Lab Power Supply Design Part 5

[firewalker]

Dave you are going to use plain C right? gcc?

Alexander.

[Nick Gammon]

Dave said here it was going to be Arduino compatible (reply #117):

https://www.eevblog.com/forum/index.php?topic=5787.msg79946#msg79946

If he uses the Arduino IDE then it is going to be compiled with avr-gcc, which supports both C and C++.

There is no particular reason to "downgrade" to plain C.

C++ lets you use classes like the Print class to make displaying things easier (like, numbers). However the internal libraries are in a mix of C and C++.

[firewalker]

I believe that plain C is much better/beautiful/portable/cleaner/understandable (at least for engineers) than Arduino's  "C" version. As far the easy printing you could always play with stdio's FILE stream and use printf.

Alexander.

[EEVblog]

Dave, I have another question about protecting the LT3080, which you may have shown now is not necessary, but I was interested to understand it.

Rather than trying to introduce various fat diodes to shunt away any voltage applied back into the output from a charging battery or something similar, could you not introduce a forward-biased diode into the output of the LT3080,  before the VSense signal? That way, the diode drop is inside the feedback loop. Would the feedback loop performance suffer? Yes, it loses us 3/4 volt, but that would not bother me...

It may be a problem, the only way would be to try it, and I don't really wish to spend even more time doing that and refining this and that, I want to get something out the door before hell freezes over 

Dave.

[Armin_Balija]

Dave, I have another question about protecting the LT3080, which you may have shown now is not necessary, but I was interested to understand it.

Rather than trying to introduce various fat diodes to shunt away any voltage applied back into the output from a charging battery or something similar, could you not introduce a forward-biased diode into the output of the LT3080,  before the VSense signal? That way, the diode drop is inside the feedback loop. Would the feedback loop performance suffer? Yes, it loses us 3/4 volt, but that would not bother me...

It may be a problem, the only way would be to try it, and I don't really wish to spend even more time doing that and refining this and that, I want to get something out the door before hell freezes over 

Dave.

Does that mean we'll be seeing a kit?!

[shebu18]

I believe that plain C is much better/beautiful/portable/cleaner/understandable (at least for engineers) than Arduino's  "C" version. As far the easy printing you could always play with stdio's FILE stream and use printf.

Alexander.

It depends on what you know better. For me the arduino C is easier to understand and i fly over it and know what it does. C is stiil something like "what the h*** is thes,....and this,...and..."(still learning)

[JohnnyG]

Hello,

Novice question: Is there any reason this couldn't be done with through hole?

I know Dave wanted this to be done with SMD, but if I wanted to do this with through-hole/protoboard, would there be any special considerations?

Thanks!

(PS sorry if this isn't the appropriate thread... I'm pretty new here)

[shebu18]

Hello,

Novice question: Is there any reason this couldn't be done with through hole?

I know Dave wanted this to be done with SMD, but if I wanted to do this with through-hole/protoboard, would there be any special considerations?

Thanks!

(PS sorry if this isn't the appropriate thread... I'm pretty new here)

Dave wanted this to be done as a KIT so only through-hole components, as less SMD's as possible.

[Rerouter]

he made it as through hole as possible, but was forced to use 3 smd parts, 2 soic 8s and another 3 pin smd part due to availability

[EEVblog]

Novice question: Is there any reason this couldn't be done with through hole?
I know Dave wanted this to be done with SMD, but if I wanted to do this with through-hole/protoboard, would there be any special considerations?

That's back to front, I wanted it all through-hole from day 1.
Can still do it all through-hole, but some parts were just much cheaper and more available in SMD, only 3 of them.

Dave.

[LaurenceW]

Homer DOH! Scratch my earlier question about a protection diode in series with the output - of course, there is no (linear) voltage feedback from the final output. So anything between pin 3 of the LT3080 and the load is only going to degrade the performance of the power supply.

There might be some voltage  feedback in the software - but only one person knows that at the moment!

OHOH- so here is another question - Since the LT3080 is operating as a voltage follower, what would the circuit do if you had tied the high end of R16 to the OUTPUT, rather then Vset?

(Dave, I know that you have to freeze the design and cut some PCBs at some point, and no doubt that there will  be versions 2, 3 and 4 of this project in the months to come!)

[metalphreak]

If you had a mosfet on the output to switch the output on/off (quite a handy feature) would you still even have to worry about voltage feedback from the output? If the PSU is powered down and the mosfet gate is tied to ground, no current should flow back into the voltage regulator?

