EEZ Bench Box 3 (BB3)

[prasimix]

I just finished new PCB layout and commited it at the GitHub. I'd going to send it to ALLPCB for manufacturing in the course of next week. If someone possibly have any comment on it please let me know.

[prasimix]

While waiting for new PCB for CF-DIC I've managed to assembly the first prototype for new MCU board based on STM32F769IGT6 in LQFP176 package. This board is intended to be mounted perpendicular to the chassis front panel with exposed sockets and connectors for Ethernet, I/O port, Micro SD card, USB, encoder and user switch (LCD TFT connector will be hidden behind front panel).
Project files are available in the new repository on GitHub. Currently MCU board take care about the following peripherals and functionality:

• Digital I/O: 2 x protected inputs, 2 x protected outputs
• Rotary encoder with switch
• 1 x user switch
• 3 x SPI channels (2 x Chip selects per channel)
• Battery backup (CR2032 button cell type)
• USB 2.0 OTG
• Micro SD card
• Ethernet PHY (10/100 Mbit/s)
• I2C EEPROM
• JTAG IDC 20-pin connector
• SDRAM (e.g. IS42S16400J)
• TFT LCD with 0.5 mm FFC 40-pin connector (e.g. RFE43BH-AIW-DNS or RFE430Y-AIW-DNS)
• TFT backlight brightness control
• Resistive touch controller
• Soft-start/stand-by AC power control
• Audio amplifier with small on-board speaker

Selected MCU is capable of driving directly TFT LCD with SDRAM used as frame buffer (one or more). I also include 100 Mbit Ethernet and still have enough pins for up to 3 SPI channels and various other I/O. PCB is 4-layer and despite that I didn't manage to provide optimal connectivity between MCU and SDRAM chip, but it seems that everything works as shown on picture that follows (Nucleo board on picture is used just as debugger with SWD interface). I have to leave it overnight and make additional memory tests to check if existing layout is usable or a new revision is required.
The MCU board has 40-pin connector on the right side where 3 SPI channels are available to communicate with up to 3 peripheral module using additional "backplane" PCB that will be introduced in coming weeks.

So far we tested TFT LCD, SDRAM and USB (we acquired USB VID:PID for free thanks to pid.codes).

We found one issue that is I'd like to believe due to our inexperience with STM32: we cannot generate 25 MHz clock from PLLI2SR that is exposed on MCO2 output that is reported here. Without that Ethernet cannot be tested. If we failed to run it with 10 MHz as master clock, it is still possible to change master clock (HSE) to something else and use that HSE as input for the MCO2 MUX to get required 25 MHz.

I'll continue to report about progress of running up this prototype PCB (r1B1) and about migration of EEZ H24005 firmware to STM32 platform. Your comments and suggestions are welcomed as usual.

[Rerouter]

I should point out your timing margins for your gerbers where on the edge of the margin, but still a smidge inside them when calculated from the setup and hold times, it only violates the manufacturers requested trace matching,

I suspect it would only really kill your memory access reliability in very humid air when the velocity of the signals reduce a little.

[prasimix]

We continued to test MCU board and found few small issues on the PCB that is reported on the GitHub's issue tracker. Situation with using MCO2 output for generating 25 MHz clock from 10 MHz master clock input is still unclear, I didn't receive so far any answer is that possible or not. Therefore I've changed master clock xtal to 25 MHz and use HSE as input for MCO2 mux and Ethernet is now functional.

The new AUX PS board is also finished and ready for testing. It include soft-start/stand-by as previous version but for up to 3 power boards. NTC instead of power resistor is used for soft-start due to smaller profile (has to be below 30 mm vertically). I have to test if it can survive charging of up to 6 x 120 uF.
Instead of Vigortronix AC/DC dual output module (5/12W, max. 5W) a VIPer35 QR flyback is employed with off-the-shelf transformer (Feryster or WE). The PCB is intentionally much wider (210 mm) since it include power AC inlet and power switch to reduce required wiring even more. It also carry standby LED and PE 4 mm banana binding post (not mounted on the picture below) that will be exposed on the enclosure's front panel.
Finally, a dedicated fan controller IC with I2C bus is used to assist MCU with driving fan with PWM and measure fan speed and ambient temperature.



