• Welcome to our site! Electro Tech is an online community (with over 170,000 members) who enjoy talking about and building electronic circuits, projects and gadgets. To participate you need to register. Registration is free. Click here to register now.

Rebuilding My Power Supply

Status
Not open for further replies.

MrAl

Well-Known Member
Most Helpful Member
I don't remember if NPN worked or not, I'll try to test it tomorrow, I'm building a negative regulator. It's interesting how easy it is to get the negative voltage, same circuit, just move the negative feed, (ground) to the point between the inductor and the capacitor. A problem I have is the lack of higher voltage power supplies. I do have a Variac and I guess I could put a bridge on that. But that has no isolation. I got to see (hear) a 22uF 100V cap fail today, haven't found the canister and there was paper everywhere. I kind of think it got hot because of it's proximity to the 100 ohm resistors.
I just simmed the circuit with a NPN and it does run, (last attachment)
Since I'm always hands on with my power supply usage and it does have definite limit, I don't think I'm going to worry to much about current limits, If I find a need for it, I'll just use a LM317

Yes be very careful for sure. See if the PMOSFET gets hot too. But then again for much higher DC buss level you'll need a higher voltage PMOSFET too. Hope you have one around.
I like to stand back several feet when i do tests like this.
 

kinarfi

Well-Known Member
When the solder that is holding the wire to the FET lead melts and lets the wire disconnect, THAT's bad, isn't it! Also means the FET is probably DEAD. OH well, I got plenty of that FET, any way, I watched what was happening with the LM2576 and changed my circuit and now it seem to be running well. One strange thing that I haven't figured out is that with THIS dual voltage power supply, if I run from the Neg out to the Pos out and bypass the common zero out, it causes the FET on the Pos regulator to get hot. I only tries it once, without melting solder, but that's what I was doing when I melted the solder. The negative side seems to be work real well, I'm very happy with it and satisfied with the positive. I have some binding posts coming and still debating what type of current monitor to use, current sense resistor or LEM Current Transducer HX 03..50-P/SP2 that uses hall effect.n It's all just for fun any way. More photos at https://www.dropbox.com/sh/wfpkr5w0hcrz7a2/RoElm-fUVF
 

Attachments

MrAl

Well-Known Member
Most Helpful Member
Hi,

Well for one thing the resistors connected to the IRF4905 seem a little too high for driving a mosfet in a switching regulator. Typically we want very low impedance drive and that's why we end up using a special mosfet driver chip sometimes. If you lower those values (provided that is possible with the driver for those) the PMOSFET should switch faster and so less heat. If any of the mosfets switch too slow they can cause lots of heat which not only wastes energy but makes them overheat and burn out.
You could check the rise and fall times at the gate (very carefully) and see what those look like.
The power loss during switching is due to the rise and fall times of the gate and thus the source/drain. The output then looks like a ramp which consumes energy and makes the mosfet heat up more. When the mosfet is fully turned on or fully turned off there is little power lost, but during the switching transient there is power lost. There is one switch transient to turn on and one switch transient to turn off, so there are two important times to consider for each switching cycle.
It helps to know how much of the total cycle time is taken up by the switching transient times so we can express that as a percentage of the total cycle time. For example 10 percent is not as good as 5 percent because more power is lost in the MOSFET.

An estimate for the switching power dissipated in a transistor that switches on and off many times per second is:
P=Imax*Vmax*p/3

where
Imax is the peak current,
Vmax is the max voltage (usually the power supply voltage), and
p is the percentage of the switch time relative to one complete cycle,
assuming the turn on and turn off times are roughly the same.

So for example if Imax=20 amps and Vmax=24 volts and it takes 10 percent of the total cycle time to switch on or switch off completely, then the power due to the switching is:
P=20*24*0.1/3=16 watts.

For the LM2576 which switches at around 50000Hz, that would mean each switch transient would take 2us and the total period is 20us, and 2/20=0.1 even though the 2us occurs twice per cycle (that's just the way the formula is formulated).
So to be clear, that means 2us rise time and 2us fall time and a total period of 20us for that example above which means the transistor would have to have a decent size heat sink.

Of course reducing the rise and fall times means reducing the power, so reducing to 1us means half that power or just 8 watts, and 500ns means just 4 watts, etc.

This is also the maximum power and in a buck circuit it could easily be less depending on the output voltage setting too. But for any given setting the power does reduce by half when reducing the rise and fall time by half.

Note that a small package without a heat sink may only be able to handle 0.75 watt without getting pretty hot. That means we'd have to use a heat sink or reduce the switching times or both.

