Lab Powersupply - Markus take

[markus_b]

Inspired by the Dave's EEVBlog power-supply series, I've designed my own. It has now matured enough, the I'm preparing to order the components and get a PCB made for a prototype. I'm putting this up here for review and critique. I've attached the schematics and the PCB layout. All input is appreciated.

Basic Premises:

• Modular construction: If you don't need all parts/gadgets you can just build the parts you need and leave the rest off. For example you can forego the switching pre-regulator or the ucurrent uamp measurement circuit.
• The construction is mostly SMD with some trough-hole. Some power parts, like the LT3083, are through-hole, to allow to add a heatsink. The micro and the op-amp are SMD, because they are not available in through-hole packages. All small stuff, like resistors, are SMD. Many components, like precision resistors are much cheaper in SMD than the same in through-hole.
• The power supply does not connect directly to mains power, but uses a standard power supply as input. This allows us to skip all mains safety related questions as those are cared for by the upstream mains power supply. The power supply input is 12V to 20V. There are plenty of laptop power supplies who supply DC power in this range. This also extends down to 12V to allow to use a 12V car battery.

Modules:

• Linear power regulator, takes two input voltages (vset/iset) and translates it into 0-40V/0-4A output with 10mV/1mA resolution. This can me modified, for example to 0-10V/0-1A with 2.5mV resolution. I use the LT3080 (1 amp) or LT3083 (3 amps), for the regulation. These regulators allow to go down to zero volt/amp. The target precision is 1%.
• Switching pre-regulator, takes 12-20V input and translates this into 2-32V. It follows the output voltage, reducing the drop-out voltage of the linear regulator to 2V. This reduces cooling requirements for the linear regulator a lot. It also allows to use any available DC power from a 12V battery to a laptop power supply and get up to 30V output.
• Microcontroller based. The micro (AVR atxmega32A3U) has an USB interface, drives a graphical LCD display and reads rotary switches. It generates the 0-1V ouput (via DAC) used as regulator input. If desired the microcontroller can be replaced with two potentiometers and a 1V reference. The USB interface does not provide galvanic isolation.
• High precision reference. If a better precision than the built-in 0.5% reference of the microcontroller is desired a separate 1V reference can be added. In this case the relevant resistors in the circuit must be of high precision too.
• uCurrent small current measurement device. This is a small shunt with amplifier to measure small currents (uAmps).
• Bluetooth interface. There is a slot to add a simple bluetooth serial interface module. This allows to interface the power supply with a computer wirelessly (galvanic isolation).
• Simple Volmeter. A spare ADC channel is wired to provide a simple 0-40V voltmeter (10mV resolution). Signal ground is the power supply ground.

Specifications:

• Output:
• 0-30 V
• 0-3 Amps (LT3083), 0-1Amps (LT3080)
• 0-30V in 10mV steps
• 0-3A in 1mA steps


• Input :
• 10-20 V


• Hybrid regulation:
• Switching regulator stage -> low power dissipation
• Linear regulator -> low noise


• LT3083 or LT3080
• Microcontroller controlled (xmega32A4U)
• Computer interface (USB, Bluetooth)
• Graphical LCD (128x65)
• Optional components/modules:
• uCurrent
• Bluetooth
• 1 Amp version by using a LT3080
• Switching preregulator
• User Interface (can be replaced by a potentiometer)

User interface elements

• Graphical LCD display
• Rotary knobs to regulate voltage and current limit
• 'Menu' and 'Cancel' button
• Speaker on port with timer output

Protection features

• Protection against user error on the output side:
• Input voltage connected when power supply is off
• Reverse polarity input voltage
  -> Diode and PPTC
• High input voltage
• Protection against user error on the input side, reverse polarity protection
  ->P-MOSFET circuit against reverse polarity

Open Issues, Caveats and Enhancements

• USB interface not isolated
• USB ground not equal to ouput ground due to ucurrent device in the return path
• Possibility to use USB as supply voltage to be developed (with reduced power)
• Performance in a very low voltage, low current environment
  The LT3083 need a small, but not zero current to flow and the LM334 only starts working a volt ot two.