Its schematics looks like this (linked from GitHub repository where BOM, Eagle and Gerbers files are also available):

[prasimix]

Yesterday, I finalized PCB layout for the new Power module that is successor of the Power board used in the EEZ H24005 power supply. The working name of this module is EEZ DIB DCP505 and its design files is now available on the GitHub (the BOM section is still missing). Its PCB form factor (160 x 95 mm) is modified in line with new EEZ DIB chassis that I'll announce soon, and comes with few new features. The major and most interesting new feature is 3-range current auto-ranging circuit. It is proposed by my friend Macola after I struggling for some time with another approach that require presettable up/down counter and require more digital I/O lines for both set and read selected range.
This one require a little bit more parts but from other side it offer simple and elegant way how to set max. output current (I_SET) and read actual current (I_MON) both with information about active range (CURR_50MA, CURR_500MA, CURR_5A outputs).

The auto-ranger is designed to cover current from 0 to 5 A with automatic switching between measurement current sense resistors to achieve max. reading resolution that will be in our case 15-bit for each range, that is theoretically 0.152 mA for 5A, 15.2 uA for 500 mA and 1.52 uA for 50 mA range. That give us enough room to provide resolution of 1 mA, 100 uA and 10 uA respectively.The simulation files are attached.
Setting output current is simple: on the I_SET input one has to apply control voltage from 0-2.5 V. The control voltage range is divided in three equal parts i.e. 0-0.833 V, 0.833-1.666 V and 1.666-2.5 V. In our case with 16-bit DAC that give us possibility to define max. current for the selected range in 21845 steps (2^16 / 3) that is approx. 0.228 mA for 5 A (5 / 2^15), 2.28 uA for 500 mA and 2.28 uA for 50 mA range.
The schematic and LTspice simulation is shown below. Simulation tested the circuit with output current (I_load) that rise over the whole range (0-5 A) and with output current set with 1.5 V that put it somewhere in the middle range (500 mA) that can be seen on the latest graph (I_ctrl) when error amplifier change its output state. As output current rise range indication outputs state are sequentially changed from 50 mA over 500 mA to 5 A.

The interesting detail is using of the D1 Schottky diode that is connected in parallel with range switching MOSFETs (M1, M2, M3). Thanks to that diode we have more freedom to choose MOSFETs with extra low R_ds,on (i.e. below 5 mOhm) that is much cheaper and comes in much smaller package for V_ds,max that shouldn't be larger then 1 V . A prerequirement that everything works fine is that voltage over D1 never exceed ~200 mV when it starts conducting and influence measurement. That shouldn't be a problem since selected current sense resistors value gives max. 50 mV of voltage drop for the full range (e.g. 5A * 0.01R = 50 mV).

The current measurement op-amp (for ADC) in this circuit is U5, that together with U6 inverter should be precise, low-offset and wired with precise resistor to achieve high precision. I'm going to use OPA4197 for testing. A small offset (above zero) is intentionally introduced to U5 scale using R47 for calibration purposes, i.e. that DAC don't need to deliver negative output for 0 A value. Also the selected gain of U5 give us full-range value below 2.5 V, again that is intentionally done for the calibration purposes that DAC don't need to deliver "overflow" value above 2.5 V for calibrated output. 
Another important op-amp that should be precise is U3 that provide measure current value to U2 and U10 that are in charge to switch between ranges.

The simulation files are attached. Please note that parts used in the simulation differs to some degree with one that I'm going to use and test on the prototype PCB.

[prasimix]

Other new feature on the Power module is "heavy-duty" OVP with triac and two fuses that should offer over-voltage protection in both direction: for the power supply and connected load. Theoretically it should manage and survive few hundreds of Amps, e.g. when 10000 uF charged to 50 V is connected on the output and discharging. How nasty such load could be is presented in the following spice simulation (zip file attached down below):



The max. current peak (440 A !) and its shape is defined with R2 and L1 and their value should be selected in accordance with what chosen triac could survive. The on-board OVP circuit looks like this:



Note that L1 inductance from the simulation is implemented directly on the PCB as spiral trace (see PCB bottom layer below). Its designed in accordance with the following calculation for number of turns and trace width:





The R2 from simulation that is R110 on the schematics is THT and could survive huge current peak (See AC01000001007JA100 datasheet).

OVP triggering threshold is set to 10% above programmed output voltage (U_SET). When triggered, triac will short power output lines and stay shorted until power is recycled or voltage above its A1/A2 pins drops down. If we presume that OVP is not triggered by Power module internal failure, we can reset it by simply set 0 V as new output voltage or by disabling OE (Output enable) circuit.

