW Wall Wart is a rectifier, maybe full wave or half wave?
Also, after the rectifier(s) do they connect the cap right up to the diodes or do they use a choke or resistor or surge thermistor in between the diodes and cap?

With this last question i am thinking about what the strain on the filter cap will be when it is being banged with plus and minus 145v pulses vs a smoothly rising sine. I suppose that if the cap is charged somewhat then the current pulse will rise sharply and then taper off every half cycle. I guess i could test one out, maybe that's the best way to find out.
I dont want to repeatedly stress the filter cap any more than it gets with the regular sine input. I know they do get pulsed even with a sine because of the conduction angles of the diodes, but i'd have to see how much worse it is with a super fast rising pulse instead of a sine.
You see where i am going with this? Maybe it would be a good idea to connect something between the pulsing 120vac output and the wall wart rather than connect it directly as with a regular ac sine line.

 

[Mr RB]

...
I dont want to repeatedly stress the filter cap any more than it gets with the regular sine input. I know they do get pulsed even with a sine because of the conduction angles of the diodes, but i'd have to see how much worse it is with a super fast rising pulse instead of a sine.
You see where i am going with this? Maybe it would be a good idea to connect something between the pulsing 120vac output and the wall wart rather than connect it directly as with a regular ac sine line.

I see exactly where you are going. It is common enough in the small SMPS supplies to have a small series resistance or even a thermistor from the bridge rect to the main HV cap, it increases diode conduction angle and reduces peak currents and power-up current (which I'm sure you know). They also tend to use undersized HV filter caps with quite a bit of ripple tolerated (again this increases conduction angle and reduces costs).

Any decent SMPS wall wart should be ok running from a small commercial inverter, there's a whole world of people out there charging their cellphones etc from small cheap inverters in caravans etc.

Of course if it bothers you then add a series resistor yourself between the inverter AC output and the wall wart input, even if just as a test to 'scope out the conduction period?

 

[MrAl]

Hello again MrRB,


Yes you're right. The only way to know for sure is to measure. So at some point i have to go ahead and hook up the wall wart to the inverter and dig out the scope. That means hooking it up at least once :-( but there's no way around it i guess, except to maybe use a somewhat larger series resistance to start and then gradually decrease the resistance until it doesnt have too much effect.
I guess i could fully test this with a dummy load and if anything goes wrong i guess i'll be hunting for a new wall wart, or else converting one of my other ones to 5v out (which i eventually have to do anyway in order to run off of 12v, so it would run off 12vdc or 120vac which would be nice anyway).

I did try it with another wall wart type but that was a transformer type. Worked pretty well really with only slightly lower output, but still worked in the application.

I was also thinking about the no load current from the 12v to the input of the inverter and if it is too high i wont want to run it that way anyway but rather from the 12v battery. I'll have to see how it goes.

 

[MrAl]

Hi,

I lucked out today when i found a schematic of a typical regulating switching wall wart. The design includes a pi filter with 10uf in 470uH and 10uf out, just after the full wave bridge rectifier. For this one they also included an extra 10 ohm resistor in series with the rectifier diodes on the AC side. It's rated for 90vac to 265vac. Only half wave rectified on the output but lots of filtering.

I also noticed that mine is rated 100vac to 240vac so it is probably a similar design, and since i'll be using it at 120vac and it can take up to 240vac it should be all right. I do a few quick tests first though.

 

[Mr RB]

Excellent. Just bear in mind they won't all have a Pi filter with inductor, they skimp on inductors on a lot of supplies for size and cost. Also a 470uH inductor at 170v DC 120Hz pulses is not going to do a lot for the incoming supply side, i think it is mainly to stop the HF energy from the SMPS getting back out to the mains.

That 10 ohm resistor though is typical and will help with that incoming peak currents.

As you said, with yours being a 100-240vAC wallwart type that is good the primary components should be good for a lot of safety margin on your 120v system.