[Berni]

That design reminded me of Daves PSU alright, not that its a bad thing since he did a great job on it.I especially like the switchmode tracking preregulator. One thing i noticed tho is that you should put a RC filter on your control voltage from the DAC because the MCU is noisy and could make the control voltage noisy, that then results in the output having extra noise.

If you want it to work better at very low voltages you could replace that current source with your own. Put a small cheep mosfet on the output that goes to ground trough a 1 ohm sense resistor and then use a opamp to regulate the voltage across the resistor. Because the opamp can have its own supply that should be able to sink the current down to 0.1V or even lower.

The USB isolation is a bit more complicated to solve but you can do it if you go get a FTDI USB to UART chip and then run that to the MCU over a optoisolator.(Btw Microchip makes the MCP2200 that does the same thing as the FTDI chip but is at half the cost)

[Rerouter]

Cons:
- seems you skimed on the feedback pin stability :/ neither of your regulators have any capacitance on there feedback pins, (1u perhaps?)
- on your switcher disable jumper i would like to see more than 1 via for 3A :/
- output jumper too close to p62,
- sil sockets are generally only rated to 1A per pin,
- same problem faced with daves design, for true almost 0V operation needs a diy current source, ($3 seems to not being doing anything important)

Pros:
- the 22uF caps with the 0.025 ohm sense resistors make a great low pass filter for the switcher, took me a while to work that one out,
- love the test point branch outs, would prefer the power supply region ones to actually be plated holes, but that is a request,
- like the way your smd parts are well spaced,

Thoughts:
- why is C2 0.33uF and not a 1uF like all of your other regulators use?
- hmm, using usb for power while being isolated... what about using a usb on the go type regulator in an isolating configuration instead, (not quite sure how that might be done :/) though that still leaves data... perhaps one of these http://www.analog.com/en/interface/digital-isolators/products/index.html

Edit: got my resistor value back to front

[markus_b]

That design reminded me of Daves PSU alright, not that its a bad thing since he did a great job on it.I especially like the switchmode tracking preregulator.

Yes, I liked Daves design a lot. Thats why I took it as starting point.

One thing i noticed tho is that you should put a RC filter on your control voltage from the DAC because the MCU is noisy and could make the control voltage noisy, that then results in the output having extra noise.

You are probably right. What values would be right for that filter ? (1k / 0.1uF)

If you want it to work better at very low voltages you could replace that current source with your own. Put a small cheep mosfet on the output that goes to ground trough a 1 ohm sense resistor and then use a opamp to regulate the voltage across the resistor. Because the opamp can have its own supply that should be able to sink the current down to 0.1V or even lower.

For now I decided to live with the limitations in that area. It is quite rare that one needs very low voltage and current at the same time.

The USB isolation is a bit more complicated to solve but you can do it if you go get a FTDI USB to UART chip and then run that to the MCU over a optoisolator.(Btw Microchip makes the MCP2200 that does the same thing as the FTDI chip but is at half the cost)

I also decided to leave USB isolation off. I have two lines of thought: 1) If you need isolation, then you can connect via bluetooth. 2) If I can get the switcher to start working at 5V, then I can use USB as a power source, this would be a great auxiliary functionality. Of course, there would be power limitations and probably no LCD backlight.

Cons:
- seems you skimed on the feedback pin stability :/ neither of your regulators have any capacitance on there feedback pins, (1u perhaps?)

You talk about the 'Set' pin of the LM3083, or the input pins of the opamps ?
You are right, the 'Set' pin should get a cap.

- on your switcher disable jumper i would like to see more than 1 via for 3A :/

You are right, I've rearranged D21 and R26 and don't need the vias anymore as I can stay entirely on the top side. In earlier designs I had four smaller resitors in parallel in place of R26 and I started out with all resistors at the top. Now it is much better :-).

- output jumper too close to p62,

Yep, moved it.