The new Power module also comes now with on-board SPI temperature sensor mounted beneath heatsink for main pass MOSFET and down-programmer MOSFET. It is thermally coupled with 5 mm thick thermal pad.
Module specific information such as calibration data or number of working hours, etc. can be now stored into on-board I2C EEPROM and makes them completely portable. Therefore, for example, it is not needed any more to recalibrate a module when it change its position inside the EEZ DIB chassis or when its moved into another chassis.   
Finally this Power module is intended to be powered with CF-DIC power pre-regulator and bias power supply that is mentioned at the beginning of this thread and its recent progress is covered here.
A whole Power module is packaged on the PCB shown on the following two pictures (thanks to the OSHpark PCB preview feature):

[prasimix]

This is reply to the question asked by one forum member which PM box is full, and I cannot reach him

The DCP505 is part of the new power supply that has to be completed in the coming months and I have to order first batch of board in coming days. It also require CF-DIC as pre-regulator and does not use Discovery but our new MCU board (STM32F7 based with larger 4.3 TFT) that is powered by new AUX PS board. The latest files can be found on the GitHub repository (currently backplane and new chassis design is missing). Feel free to subscribe on it to stay informed about progress. Please also note that when we finished with development and testing that is highly probable that a new group buy or crowdfunding will be organized.

EDIT: Nucleo board shown in post #51 is used only as debugger/firmware downloader, our MCU board is beneath TFT display. USB and/or SD card firmware downloader will be used in the final release, instead.

[prasimix]

A small update of what's going on here. The power module (DCP505) PCB has been arrived and I found instantly a stupid error on it: the power inputs on one of quad op-amp was swapped . Fortunately that wasn't a big deal to fix and I was focused on testing two new features: on-board OVP and three range current auto-ranging circuit.

The OVP behave more or less as expected, but main reason why I am going to replace it is lack of proper latching. When I've designing its PCB I still didn't know that Maxim has some interesting comparators equipped with latch functionality and which does not in the same time cost a fortune. Therefore I'd like to try their MAX9141 in the next PCB revision.

The auto-ranging circuit is first tested in CV mode, and it works as expected. Possibly the switching between ranges could be further improved by decreasing R_gate value for switching MOSFETs.
Unfortunately it doesn't works correctly when change in current range is happen in CC mode when current value is somewhere in-between two ranges. It is especially obvious on lower two ranges (50 mA and 500 mA). The problem lies in part of the circuit that is in charge for setting control voltage range (U11, U12, U13, U7 in schematic posted in #54) and which directly affect the CC control loop and switch back and forth I_SET value when output current is just in-between two ranges resulting in output current oscillations. That section was added to increase precision of I_SET (i.e. that output current can be set in 21845 steps for each range). Currently I don't have any idea how to make auto-ranging circuit stable in CC mode then to remove that section and use ~76.3 uA (5 / 2^16) as the smallest step in all ranges. That means that resolution of defining current in middle range (0-500 mA) will be 10 times worse, and 100 times worse in the low range (0-50 mA) then in high range (0-5 A). Perhaps that is still acceptable since auto-ranging circuit is primarily of interest to improve precision of measured current value.
The new auto-ranging circuit should now looks like this (available also in attachment):

[prasimix]

Since I have now all PCBs/modules defined for completing a new power supply, I can present a new chassis/enclosure that I'm discussing with enclosure manufacturer (again Varisom from Portugal) which can house up to three modules (i.e. channels). While designing it I was struggling with lots of dilemmas mainly generated by effort to make it as compact as possible that it doesn't occupy to much space it it is going to be placed on the benchtop. I spent some time trying to organize everything inside it that it can be placed in two ways: horizontally and vertically when all content on the TFT display have to be rotated for 90 (or 270) degrees. I gave up for the moment of that concept because it will require introducing of new (soft?) material for making “feet” all around, on all four sides not just one, and that will additionally prolonged firmware development.
The new enclosure should house the following items:

• MCU board
• AUX PS board (for powering MCU board, 12 V DC fans, and “non-isolated” devices on the peripheral modules that are DCP505 power module for now)
• DCP505 power module (up to three)
• BP3C backplane for connecting MCU board with peripheral modules
• Up to three power pre-regulators (presented here) for powering DCP505
• 4.3 TFT touch-screen display and
• Two 12 V DC fans