Resistors as high as 1k can easily cause rise and fall times to be as high as 1us or more, so gate drive resistor values that high could cause overheating problems.
 
Last edited:

kinarfi

Well-Known Member
Speaking of heat sinks, what I have is good, it's the heat transfer medium that I think I had a problem with, I was using that grey silicone sheet and a nylon screw isolator with a washer on top and it melted the isolator so the FET wasn't tight against the heat sink, so I tried a different type of hold down and it melted too, so I went to the old mica type and a fiber insulator with heat shrink around the screw and so far so good.
Played around with my positive voltage, you're right R collector to gate was too high, ended up taking it down to 200 and the FET stayed cool after that. That's where I'll leave it, now for the amp meter circuits. I did see that if I jumpered from collector to gate, my output voltage under load was only 11.8 +- volts.
Deleted a bunch of photos and added scope screen shots with notes.
https://www.dropbox.com/sh/wfpkr5w0hcrz7a2/RoElm-fUVF
 

MrAl

Well-Known Member
Most Helpful Member
Hi,

Some nice pics there. I can see the last pic looks like the waveforms are much more square with the 200 ohm resistor. That reduces the power consumption in the MOSFET quite a bit. The original had very poor rise and fall times so that's a vast improvement.

I also see a pic with what looks like some oscillation. I have to wonder where that came from, how that happened. We definitely want to avoid that.

MOSFET driver ic's typically can deliver a 1 amp or more pulse to the gate which turns it on or off very very quickly without wasting too much power otherwise. Low values resistors use more power but if that's not an issue then i guess it doesnt matter that much. For a bench supply i would do the same thing if i had to.

I'd like to try some of these circuits myself in real life but not sure i will get a chance too do so very soon from now. Other more pressing issues right now.
One of the reasons i wanted to try the LM2576 mods with the MOSFET is so that i could get the efficiency up much higher for a high efficiency buck circuit to convert 12v to 5v. With the present internal bipolar the efficiency using the chip alone isnt that good even with just 2 amps output (maybe around 80 percent if lucky). I'd like to get up to 90 percent or better for battery operated equipment. Since the MOSFET can be made to turn on and off so fast and with little voltage drop, i think this can be done using one fo the mods like with the PNP or the NPN or similar.
A second application would be an offline 120vac to 14vdc converter, where the MOSFET would have to handle 170vdc and step that down to around 14vdc. But this would be higher current like 20 or 30 amps too. Using an inductor rather than transformer, this would be non isolated. Using a higher frequency transformer and opto coupler feedback it would become isolated.
 
Last edited:

kinarfi

Well-Known Member
I'm finished, for now anyway, the final screen shots were taken of 10 amp from Neg to Pos with no common, IRF4805 stayed cool, I could keep my finger on it after a minute's running, heat sink was cool also. My amp meters are a bit disappointing, as the vary and not real accurate, but I'm very happy with the rest of it. Especially because it stays cool as opposed to the heat it put out before just driving some LED light that I use on my bench, one has a LT1270 driving 4 10 watt LEDs in series, ~40 volts @ 1 amp, and the other is the 4 3 watt LEDs in series with NO limiting resistor that everyone says won't work, but it does, nicely!! (I'll add a photo of it tomorrow)
https://www.dropbox.com/sh/wfpkr5w0hcrz7a2/RoElm-fUVF
 

Attachments

Last edited:

avrilcory

New Member
I have a variable power supply that had some problems, so i'm rebuilding it about the transformer is good for over 20 amps and 24 volts DC power supply. If you need to buy a power backup supply, i would just get a R9 270x, keep the parts already listed then add an SSD along with it if you can get a good deal on the rest of your parts.

Thanks for sharing with us.


Best Regards
Riello-ups
 

MrAl

Well-Known Member
Most Helpful Member
I'm finished, for now anyway, the final screen shots were taken of 10 amp from Neg to Pos with no common, IRF4805 stayed cool, I could keep my finger on it after a minute's running, heat sink was cool also. My amp meters are a bit disappointing, as the vary and not real accurate, but I'm very happy with the rest of it. Especially because it stays cool as opposed to the heat it put out before just driving some LED light that I use on my bench, one has a LT1270 driving 4 10 watt LEDs in series, ~40 volts @ 1 amp, and the other is the 4 3 watt LEDs in series with NO limiting resistor that everyone says won't work, but it does, nicely!! (I'll add a photo of it tomorrow)
https://www.dropbox.com/sh/wfpkr5w0hcrz7a2/RoElm-fUVF

Hi,

What is that oscillation picture about, is that the circuit too or a past circuit?