- sil sockets are generally only rated to 1A per pin,

Yes, I know they are a bit too small. Have a good suggestion ?
Also, the input barrel connector has the same problem. Connectors rated for 6 amps exist, but are very expensive.

- same problem faced with daves design, for true almost 0V operation needs a diy current source, ($3 seems to not being doing anything important)

As I said, I decide to live with the limitation of the design for now. If I come across another good regulator design I like I will replace that part. Richard (amspire) went on to design something, but what he came up with would mean 20-something parallel transistors, the does not appeal to me.

Pros:
- the 22uF caps with the 0.025 ohm sense resistors make a great low pass filter for the switcher, took me a while to work that one out,

This is not really intentional, but an artifact of the current sense resistor where it is.

- love the test point branch outs, would prefer the power supply region ones to actually be plated holes, but that is a request,

This is a prototype and designed to tinker, so accessible test-pints are a must. I could replace them with a single pin sil, to get a plated hole.

- like the way your smd parts are well spaced,

I have enough space and will assemble by hand, so as well use it to make it easier for me. Also, I started out with 0804 and went to 0603 later (components are cheaper and there is more variety).

Thoughts:
- why is C2 0.33uF and not a 1uF like all of your other regulators use?

That value comes straight out of the data sheet of the regulator. Also that cap is in parallel to C1 (but ceramic), so capacity is not a concern. It also has to be of higher voltage that the 1uF caps (50V instead of 10V), so it has to be a different part anyway.

Other thoughts:
I don't really like the way I supply the 6V and 3.3V power. I burn up to 1W there, but am not keen on using a switcher. I'm contemplating to remove the 6V regulator and supply the LEDs, backlight and uCurrent MOSFET directly and regulate the 3.3V also from the input voltage.
This would also open up the possibility to use the USB 5V supply as input voltage, if the switcher works down to 5V. I just have to device a good way to switch to the USB supply, if the main input voltage is missing, this without loosing much of it (no diode) and being sure not to send 12-20V down the USB line, when both are connected.

Thanks for the thoughts, helps me to proceed !

[Berni]

Well you can solve the back light thing by moving the 68Ohm resistor from the transistors collector to the emitter. That way you create a crude current source that will try to keep the voltage across that 68Ohm resistor about 0.7V lower than the pin of the MCU(So about 2.6V). That means you dont need a regulated supply for your backlight(That transistor might get pretty warm tho with 20V input).

The MOSFET should be fine with an unregulated voltage but do put a 15V zenner over its gate since most mosfets will blow up if you apply 20V on the gate.

The USB power thing is a little tricky to do in solid state so that would need a special IC for that unless you just put in a relay with its coil connected across the power jack. One other way to do it is to put in a separate little switchmode in that works from USB and then just control the enable pins on the two. Since you said this one cant work from 5V (Actually needs to work down to about 4.5V because of voltage drops trough cables) and low power switch modes can sometimes be found cheaply.

[markus_b]

Well you can solve the back light thing by moving the 68Ohm resistor from the transistors collector to the emitter. That way you create a crude current source that will try to keep the voltage across that 68Ohm resistor about 0.7V lower than the pin of the MCU(So about 2.6V). That means you dont need a regulated supply for your backlight(That transistor might get pretty warm tho with 20V input).

I've decided to go with a 220 ohm resistor directly to the input voltage (12-20V). This gives me nominal current/brightness at 12V. The backlight is connected to a timer-port on the AVR, so I'll use PWM to limit the power when the input voltage is higher.

The MOSFET should be fine with an unregulated voltage but do put a 15V zenner over its gate since most mosfets will blow up if you apply 20V on the gate.

I put the Zener back in and connected it to the input voltage too.

The USB power thing is a little tricky to do in solid state so that would need a special IC for that unless you just put in a relay with its coil connected across the power jack. One other way to do it is to put in a separate little switchmode in that works from USB and then just control the enable pins on the two. Since you said this one cant work from 5V (Actually needs to work down to about 4.5V because of voltage drops trough cables) and low power switch modes can sometimes be found cheaply.

