DIY Modular Test Equipment Project

[void_error]

1. You were looking for an alternative to DipTrace and free 3D modeling.
For schematics & layout, check out EasyEDA. From the little I've played with it, it seems to work. And it's free.
For 3D, check out OnShape -- a relative newcomer offering no-apologies parametric CAD for free. They claim to import step. Free accounts limit *private* documents. I like it, for what that's worth.

Looked at EasyEDA too but it's online which kinda sucks.
For 3D modelling I'm using FreeCAD, and download the 3D models for parts from tracepartsonline.

2. Maybe I'm just not understanding what you're doing. It sounds like a UI front panel, some stackable modules and a framework for making arbitrary other stackable modules. Then I imagine that means a UI panel can manage a mixed stack behind it. Is that the idea, or did I get lost?

Yes, the UI board is the front panel and you can stack other modules on top of it, on the opposite side of the front panel, see older posts, I have an example of how two modules stack up in 3D.

If I'm not completely lost, then here's a thought: (assume easy answers to fab questions and) arrange modules to launch their real-world connectors/terminals from one side (both?) and angled ~45deg forward. The result would be stacks with a UI panel on the "front" face, and behind that along the "side" of a stack, all the external connectors facing front-ish.
(You're placing connectors at top and bottom not sides - but maybe the constraints that lead that way apply more to stacking connections than to external connections?)
Or more simply just go straight out the side - at cost of requiring more directly perpendicular access to the side of a stack.
Something kinda like this, but with the knobs on the front and connections down the side: http://w140.com/tekwiki/wiki/File:Tek_211_front.JPG

Thought about having boards plug in at a right angle but mounting everything in an off the shelf enclosure would be more difficult, so all the boards are going to be parallel to the front panel. That way if the stack becomes too heavy to be supported by the front panel only two L-shaped brackets bolted to the bottom of the enclosure would be required.

[pmc]

... so all the boards are going to be parallel to the front panel.

Understood. I'm not understanding how boards will expose terminal connections. I do see the function gen arrangement with BNCs next to the UI/display - but that only works for the one board directly behind the UI. How would you connect stuff to a module further down the stack? Or did you have in mind just one _function_ per assembly, with one interface board directly behind the UI exposing terminals adjacent to the display, plus whatever stack of add'l boards that function requires?

[void_error]

Initially I wanted to use ribbon cables to connect the modules to the UI, that's the reason for the shrouded headers, but I figured out shortly after that the waveform gen can be strapped to the back of the UI after finding those BNC connectors that stick out.

One function per assembly is the general idea I'm sticking with but there will still be the possibility to add more functions which comes with a few small drawbacks like having to use some extra wiring if the board is located  further down the stack requires connections to the front panel or if the board is physically too large.

As an example, say we have the Lab Power Supply configuration. The connections will look like this:

[UI board] : [Digital Isolator board] - [A/D & D/A board] : [Lab Power Supply board]

":" means stacked
"-" means wires/cables are used

The main goal is to keep wiring to a minimum, I don't like things that have so much wiring they look like a spaghetti bowl.
In some configurations it's difficult or not practical or too expensive to avoid using wires/cables for some configurations.

Other options I was thinking about as far as add-on boards are concerned are the possibility to add a second display (for a dual channel lab power supply) and a keypad (quite handy for a waveform gen) but I haven't decided how they will connect to the UI.

For the waveform gen at least the third layer of the stack (first one being the UI) will consist of the power supply modules for the different voltage rails required.

[prasimix]

Hi void_error, this is a great step towards defining modular lab equipments "environment". Do you have in mind housing? I think that enclosure has to be added in the picture very early.

[void_error]

I actually do have housing in mind. I've been following your lab power supply thread and I'm probably going to use one found on modushop.biz, Economica series, 80mm tall. Shipping to Romania is quite expensive though so if anyone knows about any other places in Europe where I can find similar enclosures please let me know.

[prasimix]

I'll let you know what Varisom from Portugal could offer. Their initial price for customized solution was really appealing but lately our communication became slow (maybe we should take into account EURO 2016 factor

[pmc]

Initially I wanted to use ribbon cables to connect the modules to the UI, ... but I figured out shortly after ...

Ok, stacked packaging was never a driver for the project. Got it.

... but there will still be the possibility to add more functions which comes with a few small drawbacks like having to use some extra wiring if the board is located  further down the stack requires connections to the front panel ...

That's what puzzled me on first reading -- stacking functions behind a flexible UI sounds interesting, but then extra wiring and loss of generality for front panel arrangement sound less interesting.