To get an idea how parts mentioned above are related to each other, I'll post first pictures of the working prototype that is currently used for firmware development:





The MCU board is placed horizontally just as backplane connected using the 40-pin (2x20) 0.1” connector. Such separation lowering the cost of PCBs production since backplane can be now 2-layers, while 4-layers can be used for more demanding MCU board. I've gave up from idea that backplane has to be mounted on the enclosure's rear panel, and decide to mount it on the bottom plate where modules have to be inserted as in a same manner as in PC (i.e. that require that top/cover plate has to be removed before module is inserted/removed).
The backplane allows connection of control lines between MCU and peripheral module using the 26-pin (2x13) 0.1” connector (black receptacles visible on the rear part of the backplane). I have also decide to keep possibility for coupling power outputs in series or parallel for modules on first two slots. Hence 20-pin (2x10) 0.1” connectors and place for two power relays on the backplane (currently jumpers are used instead of them). That will allow MCU controlled coupling for up to 100 V (in series) or up to 10 A (in parallel) that will be available on the output terminals on the module inserted into first slot. OE (Output Enable) LED is now bi-color for indicating that output is active and not coupled (Green) or active and coupled (Red). That 20-pin connector on the third slot is “dummy” used just to short post-regulator power outputs with output terminals on the PCB front side (4 mm, 19.05 mm spaced red/black connectors).

The power module (DCP505) has exposed all LEDs and connectors on its PCB front edge where front panel metal plate will be mounted with two M3 bolts to the PCBs and another two to the enclosure front panel. The cooling of power (pass) MOSFET and down-programmer MOSFET is achieved with mounting heatsink directly on the PCB fixed with three M3 bolts. Heatsink on the picture is “off-the-shelf” SK 624 50 AL. I expect that will be enough for up to 2 A of output current (~ 5W of dissipation) without need to turn on fans at all. An “issue” with this heatsink is that one has to make five M3 threaded holes that is more challenging than expected since fins are very dense (I was ruined one heatsink because drill tip broke while drilling and remain blocked in the hole). Therefore I have to check with Varisom if they can offer similar solution as one used for EEZ H24005 power modules).

The AUX PS module as shown on pictures are not on the right position. It will be mounted (vertically) on left from TFT display (and MCU board). Hence the 16-pin cable can be much shorter. Its PCB width is as wide as enclosure inside depth (230 mm) and that will reduce needed wiring since power (AC mains) switch, standby LED and PE 4 mm connector are exposed on the enclosure's front panel and IEC inlet (with two 20x5 fuses) on the rear panel.

Now we can see that the final width for three modules/channels enclosure is defined, starting from the left, with thickness of the AUX PS module, TFT display/MCU board width and 3 x 35 mm peripheral modules. That give us 290 mm in total. The height of the enclosure will be 119 mm (that is within 3U height). Enclosure can be carry around using 3248.1001 handle mounted on the top.

The first proposal for enclosure's front and rear panels looks like this:





The following items mounted on the MCU board will be exposed on the enclosure's front panel (its working name is Bench Box 3):

• RJ-45 for 10/100 Ethernet (this is not optimal, a rear panel is perhaps better place for it but that would require additional wiring)
• 2x digital inputs i 2x outputs (protected) used for external triggering and syncing
• Micro SD card 
• Mini USB OTG
• Encoder with Ø31x16mm knob
  
• User SW (programmable for various purposes)

I'll continue commiting changes about all parts of the project on GitHub repo here and your comments and ideas are welcomed as usual.

[alex-sh]

@prasimix Very nice layout! I like it.
What touch screen are you using please?

[prasimix]

The idea is that various (any?) display with 40-pin 0.5mm connector (RGB interface, 12 o'clock viewing angle) can be used. Currently I've tested Raystar Optronic's RFE43BH-AIW-DNS. Display on the picture is Riverdi's RVT4.3ATFWR00 but unfortunately that is a 6 o'clock viewing angle display and cannot be used in final product.
I got good recommendation for EastRising displays available on buydisplay.com and eBay. The price is fantastic, and sufficient reason to test some of them like ER-TFT043-8 or ER-TFT043-3.

It is important to mention that most manufacturers offers the same TFT displays with both resistive and capacitive touchscreen. Currently we are using resistive that is only supported by selected TSC2007 controller.

