Power Supply Part 3

[FreeThinker]

Ok part 3 is out and we have the basic design running on the bread board. Works well enough with plenty  of room for improvement at a later stage (MPU control is the final goal I think). It would seem like the next segment will be the PCB layout. Any bets on the package he uses? Got a strong feeling it may be DipTrace. Why has the original spec gone from 6v to 5.5v? no good reason for it at this stage so probably got some plans in the pipe line to justify it but what? (Answers on a post card to...... ) Seems to me Dave is learning fast the art of keeping them coming back for more Lol. Can't wait for the next one.

[Bored@Work]

Any bets on the package he uses? Got a strong feeling it may be DipTrace.

Altium.

Why has the original spec gone from 6v to 5.5v?

Maybe for the same reason the min spec got up from 0V: Good enough and easier to do.

[sonicj]

Bang!

[EEVblog]

Why has the original spec gone from 6v to 5.5v?

Maybe for the same reason the min spec got up from 0V: Good enough and easier to do.

It still goes down to 0V, it just need the load to do it.

Yes, I will use Altium, I haven't got the time at the moment to learn a new package.
Not that it should matter, if there is anything PCB based, then it will be as package agnostic as possible. Otherwise you aren't learning PCB design, you are learning a PCB package, and that ain't nearly as useful.

Next video is PWM control, uploading now...
Wasn't going to do that, but many people asked for it.

Dave.

[Psi]

To get down even closer to zero volts output could you not just put a diode or two in series with the output and then modify/add feedback to use the new output voltage and compensate for the diode Vdrop.

That way the output could be zero volts while the regulator is producing 1.4V (two diodes).

The feedback should take care of the difference in diode Vdrop at different currents/temperatures.

Of course, its not going to be as efficient.

[EEVblog]

Why has the original spec gone from 6v to 5.5v? no good reason for it at this stage so probably got some plans in the pipe line to justify it but what?

All will become clear once the final version rolls out...
The voltage "change" was simply a result of the 12K feedback resistor used. It isn't that important, final version might go up to 10V

Dave.

[Lightages]

Going down to 0V would be useful with very little external load. Perhaps a comparator or some other way of switching in the extra load resistor internally when getting down to 0.7V would be a useful mod.

[mstevens]

Dave,

Can you test the efficiency of the supply at rated voltage and load.

This brings me to another question probably best posted in the General forum, I know the Gossen MH Energy will do power but any ideas on an inexpensive power meter?


Thanks.

[amspire]

Dave,

Can you test the efficiency of the supply at rated voltage and load.

This brings me to another question probably best posted in the General forum, I know the Gossen MH Energy will do power but any ideas on an inexpensive power meter?


Thanks.

It is a linear supply, so efficiency is not really an issue. If you are outputting 1A at 0.1 volts, it will have an efficiency of less then 0.1%.  Efficiency is really only a concern with switching power supplies.

Richard.

[Hobgoblin612]

Dave, would a MC3303 be a suitable higher voltage op-amp to use in the power supply?

[EEVblog]

Dave, would a MC3303 be a suitable higher voltage op-amp to use in the power supply?

It'll basically work. Not exactly low offset voltage though, if you need that.

Dave.

[Hobgoblin612]

as i am controlling it with a pot and adding a voltage panel meter on the output then i don't think it will matter for the voltage control. However it may matter for the current differential amp because i am using a 0.1ohm sense resistor so i can feed the output of the opamp directly into a 200mV FSD digital panel meter. Can you suggest an alternative op-amp or the crucial specs for choosing one myself (it is so confusing).

[Nick Gammon]

Dave, thanks for much for this project, very educational!

I've built it up, hopefully correctly, however there is something I don't get.

Connected to a 8V lab supply, and with a 100 ohm load across the output ...

I am measuring the output voltage and current with two multimeters.

Dialing up 5V output using the 10-turn pot, and cranking up the current pot (so the LED doesn't light) all is well. I read 49.63 mA current which is about according to theory, bearing in mind I didn't take a huge amount of care to check all the resistor values were exactly to spec.

Now this is the funny bit ...