So, the idea thrown out for your consideration was basically: keep the UI in the front panel of some box and stick the T&M connections out one side of the box, directly off the edge of each module PCB. More modules -> deeper box/longer side behind the same front panel.  (thinking one side to avoid locking in other dimensions/board area)

That said...

Requiring a lot of clearance for that side -- to see, reach and push/twist/pull whatever connector on the 4th card back -- seems like a strike against. Aiming each card's connectors ~45deg toward the front might make side connections more usable. So that side would be, um, louvered. And you could see connectors/jacks/sockets/whatever without clearance for your skull + near point distance alongside the unit. Plausibility requires some design idea for simple construction -- IFF you think the result would be desirable. (have half an idea. or just short wires.)

And...

Out-the-front is still attractive for a single-function assembly like the freq gen arrangement you've modeled. PCB pads arranged for attaching either "vertical" or right-angle connectors (or angle adaptificator) would give the option to build a first/only module either way.

Makes sense?

Just some thoughts. Your ball.

[void_error]

So, the idea thrown out for your consideration was basically: keep the UI in the front panel of some box and stick the T&M connections out one side of the box, directly off the edge of each module PCB. More modules -> deeper box/longer side behind the same front panel.  (thinking one side to avoid locking in other dimensions/board area)

That said...

Requiring a lot of clearance for that side -- to see, reach and push/twist/pull whatever connector on the 4th card back -- seems like a strike against. Aiming each card's connectors ~45deg toward the front might make side connections more usable. So that side would be, um, louvered. And you could see connectors/jacks/sockets/whatever without clearance for your skull + near point distance alongside the unit. Plausibility requires some design idea for simple construction -- IFF you think the result would be desirable. (have half an idea. or just short wires.)

I think I get what you mean.

Say you're looking at the front panel on the Z axis, perpendicular to the plane formed by the X (horizontal) and Y (vertical) axis.
Say you want the connectors on the right side, so you take the connector which would normally be on the X axis (sticking out to the side) and twist it clockwise on the Z axis by 45 degrees.

That would be quite neat but on mechanical side of things there's going to be more work... and wires.

Let me know if I really get it.

And...

Out-the-front is still attractive for a single-function assembly like the freq gen arrangement you've modeled. PCB pads arranged for attaching either "vertical" or right-angle connectors (or angle adaptificator) would give the option to build a first/only module either way.

Makes sense?

Just some thoughts. Your ball.

The UI can also work as an (expensive) alarm clock or a development board. Some other modules I have in mind are a mains Watt Meter and a Frequency Counter.

If you use right-angle connectors you could use stripboards for board interconnects as all the signal/power connectors are aligned to a 0.1 inch grid - breadboard/protoboard compatibility.

Trying to keep it quite simple, mostly for time reasons. I want to have something done by the end of this year.

80mm tall

Different enclosures for different functions. As long as it has an easy to drill front panel.

[rollingstone]

Hi, Nice project. Sorry, I didn't followed Your post from the begginning. Just remembered I did once something like this. It was 8 HF generator modules and 8 HF amplifier modules using standardised EUROCARD subracks. What case you going to use for modular design? I used this one : http://lt.farnell.com/schroff/24563-191/subrack-3u-175mm-84hp/dp/1455794
http://www.farnell.com/datasheets/95177.pdf?_ga=1.224072792.2973774.1448267837
Only 40 eur and does it's job. Then you can use plug in units to slide in:
http://www.farnell.com/datasheets/95306.pdf?_ga=1.216596437.2973774.1448267837
or shielded module:
http://lt.farnell.com/schroff/20809-537/card-frame-3u-227mm-10hp/dp/1370403
You just need to drill or to cnc front plate to make holes for indicator LEDS, switches and BNC connectors and stuff like that.
You can make one universal power with many regulated voltages (3.3, 5, +12, -12, 24 volts etc.) as plug in unit for entire rack, and then many other modules that you can plug in and plug out into rack and make it configurable and very universal.
I don't have any photos but it looks something like this:
http://dinrackis.com/euro-card-racks.html
http://mulogic.com/UCF-3HE-rack.html
 (immages from internet).
 

[void_error]

You got it a bit wrong, it's not the case that's going to be modular, it's the stuff inside the case.
Those rack-mount frames do look nice but the plug-in units are bloody expensive. I've already decided what kind of enclosures I'm going to use and I won't have anything standardized across all the different bits of test equipment I'm making, the decision to have everything share a common front panel was made mostly because of time available for this project.

As far as progress goes it seems that I've reached that part where I tweak stuff - quite happy with that.

