Help with step down 48V to 12V!!

[gumattos]

Have you implemented all changes as per my last circuit? There are some differences with the circuit and values to operate with 48 V supply. Look carefully at value of R1 and connection point of R7.

E-Design, you guessed right, I forgot the change in the connection of R7, when I made the changes, something pretty interesting happened, and I think we are close. After making the changes on the Resistor from 10k to 100k, adding the 100 Ohm after the 22nF capacitor and changing the connection point of R7, the circuit apparently works for a few seconds. In the first maybe 10 secondsI can hear the inductor making a noise (probably because of the switching), and the NPN transistor don't get hot, and the Vout is 12V. However after some seconds, the nois stops, the voltage drops from 12 to 10,9 approx. and the NPN transistor starts to get hot again. Nothing is damaged apparently because I repeated the test several times and it behaves the same always.

Details: I am using E-Design's latest version of the circuit, with the two series Zener and one Zener on the collector of the NPN.

Now I am sure the circuit is identical to the last circuit that E-Design made.

UPDATE: The test above was using approx. 100mA, when I tried with the 0,5A load the result was an oscillating system, very lound noise, but this time the MOSFET got extremely hot, and the multimeter can't show a regulated output, it gets crazy with a voltage that vary between 38V and 2V, nothing realiable.

 

[E-design]

What is the value of your inductor?
Do you have C4 connected? It will affect the start-up. It may even cause some instability if not connected.

 

[gumattos]

What is the value of your inductor?
Do you have C4 connected? It will affect the start-up. It may even cause some instability if not connected.

Yep, C4 is there, 470nF, and the inductor I am using is 220uH.


UPDATE: E-Design, if you think that different components will work, I can buy them, it will take some time to arrive but if you think the test is worth it, no problem. I think it's unlikely I did some connection wrong, because the circuit works for some seconds then stop oscillating, I don't think bad connections would do this, but if you think it does, I can remake the circuit too.

UPDATE2: Added an input capacitor, 50V 470uF, no changes, while the inductor is making noise, 12V NPN OK, when the sound stops after some seconds, NPN transistor gets hot.

I am annexing a photo of the circuit, you can see the inductor inductance, the 470nF is the red one, the green is the 1n.

 

[E-design]

You can try to make C2 maybe a 2n2 and see if it helps sustain switching operation. You don't want this too large as it may force the frequency overly low.

 

[gumattos]

You can try to make C2 maybe a 2n2 and see if it helps sustain switching operation. You don't want this too large as it may force the frequency overly low.

I placed two 1n in parallel, but the symptom changed. Now without load the circuit is stable at 12V, but put any load and it gets crazy, you can hear clicks every half a second maybe and crazy voltage, after some time it stabilizes at 11,5V and stops the clicks, then the NPN transistor starts to get hot. I will make a new board with new components and two 1nF caps in parallel to see if it changes comething now. E-Design, I am searching for a 48V to 12V converter that has a cost close to this circuit (or lower) and have more than 80% efficiency, if you have any alternative circuit to suggest, before creating this thread I searched a lot and coundn't fit a circuit to my case, this one, thanks to all of you helping, is the one I have more hope, so I will keep trying here.

 

[gumattos]

I was thinking, is it possible to use darlington to make the transistors have more gain or comparators in any way? If not, maybe not making a oscillator to force the circuit to not stagnate. Could one of these ideas work and be efficient?

 

[E-design]

When you take it back to 24 V operation without the extra zener and using a 12 V reference zener, is it operating stable?

 

[gumattos]

When you take it back to 24 V operation without the extra zener and using a 12 V reference zener, is it operating stable?

Yes, it worked with 86% eff. with an 0,1A load, didn't tested with 0,5A but the result should be better since the frequency will rise. However, with the 24V input, the inductor still make a noise, so I know it's oscillating.

The problem is with the 48V input, somehow the higher voltage is making the system stagnate (I guess). The 100 Ohm resistor between the collectors of the transistors make it oscillate slower (at least in the simulation), but I don't know if it helps stagnation or not.

One thing that I still can't understand is why the circuit oscillate for about 7 or 8 seconds before stopping and the transistor starts heating up. 7 seconds is way too much time for this type of circuit, what is taking that time to change ta makes the circuit stagnate? That's my main doubt.