Gradually turning down the current pot, I get this:

48.11 mA gives 4.84V
45.04 mA gives 4.53V
41.09 mA gives 4.15V
38.93 mA gives 3.93V
37.42 mA gives 3.78V

So far so good ...

But now carefully reducing the current very slightly, I get:

0.636 mA giving 0.064V

It's like it has hit some threshold condition, where the voltage regulator doesn't turn on.

Can this be explained? I could accept "construction error" if it wasn't so well behaved up to that point.

[amspire]

But now carefully reducing the current very slightly, I get:

0.636 mA giving 0.064V

It's like it has hit some threshold condition, where the voltage regulator doesn't turn on.

If you have a single turn current pot, and it is adjusting from 0 to 1A, then it could be that the wiper is hitting the metalized ends at just below 4% of its range. Check the voltage on the current pot wiper and see if it suddenly drops.

Are you using the same opamps as Dave, or have you substituted something else. A substitute for the current control opamp may have inputs that cannot work down to the negative rail.

If it is not either of these explanations, another possibility is instability. Check for any AC voltages around the current control circuit. An oscilloscope would be better then a multimeter to check for this, if you have one. Many multimeters have lousy high frequency response.

Richard.

[Nick Gammon]

It's a 10-turn pot, a long way from its end point.

Further information is, measuring the output of the current-adjust op-amp, I get interesting results ...

With no current limiting: output 0V
With "linear" current limiting (range 37 mA to 49 mA): output 0.85V
With "cut out" mode: output: 8V

Measuring the collector of the transistor, I see oscillation (as Dave described in the video) during the linear phase, where it appears to be quickly turning on and off when the current is exceeded, so the average is the set current. Above the current limit amount the line is straight (as you would expect), and below the "cut out" point, it is straight again. It's as if when the op-amp output goes to 8V it stops the oscillation of the voltage regulator, so it stops seeking the current/voltage that would bring it equilibrium.

[Nick Gammon]

Are you using the same opamps as Dave, or have you substituted something else. A substitute for the current control opamp may have inputs that cannot work down to the negative rail.

I'm using the same op-amps as Dave. I note that at the problem point the (+) input to the op-amp is about 60 mV.

[amspire]

Interesting.

Something does not sound right.

The differential amp from the current sense resistor may not work at low output voltages.

Measure the voltages on the current limit opamp inputs and output above and below cutoff.  See if there is a point that does not seem to make sense - such as the negative input is above the positive input, and yet the opamp output is at 8V. Or see if the differential amplifier output is not right.  How well did you match the 10K resistors in the differential amp?

How are you going to provide the power to the supply? If you are using a mains transformer and rectifier, then I would really encourage you to organize a negative rail for the opamps, and to attach the current sink IC to. If the opamps have a negative rail, they will work much better near 0 volts and 0mA output and all you need is a few 1N914/1N4148 diodes (or any of the standard 1N40xx diodes are fine), a few caps and maybe a -5V regulator chip. Whatever the transformer and rectifier arrangement, you can always get an extra low current negative rail if you need it.

Richard

[Nick Gammon]

The power is just from a lab power supply set to 8V.

It is starting to look like there is some limitation (some component?) that you can't reduce the current below about 37 mA - regardless of the voltage. I suppose if there is a lower limit for voltage under light load there might be a lower limit for current.

Increasing the voltage pot to get 5.5V, the current still cuts out at about the same point (37 mA). Conversely, reducing the voltage to 4V, it still cuts out at 37 mA.

As a further test I changed the load to 50 ohms (two 100 ohms in parallel). As expected at 5V it now draws close to 100 mA. Turning down the current pot, the limiting kicks in earlier (eg. you can limit current to 70 mA) but at the 37 mA point it cuts out again.

So I'll be happy if Dave says "well of course it won't work below 40 mA for reason xxx". Then at least I know I built it OK. Then I can get on with experimenting with computer-controlled current/voltage.

My measurements of the op-amp inputs and outputs were inconclusive.

[amspire]

Can you dig up a 1.5V battery and use it to make a negative 1.5V rail. Connect the negative power rails of all the opamps and the LM334 to it. I think you will find it makes a huge difference. Personally, I think it is really important for a power supply to go to 0V, and the main thing that might potentially stop Dave's design from doing it is poor opamp performance near the negative rail.

