DIY SCPI programmable dual channel bench PSU 0-50V/3A (now EEZ H24005)

[prasimix]

List of important changes

Update 2015-02-14: New power pre-regulator schematic is posted here.
Update 2015-02-26: Corrected bias power supply schematic (Sheet 1) and simplified post regulator schematic without TLC555/relay based bias control (Sheet 2) is posted here.
Update 2015-03-21: Spice models added for redesigned power post-regulator.
Update 2015-04-06: Modified post-regulator Spice model with new OE circuit is added.
Update 2015-04-08: New schematics for pre-regulator and post-regulator and new PCBs are presented here.
Update 2015-05-15: Mosfet pre-regulator board is assembled and tested.
Update 2015-06-20: THT Eagle files, BOM and Gerbers for Pre-regulator PCB r2B5 (06/2015).
Update 2015-07-29: New PCB for front panel designed as Arduino shield for Mega2560 or Due boards is introduced. An 3.2" color TFT touch-screen display replaces monochrome 128x64 LCD. This PCB makes obsolete previous MCU board, 4x4 keypad and BP PCB.
Update 2015-07-30: New AUX power supply with soft-start/stand-by is added that make previous obsolete.
Update 2015-10-03: SMD version of Mosfet pre-regulator board and post-regulator are assembled and tested.
Update 2015-10-21: SMD Eagle files, BOM and Gerbers for the Arduino Shield latest revision
Update 2015-11-01: A new thread about software support for this power supply is opened here
Update 2015-11-10: A repository for hosting hardware design files and source code is created and now available on https://github.com/eez-open
Update 2016-05-02: SMPS LTC3864 pre-regulator module

Redesigned PSU with smaller enclosure and SMPS pre-regulator
Update 2016-03-04: AUX Power supply board with fan controller, Ethernet and USB connection
Update 2016-03-04: New power board (pre- and post-regulator merged into one)
Update 2016-06-17: Redesigned set of PCBs for lower profile enclosure and hybrid power modules is available on the GitHub. This revision will be used for planned group buy.
Update 2016-09-03: Custom enclosure made by Varisom, Portugal arrived. Here you can see how it looks like.
Update 2016-09-06: Timgiles added Mouser parts into BOM. Please check GihHub here.
Update 2016-10-07: Firmware M2 that support PSU H24005 is released.
Update 2016-12-29: Firmware M3 is released (comes with support for channel coupling, tracking mode, etc.).
Update 2017-02-23: EEZ H24005 became a fully funded campaign on the CrowdSupply.
Update 2017-03-23: Firmware M4 is released (triggering system, list programming).
______________________________________________

Introduction
I’d like to present what I did until now trying to make a bench power supply that should have a solid “analog” core that can be enhanced by adding MCU for on-site or remote programming, monitoring and data logging. My understanding of analog circuits are basic and I tried to compile many ideas found on various places. Initial document was Agilent E3634A Service guide (Part No. E2634-90010, April 2000) and presentations found on i.e. Gerry’s and Ian’s blogs.
Presented work is still not backed by any simulation (I still wonder which one to choose, it that LTspice or something else, your recommendations are welcome). I can imagine that such simulation could save considerable amount of time and material.
There is many DIY power supply projects available (as cheap yet solid enough PSU that comes from various manufacturers) but I decide to try to make one by myself to learn a little about various circuits and parts that are now available on the market. The result of such quest is that I include many components that goes far beyond 2N3055 or LM317 based solutions but since we are in 21 century and logistics is amazing and since my budget is not so tight I think that I should proceed in that way. As a “reality check” to not end up with some really “esoteric” parts all what was selected has to be available through at least one of two European supplier (see below).
It’s also interesting that in DIY projects Vout of 30V is as some kind of magic threshold which is not cross over very often. Maybe it’s just a practical reason for that since most of today equipment do not need anything beyond 24V?
Situation with ideas about digital part and programming is better and many parts that is added on MCU board was inspired by my own view what kind of functionality programmable bench power supply should have. But this is a next step. The first one is to create solid core that can survive various testing torture and do not call into question safety/integrity of the various loads/equipment that could be connected somewhere in the future.
I spent many hours “breadboarding” and end up with complete mess (see picture of my desktop). Thanks to that many components was burned when I touch wrong wire or components or apply wrong voltage So, it’s high time to put what is done on PCB. PCBs (see below) will be successors of the first batch where I mainly tested SMPS functionality that was previously unknown to me. On the beginning I tried to test LM5005 for bias, LM5118 for pre-regulator and LM5119 for both functionality since it is dual channel. Unfortunately I never succeed to do any testing with the latest for the simple reason that soldering QFN without reflow oven which I still don’t have became mission impossible despite some encouraging videos such as . Maybe the reason for my failure is not using of a solder paste but even if I succeed I start to wonder what kind of nightmare that may be for the other people who eventually decide to follow this project. Fortunately soldering (with the right soldering tip) of even tiniest SMD ICs (such as 5-pin SOT-23) become surprisingly simple.
In this first post I’d like to present some basic facts about projects such as goals, features, schematic  and PCBs that I’m going to order soon. I hope that will generate discussion that will further improve this project even if that not necessary simplify something that one can already found too complex Maybe we even ends up in not so distant future with group buy, who knows. Here we go:

Project goals list

• “DIY friendly” – use THT components as much as possible, sockets for some ICs for experimenting with various solutions. Combine pinouts on PCBs of some components to host both SMD and THT variants, use jumpers and test points for simplified measurements
• Dual or single channel version
• Regulated voltage (CV) 0-50V, resolution 10mV
• Regulated current (CC) 0-3A, resolution 10mA
• SMPS buck-boost pre-regulator with output tracking control for lower power dissipation
• N-ch MOSFET based post-regulator
• Flexible main transformer output voltage rate (from 36-48VAC) using buck-boost mode of operation (output power when Vout > Vin limited mainly by allowable power inductor current)
• Main power inductor could be from various brands or custom made
• Over-current protection (OCP) for short circuit condition
• Easily modified for other ranges i. e. 0-30V or 0-5A
• Basic functionality has to be achieved with manual control without MCU board (set voltage and current, CC/CV indication, output enable, OCP reset)
  
• AC input protection (SAR, MOV, TVS)
• DC output protection (SAR, MOV, TVS)
• Simple serial (0-100V/3A) or parallel (0-50V/6A) connection of two power modules with tracking and LED indication (MCU controlled)
• Simple connection of Sense+/- outputs with LED indication
• Modular design that modules can be used in other projects
• Completely isolated MCU board signaling from analog section (Main power board) and PC
• SPI communication
• USB interface with ESD protection
• Ethernet interface with ESD protection
• Programming by Arduino sketches
• PC console for settings and monitoring
• Not using fan or other source of audible noise
• ADC and DAC resolution of min. 14 bits
• No power up or power down overshoot on the output (with or without load)
• Generate bias supply from main transformer (no aux transformer required)
• Single LCD 128x64 dot matrix for displaying all information for both channels
• 4 x 4 keypad + rotary encoder when controlled by MCU
• All components have to be available on Farnell element14 UK and/or (Transfer Multisort Elektronik (TMU.eu)). Presumably all selected components could be also found on Digikey and Mouser.
  
• Fit into metal case such as Modushop.biz (Hi-fi 2000) Economica EP1152820 L 280 x P 200 or Economica EP1153220 L 320 x P 200). Another possible source is Schaeffer (offer customized front panels and Alu cases).
  

PCB list

• Power board 0-50V/3A (two is required for dual channel version) – one PCB is mandatory for basic operation
• MCU board (piggyback on LCD) – offer MCU controlled mode of operation and PC connectivity
• Aux power (+5V) and AC input protection – supply for all MCU related PCBs
• Binding posts with connection LED indicators and DC output protection – optional, require MCU board
• 4 x 4 Keypad – mandatory for MCU mode of operation
• Ethernet + USB sockets board with protection – optional (can be replaced with some panel mounted sockets)

Modules feature list
Power board
Schematic on Sheets 1/10 to 4/10

• High voltage TI LM5118) buck-boost pre-regulator
  
• Support for the following power inductors: Vishay IHTH1125 series (THT),  Bourns PM2120 series (SMD), Bourns 2300 series (THT), Murata 1400 series (THT), WE-HCI Flat Wire High Current Inductor (SMD) and vertical ETD 29/16/10 ferrite core (i.e. Epcos) for custom made inductor (use i.e. ExcellentIT inductor calculator, Inductor5200.exe)
  
• High voltage TI LM5574) bias SMPS and low noise positive LDO LTC (LT3030) for generating +15V and +5V and low noise negative LDO TI (TPS7A3001) for generating -15V (Ripple rejection should be around 50dB for selected SMPS working frequency of ~215 KHz!).
  