Another change to be made to the UI is replacing the 74HC4094 shift register used for driving the backlight LEDs of the pushbuttons with a 74HC595 as it's also used for multiplexing the Chip Selects and other things on the Waveform Generator speaking of which... the PCB is quite a pain in the a$$ to route, taking longer than expected, and I want the whole assembly to mount into a 100mm tall enclosure - that makes the Waveform Generator board size have a limit of 90mm. There's the option to make the board physically larger at the cost of... well... cost. It'll take some time until I figure out the best compromise although I've got quite a few ideas... stay tuned.

[void_error]

Finally managed to find the best component placement for the waveform gen board, final dimensions are 60mm wide by 90mm tall, front view.
All that's left to route is the power traces and tweak everything (tedious), which means I'll leave that for another day and work on another module - the USB - UART board... and it's going to be modular too, it will be able to plug into the UI board or accept a plug-in board with RS-232 and/or RS-485 (half duplex or full duplex). Some changes are required though, since the PIC16F1454 doesn't have enough pins for all the control lines, so I'm replacing it with its bigger brother, the 16F1459, which is also used on Microchip's Low Pin Count USB Development Kit so it's just a matter of modifying the USB CDC Emulator demo code as far as firmware goes.

Power Supply boards are next on the list. No big deal there, see first page. A 3D model of the Waveform Generator assembly will follow soon after.
After all of the above are completed it'll be PCB ordering time, I'm going with PCBWay, they seem decent and cheap.

Since the waveform gen is almost finished, I'm posting the schematic just in case anyone notices anything wrong, call it revision 1. The posts on the first page will be updated with schematics as soon as I have them built and tested.

[void_error]

The USB-UART module is almost finished. I only have to figure out what I can do with the two free pins I have left on the PIC16F1459... any ideas? Already have RX/TX/USB LEDs. What else can I do with with two extra I/O pins? Monitor USB voltage with one of them? RS-232 flow control lines * and RS-485 direction control are already in place.

* do I even need those? if not then I might as well get rid of them and use the 16F1454 but it would require porting the USB CDC Emulator demo rather than just modifying it and that would take more time which I have very little of.

Other than that, I wanted to cheap out on the UI board and ditch those two MCP23S08 I/O expanders by using the PIC16F18857's bigger brother, the PIC16F18877 but haven't managed to assign all the peripherals to pins while maintaining an easy-to-do PCB layout even with all the flexibility the PPS has to offer so nothing changed there.

I'm starting to dislike the idea of the Aux Power Supply modules being held only by two screws on one side so I might have another go at making everything fit together with nothing "flapping in the breeze". I might end up with more mounting holes than you can poke a stick at... but if it works...
All that while trying to keep the 0.1 inch headers on an imperial grid while everything that's going to poke through the front panel (buttons, encoders, display) is going to be metric.

[void_error]

Other than that, I wanted to cheap out on the UI board and ditch those two MCP23S08 I/O expanders by using the PIC16F18857's bigger brother, the PIC16F18877 but haven't managed to assign all the peripherals to pins while maintaining an easy-to-do PCB layout even with all the flexibility the PPS has to offer so nothing changed there.

Wrong. It can be easily done. Found a way around to allocating some pin functions to locations not supported, through the CLC module. I've been looking into that lately and if I use one of the CLC modules I can move some PWM outputs to PORTA. Only CCP5 can be assigned to PORTA but also CLC1OUT and CLC2OUT and they can accept inputs from other CCP/PWM modules.
The control pins on the display are input-only so they can be driven by a 74HC595. There's already one driving LEDs, it's just a matter of daisy-chaining the second one.

I've also had a look at the NCO module for square wave generation. The MCC comes in handy here for checking stuff, removes the need for spreadsheets to calculate NCO output frequency. Using Timer0 as a clock source for lower frequencies gives quite accurate results. Accuracy should be less than 1% +/- clock generator accuracy. The square wave / PWM output was never intended to be high precision anyway.

So it's almost back to square one... but at least I won't have to write code for the MCP23S08 anymore.

[void_error]

It's looking good so far. There's one MCU pin available which is going to be used for something eventually.
Board size will be increased from 120x70mm to 130x70mm to make room for extra mounting holes. This is actually a good thing since the waveform gen board could use some extra width.

EDIT: Got it slightly wrong. I actually have two free pins. I'll find a use for them.

[void_error]

Looks like I ended up with 16 mounting holes on the UI board.

Got some work done on the Aux Power Supply modules and ran into some issues again because of the board size limitations, both modules are going to be 34x70mm in size.
On the analog supply board the space issues were dealt with by moving the SMD parts (diodes, resistors, MLCCs) to the bottom side of the board, the top side being occupies by connectors, bulk filter caps and the +/-15V linear regulators mounted horizontally on heatsinks (no TO-220s flapping in the breeze here).