I'm really itching to see the overall efficiency ie measured 12v current in vs 5v current out to your tablet. As the final solution I'm still voting for a 34063 SMPS IC and a decent FET, which I think yould give you >90% efficiency at 12v->5v, small and neat and very little to go wrong.

 

[MrAl]

Hello MrRB,

Yeah i was hoping that the 240vac rating would mean it would put up with 120vac input much easier.

I have a feeling that the input on the 12vdc side of the non sine inverter will be 300ma PLUS any additional current caused by the device itself plugged into it. So looking at this from the point of view of 120vac output, that would mean the input side has 10 times more current than the output. So a 10 watt device at 120v draws 1/12 amp, so the input side sees 10/12 plus maybe 10/120, and add to that the other current of 300ma and we are up to about 1.2 amps on the input, which is around 14 watts. So the eff is not going to be too good, maybe 70 percent. So that makes a 100AHr battery look like a 70AHr battery, or in my case, a 20AHr battery looks like a 14AHr battery (as compared to a 12vdc to 5vdc 100 percent eff converter). A 12vdc to 5vdc 90 percent eff converter would be like a 100Ahr battery looks like a 90AHr battery, and a 20AHr battery looks like a 18AHr battery, a significant gain i think. So that means roughly an additional 4 hours run time with the pure dc to dc rather than the inverter. That's significant i think, but then again this other setup (120vac converter) would only be used if something went wrong with the dc to dc converter that will eventually be constructed for the task).

Once i get up the guts to actually connect it all up (ha ha) i'll do some measurements and see what we actually have, but i dont think it will be too much different than that, unless the quiescent current goes down from 300ma to maybe 100ma, in which case i would see less gain from the pure dc to dc setup. Comparing to a 20AHr battery again i might see another 2 hours run time, bringing the total up from 14 to 16 hours run time instead of 18 hours run time with the pure dc to dc. Think it is worth it?
Well there is another side point too though, and that is what happens if the inverter poops out in the middle of a storm? Then i have nothing unless i have the pure dc to dc ready to go anyway.

Im not sure if you are aware of this or not, but the 34063 chip has more ripple than a conventional switching controller, for the same component values and operating load and such. That's because it uses a partly digital regulating technique rather than a pure analog technique that almost all other converter chips use. Does that matter that much? Well, maybe not as much if the inductor size is increased somewhat, like 2x times. So an inductor that is 500uH in a pure analog design would require an increase up to about 1000uH for a 34063 design. And i think the inductor has to be somewhat large too as those numbers indicate, like 1mH as compared to 100uH for other somewhat simple designs.
But, after all is said and done, i do happen to have a couple of these chips laying around That means i might just give them a shot anyway. I was thinking of a more custom design at would get up above 90 percent eff but that would take more board space and a bit more effort to construct. So for now i guess i am undecided.
But you're saying that we can get above 90 percent with this chip too, so i guess it is worth looking into. My only doubt is can we get an inductor that has enough inductance and still low esr, where the inductance meets the output ripple requirement (which isnt exactly decided yet either though).
I'll have to look at a couple reference designs for the chip, it's been quite a long time since i worked with that one.

Yeah that 34063 chip puts out a nasty ripple, and it dosent matter if we increase the inductance or output cap value because is is very possible that it will regulate super cycle. So if we set the oscillator frequency to say 50kHz, the output ripple could be 50kHz and 25kHz and 10kHz and 5kHz for example, and that lower frequency is harder to filter. Increasing the inductance and capacitance means lowering the output filter frequency which means the switch cycle frequency could lower too, so it may not help to make components larger. Maybe at a nice large value for the inductor, but that's hard to get a lower esr with too. That's the drawback to using a digital regulating technique that might be ok with loads that are not too ripple sensitive.
If you'd like to see the effects you could do a simulation and note the output is nasty. I might be able to set up the simulation if you want to use LT Spice, or if you dont mind setting it up you could see what happens.