Also, it is not that an LED can not be driven from a voltage source. Of course i can connect an LED to a power supply, turn up the voltage slowly, and watch the LED current and see it light up nice and bright at 20ma if it is a small LED or 300ma if it is a high power LED. It's not that we can never do that.
What is is about is that if we leave the power supply set at the 3.5v (or whatever it comes out to for that LED) the LED characteristics may change over time with aging. This could also change while the LED is heating up. The LED's i have measured have a 2 to 3mv per degree C change in voltage with nearly constant current drive, so once they heat up they need a different voltage if they are to operated at the same current level. If the power supply output impedance is low enough, this means the LED either gets too much current or too little current after a while which makes the output dim somewhat. If it does not dim too much then we're ok but if not then we might not be happy with it.
Even a small series resistance can help with this that's why we like to do that too.

So it's not that it can never work like that, it's just that it is not as good as if it is done in a more reliable and consistent way.
 

kinarfi

Well-Known Member
I have a variable power supply that had some problems, so i'm rebuilding it about the transformer is good for over 20 amps and 24 volts DC power supply. If you need to buy a power backup supply, i would just get a R9 270x, keep the parts already listed then add an SSD along with it if you can get a good deal on the rest of your parts.

Thanks for sharing with us.


Best Regards
Riello-ups
You're welcome, it was My pleasure,
Kinarfi
 

kinarfi

Well-Known Member
Hi,

What is that oscillation picture about, is that the circuit too or a past circuit?

Also, it is not that an LED can not be driven from a voltage source. Of course i can connect an LED to a power supply, turn up the voltage slowly, and watch the LED current and see it light up nice and bright at 20ma if it is a small LED or 300ma if it is a high power LED. It's not that we can never do that.
What is is about is that if we leave the power supply set at the 3.5v (or whatever it comes out to for that LED) the LED characteristics may change over time with aging. This could also change while the LED is heating up. The LED's i have measured have a 2 to 3mv per degree C change in voltage with nearly constant current drive, so once they heat up they need a different voltage if they are to operated at the same current level. If the power supply output impedance is low enough, this means the LED either gets too much current or too little current after a while which makes the output dim somewhat. If it does not dim too much then we're ok but if not then we might not be happy with it.
Even a small series resistance can help with this that's why we like to do that too.

So it's not that it can never work like that, it's just that it is not as good as if it is done in a more reliable and consistent way.
https://www.dropbox.com/sh/oir6ah6upe9jdyl/KuFYU3XU5E
As I was preparing to post this, I noticed that I had a resistor in the first photo, so I had to check, it is in series with the fan, look at the name on each photo, it has the volts and current, if applicable, I have noticed lately that the LEDs seem to be dimmer and require more current than when new, Maybe it's the ambient light or the fact that they have been over driven on multiple occasions, but the ones on my OHV are still going strong. I still think that there are times that a series resistor would cause problems, to add any resistance at all would require the removal of one LED to create the head for resistor to work in. In the case of the 6 strings of 4 on each light on my OHV, I would have to change it to 8 strings of 3 & a resistor. There are just times when the voltage is right and LEDs are right that there is not need or place for the resistor.
 
Last edited:

kinarfi

Well-Known Member
What am I doing wrong with my instrument op amps. I down loaded the zip file for INA122 files from TI, unzipped it and then copied the ina122.lib and pasted it into the ina.lib so it would show up in the [Opamp Add] list in spice. The bad thing is that it's actions in spice are about the same as I get for real. To me, what I have is just like the data sheet shows in fig. 6 page 9. With the setting I have, one amp through the current sense resistor, I expect 1 volt out, not a - .52v and it doesn't change as the current is changed.
Thanks,
Kinarfi
 

Attachments

MrAl

Well-Known Member
Most Helpful Member
Hi,

If you have everything hooked up right then it must be a problem with the model. The model is a little complicated so it's not easy to spot anything wrong or not compatible with LT Spice.

If you have a real device at home you could try this on the bench and see what happens. If it works better, we could build a simpler model out of regular op amps or an ideal model just to get it working in spice.
To make your own model, try using the internal diagram with two low input offset op amps as they show on the data sheet. Use the same value resistors.
If that works then there is something wrong with the model in the zip file.
From what i read it should give you a gain of around 200. That means with 1 amp through 5mOhms is 5mv and that times 200 is 1.000 volts which i think is what you were looking for and that does sound reasonable.