I will try to get an oscilloscope, so I can analyse the circuit, but only Monday I guess. Until that, we can continue trying stuff. I have some LM741 and some LM393 here, in case anybody thinks os anything using them.

 

[BradtheRad]

I can't claim my input leads to a better solution than the help you're getting from E-Design, but here for comparison is a buck converter with snap action as it comes from a single logic gate. The feedback loop responds to voltage. Operation is easy to grasp.



A capacitor is at input of the logic gate. This delays feedback, bringing hysteresis into the feedback response. (A resistor is necessary, or else the resistance in the potentiometer.)

The logic gate can be replaced by an op amp. You can give it a power supply by dropping 48V via resistive divider, or zener diode. The NPN does not need more than a volt or two to drive it.

I've been playing with a hardware version (low power, light load). It's not ideal since I usually only get oscillations within a narrow range of supply voltage and/or load. It may be due to characteristics in the inductors I've been testing it with.

 

[gumattos]

I can't claim my input leads to a better solution than the help you're getting from E-Design, but here for comparison is a buck converter with snap action as it comes from a single logic gate. The feedback loop responds to voltage. Operation is easy to grasp.



A capacitor is at input of the logic gate. This delays feedback, bringing hysteresis into the feedback response. (A resistor is necessary, or else the resistance in the potentiometer.)

The logic gate can be replaced by an op amp. You can give it a power supply by dropping 48V via resistive divider, or zener diode. The NPN does not need more than a volt or two to drive it.

I've been playing with a hardware version (low power, light load). It's not ideal since I usually only get oscillations within a narrow range of supply voltage and/or load. It may be due to characteristics in the inductors I've been testing it with.

Interesting, I can change the NPN for a P Channel MOSFET for better switching can't I? I will make some simulations and I hope it works, or I will have to use darlington maybe, since I will need to put up to 1A load. Thank you BradtheRad, I will try this way and see if I get any good results!

 

[gumattos]

So, I did these two circuits, none of them worked on the simulation, all the voltages probes showed constant voltages and far from the result I was expecting, what you guys think?

Project 1 is the circuit BradtheRad suggested, but with an OpAmp and the other circuit is one I saw on the internet somewhere sometime ago. Any chances these circuits will work?

 

[E-design]

You can try other switching regulator circuits, and many of them may work, but achieving high efficiency is another question.

In the LM7812 data sheet, you can find a simple circuit to use the regulator as a switching regulator. The switching frequency won't be very high. The turn-on overshoot may be problematic as well.

The circuit I originally suggested worked very well on 24 V at the time, I tested it, with high efficiency. I would like to test it on 48 V, but unfortunately I don't have that test circuit any more (or rather I can't find it).

 

[BradtheRad]

Here is the op amp version I've played with. Voltage regulated by means of hysteresis. The op amp has positive feedback to the non-inverting input (which is the common way to generate hysteresis). It should be more predictable and easier to adjust, compared to the capacitor on an input.

Interesting, I can change the NPN for a P Channel MOSFET for better switching can't I?

The N & P arrangement makes it possible for a low voltage device to switch a high voltage. However the entire circuit can be moved from the low side to the high side (if you wish to take output which is referenced to the positive rail). Then you can switch it with just one Nmos referenced to ground.

 

[gumattos]

Here is the op amp version I've played with. Voltage regulated by means of hysteresis. The op amp has positive feedback to the non-inverting input (which is the common way to generate hysteresis). It should be more predictable and easier to adjust, compared to the capacitor on an input.





The N & P arrangement makes it possible for a low voltage device to switch a high voltage. However the entire circuit can be moved from the low side to the high side (if you wish to take output which is referenced to the positive rail). Then you can switch it with just one Nmos referenced to ground.

I mistakenly said about changing the NPN transistor, I meant changing the PNP transistor for a P Channel MOSFET to make the switching better, I would work right? And the reason I used a voltage divisor was beacuse the 48V would be too much for the OpAmp so I divided the output voltage just in case it reaches 48V in some moment and damage the OpAmp, I supose that using the voltage divisor nothing would change I guess.

 

[gumattos]

Here is the op amp version I've played with. Voltage regulated by means of hysteresis. The op amp has positive feedback to the non-inverting input (which is the common way to generate hysteresis). It should be more predictable and easier to adjust, compared to the capacitor on an input.