• TO220/FP power N-ch MOSFET and diode for SMPS for simple mounting on shared external heatsink
• Op amp based CC and CV control loops (TL074 is default, can be tested with other quad op-amp in DIP14 package)
  
• LTC6102) Precision Zero Drift Current Sense Amplifier
  
• Op amp based tracking control (inspired by the following design: Simple switchers make simple pre-regulators)
  
• DAC, 2-channel, 16-bits (TI DAC8552) provide U_SET and I_SET values simultaneously
  
• ADC, 4-channel, multiplexed, 15-bits (TI ADS1120). Two channels are used for measuring U_OUT and I_OUT another two are connected to U_SET and I_SET that comes from DAC and what can be used for some sort of auto-calibration.
  
• Precision +2.5V voltage reference with REF01 industry standard DIP8 socket for LTC LT1019-2.5) (comes with SMD adapters for LTC LT1790-2.5 or  LT6654-2.5, TI REF5025 and TI LM4132)
  
• SMPS controllers frequency synchronization with isolation (Silabs Si8610) between Power boards and Aux power board for lowering EMI
  
• Test points for accurate measuring of input and output voltage and pre-regulator control loop and post-regulator CV control loop as suggested in TI snva364a
  
• Connector for U_SET and I_SET potentiometers (i.e. 50K multiturn) are provided in the fashion that manual or MCU control cannot be done simultaneously
• Connectors for OUT_ENABLE and OC_RESET switches
• Connector for OC LED
• Connector for CC and CV LED
• Connector for digital panel meter (eBay grade or perhaps from here) bias and U_OUT for manual operation (when MCU is not used)
  
• Buzzer for audible notifications and warnings
• SPI I/O expander (Microchip MCP23S08) is used for minimize required number of I/O lines (which has to be isolated!) between Power board and MCU board. Currently 7 out of 8 lines are occupied for CV mode (input), CC mode (input), OC reset (I/O), Power good (I/O), Pre-regulator enable (I/O), ADC DRDY (input) and OCP_ACTIVE (input).
  

Current status:
Most parts are tested. That also include basic ADC/DAC functionality (set Vout and Iout using DAC and measuring Vout/Iout using ADC). Require additional testing for pre-regulator tracker when pre-regulator enter buck-boost mode (Vin ? Vout). Currently design include a silly way of achieving power up and power down without overshooting regardless of the fact that load is connected or not. That is mainly because a bias supply with separate transformer (LM317/7805/LM339L) is used when experimenting and power-up sequencing is not correct (correct biasing comes too late despite of fact that pre-regulator has huge soft start of ~40ms).
During testing a LM5118 based pre-regulator was on the first PCB version. Power inductor in buck-boost can be under real stress and some of them (Bourns PM2110-820-K-RC1240) can reach, I believe due to the saturation (tested with Vin=40VAC and  Vout=50V/3A), a temperature of 90oC or even more! The most flexible solution is playing with own inductor winded on the ETD ferrite core.
Low noise LDO (LT3030, TPS7A3001) in combination with SMPS pre-regulator (LM5574) was tested on breadboard only. That was pretty challenging and real performance (ripple rejection) was not possible to test.
For load/stability testing the following parts are used: 2 x 8R2/100W, 2 x 33/25W power resistors and 4 x 12V/35W halogen lamps. Tested with full 150W and beyond that (approx. 200W). Testing of two channels will be started when PCB arrived (then will be possible to test SMPS frequency sync and serial and parallel connection). I’d like to extend stability testing using 100R/25W resistor and DC motor (~100W).

(Removed description of obsolete MCU board, 4x4 keypad, Ethernet + USB sockets, Binding posts board and Aux power supply due to message limit of 20000 characters)

[prasimix]

Sheets from schematic (PDF) as separate image files ...

[prasimix]

Here you can find some measurements done on pre-regulator (assembled on first PCB revision) with pure resistive load (two 8R2/100W in connected in series). Vin=40VAC therefore for the upper limit LM5118 is in buck-boost mode that generate more output ripple then is buck mode. For measurements are used simple short ground probing (no differential probing with two channels).

Channel color legend:
Yellow - Vout ripple (AC coupled)
Cyan - Vout (DC coupled)
Magenta - HI mosfet Gate
Blue - LO mosfet Gate

Power inductor is approx. 100uH (custom made on ETD29/16/10 ferrite core).