LATER:
Disappointment: with 1 amp output from the little switching wall wart (i did connect it finally for the test) the inverter draws 0.84 amps, which makes the whole setup only 50 percent efficient when powering the Tablet. I used a dummy load of 5 ohms instead of the tablet, but the output was a very clean 5v.
One thing i didnt do was measure the efficiency of the wall wart itself. It could be that it alone isnt very efficient. But together the combined efficiency is only 50 percent, 10 percent less than i had guessed, but then again i did not include the efficiency of the wall wart when i guessed
So that makes a 20 AHr battery at 12v provide about 24 hours of run time instead of about 44 hours (90 percent eff dc to dc), a very significant difference.
So now i've realized that the wall wart eff will take a toll too, and even though it isnt bad it cant be better than 90 percent eff.

About the peak currents into the wall wart (with 1 amp load on the wall wart)...
With the wall wart plugged directly into the 120vac mains line, the peak is up around 400ma, and it is a very sharp peak and there is no conduction for part of the cycle.
With the wall wart plugged into the inverter the main peak goes down to about 125ma, but there is a very sharp peak that lasts for a very very short time that could be 500ma or so. That pulse is so short though i have to wonder if it matters. The main current wave shape into the wall wart looks almost like the inverter output, very unlike the 120vac mains line current into the wall wart which is a sharp rise and then lowers down followed by a longer time where it ramps down more slowly, then goes to zero and stays there until the next cycle. So it could be the inverter is actually more gentle on the wall wart than the mains line is.

 

[Mr RB]

...
Disappointment: with 1 amp output from the little switching wall wart (i did connect it finally for the test) the inverter draws 0.84 amps, which makes the whole setup only 50 percent efficient when powering the Tablet. I used a dummy load of 5 ohms instead of the tablet, but the output was a very clean 5v.
...

Darn, I had hoped it would do better than that, although I did suspect a poor result.

...
Im not sure if you are aware of this or not, but the 34063 chip has more ripple than a conventional switching controller, for the same component values and operating load and such. That's because it uses a partly digital regulating technique rather than a pure analog technique that almost all other converter chips use. Does that matter that much? Well, maybe not as much if the inductor size is increased somewhat, like 2x times.
...

Yep, I knew that but the 340963 is not really that bad, it's a nice reliable design that is a bit "old fashioned" in a way that it works well with good quality larger inductors (which I prefer because I like good efficiency and good reliability, both are better at lower SMPS freq and with larger inductor).

I still have a couple hundred 34063's in stock and use them from time to time, and always been very happy with the result. If you prefer another IC that's fine, but for really good efficiency and reliability you should go for a nice powdered iron toroid about 500uH to 1mH and lowish switching freq (20-40kHz), which are also ideal for a 34063.

Also re the ripple I'm still assuming your tablet has internal linear voltage regulators on all the rails (which it needs being battery operated!). So a couple of mV ripple on the input would be insignificant, and is removable anyway with a Pi filter on the SMPS output if you like.

All you need to get >90% with the 34063 at 1-2A output is to use low drop switch and diode. For a diode I would use a dual 10A or dual 15A schottky diode TO220 pack, you can rip one of these from any old SMPS supply like an old PC PSU if you don't have a new diode. Just parallel the two diodes and it will have a very low forward drop at 1-2A.

For the FET it depends how you like to drive the FET fromt he 34063 but a PFET is probably easier. Luckily your Vin of 12v is low, and Iout 1-2A is low, so you should easily find a PFET that will have a very low forward drop under those conditions.

If you use a good powdered iron inductor with minimal turns of thick wire (like about 30 turns of 1mm wire on a 30mm dia toroid) the I2R losses will be very low so you can disregard inductor losses.

