DIY Modular Test Equipment Project

[void_error]

GitHub repo is now up . Check first post.

[void_error]

Waveform Generator + Frequency counter project has also been uploaded to GitHub. Link on first page. Don't be afraid to point out any mistakes you may or may not come across.

I'll also have to update the individual posts for each module. They're getting outdated.

[void_error]

The first of the two voltage regulator modules is up on GitHub, the one for the digital part. It also has a... hmm... no idea how to call it... but it's basically a step-down regulator power stage with a PWM input and it's used for fan control. I'll upload the one for the analog part sometime later today if I don't get caught up in other stuff.

On another note, most of the 4 PCBs that make up the minimum setup for the Waveform Generator + Frequency Counter are being worked on and their sizes set. I'm still having a few issues with connector placement but those will be sorted out. I haven't included the isolated USB-UART module since it's tiny in comparison but it also has to do with connector placement. In case I haven't mentioned already, it's going to use either a PIC18F13K50 or a MCP2200, they're pin compatible with the latter requiring no firmware.

Almost forgot, I updated the first post. Again.

EDIT: Forgot to add that regulator modules will also be used by the Electronic Load Module.

[prasimix]

I'd like to encourage you to use GitHub Master branch as much as possible, and that important milestones, prereleases or releases simply mark by publishing it where you can include some description or list of accomplished tasks. Making multiple branches especially in early phase is a good source of confusion for one how is following you, want to fork a project and stay synced.

[void_error]

Will switch to master branch as soon as I get to past Rev 0. I'm still making small changes to the schematic as I work on PCB layout with one example being multi-part components which will be swapped around quite a lot. Rev 1A will be the first to be merged into master.

[void_error]

Finally had some spare time to add the voltage regulator module for the analog sections to the draft branch of the related repo on GitHub.
As always, the link has been included first post.

[void_error]

Wasted the whole weekend on PCB layout. Results in first post / on GitHub. There's room for improvement. I haven't included the footprint libraries although I will since there are some footprints I made myself. Not all parts have 3D models yet.

[void_error]

There might be a problem with the +/-5V regulators on the waveform generator schematic, thought I'd redo some math on the thermal side of things and it looks like those LM317L/LM337L won't be too happy at a little under 50mA with a 15V input and I don't want to 'Rigol it' .

I have two options:
1. Use physically larger packages for the regulators
2. Move the analog supplies of everything using +/-5V apart from the AD603 to other rails.

I'm inclined towards number 2 since the DACs can take up to +/-15V and the LMH6642 can easily live off +/- 3.3V. Only the DAC for adjusting the offset will need to be run off +/-15V though.

As for option 1, I'm looking at D²PAK packages for LM317 / LM337 but those take up considerably more space compared to SO-8.

[void_error]

I have two options:
1. Use physically larger packages for the regulators
2. Move the analog supplies of everything using +/-5V apart from the AD603 to other rails.

I'm inclined towards number 2 since the DACs can take up to +/-15V and the LMH6642 can easily live off +/- 3.3V. Only the DAC for adjusting the offset will need to be run off +/-15V though.

As for option 1, I'm looking at D²PAK packages for LM317 / LM337 but those take up considerably more space compared to SO-8.

Yet again I realize I didn't have my head screwed in properly and while both of the above would work the voltages need to be +/-5V so there's no need for adjustable regulators. 7805/7905 will do just fine. As far as the case goes, D²PAK is HUGE and SO-8 can't handle half a watt... at least with a temperature of 50C inside the case which isn't likely since there's going to be a fan in there.

Luckily the 7805/7905 also come in DPAK cases which are considerably smaller than the chopped up TO-220 which is the D²PAK, they're roughly the same size as an SO-8 and have significantly lower thermal resistance, under the form of 78M05 and 79M05. Junction to ambient thermal resistance doesn't exceed 100C/W so that's only a 50C rise for 0.5W (10V drop @ 50mA) which is well within acceptable limits and real world thermals will be significantly better since it's mounted on a PCB. Problem solved. I choose option 3.

[void_error]

waveform generator 115x95mm.

Nope. That was a bad idea. Plus the case which was my initial choice would have ended up being mostly filled with air so I'm swapping it with this one and making the waveform gen PCB 175x70mm.

Just so this is more fun, it looks like differential signalling might be required since the traces are getting quite long, not to mention that adding a ribbon cable would make things a lot worse. That means adding some differential transceivers and have the UART over RS-485, use RS-485 transceivers for SPI like metioned here, same for the GPO and GPI, differential signalling for the clock (haven't looked into that but it should be able to handle 16MHz), should get away with single ended for the 400kHz regulator sync signal as long as it's buffered on every board I think, the 1-wire bus should be fine since it's really slow anyway.

The only thing I have no idea what to do with is I2C. Any suggestions?

[void_error]

Small progress update:

After messing around with the idea of using RS-422/RS-485 transceivers I ditched it in favor of LVDS/M-LVDS. Since I also want to reduce crosstalk between pairs I'll be upping the pin count of the bus connector to 40 pins to allow for ground between pairs, although I'm not sure if the differential impedance will affect signal integrity and/or reflections across the ribbon cable too much at the rather low speeds the SPI will be running at - 16MHz clock or lower.
Also, the glue logic used to switch between SPI and the shift register lines will be moved to each individual board so there's only going to be one differential SPI line with an extra line to select who gets clock and data.