[diyaudio]

Good job I admire your efforts. 

Why the over complex design? a goal with the same specifications can be achieved with a quad op-amp, some regulators and references for the error-amplifiers and power mosfets.

[prasimix]

Additional possibly interesting measurement what I can provide in this moment is how PSU follow DAC instructions. Please note that measurement was performed with post-regulator section build on breadboard. DAC command was sent using Arduino in infinite loop sequence of three U_SET value each 5, 20 or 50ms long. Connected load was again 16R4 (2 x 8R2). As you can see there is no overshooting nor undershooting presents. Falling edge is possibly too slow. This is done with 100uF capacitor connected on the output. Channel color legend is the same as in the previous post. HI and LO gate signals frequency is too high to be seen correctly.

[prasimix]

Good job I admire your efforts. 

Why the over complex design? a goal with the same specifications can be achieved with a quad op-amp, some regulators and references for the error-amplifiers and power mosfets.

Thanks for your feedback. I'm open for any suggestion as long as it can provide 0-50V/3A with isolated low voltage transformer (no AC-DC pre-regulation, I'm still not ready to cope with 230V or 400V from PFC), stable operation (no overshooting regardless of connected load), quit operation (no cooling fan) and isolated and simple communication with digital part.

[Alex30]

Very impressive prasimix. Way beyond my level of understanding but based on your specs it sounds as though you have made a top of the range PSU. I will be following this with interest!

[prasimix]

Very impressive prasimix. Way beyond my level of understanding but based on your specs it sounds as though you have made a top of the range PSU. I will be following this with interest!

Thanks. Maybe it looks a little bit too complex but for basic functionality you need only what is presented on the first three sheets. Even there Pre-regulator and bias circuity could be far simpler (Sheet 1) but I really want to play a little with new stuff and hopefully get i.e. low noise solution despite of inherently noisy SMPS environment. LM5118 based pre-regulator alone in buck mode already shows very nice results (at least far better then what I heard from a friend who told me that I cannot expect anything better then 100mVpp of Vout ripple with SMPS alone).
I'll add more information when I continue works on PCBs and make some real testing with MCU board activated.

[steve_w]

Are you doing the boards?

regards

SW

[prasimix]

Yes, I posted PCB layout by mistake in separate thread and asked moderator/admin to move it here.

[prasimix]

While waiting for PCBs I spent some time to organize my halogen lamps based dummy load in a proper way (holder of the first one was almost melted by high temperature). Here are some pictures if some find such approach useful.

[prasimix]

I spent some time to test how is working Front Panel Designer (v4.41 for 64-bit linux) from Schaeffer AG. It's really nice and simple tool covering everything that I can imagine in this moment. Great feature is built-in price calculator. Unfortunately not so nice "feature" is the total cost of simple design . Here is what I have in mind for MCU controlled and manually controlled dual channel unit. I took as a reference dimensions of Modushop's Alu-cases (with front panel's of 320x115mm and 280x115mm respectively).

The first design costs 73.91EUR (87.95EUR incl. VAT, excl. S&H) and the second one 109.74EUR (130.59EUR incl. VAT, excl. S&H) . Does anybody recently made an order that can tell me how accurate is that price calculator?

[prasimix]

Here is a few suggestions that I got so far as possible alternative route for excellent but pricey Schaeffer solution. The first two are addressing cover labels, the latest one looks as ingenious total solution (at least for the smaller front and back panels):

• Laser printing with polypropylene (PP) folder
• Waterslide decal (described as Method 3) with Lazertran paper

[ali80]

interesting project, keep up the good work

[Alex30]

Those decals are really great but if it was me I would go the cheaper route and get some of those translucent sticker sheets off ebay and then either cut out the label region or carefully cut out the individual letters\symbols depending on what look you wanted, either that or go with option 2 which is a bit more expensive. But yeah, finding cases and workable front panels is an absolute pain, which is why most of my projects are not even boxed because I lose interest! Since I've got into electronics I have always regretted the fact that my mate's dad at high school sold his industrial CNC machine that he used to make cool metal designs on. I've been thinking of getting a 3d abs printer just for easy case designs because it's all just too hard basing your needs on random case sizes you find on these random websites.

[senso]

Following this thread, seems interesting.