Your other main efficiency loss is that series resistor on the Vin to the 34063 used for overcurrent protection. I think it needs 0.3v drop to activate the OC protect, so in use you will lose maybe 0.2v on that resistor. That's probably about a 1.5% efficiency loss with a 12v input and 50% duty cycle. You could play with that by using a PTC thermistor there? Or some other trick?

...
About the peak currents into the wall wart (with 1 amp load on the wall wart)...
With the wall wart plugged directly into the 120vac mains line, the peak is up around 400ma, and it is a very sharp peak and there is no conduction for part of the cycle.
With the wall wart plugged into the inverter the main peak goes down to about 125ma, but there is a very sharp peak that lasts for a very very short time that could be 500ma or so. That pulse is so short though i have to wonder if it matters. The main current wave shape into the wall wart looks almost like the inverter output, very unlike the 120vac mains line current into the wall wart which is a sharp rise and then lowers down followed by a longer time where it ramps down more slowly, then goes to zero and stays there until the next cycle. So it could be the inverter is actually more gentle on the wall wart than the mains line is.

Thanks for that info that's interesting! I'm guessing the little inverter is built with a high-ish output impedance compared tot he AC mains? That makes sense as it would be better for the inverter to supply less current over longer periods.

 

[MrAl]

Hi again MrRB,


Yeah, i was forgetting that when i plug the wall wart into an inverter what i am
actually doing is plugging one inverter into another inverter which as you know
reduces efficiency even more, so that's what i figure is causing the very poor
efficiency.

Oh you have a couple hundred in stock? No wonder you like them so much
What i did was i purchased a couple (i think 2 of them) before doing a simulation.
When i got them then i did a simulation and found the output ripple was so bad
i just couldnt believe it, and so i investigated and that's when i found that the
digitial technique is not such a great idea unless we really have to do it that
way. An analog approach works so much better it's like night and day. For example,
a design with that chip might require a 1mH inductor even at 50kHz (as you know)
but with an analog based switching controller it would do fine with 100uH. That's
too large a difference to ignore because it's much easier to get high efficiency
with a 100uH inductor than with a 1mH inductor. And although adding a pi filter
on the output is certainly a good idea, that means yet another inductor in series
with the load current which further reduces efficiency. See what i mean now?
I would definitely consider using this for a motor drive where it doesnt matter
though. I would also consider not using the overcurrent mechanism and go with
an independent external protection circuit as i was talking about earlier in
this thread.
The investigation i did led me to see that the output ripple depends highly
on the combine L and C on the output. That's because the L and C have a
combined response which forces the controller chip to skip switch cycles. It
could miss as many as 10 cycles with some values of L and C, and possibly change
with load current. That's the main cause of the high ripple, because of the
random skipping low frequency components are generated and of course dont
filter out nearly as well as the original switch frequency components do.

I dont think i have any 1mH inductors around either except for low current ones
with somewhat higher than desired esr. I do have lower values like 50uH, 100uH,
150uH, etc., which can take up to 3 amps.

But i havent settled on an approach yet, and as you say it does work, i'd just have
to purchase an inductor or two with low esr and high value inductance (try that).
I also have the chips already. And yes if it has an internal linear regulator that
would clear up the ripple, but i cant be sure of that yet. I would hope it uses
a switcher because i think the battery is 7.2v and the chips run on maybe 3.3v
(i havent looked into this too much yet though).

I guess i'll have to think all this over.

In the mean time, if you care to try to find an inductor with a value of around 1mH and low low
esr (for efficiency sake) see what you can find. I think it might be hard to find one within reason.

 

[Mr RB]

Oh you have a couple hundred in stock? No wonder you like them so much
What i did was i purchased a couple (i think 2 of them) before doing a simulation.
When i got them then i did a simulation and found the output ripple was so bad
i just couldnt believe it, and so i investigated and that's when i found that the
digitial technique is not such a great idea unless we really have to do it that
way.