The digital supply board had a few changes. Initially I wanted to use LTC3824s but there wasn't any room for all the external components needed so I ditched the chip and went on to look for something else. Found the TPS54360 which comes with a really handy spreadsheet which does all the math for you. Now I can fit both regulators on one board quite easily.

[void_error]

The digital supply board had a few changes. Initially I wanted to use LTC3824s but there wasn't any room for all the external components needed so I ditched the chip and went on to look for something else. Found the TPS54360 which comes with a really handy spreadsheet which does all the math for you. Now I can fit both regulators on one board quite easily.

Nope... even more changes. The way I want to sync the two regulators, one with an inverted sync signal, was a bit tricky with the TPS54360 as the timing resistor pin also serves as a sync pin and needed more parts than I had board space so it was back to suitable part hunting.
LT3680 seems to do the job - 12V @ 2A for one of the regulators and 5V @ 3A for the other one. And everything fits. (Note: @ means at up to here)
A dedicated sync pin makes life much easier.

Of course, the component values can be recalculated for different voltages, who knows where else I might need one of these boards. The sync logic runs off a separate 3.3V LDO (LM3480) and you might be wondering why I chose to run it off the 12V rail. Here's why: 12V also powers the fans on the DC Load and on the Lab Power Supply and those two don't need a 5V rail.



I'll be playing with these a bit more in LTspice to find the proper values for the compensation network components. Would have been easier if LT just made a spreadsheet like TI did.

Back to the dreadful-to-route waveform gen board... nothing changed there except for the LDO for the 50MHz oscillator which was a MCP1700, now it's a LM3480. Once I finish moving component groups around to get the best possible placement and finish the layout I'll have a 3D model of the whole assembly. Finally.

The other power supply board required by the waveform gen is not really worth talking about too much as it's just a pair of 7815/7915 linear regulators, a pair of 78L05/79L05, a bunch of filter caps and some diodes. Oh, and two small heatsinks.

Should I add some PTC fuses (where?) just to be on the safe side or is it a bad idea?

[void_error]

Eye candy time!

Although I can't call anything finished yet I'm close enough. Lots of minor tweaks and some double-checking is still required for every single PCB but I finally got as far as being able to slap everything together in 3D with all the connectors in place. There's still a long way to go.

Color coded for visibility reasons:

The white board is the front panel user interface. The waveform generator board is the blue one with the BNC connectors sticking out.




The 5-pin connector on the green board (USB-UART module) goes to a USB panel mounted connector. One side of the optoisolators is powered by USB +5V while the other gets power from the 3.3V rail on the user interface board.




The purple board provides the analog supply rails: +/-15V and +/-5V. The two TO-220 regulators will have some small heatsinks between them and the board. Filter caps placed below the heatsinks. There's still room for a frequency counter module




There's more than enough space for those bulky inductors.




The two black connectors on the top accept terminal block headers(?). The wires will be coming from a 18V center-tapped transformer. I managed to keep wiring to a minimum




Connectors on top for easy probing and firmware upgrade. Might include a serial bootloader at some point...

[void_error]

Haven't worked on any part of this project at all for a few days and that's when ideas start coming and it's time to start tweaking.

I was looking for parts on Farnell/element14 website and found out they have a lot of quadrature rotary encoders to offer which made me want to change things a bit, forget the Bourns PEC11-L which is available at distributors which have high shipping cost to my country. Alps seem to have some nice encoders at a decent price also with included push option so I'm going with one of those. $5 shipping is the same as TME charges. If I can stick to these two distributors I'm happy and so far they seem to have all the parts I need.

Speaking about tweaking the designs I'm starting to think the UI board doesn't need a SEPIC converter if I use a charge pump to generate the -5V rail on the lab power supply board instead of an inverting DC-DC converter which would have used the LT3580. The charge pump wouldn't be a big deal as I could just use some logic gates driving small dual MOSFETs and clock them with the SYNC signal. That would eliminate the requirement for coupled inductors which aren't as easy to get as regular ones.
The LT3580 would be then replaced by the LT3680 which is also used on the aux digital supply board, reducing the BOM count.

With the waveform gen configuration being almost complete it's time to focus on the next two groups of modules I have planned, the DC Load and lab power supply and sort out the mechanical aspects.
Since I want the DC load itself to be modular there are two possible setups:

• Have the modules in parallel for more power
• Have the modules grouped and isolated from each other, like a multi-channel power supply

The first setup can contain a local microcontroller (on the DC load PCB) or not. The micro would be used for constant resistance or constant power modes. Paralleled modules will share a single DAC .
The second setup will include some form of isolated DC-DC converter and logic isolators.