[Blue]

Dave,

The sense resistor network R30-R33 divides the sense output only 4 times. Not 5 times as you mentioned in the videoblog at 31:19. You might blow up the adc...

Cheers,

Nick

[LaurenceW]

Nick, I disagree. 5K (R31//R33) of a total of 25K (R30+R32+R31//R33) is one fifth...

And anyway, the DAC inputs will be diode clamped to supply and ground (see Microchip spec sheet, page 18). Anything much less than a lightning strike coming via 20K of R30+R32 isn't going to bother the DAC, I would suggest (though i am NOT about to try that!)

[Blue]

Hi Laurence,

You are absolutely right. I must be sleeping..

[ejeffrey]

If you had a mosfet on the output to switch the output on/off (quite a handy feature) would you still even have to worry about voltage feedback from the output? If the PSU is powered down and the mosfet gate is tied to ground, no current should flow back into the voltage regulator?

MOSFETs are not ideal for this.  Power MOSFETs have a body diode so that under reverse bias they conduct regardless of the gate voltage.  A mechanical relay is still the best way to switch something actually off, although an SSR is pretty good as long as you want normally open contacts.

It isn't really a big deal though.  The protection diode is only needed if the input can be shorted.  If the total design does not have that as a failure mode you don't have to worry about it.  If you do decide to care it should be no problem for a builder to wire a 5 amp discrete diode across the entire supply.  Keep in mind that this might still damage the "load" that is supplying the fault current.

[koch777]

I wonder if anybody noticed current regulation limitation of current design?
I will refer to Dave's schematics posted someplace (pdf file)
Suppose we have 10V at the output and current limit is set to 0.1A. Loading power supply with 1 Ohm will activate current limit circuit U7 - U3B - Q2. Q2 will effectively short SET pin of LT3080 to ground via 100 Ohm resistor R12 (actually it is C-E of Q2 and R12 in parallel). At the same time output of opamp U3A will stay at 10V. These 10V will source ~9mA through R11 + (R12 || Q2). This current will develop voltage drop across R12 of approximately 0.4V-0.5 V meaning that output of LT3080 will not go to zero. Out load of 1 Ohm at 0.4V output is 0.4A, not 0.1A we want. Current regulation is limited to a certain degree. The problem is worse at high output voltages and lower end of current limit and virtually doesn't exists at low voltage/higher current settings. Not a big issue, but something to be aware of.

It is purely theoretical and I can't be 100% sure it behaves as described. At least on a LTSpice model it is.
That is not to say that design is bad. Actually it is Dave's blog #221 that brought my attention to LT3080 and I think it is a brilliant device. I never seen such fast reaction to load transients without going into oscillation. And addition of pass-through PNP transistor(s) could really bring this stuff to the limit (like "standard" 30V/5A). On a model at least it really shines.

[currentflow]

Hi Dave,

Thank you for an excellent project. Regarding freeing up some microcontroller pins, you could consider the following:

1) Share ADC-CLK with DAC-CK and share DAC-D with ADC-DIN. This saves two IO pins and retains functionality through the relevant chip-select control of the DAC and ADC chips.
2) A further IO pin may be saved by Charlieplexing the keys, allowing up to six keys using just three IO lines. See http://pcbheaven.com/wikipages/Charlieplexing/
3) Instead of 2) above, use one IO (ADC input) line for all keys such as: Interface multiple keys with one wire. This has a disadvantage in that only one key may be pressed at a time, but it could free up two more IO pins.

Now there are five spare IO pins for further functionality

Regards,
Steve

[Ajahn Lambda]

If you had a mosfet on the output to switch the output on/off (quite a handy feature) would you still even have to worry about voltage feedback from the output? If the PSU is powered down and the mosfet gate is tied to ground, no current should flow back into the voltage regulator?

MOSFETs are not ideal for this.  Power MOSFETs have a body diode so that under reverse bias they conduct regardless of the gate voltage.  A mechanical relay is still the best way to switch something actually off, although an SSR is pretty good as long as you want normally open contacts.

Agreed, but it will also depend upon what's 'upstream' of the MOSFET.  Peronally, I don't depend upon the body diode to provide any sort of real-world protection; I know you were referring to its intrinsic property, so I'm not trying to nail you on this!   

Then again, as long as there's an 'upstream' protection, i.e. a rectifier of some type, and the filter cap is protected/rated appropriately, there's nothing to worry about.  The output should appear as an open, in 'default' or OFF condition.  But again, we're back to your scenario of using a relay or SCR to switch the output connection.