It's not so bad with a constant load like 1A, you can run a sim test if you like of a 12v->5v buck at 1A output, the oscillation should settle out very nicely at a steady frequency (it will just change the OFF period to give regulation). The thing you mentioned re skipping cycles etc generally occurs at very low currents and is the way the IC is meant to operate (which is more in line with a "discontinuous" mode that most SMPS ICs will use at low currents. Choosing the right RC timing cap value is critical running the 34063 at low frequencies too, to stop it going discontinuous.

... That's
too large a difference to ignore because it's much easier to get high efficiency
with a 100uH inductor than with a 1mH inductor.

That one I will argue! Once you have the static state ON and OFF losses to a minimum, by using very low forward drop FET and diode, then all the remaining efficiency losses come from swithcing losses. Each switching event incurs a fixed "parcel" of losses, so the lower you can run the switching frequency the lower the overall losses are. I did a commercial buck SMPS for solar use and ended up running at 17kHz, for exactly that reason.

Also, a 1mH inductor will have less current ripple at lower frequencies, and reduced current ripple again means better efficiency. The only area a 100uH iinductor will be better than a 1mH inductor is that it likely has lower DC resistance. However with the right toroid wound with a small number of turns of thick wire (see my post above) you can get an inductor of 500-1000uH with a DC resistance as low as 10 milliohms or so. At your 1A output that will be only 10mW lost (assuming current ripple is low), only 0.2 of a percent of the 5W output. This will be insignificant compared the the massive efficiency savings from running at low freq close to 20kHz and having low current ripple.

...
I dont think i have any 1mH inductors around either except for low current ones
with somewhat higher than desired esr. I do have lower values like 50uH, 100uH,
150uH, etc., which can take up to 3 amps. ...

Those are general purpose inductor/chokes and won't be very good for a high efficiency buck SMPS. You need to wind one on a powdered iron toroid with good core properties and very few turns, 25-35.

Anyway although it probably looks like it I'm not trying to convince you to use a 34063, the choice of IC is not that important. The important details will be the FET, diode and the right inductor. The point I was making is that once you have chosen the best parts for those and low freq operation, the 34063 will work pretty good.

If you wanted to simulate a 34063 buck using an external PFET (you can use similar circuit the datasheet example for external PNP) and TO-220 20A diode pack and a 500-1000uH inductor with 10 milliohms I think you'll find a pretty good result.

 

[MrAl]

Hi MrRB,

Yes you're right if you can find an inductor that has low esr and also low Rac then it doesnt matter as much what the inductance is, but i automatically figure in the cost too when i talk about these. For example, the 100uH inductors i can get cost at most around 1.75 USD, usually less, and handle up to 3 amps or more (5 amps perhaps). Im shooting for a 2 amp output (max) so i want something that can handle say 3 amps before the inductance decreases significantly.

If you can find me a ready made inductor from a reputable source (not eBay please ha ha) that is 1mH and has 10mOhm series resistance for a comparable cost i'll definitely consider using the 34063, but i havent had any luck finding them with that inductance and that low of an esr that dont cost a small fortune. Maybe im not looking in the right place Digikey? Winding my own means not only achieving the inductance but also keeping the core from saturating with DC current.

BTW the problem with the said chip is NOT that it goes into discontinuous mode, that's entirely different, it's that it skips switch cycles because of the LC resonance. The LC "plant" reacts as it is driven, it is not driven to a predetermined output by a predetermined input because it has it's own idea of what the output voltage should be based on the energy that it took in. There's no way to stop it from reacting and get high efficiency at the same time without sync'ing the resonance to the LC natural frequency, which is not reliable.
If the above doesnt make any sense or seems strange, i'll post some wave forms which would show exactly what is happening and that clearly it is not entering a discontinuous mode of operation but is another 'mode' altogether not seen in more conventional (pure analog control PWM) switchers.