For the lab power supply things will be a bit simpler, each power supply module is going to have an isolator+ADC&DAC board on top of it.

Another piece of test equipment that I intend to design and build, another module in this case, is a mains Wattmeter.

[void_error]

Over the last few days I've been trying to figure out how to make the DC load modular and spent a lot of time looking for suitable parts.
Had to stick with SPI because there are not many I²C ADCs available.

The tricky part is making the modules daisy-chainable for the two main configurations - with or without an on-board micro.
Why? Because there is only one Chip Select line available.
With an on-board micro it can be done in firmware, each microcontroller will have an address. The peripherals will be on a second SPI bus.
When there is no on-board micro I'll be using cascaded 74HC595s to multiplex the Chip Select signals for the DAC, ADCs & Temperature sensors.
In the end it all boils down to a lot of jumper links used depending on the board configuration.

I still have no idea if passing the analog signals from one DC load board to another is a good idea for "slave" boards containing only the analog stuff and temp sensors with the "master" board having the optional micro and DAC/ADCs.

[void_error]

Before thinking of finalizing the DC load design I thought I'd check if the load modules scale up nicely and are stable in parallel with the help of LTspice.
The actual control loop will be using an OPA2727 but for the simulation I used the closest LT part I could find within 5 minutes which is the LT1213.
Everything seems to scale up nicely with no parasitic inductances/capacitances so a 400W DC load comprised of 4 DC load modules appears to be possible. Many thanks to Jay_Diddy_B for thoroughly explaining how to get the loop stable.

On the digital side I had a few issues with designing the module for operation with or without an on-board microcontroller but most of them are now dealt with.
One of the issues was multiplexing the fan rpm signal on the non-micro configuration but I had some free pins on the 74HC595 shift register used for multiplexing the ADC/DAC/Temp sensor chip select pins and used one to enable/disable the output of a 74LVC1G126 3-state buffer. A schematic would probably explain this better but it's not finished yet.

[void_error]

At this stage I'm jumping between the (still) nearly completed UI/aux psu/waveform gen/USB-UART boards tweaking stuff and thinking of things I might want to add.
Found one: add some energy storage caps to the UI board so they can power the EEPROM, micro, oscillator, and the optional RTC for an EEPROM write after power off as there will be no standby power supply and I'd like to have the previous settings saved until the next power on. Yes, it's going to have a "big clunkin' power switch".

There are two things that come into play here:

• Detect the 3.3V rail voltage decrease
• Write to the EEPROM before the capacitor voltage drops below a certain value, about 2.7V in this case, the micro can run at 32MHz down to 2.5V

The first one will be handled by one of the comparators, input via an externally available pin.
The second will be dealt with using the same comparator but with the input connected to the internal DAC acting as an internal voltage divider (referenced to Vdd).
The other input of the comparator will be internally connected to the internal voltage reference.

[void_error]

Found one: add some energy storage caps to the UI board so they can power the EEPROM, micro, oscillator, and the optional RTC for an EEPROM write after power off as there will be no standby power supply and I'd like to have the previous settings saved until the next power on.

Good thing I remembered to RTFM for the oscillator, which is an ILSI ISM95 or ISM97 because it doesn't work down to 2.7V so I'll have to switch to the PIC's internal RC oscillator when a power failure is detected and that's possible according to the datasheet.

Another possible issue which I'm glad I didn't overlook could have been caused by the input protection diodes within the micro but another P-channel MOSFET which disconnects the resistor divider's positive side solved the problem.

With that said, the schematic for the UI seems to be complete... except a few part values...

EDIT: Forgot to change the nets for the encoders from +3V3 to VDD... and maybe a few more pull-ups. Corrected everything. Even had some MOSFETs backwards...

[max_torque]

maybe silly question, but wouldn't it be worth doing all the different modules on a standarised card, with a std backplane connector?  therefore you could add extra modules, or change the allocations easily.  Even better, build them into a std rack mount enclosure like this sort of thing:



and it would look pretty professional too 

[void_error]

Someone else also suggested that, but it would be too expensive. Rack mounted enclosures aren't cheap either.

[max_torque]

whilst proper eurocard rack enclosures are expensive, i bet you could do a simple DIY solution using std back plane connectors and some laser cut acrylic or simple ally frame etc??

Design you "processing" pcb as the back plane, with the IO and numerous serial bus connections, and have the various cards slot into that!