The N & P arrangement makes it possible for a low voltage device to switch a high voltage. However the entire circuit can be moved from the low side to the high side (if you wish to take output which is referenced to the positive rail). Then you can switch it with just one Nmos referenced to ground.

So, I switched the PNP transistor for an P Channel MOSFET, now I can hear the switching noise but it is a very high-pitched sound, however with a 0,6A load I got 0,6A on the input, so my efficiency is something around 25%. This is the actual circuit I am using. I will try now to change some numbers see if I get any better results, anybody have any idea?



Another thing, E-Design, the first circuit you suggested for the first time, you think with faster and with higher gain transistors it may work?

UPDATE 1: Changed the voltage divider on the MOSFET gate from 10K/20K to 1K/2K and now I am getting 0,5A on load and 0,2A on input, so I now have 62% efficiency. The 25% eff. was because of a bad connection, after it was about 40%, and now 62%. I will continue with the changes and I am open to new suggestions, the sound stopped too so it might be in a much higher frequency now.

 

[gumattos]

I couldn't edit my last post, I don't know why, but I made some changes and got new good efficiency, 48V and 0,15A on the input and 11,8V 0,5A on output, 82% efficiency approx.. Can be better I guess, changing for a comparator and using a better transistor as pull-down (I only have one 2N2222 that's why I used the BD137). I will keep on making changes.

 

[Akanimo]

gumattos,

I understand that you may not have much time on your side but if you have to be using converter of different specs often I recommend learning to design one yourself...

If you have a little time to spend on learning flyback design from scratch, I believe it will pay off.
There is not much you can achieve with having to change components in a trial-and-error fashion.

 

[E-design]

I couldn't edit my last post, I don't know why, but I made some changes and got new good efficiency

Can you show your simulation waveforms of the gate signal, input and load output current?
I marked the points of interest on your schematic.

 

[gumattos]

Can you show your simulation waveforms of the gate signal, input and load output current?
I marked the points of interest on your schematic.

Yes I can, but I don't think you will find what you are searching for. I downloaded TINA-TI and some simulations worked, this especific one I don't think so, maybe the waveforms are fine but not the frequency, I will send the pictures so you can understand what I am talkling about. The simulation shows some bad outputs but when I tested in real life, I get 83% eff.. I think I will get an Oscilloscope tomorrow to see how things are working.

Voltage on gate:

As you can see here, in one entire milissecond only one oscillation, but I don't think that's what's really happening, you will see more strange behaviour soon.

OpAmp out:

As you can see, the output is no even close to the maximum ou minimum of the AmpOp, and since it's working as a comparator, this should be different.

Voltage on AM1:


Voltage on load:

As you can see here, the output is not even close to 12V, however that's exactly what I get when I tested the circuit.

Current on AM1:


Current on AM2:


Probably I am doing something wrong with my simulations, maybe some configuration. The circuit makes a lot of sense in my head and it is working on real life, and with the best efficiency and stability of all. Tomorrow I will place a comparator to replace the UA741 see if I get better results, and maybe a Cap between non-inverting and the output to get some feedback and see if I can get better results, I will try to get a Oscilloscope too and see what's really going on.

- - - Updated - - -

gumattos,

I understand that you may not have much time on your side but if you have to be using converter of different specs often I recommend learning to design one yourself...

If you have a little time to spend on learning flyback design from scratch, I believe it will pay off.
There is not much you can achieve with having to change components in a trial-and-error fashion.

Akanimo, I guess you are right, actually this last circuit is the one I most studied about and have solid ideas of how to get it better, but I have to admit that the others I could only understand, but not knew how to adapt. You are right with the problem of time, but I will get some deep knowledge of how to make step-down converters as soon as I get everything working!

 

[E-design]

Well, I have no idea why you could not get the circuits I posted to work.

I tested the exact circuit below on a plug-in breadboard which is not optimum, and it worked the first time. I changed a few values to maximize the efficiency at 500 mA out and had it running for more than an hour. The only thing that slightly heated up was the 470 uH choke.

The Simulator predicted an efficiency of 87.65 % and I measured it at 86.1 %.
I suspect that the difference is being consumed in the choke. I will try a larger current choke later and see if it improves things.

The switching waveform on the drain of Q1 is nice with rise and fall times under 50 nS.

I will post actual scope plots a bit later.