Actually I don't know what the minimal voltage is, for the switcher. It may well work at 4.5V. I'll have to test.

But the switchover is not easy to do reliably. That's why I'll let it off for now. The problem probably merits its own thread anyway.

As last mayor change I swapped the MCP1700 against a LP3951 as 3.3V regulator. The latter supports 30V and can dissipate 1W, this is sufficient to feed it directly from the input voltage. So I can do away with the kludge of the two regulators in series.

I've incorporated some of your suggestions and modified the schematics and the PCB, as necessary.

[poorchava]

I think you should add some small chip inductor in series with any analog supply (especially reference and mcu analog power) and probably add some spots for decoupling capacitors (like 2 in parallel). That doesn't cost much and may reduce any noise significantly.

[markus_b]

I have done quite a bit of cleanup work, I'll name this version 1.1a.

I've also run the BOM through pricing. Grand total (sans PCB) is about $90 (Mouser). Not too bad. The most expensive part is the LCD ($15), second is the box ($8.50).

I think you should add some small chip inductor in series with any analog supply (especially reference and mcu analog power) and probably add some spots for decoupling capacitors (like 2 in parallel). That doesn't cost much and may reduce any noise significantly.

I was wondering if some inductors are necessary. Anybody has some real world experience ?

[Rerouter]

if you really wanted to get into that, then most of the noise you will be aiming to mop up will be over 200Khz from the switcher and in the low Mhz for your micro, i would reccomend fitting one of these inline of your AVCC connection to 3.3V or simply replacing R92 with it, , http://au.element14.com/murata/lqh32cn101k23l/inductor-1210-case-100uh/dp/9522220RL

combined with C91 a 100uH inductor effectivly removes all digital noise from your analog supply, and if you really wanted to try and kill any noise here, another on AGND,

while for your voltage reference your down at -30dB when you hit around 100-300Khz which is nothing in the scheme of things, so that should be fine to be left unfiltered,
out of curiousity are the 1uF capacitors tantalum? or ceramic?,

also the ucurrent op amp, its power supply rejection falls over at 100Khz, so i would reccomend increasing its capacitor to 1u or having a 1u in parrellel with C71, and using another 100uH inductor inline with its supply line, to kill off the switcher and digtal noise from its output,

this should only leave the transformer emi and the usb port feeding in nasty noise, everything else above 100-200K should either be filtered or not matter,

considering you would otherwise be filtering AVCC and the ucurrent seperatly i can see no major issue to leaving IC2 unfiltered, as the rest is digital,

also your schematic looks to have a bug with the output of the reference, and your board got a tad wobbly near the op amps, other than that its good to see less and less of the power traces swapping sides,

[fmaimon]

Markus,

Do you have the component list, from digikey or any other vendor, for the pre-regulator section? This is the part I'm most interested.

[markus_b]

if you really wanted to get into that, then most of the noise you will be aiming to mop up will be over 200Khz from the switcher and in the low Mhz for your micro, i would reccomend fitting one of these inline of your AVCC connection to 3.3V or simply replacing R92 with it

I've replaced R92 with a inductor and added one for the ucurrent supply. Is a goos idea and does not cost much.

out of curiousity are the 1uF capacitors tantalum? or ceramic?,

Most of the caps, except to 22u aluminium, are ceramic. I like ceramic, because they have good ESR, age well and have no downsides I know of. My knowledge is limited, though. Unfortunately for the larger capacities and voltages they are not available.

also your schematic looks to have a bug with the output of the reference, and your board got a tad wobbly near the op amps, other than that its good to see less and less of the power traces swapping sides,

I don't see what you mean with the 'a bug with the output of the reference'. But there is a mess in the PCB in some areas because I've moved components and did not fix the traces afterward. Will have to clean up some points. Also I have not yet re-run the ERC.

Do you have the component list, from digikey or any other vendor, for the pre-regulator section? This is the part I'm most interested.

I'll attach the current part-list, all parts are from the Mouser Europe. For me Mouser works out best, because they ship locally, that means I don't get additional customs fees. I'm using the bom-ex.ulp script to manage the BOM, recommended !