LATER:
I found an interesting trick to help fix the situation above. It stems from the theory of boost converters where the output becomes unstable when the filter capacitance ESR becomes lower than a certain value. It's the one time when we dont want too low of an ESR for a DC power supply filter cap. Well the trick is to increase the ESR on this converter too, and that helps stabilize the output significantly. There may be a way to use this after all
I'll have to look at this in more detail first though.

EVEN LATER:
The required ESR has to be a little bit highish for my taste, but i havent evaluated the effect on efficiency yet. If it doesnt hurt too much i'd be all for it. The required ESR (or should i say the value that helps get this thiing much more stable) is 0.2 ohms. That's usually considered too high but not here unless it bothers the efficiency. I'll try to get to that soon to see what the effect is.

In the mean time, if you can find a 1mH 10mOhm inductor somewhere for a reasonable cost i'd be quite happy. Perhaps even 500uH ?

I had another project in mind this might work with too eventually.

 

[Mr RB]

...
Yes you're right if you can find an inductor that has low esr and also low Rac then it doesnt matter as much what the inductance is, but i automatically figure in the cost too when i talk about these. For example, the 100uH inductors i can get cost at most around 1.75 USD, usually less, and handle up to 3 amps or more (5 amps perhaps). Im shooting for a 2 amp output (max) so i want something that can handle say 3 amps before the inductance decreases significantly. ...

I knwo those commercial Asian inductors and have used them. A 3A inductor is not ideal for a 2A high efficiency SMPS in my opinion, I would use an inductor good for 8-10A or so, mainly to get the large core size and thick low resistance windings. The extra dollar or two does not matter for your one-off high efficiency project. Those 3A pre-wound ones are also a bit lossy on the cores as they are for "choke" use also.

...
If you can find me a ready made inductor from a reputable source (not eBay please ha ha) that is 1mH and has 10mOhm series resistance for a comparable cost i'll definitely consider using the 34063, but i havent had any luck finding them with that inductance and that low of an esr that dont cost a small fortune. Maybe im not looking in the right place Digikey? Winding my own means not only achieving the inductance but also keeping the core from saturating with DC current.

I doubt you'll find a really high performance inductor off the shelf. However you can make one easily by winding some small number of thick turns of wire on the right toroid core. A core of about 25mm-30mm will be plenty large and will not saturate from your 2A max current, so you can ignore saturation. Likewise the large core is big enough to take thick wire 1/1.25/1.5mm diameter so you can get low DC ohms.

Getting the core is not impossible, many of the hobby outlets sell powdered iron inductor toroid cores, usually in one material chosen to be about the middle of the range. If I get a chance I'll take a photo of some of my toroids, it might help if I can show what I am talking about instead of just a description.

...
BTW the problem with the said chip is NOT that it goes into discontinuous mode, that's entirely different, it's that it skips switch cycles because of the LC resonance. The LC "plant" reacts as it is driven, it is not driven to a predetermined output by a predetermined input because it has it's own idea of what the output voltage should be based on the energy that it took in. There's no way to stop it from reacting and get high efficiency at the same time without sync'ing the resonance to the LC natural frequency, which is not reliable.
If the above doesnt make any sense or seems strange, i'll post some wave forms which would show exactly what is happening and that clearly it is not entering a discontinuous mode of operation but is another 'mode' altogether not seen in more conventional (pure analog control PWM) switchers.

I understand what you said, and have seen the same instability at times. The important thing to stabilise a 34063 is to choose the right timed period, (ie; the right C value for its RC). The cap is on the TC pin for the internal oscillator, and sets the fixed ON period. If the cap is too small for the inductor value, the osc freq is too high and it will make 2 or 3 ON pulses before regulation occurs, then an abnormally large gap after the "set" of pulses.

What I do is tune them backwards. If I know the inductor will work well at 20kHz at 1-2A, and it is a 12->5v converter at say 85% efficiency it needs about a 49% duty cycle; (5 / (12 * 0.85))

So then set the TC cap to a value to give 49% of 20kHz period, or 24.5uS; (1/20000 * 0.49). Generally I just fiddle with inserting cap values and pick the one larger in size if it is not exact. Obviously looking at the fixed ON period with the 'scope.