When it is complete, what will be the power source then? A transformer, or something like a notebook power pack? Even if you just have a single voltage supply source, adding a couple of forward-biased diodes on the negative input to the 0V could provide a -1.2V negative potential that you can use for the opamp negative supplies.

Richard

[Nick Gammon]

OK, I'll do that tomorrow. I have power supplies coming out of my ears right now, so to speak.

I was sort-of planning to plug in a plugpack supply, why reinvent the wheel?

Next I am planning to connect it to an Arduino and have some sort of cool remote-controlled adjustable power supply for things like testing projects where you want precisely controlled voltages and currents (within reason). By "precisely controlled" I mean "I hope I don't blow this thing up when I test it".

I don't really mind if it doesn't go below 40 mA. After all, you don't blow that much up with that. Or maybe you do...

[Nick Gammon]

More information ...

I removed the 22uF capacitor between pin 2 (the Set pin) of the LT3080 and earth, and with that gone, could adjust the current smoothly down to almost zero (although the pin was very noisy, like as seen on Dave's video - lab power supply video part 3, around 33 minutes in).

Putting in the 0.1 capacitor instead (as he did) I could adjust the current down to about 5 mA.

My guess is that the higher capacitor value stops the high frequency switching needed to turn the voltage regulator on and off, effectively reducing the average voltage out.

So it would appear that there is a compromise between the amount of filtering, and the lowest you can set the current limiting to.

Has anyone else found this? Does it sound a likely explanation?

[EEVblog]

That is not what I see.
I checked mine and did not find this, in fact the opposite.
With a shorted load (just the current meter) and max Vout setting I could adjust the current down to 11mA with the pot at zero.
With a 100ohm load I could adjust down to 0mA no problems. The single turn pot was a bit dicky, but I could adjust for 2mA, 5mA or whatever.
Under 10mA I found the output suddenly oscillated at around 30mV max using a 22uF cap on the adjust pin.
A 47uF cap fixed the oscillation entirely.

Dave.

[Nick Gammon]

Hmm. Thanks for checking that. I tried with a 47uF cap with no noticeable difference. Is that a 2n3904 transistor there? I haven't found a circuit anywhere apart from trying to read what you hold up to the camera.

And the cap between pins 4/5 of the regulator and Gnd - is that 470 nF?

[EEVblog]

Sorry, circuit is at the lab at the moment.
I'm using a 10uF and 470nF output cap, but it should matter, 1uF low ESR minimum.

Dave.

[Nick Gammon]

Your caps look a LOT larger than mine. Is that because they are low ESR? And I suppose that is very important? If so, that probably explains what is happening.

[EEVblog]

Your caps look a LOT larger than mine. Is that because they are low ESR? And I suppose that is very important? If so, that probably explains what is happening.

They are polyprop caps. Any ceramic cap will be low esr.
Mine was stable with just an old 10uF electro.

Dave.

[justanothercanuck]

Sorry for the ancient bump, but I had a few questions about the PS design (pre-microcontroller)...  I figured I might give it a go, and see if I can tweak it to my liking...

1) Does my schematic look right?  I scraped it off a framegrab, so I'm not sure if I missed something...
2) Are all the op-amps used TLC2272's (U1, U2)?
3) How are the op-amps for current (U1B, U2B) powered?  Straight off VIN (P1)?
4) How important is the 10 milliohm resistance if I were to parallelize the LT3080?  With the resistance being that low, could I just tie all the outputs together and ignore it?

Thanks

[c4757p]

Read this thread and ditch the 3080.

[justanothercanuck]

Read this thread and ditch the 3080.

I'm not sure I follow...  All I got from that thread was that it doesn't like return power on the output...  That can be worked around. 

Care to elaborate? 

[Rerouter]

set and output cannot stray more than 1V from one another without risking permanent damage, or at least that is what the datasheet and some testing have strongly implied until the thread hears back from the company,

[justanothercanuck]

I'm still not quite sure what that means for this particular usage, but I guess I can always step up to the LT3083...  Saves me from buying 5 LT3080's anyways (I can buy two LT3083's and get 6A max output)...  $10 a pop sounds like a lot, but what can you do I suppose...