Lab Power Supply Design Part 5

[metalphreak]

alm: Yeah all the outputs will have the same ground reference. So there'll be no option to do series or dual +/- rails. I was thinking more along the lines of simply having more separate outputs where the voltage and current can be monitored. It wouldn't be hard to add USB/RS232 datalogging capabilities, or even some onboard memory for logging (download it later).

[LaurenceW]

Dave, 500 reads already! You have hit a "nerve" (in a good way) with this one. I have followed you through from the very first videos, and follow what you are doing well enough (sign of a good teacher )

You are no doubt going to get DOZENS of "how about this/that?" options. Here are mine.

>> 10V limit? That's fine, but for a number of projects I would like to look to at least the "industry standard" of 30V, even though I won't use it often. I will go though the circuit myself to see what limits higher voltage working (I can sacrifice some of the 2.5mV resolution!)

>> I would like to give some though to a twin channel supply, that the outputs can be paralleled/Serialed up. Hmm.. might need two complete identical PSUs, with some (digital opto) isolation between the two... That'll be "Power Supply Design part 17", then!

>> "I agree with Nick" (famous slogan in the UK General Election, last year; may not translate down under!) You can save three of your four switch pins PB1-PB4 using a simple switched resistor ladder. Make the resulting analogue voltages 2, 1, .5 and .25V. Then if some smart arse presses two buttons to produce an analogue output of 0.75V you will have the A>D resolution to detect and reject it. I have used this to good effect before - it is perfectly reliable with up to 5 switches.

>>With one of those freed up pins... how about a little bit of temperature feedback from the heat sink? (or is the LT3080 able to look after itself?)

Some really nice touches in your design, and very clearly explained - thanks again.

[PStevenson]

I'm just hoping you do a good production run of these things cause I don't want to miss out on this like I do the uCurrent!
I'm a power supply psycho!

[metalphreak]

>> 10V limit? That's fine, but for a number of projects I would like to look to at least the "industry standard" of 30V, even though I won't use it often. I will go though the circuit myself to see what limits higher voltage working (I can sacrifice some of the 2.5mV resolution!)

I'd like atleast 12V on mine. 2.048v ref voltage with a gain of 6 is 12.288v with 3mV resolution. This may not be enough to correct for losses to get 12v output at high current loads, but then you could just do a gain of 7, 8, etc You just need to adjust the resistor dividers on both the VSET opamp and the VSENSE input.

If I use the MCP4728 with its internal 2.048 Vreg and the switchable x1, x2 gain, it would allow me to dynamically change the output range. Any error could be calibrated using the ADC which will have a proper voltage reference source. Trade-off between an easy digitally controlled voltage range switch, and having to calibrate stuff in software.

[Psi]

Hi, this might be a wacky idea, but how difficult would it to add a function generator option that turns the dc output specs into an ac output with the same specs (like voltage 5 VDC to 5 V peak-to-peak at a set current).

Rutger

If ya wanted that sort of extra functionality it would be easier to use a 3A opamp direct.

[LaurenceW]

Another question!

What about start up, or if the software loses the plot, for some reason? what are the output states of U5 VoutA and VoutB? Do these default to a nice safe 0? at low impedance? We don't want the op amp inputs flapping about at hi-Z if the controller isn't, well, controlling...  worth a couple of 100K pull down resistors on Vset and Iset?

I am thinking that the output voltage could be scaled up to 25V, while keeping an eye on V+, which mustn't go over 36V,  or the main reg will be TOAST.  Then it will also be necessary to put another (say) three-pin 12v reg down from V+ to run the op amps. Need to recalc voltage feedback gains, of course.

Tinker time!

[Rerouter]

For this design, being how i would like to beef it up to 0-~30V (3x the gain), firstly, can you forsee any obvious issues that would come from this, and secondly, would there be any issue in using fets to switch tappings on a transformer, being how i would loose the majority of my current capability with over 12V difference in to out,

[RJSC]

For this design, being how i would like to beef it up to 0-~30V (3x the gain), firstly, can you forsee any obvious issues that would come from this, and secondly, would there be any issue in using fets to switch tappings on a transformer, being how i would loose the majority of my current capability with over 12V difference in to out,

You can also charge a big capacitor until it is like 5 volts over the output voltage and then disconnect it from the transformer until the voltage of the rectified sine wave comes down to the same value, effectively making a very low frequency switching pre-regulator with a large ripple (4 V) which in turn is smoothed out by the rest of the circuit.
Consider maximum output current, maximum dropout and the maximum power dissipation on the regulator you want to be energy efficient to calculate the capacitor and the voltage over the output to switch off the transformer.

You can also use a pre-built switching power supply like one of those adjustable laptop adapters and replace the switch and associated feedback divider resistors with a digital potentiometer under control of the microcontroller in order to dynamically keep the input voltage a just a couple of volts over the output voltage.

That is what i am going to try to do with mine, however, the laptop power supplies I have seen until today are adjustable up to 24V and I do not have any to open and check the circuit to see if if can go up to 32V or so in order to get 30V output.
Since the DAC is 12 bit (4096 steps) I'd like mine to go up to 40.96 V in 10 mv steps to use the DAC's range accurately and fully, but i doubt I wil find any cheap switching power supplies on the generic market that can go up to 43V and supply at least 3 A (ideally over 4 amps to be able to make it go to 4.096 A in 1 mA steps).

[Rerouter]

RJSC

the reason why i stated 30V and not higher is, the low dropout lines current drops to a fickle 250mA when you want to feed in more than 36V

[JimmyM]

I have an HP printer power supply that does 32V. About 2A or so. You might look at those.