Now because the TC oscillator is running slowly, you get more current in the inductor with the longer fixed ON period, and a longer natural buck (OFF period) so it should stabilise quite well oscillating around the Vref regulation (or at least a lot better stability than using a cap which is too small!).

...
LATER:
I found an interesting trick to help fix the situation above. It stems from the theory of boost converters where the output becomes unstable when the filter capacitance ESR becomes lower than a certain value. It's the one time when we dont want too low of an ESR for a DC power supply filter cap. Well the trick is to increase the ESR on this converter too, and that helps stabilize the output significantly.

That's exactly to be expected as adding resistance to the output cap increases the amplitude of the voltage ripple, making it easier for the oscillator to stbailise even if the oscillator is running too fast. Using the larger TC cap (see above) does the same thing, tuning the 34063 timed-ON period with the ideal ON/OFF cycling periods of the inductor at your main load current.

The 34063 is an older part normally used with an electro output cap, like you said with a higher cap ESR. A 470uF electro (like in the datssheet) should normally have 0.2 ohms ESR maybe more at SMPS frequencies. Anyway like I said using a larger TC cap to slow the osc freq will increase voltage ripple at the inductor cap junction and stabilise oscillation.

 

[MrAl]

Hi again MrRB,


Yes i'd like to see some of your inductors and what inductance they have and if they are hand wound how the inductance was tested, what core material and mu and how many turns.
They do make good inductors like we are talking about, but they are quite expensive. If you look on Digikey you'll see what i mean.

 

[Mr RB]

I just looked on Digikey, they don't seem to sell powdered iron toroid cores! Maybe their business is based more on people who want ready made parts (manufacturers etc) who don't want to spend 5 minutes winding some turns. Their ready made toroids don't impress me much.

Jaycar in Australia still stocks powdered iron cores;
**broken link removed**
(that size would be ideal although I don't know their core material)

I'll see if I can take a photo later today and post it up.

 

[Mr RB]

I dumped out my box of iron toroid cores, here's the photo. There are a few commercial wound ones in there too but generally I keep them in another box, these ones are for winding and experimentation.

If doing a "proper" project (ie for production), I buy in a range of core materials in the right size, and experiment with a prototype. I haven't done magnetics calcs since the classroom about 30 years ago as I find it better to work on the real inductor under load on the testbench and 'scope it, calcs only give a ballpark as to core size and material and I already have that from experience.

I've also got a similar quantity of ferrite toroids and potcores but prefer iron types when the goal is very high efficiencies due to the benefits at low freq.

(edit; forum attachments are broken so I put the photo up manually

**broken link removed**

I'm thnking of making a SMPS automatic SLA battery charger to suit a 6v 10Ah battery for a robot project, i'll probably use a 34063 for the buck SMPS and a PIC to monitor charging voltage/current.

 

[MrAl]

Hi MrRB,

Hey nice pic there, thanks for posting that.

Well i know i have had pretty good success with off the shelf parts when the inductance is lower, like 100uH or so. That seems a lot easier doesnt it?
But you are getting me curious as to how i can get this to work with the 34063 part too. I'll have to put some more effort into it i guess.

Do you have any reference designs using say 1mH and up to 2 amps output at lowish voltage like 5v or 10v etc.?

 

[Mr RB]

The off the shelf parts are a bit high in DC resistance and less than perfect core material (ie they saturate). The last commercial design I did was at 10-15A continuous output and it was much easier to wind cores to get the high perf spec.

But at your 1-2A the commercial ones are likely to be ok as that is a pretty easy spec.

If I get some time between family visits etc I'll throw together something for this 6v robot battery charger I wanted to build, which will be about 7.2v output at 1.5-2A, if I can find a decent PFET in my junkbox it should be easy enough to get >90% eff from a 34063 even with one of those commercial 100-220uH type toroids.