Another problem I ran into is having to sample voltage and current simultaneously and while there are simultaneous sampling ADCs out there I couldn't find anything suitable at a reasonable price so I'm opting for a bit of a hack here: clock both AD7683s at the same time, with one outputting it's data directly into the MISO line, while the other outputs into a 24-bit SRAM made out of D-type flip-flops. To get the data from the makeshift SRAM there's going to be a pin used in the same way as the ADC Chip Select. And in case you were wondering, there's a pin-compatible replacement from TI for the currently used ADC which is slightly cheaper - ADS8326.

[void_error]

Minor setback: graphics card RAM died and I have no onboard graphics. Result: Windows commited suicide while attempthing to repair itself.
Took this as an opportunity to boot linux mint in text mode off a USB flash drive and use lynx to navigate the interwebz and also post this simply because I can and I hate typing on my smartphone.
The good thing is that the card's under warranty for another 3 months so I should be getting a replacement within the next 2 weeks so I can get back to adding LVDS to all the boards...
Speaking of which, the GPO and GPI lines will be gone and a trigger line will be added.

[prasimix]

Another problem I ran into is having to sample voltage and current simultaneously and while there are simultaneous sampling ADCs out there I couldn't find anything suitable at a reasonable price so I'm opting for a bit of a hack here: clock both AD7683s at the same time, with one outputting it's data directly into the MISO line, while the other outputs into a 24-bit SRAM made out of D-type flip-flops. To get the data from the makeshift SRAM there's going to be a pin used in the same way as the ADC Chip Select. And in case you were wondering, there's a pin-compatible replacement from TI for the currently used ADC which is slightly cheaper - ADS8326.

Did you possibly consider ADS131A02? It's 24-bit but with very attractive price.

[void_error]

Thanks prasimix, I had a look at the datasheet and it looks like a candidate for for the Power Supply Module, the Electronic Load Module and the Mains Power Meter Module, the only tricky part id the fact that it requires a 16.384MHz clock, which I could provide via the clock line on the bus and will also be used for switching regulator sync after division, the problem being that is I use 4x PLL on the main MCU I'd have to use a 8.192MHz crystal due to the 64MHz maximum clock limit and send the x4 multiplied by the PLL out the CLKR pin after it's divided by 2.
Since the same clock is used by any other MCUs (PIC18F25K40) they'd have to run at 16.384MHz (4.096MHz instruction clock) unless I divide it by 2 and use the 4x PLL multiplier which is not that big of a problem.

On the bright side, my PC is back up and running. Work will resume as soon as I have enough time and energy to put into this which might be totally random.

I managed to mess around with adding LVDS and changing the bus connector yet again to one with more pins (40) and changing the PCB layout for the UI. I've also been looking into the possibility of keeping the ribbon cable for LVDS and it seems like a viable option as long as I keep stub lengths short (>30mm at the moment) and the clock speeds low, with the highest frequency being 16MHz. If anyone knows why this wouldn't work please don't hesitate to let me know.

[void_error]

After being stuck for quite some time with having to use two 40-pin connectors on each peripheral board due to the daisy chaining requirement for the 595s which makes PCB layout a complete mess, I've ditched the shift registers and quite a bit of the glue logic in favor of a some of 3-to-8 decoders and a MCP23S08 I/O expander.

The lines formerly going to the 595s are now address lines, a sort of bank select for the MCP23S08s. The chip select lines for the SPI stuff (ADCs, DACs, etc) on each board will be the decoded outputs of another 3-to-8 decoder with the 3 inputs connected to the outputs of the MCP23S08 which also offers a level of idiot-proofness (is this a real word?) in the sense that only one chip select can be active at one time in addition to the reduction of the number of lines required on the I/O expander.

I won't be putting any of these changes on GitHub just yet as I still have to fiddle with the pins on the micro when I start working on changing the PCB layout.

Speaking of which... is anyone interested of having the possibility putting your own favorite micro (on a daughter board) so I can add a bunch of pin headers to facilitate that?
EDIT: That would include all of the boards which will have a micro, except for the UI board.

[void_error]

Speaking of which... is anyone interested of having the possibility putting your own favorite micro (on a daughter board) so I can add a bunch of pin headers to facilitate that?

No more need for that. Switching to STM32s which have an integrated ROM bootloader. I've been looking into them for quite some time but until January this year all the decent development tools weren't free. Now there's this which is basically eclipse with GCC & GDB and works out of the box. No more slapping development tools together.

The MCU of choice is going to be STM32F101VC for the UI. 100 pins LQFP, has a RTC with battery backup so no need for a separate RTCC chip.
ROM bootloader is accessible via USART1. At the moment I'm not too sure about how I'll update the firmware on the MCUs of the other modules so I might have to abandon the half-duplex RS-485 style of communication via UART and think of something else just so I (or you) can flash everything after it's been all put together.

[void_error]

Updated all of this project's github repos, have a look if you want. PCB layouts aren't finished and schematic need some cleaning up. I'll do it when I have the time and energy.

[void_error]

Waveform gen is almost finished, everything's now up to date on Github as I finally found some time to work on it. The PCBs need a bit more work, via stitching in some places and spme traces need tweaking. Everything stacks up nicely.