[Loki]

I`v done some research and found that Atmel has a well suited for this project, in my opinion, microcontroler. There is ATxmega16/32/64/128A4 family. It has build-in 2-channel 12-bit DAC and 12-channel 12-bit ADC. It cost about 6 dollars. The only problem is that it is in 44-TQFP package. How about that?

Sorry for my english

[shebu18]

The ATxmega would be good if you make it only for you, but the idea was that the PSU was a kit version, so the logic is to use a majority of THD parts. In some cases you can use only smd parts, but in a large package for easy soldering.

[LaurenceW]

Watch out! You can't push the LT3080 beyond 36V Vin, and expect it to last you a long time.

I have tried "slow speed" switching  of the mains transformer via a thyristor onto a big main supply cap before, to reduce the average voltage on the cap. Two mains cycles on, two off; that sort of thing  All that happened is that the one or two mains cycles when the transformer was connected to the load caused it to get DAMN HOT (suspect the core was saturating).

Best bet is to have a transformer with a tap at around 2/3 of the total output voltage, and relay-switch between the 100% and 60-70% tap. This will reduce the peak power dissipation of the series regulator. Software (if you have an intelligent controller) or a simple comparator switches to the higher transformer tap, whenever the output voltage is higher than a pre-set point. Build in about 1 V hysteresis to prevent the relay chattering.

[JimmyM]

I'd be happy with ~15V output. I figure I can do that with a few gain resistor tweaks, a couple of voltage divider tweaks. I just have to watch out that the higher input voltage doesn't burn-up any components running directly from Vin. Then I'd just have to add an LM7812 to run the electronics Vin rail.

[Anders]

Just wondering;

I’m not an expert, far from it but why reinvent the wheel?
Why use so many peripheral chips when you can save space and money by using for example a reliable PIC18F46J50 or similar which got most of it onboard?
Just drop the “full of nasty surprises” Atmel chip.

[metalphreak]

Just wondering;

I’m not an expert, far from it but why reinvent the wheel?
Why use so many peripheral chips when you can save space and money by using for example a reliable PIC18F46J50 or similar which got most of it onboard?
Just drop the “full of nasty surprises” Atmel chip.

PIC18F46J50 only has 10bit ADCs. Atmel does the ATxMega series with 12bit ADC and DACs, but 1) they aren't available in a DIP package (Dave's making kits that customers assemble themselves) 2) Have you ever tried to reliably get stock of some of the Atmel chips.... its a bit of a joke sometimes. Not good for an open source design if nobody can get parts!

[IanB]

2) Have you ever tried to reliably get stock of some of the Atmel chips

Do you mean PIC chips?

[Bored@Work]

Why use so many peripheral chips when you can save space and money by using for example a reliable PIC18F46J50 or similar which got most of it onboard?
Just drop the “full of nasty surprises” Atmel chip.

Why don't you just do your own design and kit, if you know how to do it right?

[McMonster]

There's one more AVR catch I've noticed in the schematic, AREF pin tied to Vcc. You're not supposed to do that, you can select internal reference to be AVcc in software (REFS bits in ADMUX register) and then you should put a decoupling cap from Aref to ground or just leave it disconnected if you're not using micro's ADC.

If you (or someone plaing with the kit) accidentally switch reference to internal 1.1V it'll appear on the AREF pin and get shorted with Vcc.

[slateraptor]

Why use so many peripheral chips when you can save space and money by using for example a reliable PIC18F46J50 or similar which got most of it onboard?
Just drop the “full of nasty surprises” Atmel chip.

Why don't you just do your own design and kit, if you know how to do it right?

Freeloading and a general lack of appreciation for the contributions that Dave has graced the community with, obviously.

[shebu18]

Why use so many peripheral chips when you can save space and money by using for example a reliable PIC18F46J50 or similar which got most of it onboard?
Just drop the “full of nasty surprises” Atmel chip.

Why don't you just do your own design and kit, if you know how to do it right?

+1 from me!

[slateraptor]

There's one more AVR catch I've noticed in the schematic, AREF pin tied to Vcc. You're not supposed to do that, you can select internal reference to be AVcc in software (REFS bits in ADMUX register) and then you should put a decoupling cap from Aref to ground or just leave it disconnected if you're not using micro's ADC.

If you (or someone plaing with the kit) accidentally switch reference to internal 1.1V it'll appear on the AREF pin and get shorted with Vcc.

Can you direct us to the reference in documentation, please?

[alm]

RTFDatasheet .

See page 258/Section 24.5.2 of the ATmega328P datasheet.

[slateraptor]

RTFDatasheet .

See page 258/Section 24.5.2 of the ATmega328P datasheet.

I said please.

Good catch and thanks for the reference. I'm simply of the disposition that any device-specific claim ought to be backed up with an appropriate reference in documentation.

If I do end up building one for shits and grins, I wouldn't use an Atmel uC anyways. I'm really more interested in Dave's analog thought stream, general system integration methodology, and how he intends to mitigate conducted susceptibility on board layout.

[metalphreak]

2) Have you ever tried to reliably get stock of some of the Atmel chips

Do you mean PIC chips?

http://au.element14.com/jsp/search/browse.jsp?N=2101+203063&Ntk=gensearch&Ntt=ATxmega&Ntx=mode+matchallpartial

I've never had problems getting stock from MicroChipDirect, and shipping is always a flat ~$17 fee. Most of the time I consider getting an Atmel chip I get roadblocked with no stock at AU suppliers, and its $30+ shipping from the USA.  MCD usually have pretty good pricing too. PIC18F4550 TQFP is $4.38 for 1-25 chips. Unless you need single units or parts overnight, MCD is always a better option for me in Australia

Looks like Digikey has plenty of atxmega's, so my quick search of local suppliers gave me the wrong impression of availability.