About the front-panel, maybe 10mm thick acrylic laser cut and then a vinil decal on top for lettering, would be cheaper, as almost every big printing services also have acrylic laser cutters.

[prasimix]

Thanks Senso for your input. Acrylic or any other non-metal material has a small disadvantage when we comes to the question of EMI. The PSU is using SMPS as a pre-regulator and good shielding has to be in place. Ok, acrylic panel can be backed with simple single layer PCB from inner/hidden side.

[senso]

If you can export/re-design your front-panel in SVG/vectorial format there are some online laser/milling services, but usually they are all US based, and shipping to Europe as usual kills the deal.
There is also www.mfg.com , it might look sketchy, but I know people that used the service and where satisfied both in quality and good prices, it might be worth a shot.

[Alex30]

Thanks Senso for your input. Acrylic or any other non-metal material has a small disadvantage when we comes to the question of EMI. The PSU is using SMPS as a pre-regulator and good shielding has to be in place. Ok, acrylic panel can be backed with simple single layer PCB from inner/hidden side.

I don't want to join the tin foil society just yet but I was wondering if tin foil glued onto the interior of a plastic case could provide effective shielding?

Edit
I also envisaged the shielding to be bonded to ground btw

[codeboy2k]

I don't want to join the tin foil society just yet but I was wondering if tin foil glued onto the interior of a plastic case could provide effective shielding?
Edit
I also envisaged the shielding to be bonded to ground btw

I've never tried it with Al foil, but I think it could work. I've effectively used paintcans and tin cookie containers with good success. They are tin plated steel and thicker than Al foil, though.



See Wiki here too.
http://en.wikipedia.org/wiki/Aluminium_foil#Electromagnetic_shielding

[uChip]

[editing for clarity] Regarding how to create nice enclosures and front panels for this (or any) project.  Not too long ago I purchased and built the function generator kit shown in the first picture.  Let's not get into the quality (or lack thereof) of the kit.
first picture: Function Generator from JYETech.com
The kit did not come with an enclosure but did come with a pretty nice front panel (silkscreened acrylic) so I designed and printed and enclosure for it.
second picture
I wanted to design nice looking front panels for some other projects so I did some research.  Most of the solutions I found were too expensive, too error prone or required too much artistic crafting (soldering is about as skilled a handcraft as I can do).

What I came up with looks like this:
third picture
The front panel is a sandwich of laser cut acrylic sheets.  Cost is about $15 for the laser cut acrylic.  What is perhaps unique is that the front panel design is created in Eagle CAD.  The front panel is extracted from the PCB design.  The key is an Eagle library that has the front panel mechanicals included as part of the component definition.

For shielding you could 1) paint the interior of the enclosure with a conductive paint 2) apply tinfoil (sort of like gold leaf) to the interior of the plastic 3) enclose appropriate parts of the circuit in a metal box fastened to the floor of the enclosure, or I'm sure there are other solutions.

If you'd like more info or to download the library see this forum thread https://forum.sparkfun.com/viewtopic.php?f=5&t=39492 and my github repo https://github.com/StdBx/Series100.

  - Chip

PS. Rereading this, it comes off as a sales pitch.  I'm not selling anything, but I am publishing my work so that others might have an easier time doing something similar.

[prasimix]

Thanks uChip for contribution. I'll keep it in mind as time to make a final decision about housing comes.

[prasimix]

PCB's arrived yesterday or within 2 weeks from moment of ordering. A nice supplier (printed.cz) which I'd like to recommend.



Here is the main board where is located all circuit from Sheet 1 to 4 namely input bridge rectifier and capacitors, power pre-regulator, power post-regulator (CV and CC), bias supply pre-regulator and LDOs, OCP, CC/CV indicators and isolated sync for pre-regulators. That is a functional analog part of the PSU. On the same board is also located digital part that is used to communicate with MCU board via SPI. That mean DAC, ADC and I/O expander.

[prasimix]

Companion PCBs to the power board looks in this revision like this:

[Kevin.D]

Nice project. I can see you put alot of effort into it, is it your schematic or from an already  published/tested project ? The reason I ask is that I think there may be a stability/CC performance issue due to the way Q4 is driven. Do you have a spice sim available to post (just the basic main control loop components is all thats needed to show if this is so Q4,Q5,Q6, IC7A-IC7D and associated small  stuff) .

Regards

Edit .I should have read the other few posts above I  see someone says it's a kit from  JYETech.com .Still do you have a sim available ?