I guess i will have to try to add that model to my LT Spice too and see what happens. I'll try to do that today sometime.
 

kinarfi

Well-Known Member
Maybe this is my spice problem, PSpice model in LTSpice, not sure how compatible they are.
 

Attachments

MrAl

Well-Known Member
Most Helpful Member
Hi,

Yes maybe that is the problem. I went over the data sheet and i see the offset can be as high as 250uv but that would not cause such a large error like you are seeing.

Are you sure you got the pin assignments all correct?

I tried to enter the spice model into MicroCap simulator and it messed up all the models. Something must be wrong with the file or it is not compatible. Not sure what is causing this.
 
Last edited:

kinarfi

Well-Known Member
I have be trying to may a 200 gain amp all day, I used INA122s, INA128, TLC2274, LM324, and failed miserably with all. Nothing seemed to work. What I am doing is trying to use the .005 ohm (how do you use the symbols at the top of the page?) I bought to get the amperage, one end is grounded and that seems to mess with every thing, so if you have a schematic for that, please let me know.
 

MrAl

Well-Known Member
Most Helpful Member
Hi,

You should definitely be able to get this to work with an LM324 but with that op amp you'll have to modify the spice file to set the offset to zero or much lower than it may appear in the downloaded file. That's because the LM324 spice model probably has a 2mv offset which is WAYYY too much for your application that needs to sense 5mv.
But the circuit shown on the INA122 data sheet will work if those 25k and 100k resistors are added to two op amps as shown (somewhere on the data sheet).
If the LM324 spice file is not changed however then it may look like there is a huge error in the measurement. Alternately find a very low input offset op amp in the LT Spice files.
I could post a circuit but it would be the very same one on the data sheet for the INA122.

If all else fails you could use a voltage controlled voltage source with a gain of 200 and assume that the INA122 device works properly in real life. That will at least get you going. That wont simulate frequency response though.
 

kinarfi

Well-Known Member
Earlier you said something about oscillations, I think it's just a spice error, rock solid in real life. Had my first failure today, voltage was turned up to near max, + side, and the leads shorted momentarily and the 4905 died shorted. I also think part of my problem is the use of TLC2274, it's only rated for +- 8v or 16. Found a LM358 (half of LM324) and fed it with + - 13 derived from zeners and had my best results from that. Occasionally I pull more current than can be indicated by the just under 13 volt supply, so I'm thinking to unbalance the feed for the LM358 to 24+ & 5- . Any idea what I should expect? spice says OK
 

MrAl

Well-Known Member
Most Helpful Member
Hi,

It's hard to answer your question because i am not entirely sure what you are doing with the LM358. Are you just using one section or did you connect the two sections as the INA122 device shows? If you can make a quick drawing, all the better.

It is possible to use the LM358 alone but then we have to find a way to adjust for the offset voltage when we have to measure down as low as 5mv because of the possible 2mv offset of the op amp itself. There might be a trick we can use if you want to try it. Some op amps have an offset adjust built in so we can use an external pot to adjust offset.
 

kinarfi

Well-Known Member
Spent the first part the day getting the amp meter running, NOT exactly linear or real accurate, but good indicators. Then I was testing a transistor and shorted the leads again, pretty sure I lost another IRF4905, so I'll add a few more pieces to the mix tomorrow to protect the positive half of the circuit, the negative side has it's own short circuit built into the LT1270A. I added a diode, an SCR and an NFET. When the current gets too high and the voltage across the current sense resistor (R19) gets high enough, it trips the SCR (U13) taking the gate of the NFET (M1) to ground turning it off which stops the PFET. I also penciled in a reset switch and an indicator light.

It's a good thing I screw up a lot and have to do thing two or three or four or more times or I might run out of things to do. :)
Spice says trip voltage is
683.634mV no diode equivalent to 136.726 amps through sense resistor
1.08622V with diode equivalent t 217.244 amp through sense resistor
Think I'll try it with out.
 

Attachments

Last edited:

large_ghostman

Well-Known Member
Most Helpful Member
I got lost ages ago with what your upto, But I do enjoy reading it :D 217A:nailbiting: all a little extreme for flashing LEDS:happy:, I have visions of slightly twitching components writhing around the floor at your place, legs in the air with small wisps of smoke still escaping. and the sounds of manic laughter from behind a closed door:hilarious:. You kill fets without fear,remorse or regret. I wish I was as callous, I might learn a bit quicker :woot:
 
Status
Not open for further replies.

Latest threads

EE World Online Articles

Loading
Top