[alex-sh]

The idea is that various (any?) display with 40-pin 0.5mm connector (RGB interface, 12 o'clock viewing angle) can be used. Currently I've tested Raystar Optronic's RFE43BH-AIW-DNS. Display on the picture is Riverdi's RVT4.3ATFWR00 but unfortunately that is a 6 o'clock viewing angle display and cannot be used in final product.
I got good recommendation for EastRising displays available on buydisplay.com and eBay. The price is fantastic, and sufficient reason to test some of them like ER-TFT043-8 or ER-TFT043-3.

It is important to mention that most manufacturers offers the same TFT displays with both resistive and capacitive touchscreen. Currently we are using resistive that is only supported by selected TSC2007 controller.

Raystar or Riverdi are not cheap. At this price level, Nextion (Serial interface) maybe a good alternative.
The other displays you mentioned (EastRising?) are really well priced. At < $10 for 4.3" display is no brainer.

[prasimix]

At this price level, Nextion (Serial interface) maybe a good alternative.

Right, if its interface/controller works flawlessly, but still that approach make a whole thing less open source .

[alex-sh]

Yes, Nextion does work flawlessly and extremely fast (has got its own processor)

[bloguetronica]

That is very nice! Thanks for sharing!

Kind regards, Samuel Lourenço

[LapTop006]

The first proposal for enclosure's front and rear panels looks like this:

Neat. My initial response is "can you make it a touch shorter and do a full rack width case", but then I thought again. On my home bench at the moment is a gap where a HP E3631 used to be until it died. I actually have an Agilent N6705A sitting under the bunch to be a replacement, but haven't put it in place, in part because it's so big.

Looking forward to when I can put money down, your UI is better than the N6705A.

[prasimix]

Thanks, I believe that with new MCU and bigger display we'll make UI even better. We also learn something about usability of the existing PSU (i.e. EEZ H24005) that can be further improved.

[prasimix]

A long time ago a request for multiple fixed outputs for digital design was posted on EEZ H24005 GitHub repo. I'm wondering how much outputs and what voltages/current such power supply module should provide? Does some of them should be variable to some extent? Does current auto-ranging for stand-by consumption measurement makes sense?
The newly proposed chassis could carry up to three modules. I believe that two fully featured (DCP505) modules is enough and that leave one place vacant for a multiple outputs module.

Thanks in advance for your valuable inputs.

[prasimix]

The 3D model for new chassis is more or less finished and a couple of screenshots is shown below (top cover is set as transparent). A 3D PDF file can be found in attachment (Adobe reader, as far as I know is still the only one who can display 3D drawings):

[Aigor]

Cool! Fusion360?

[prasimix]

No, SolidWorks made by great guy Mario from Varisom, Portugal (based on my basic DXF drawing made in Eagle). Hopefully they'll make it right in the first try, there is so many details that we tried to check via Skype .

[nimish]

A long time ago a request for multiple fixed outputs for digital design was posted on EEZ H24005 GitHub repo. I'm wondering how much outputs and what voltages/current such power supply module should provide? Does some of them should be variable to some extent? Does current auto-ranging for stand-by consumption measurement makes sense?
The newly proposed chassis could carry up to three modules. I believe that two fully featured (DCP505) modules is enough and that leave one place vacant for a multiple outputs module.

Thanks in advance for your valuable inputs.

Make it a usb 5v / 3.3v and or type c pd


Sent from my iPhone using Tapatalk

[prasimix]

Thanks, I know nothing about USB type C, just information that it can be used also for "high power" supply, and have to investigate that further.
For 3.3/5 V outputs what current could be appropriate: up to 2-3 A?

[LapTop006]

Thanks, I know nothing about USB type C, just information that it can be used also for "high power" supply, and have to investigate that further.
For 3.3/5 V outputs what current could be appropriate: up to 2-3 A?

Whilst I would absolutely love a type-c USB-PD output, they might be a bit of a pain for an early module.

USB-PD rules say up to 3A at anywhere from 5-20v (with 5, 9, 15 & 20 being the standard voltages that should be supported), and 5A at 20v only if connected to a cable that declares itself safe for 5A.

[prasimix]

I found that the USB 3 PD functionality can be provided using MCU from STM32F0 family for which ST offers USB-PD stack. I believe that could speed up a whole development of something that looks rather complex.