One word of caution about the switching regulator: The schematics and component sizing comes straight out of the TI Webench tool with some untested conjecture of mine. I think the tool is fine, but it does not take variable output in account. So essentially I let it calculate the worst case (12V to 30V,3Amps), presuming it can do less (lower voltages and currents). That is one of the main areas I'll have to stress-test my prototype. At this time this is just armchair electronics :-).

[fmaimon]

Markus,

How are you going to vary the output of the pre-regulator with this design?

[markus_b]

How are you going to vary the output of the pre-regulator with this design?

The preregulator will follow the output of the analog regulator with 0.7-1.4V difference. This is realized with Q31 whose base is connected to the analog output. This design is actually suggested in the LM3083 datasheet to control a switching preregulator. The base-emitter voltage of Q31 will define the difference (0.7V). The diode D31 is there so I can add another 0.6V difference if testing shows that it is beneficial. The LT3083 claims to be happy with a 0.7V drop-out, so we'll see.

[fmaimon]

Don't you need a resistor from the FB pin to ground? Maybe set up both resistors for the voltage feedback so, if Q31 fails open, then the output voltage don't go too high.

[markus_b]

Don't you need a resistor from the FB pin to ground? Maybe set up both resistors for the voltage feedback so, if Q31 fails open, then the output voltage don't go too high.

Wohoaa ! That's a catch ! Thanks !

Yes, there is supposed to be s resistor to ground. Needs to be to counter-effect Q31. Somehow I've overlooked this when combining the feedback mechanism with Q31 from the LT3083 datasheet with the switching regulator schematics from the webench tool.

I've done some more cleanup and added an additional choke in the supply line of the analogue op-amp. Noise there will translate into noise at the output too.

[markus_b]

I've decided that it is now the time to test the entire setup. I've placed an order for the components with Mouser and for the PCB with Seeed. The components will be here next week, the PCB will take a week more.

We'll see, hope I don't find too many errors as bad as the one caught by fmaimon.

I'll get 10 PCB, and may be ready to part with some.

[Rerouter]

just want to point out you need to swap out those 100nF on your ucurrent AND op amps to 1u if you want it to filter all the way down to below 200KHz, with a 100n you would only just knock out some of the digital noise, so far thats the only problem i'm seeing, and hope it goes well, (i could use 1 of those boards )

[poorchava]

OR you can add 1u, 100n and 1n in parallel. Ceramic caps cost next to nothing, your design is not very space constrained and they can decouple particular frequencies to GND. I usually tend to add such footprints if I have space for that. Then when some stuff comes up at particular frequency i just experiment with additional decoupling caps until i find suitable value that lowers noise to acceptable level. If I don't have space i just end up soldering ceramic caps one on top of the another .

Such stuff is rather for typical mixed-signal circuits (eg. containing thermocouples, RTDs and then FPGA/CPLD+microcontroller) where high frequency digital-born noise is always an issue. It might not be the case for power supply though.

[markus_b]

Or you can add 1u, 100n and 1n in parallel. Ceramic caps cost next to nothing, your design is not very space constrained and they can decouple particular frequencies to GND.

I'm using ceramic caps in all these places, is there a point to put a 100n and a 1n ceramic cap in parallel (is the 1n sufficiently better at high frequencies to make a difference) ?

I understand that placing a electrolytic and a ceramic cap in parallel makes sense (the electrolytic is not great at high frequencies). But does the same apply between caps of the same kind ?