It isn't really a big deal though.  The protection diode is only needed if the input can be shorted.  If the total design does not have that as a failure mode you don't have to worry about it.  If you do decide to care it should be no problem for a builder to wire a 5 amp discrete diode across the entire supply.  Keep in mind that this might still damage the "load" that is supplying the fault current.

Yup! 

[NiHaoMike]

AFAIK all Arduino compatible Ethernet variants use the Wiznet W5100 chip which is fine pitch SMD.
Hardly suitable for a DIY kit.

Dave.

The microchip ENC28J60 is available in a 28pin DIP package. I'm sure there's an arduino library for it somewhere.

Look up EtherCard. There's an ENC28J60 in my Arduino-based power meter and it works nicely. It even has a clock output that can be used to clock microcontrollers. (In my design, it clocks the dsPIC that's doing the hard calculations.)

For a way to protect the low range current shunt, you could put in a diode to restrict the drop to 0.6V, or you can just turn the MOSFET around for a zero cost equivalent.

[mobbarley]

For a micro I was going to go with a PIC18F4550 for a few reasons.

2) Hardware i2c, serial, and USB2.0 (all usable at once I believe)

The EUSART shares a pin with SPI on these chips, not sure about I2C - but you could replace the serial with CDC

[LaurenceW]

Another point about the suggested volts/amps range of Dave's original design (which has certainly got us all talking!).

I think the original design, while admittedly stretchable, was going to top out at 10-12V and at 1A... the argument being that nobody needed more than that.

Then right along comes this Rubidium frequency gizmo - what does that need? 15-18V at a couple of amps (Start up), AND 5V.

Ah...

This has been a fantastic tutorial, it really has. I may yet build one up "as-is", as a further part of the learning curve. But I remain convinced that I am going to need to push the specs a little further than the current design.

Another GOTCHA to watch out for, for those pushing the voltage up a bit... the LT3080 has some sort of "safe area" operating current, and will NOT allow you to get 5V regulated at 1A, while at the same time running a primary side high enough to give you 25V, output, say. Check out the "Current Limit" diagram in the centre of page six of the spec sheet. To get both high (1A) current AND high (25V) voltages ranges, it is going to be necessary to switch the primary voltage up and down, to minimise the voltage (and so power) drop across the LT3080.  That could be done under software control, but that isn't really necessary. A voltage comparator and a relay to switch winding taps from the transformer will do it.

We just can't stop this armchair tweaking, can we?!

[IanB]

I don't think its the rubidium frequency standard.

My guess is something to do with battery charging. Various battery charging algorithms for lithium ion or NiMH need accurate control of current and voltage and accurate measurement of current and voltage to decide when charging should finish. You don't need more than about 5 V to charge a single li-ion or li-poly cell (or parallel bank), and 12 V DC input caters for charging in the field when a car is nearby. Linking the whole thing to a micro permits accurate programming of the charge algorithms, and none of the commercial RC hobby style chargers are open enough to permit complete flexibility of the program.

So I'm thinking it might be something related to the quad-copter and battery power.

[EEVblog]

Another point about the suggested volts/amps range of Dave's original design (which has certainly got us all talking!).

I think the original design, while admittedly stretchable, was going to top out at 10-12V and at 1A... the argument being that nobody needed more than that.

No, that is not the argument.
This is not meant to be the be-all end-all lab bench power supply, not even close.
This is a niche power supply, with the design limitation there for reasons that will become obvious in future videos.
Whether or not anyone agrees they are good or reasonable limitations, I don't really care, this is what I wanted to do, I'm doing this for me.
If you don't like the limitations, feel free to go modify it or design your own.
There may be future builds with greater capability, but not this time around.
No, it's not designed to be a battery charger. What's the point in that when you can buy intelligent multi-chemisty multi-cell high current balanced battery chargers for $20-$30?

I could tell you now, but that would ruin the "Ta-Da, here it is" bit...
It's not really that hard to guess anyway...

Dave.

[IanB]

No, it's not designed to be a battery charger. What's the point in that when you can buy intelligent multi-chemisty multi-cell high current balanced battery chargers for $20-$30?

Fair enough, but you are making this a but of puzzle. You can't complain if people keep guessing 

The cheap battery chargers from China are closed program and limited in their range of adjustments, though of course adequate for the purpose intended. Higher quality and more flexible chargers easily exceed $150 putting them in a different price bracket.