 

[MrAl]

Hi MrRB,


Well i did see some nice ones, very low ESR, high inductance, but they were quite expensive ($20 USD or more, one was $75 USD). That's why i was so curious to see what you could find. The lower current models (the $20 type were higher current) all had quite high ESR.

But i am still curious as to what mu the core material had that you were talking about and what size the cores were. If i could get the technical information i could runs some tests using my own software. I wrote some software a while back for designing inductors from scratch, but i need the core material data and the possible size (dimensions).
What i found in the past by calculation and by testing was that the smaller cores saturate too easy and the higher the mu the easier they saturate. Also, with a certain number of turns i might get the required inductance but then it saturates so i have to either use a gap and then increase the number of turns or else go to a lower mu core and they again add more turns. So either way it seems to increase the number of turns which raises the series resistance. Maybe a big enough core where it would not saturate as easily might work, so knowing the dimensions could help here.

So i guess what i am asking is if you have an inductor you wound yourself that is 1mH and can handle say 3 amps i need to know what mu the material is and what the dimensions are. That way i can go from there. Either that or the manufacturer if they publish the data sheet for the core and/or the material.

Here's an example of what i am seeing on the web (all 1mH):
$10.67, 0.212 ohms, 3.1 amps, 1.2x1x1.4 inches
$24.99, 0.120 ohms, 5 amps, 1.6x1.6x1.2 inches, weight 0.32 pounds
$18.06, 0.038 ohms, 10 amps, 3.2x1.8x2.0 inches, weight 1 pound

notice the low ohms model is 18 USD and is quite large really.

 

[Mr RB]

...
What i found in the past by calculation and by testing was that the smaller cores saturate too easy and the higher the mu the easier they saturate. Also, with a certain number of turns i might get the required inductance but then it saturates so i have to either use a gap and then increase the number of turns or else go to a lower mu core and they again add more turns. So either way it seems to increase the number of turns which raises the series resistance. Maybe a big enough core where it would not saturate as easily might work, so knowing the dimensions could help here.
...

Yeah it's a bit of a "black art" for sure. Sorry I can't offer much info re core materials, I prefer powdered iron and generally pick a core material (or a few samples) from near the middle of the available range that the manufacturer has.

...
So i guess what i am asking is if you have an inductor you wound yourself that is 1mH and can handle say 3 amps i need to know what mu the material is and what the dimensions are.
...

I was probably a bit over enthusiastic re 1mH, at that point you were showing a lot of concern re the output voltage ripple so I went for a high value. A 220-470uH part might be better suited based on tests I did today.

As you wanted off the shelf item I threw one in a 34063 circuit and wrote up the results, please check the project I just posted here;
https://www.electro-tech-online.com...cy-smps-buck-converter-using-34063-ic.132064/

The inductor was an off the shelf item that i have unfortunately lost the paperwork for, but it was from Altronics Australia hobby catalogue and is (I believe) their 3A model in somewhere near 220-470uH range.

I just did the tests and wrote it all up, and have not looked up the calculation yet but if you like I think you can find the uH value from these specs I just measured;
applied voltage across L1 = 7.35v
current rise = 0 to 1.19A in 31uS

L1 overall size is 24mm diameter, and it has 51 turns of 1.0mm diameter wire. Photos are in the link above. I tested a fewe off the shelf 3A inductors, that one had the most turns so it's likely 330uH or 470uH official value.

Note the Altronics cataloge lists these inductors as 100-200 milliohms, this one I tested at exactly 1A DC and it dropped 32mV so it is an actual 32 milliohms DC resistance. So much for listed specs, now you know why I build and test and rarely do calcs on inductors.

 

[MrAl]

Hi MrRB,

I took a look at the circuit and it looks promising so i'll give it a shot. Might have to order a couple parts first though.