[Rerouter]

i suppose the golden question is where does filtering stop and decoupling start, in my own mind they are the exact same thing, thus the 1n and 100n would serve only to slightly lower the corner frequency of your filter and thus kill a little more noise in an area where not really expecting it, though i do understand how using parrellel capacitors of different values can fill in points of self resonance, but the 1n still has me a tad confused, as its for decoupling around the 190Mhz range

i agree to the reasoning of leaving the 100n in parrellel if you wish, but i see no point to a 1n what so ever, (1uf ceramic is generally for 10Mhz, 100n is usually for 22Mhz, so if you want to get crazy it might help loose 0.5db of noise, but the filter should already kill it for you)

[Rerouter]

hmm, i have started digging through the modules in detail, i have to wonder how you came up with your resistor values in your op amps, as the values make no sense to my simulations,

as far as i can come up with, R45 and R47 should be 1.91K, with R46 and R48 being 39K so 0-4A corresponds to 0 - 2.048V or atleast well within tolerance,

also i dont undestand your gain of 40 on Vset, should it not be a gain of 20? in which case swapping out R53 with 2.05K  or parrelleling R54 with another 39K would make a rather accurate fix,

and IC42G$1 seems to have its inputs upside down,

there is also one or 2 more tweaks i'm looking into to better stabalise the current limiting but thats for later,

[markus_b]

as far as i can come up with, R45 and R47 should be 1.91K, with R46 and R48 being 39K so 0-4A corresponds to 0 - 2.048V or atleast well within tolerance,

also i dont undestand your gain of 40 on Vset, should it not be a gain of 20? in which case swapping out R53 with 2.05K  or parrelleling R54 with another 39K would make a rather accurate fix,

and IC42G$1 seems to have its inputs upside down,

there is also one or 2 more tweaks i'm looking into to better stabalise the current limiting but thats for later,

There might be errors and problems lurking there. I decided to go start anyway because changing the value of a resistor later in not a problem.

The internal reference voltage of the xmega is 1V. The external LM4140 is also 1V. That means all my analog stuff coming/going from the micro is based on that 1V. The DAC will go from 0-1V , the ADC will measure from 0-1V. All voltages are referenced on that volt.
In order to have some headroom and because I have 12bits (4096 steps) DAC/ADC resolution I decided to multiply the reference voltage by 40. So the digital part can go from 0 to 40 volt. Ideally this should be tuned to 40.96 Volt, then we can regulate the voltage in nice simple 10mV steps. But, having been in a hurry to get started, I have not gone through the calculation again to make sure that all values mentioned in the schematic are correct for this.

As some remarks: about the use of the different OP-Amps:
G$2: Amplify the (up to 1V) DAC output and feed it into the set pin of the LT3083
G$3: Decouple the current sense resistor (prevent R45/R46/R49 from drawing load being measured)
G$4: Amplify the (up to 100mV) current sense output to 1V to be the same as the DAC output
G$1: Comparator: As soon as the current sense output exceeds the DAC current voltage, limit the output via Q41. I think the polarity is correct, the ouput has to go up, when the positive input (current measure) is bigger then the negative input (current limit).

I've got the parts I've ordered last weekend form Mouser. It all arrived Wednesday. There was a problem with the transformer for the switcher, I got some other inductance. Probably a mix-up on my part. I'll order the correct one after the PCBs get here. There are probably more problems requiring parts.

[Strada916]

Hi Markus, Just wondering what IC51 LM334M is doing in the circuit? It only has pin one connected to R59 and pin 4 going to Vout no ground connection and also TP4 V-SW does not seem to be connected ?

Otherwise great little project I am interested to see how it turns out and when it is up and running maybe do a video blog on it? Once you have ironed most of the bugs I am interested in buying a board and software to build my own. Keep up the great work.

[markus_b]

Hi Markus, Just wondering what IC51 LM334M is doing in the circuit? It only has pin one connected to R59 and pin 4 going to Vout no ground connection and also TP4 V-SW does not seem to be connected ?

The LT308x has a small drawback: It needs a minimum load to be able to regulate correctly. The LM334 ist there to draw that minimum load. The LT334 is connected to Vout and ground (and R59).

TP4-V-SW is connected to the output of the switch (or the input of the LT308x regulation).

I don't know if I'll make video about it, maybe. But you'll definitely hear more from me as progress allows. Now I'm waiting for the 1st round of PCBs, but I suspect they will arrive while I'm gone for two weeks, so it may be mid-August before there are any news.

[Strada916]

ahh ok that clears things, thanks.

Good things come to those who wait.