BlackEdge, my* and mixed signal interfacing support

Folknology 2019-02-01 11:42:54 UTC #1

BlackIce and planned BlackEdge and my* products all have mixed signal support, both analogue and digital, this is an important feature for all myStorm products and matches many real world application requirements. Therefore one of the elements we need to decide how to proceed on is how to expose these interfaces in standardised ways.

I have a none Pmod based carrier design (more later) aimed at more robust forms of expansion for robotics, automation and industrial applications, but it would be good to build on existing Pmods for general purpose development. Thus if we are to focus on a Pmod based carrier for example how could we expose the analogue signals to make them useful? There are a couple of possibilities I have thought about so far :

Use dedicated analogue Pmods either single or double where all non power pins become analogue pins.
Use the space in between a pair of double Pmods to insert some extra analogue pins.

The first is self explanatory but the second requires a bit more enumeration; if 2 rows of 3 x 0.1" headers are inserted between 2 existing double (adjacent) Pmods you end up with a dual row (2 x 15) 0.1" 30 pin right angled female header, with a pinout looking into it as :

+3v3,GND,D01,D02,D03,D04 A1,+5V,A2,3v3,GND,D05,D06,D07,D08 +3v3,GND,D09,D10,D11,D12 A3,GND,A4,3v3,GND,D13,D14,D15,D16

I.e. 16 digital, 4 analogue signals, +3v3, +5v, and ground.

This would still accept standard dual double Pmod peripherals, individual Pmods (single or double) or the new mixed signal peripheral lets call them Mixmods. This would enable one to build new mixed signal adaptors and plugins whilst maintaining backward compatibility with Pmod peripherals on the carrier, It also provides +5v which many think is missing from the Pmod standard.

So consider this the start of the conversation on Mixed Signal Pmod interfacing, please let us know your ideas to solve this challenge and any thoughts on our suggestions here..

daveshah 2019-02-01 13:04:18 UTC #2

This looks good, the 5V in the middle is now becoming a semi-standard too (also used on the icebreaker, and I think some other recent boards).

I think this would be very useful for things like motor controller boards, etc, where you want a few analog signals for temperature/voltage monitoring and the likes.

lawrie 2019-02-01 13:07:42 UTC #3

Yes, I like the second suggestion, too.

Folknology 2019-02-01 14:49:54 UTC #4

Just to add some imagery to the discussion here is the layout of ModMixer Carrier that would support such an idea(2).

This sports 2 x MixMods (west & east) and a pair of Dual Pmods (north). Although not show here PM1/2 AM1, and PM5/6 AM2 would be populated with 2 x dual row 30 pin right angled female 0.1" headers

ken.boak 2019-02-01 17:05:57 UTC #5

Al,

My only concern is that the analogue inputs on the STM32 are limited to 4V max - or 3.3V in all practical terms.

Pins are classified either FT - 5 volt tolerant or TT - 3.3V tolerant. The datasheet is somewhat confusing in that some ADC inputs are FT and some are TT.

A rule of thumb would not to exceed 3.3V on any pin that has been enabled as an ADC, op-amp or comparator input.

Having a +5V so close to the analogue inputs might lead to some unplanned incidents.

You could always fit external 3V3 clamp diodes and a series current limiting resistor, or have a jumper/switch that sets the voltage of the +ve pin to 3V3 or 5V.

Frying a board (or STM32) because of an accidental plugging of 5V seems a bit risky.

Folknology 2019-02-02 13:57:25 UTC #6

Yup this is definitely a risk, on the positive side the connectors containing the 5V pin are female and therefore reduce the risk of accidental contact. One other possibility is to block the centre ground pin socket for polarisation purposes (ground is available on other pins), This may reduce the risk even further?

+3v3,GND,D01,D02,D03,D04 A1,+5V,A2,3v3,GND,D05,D06,D07,D08 +3v3,GND,D09,D10,D11,D12 A3,NCB,A4,3v3,GND,D13,D14,D15,D16

Not full proof but lowers the risks significantly...

Folknology 2019-02-24 14:44:09 UTC #7

My latest favoured pinout for mixed mods

3v3,GND,D01,D02,D03,D04,C0,5V,C1,3v3,GND,D05,D06,D07,D08 3v3,GND,D09,D10,D11,D12,A1,A2,A3,3v3,GND,D13,D14,D15,D16

Key changes:
1) All analogue pins on different row to 5v reduces risk of shorts between them.
2) Added common digital signals C0,C1 great for things like I2C and MDIO (Anyone tried to make an Ethernet PMOD, you would have probably gained from this idea..)

daveshah 2019-02-24 21:10:28 UTC #8

Will the C0/C1 pins be going straight to the FPGA? That's probably even more vulnerable to 5V than an STM32 analog pin...

piers 2019-02-24 23:22:58 UTC #9

Would something like the active clamping described in tip 11 of http://ww1.microchip.com/downloads/en/DeviceDoc/chapter%208.pdf be a suitable way to protect the FPGAs i/o pins from 5v shorts?

piers 2019-02-25 07:46:58 UTC #10

Though for bi-directional pins a level shifter chip may be a better option

Folknology 2019-02-25 08:09:34 UTC #11

I don't think we need level shifting as these are 3v3 signals, the issue is primarily the risk of damage to signal lines if accidentally shorted to the 5v. I am starting to think of maybe omitting the 5v entirely..

Replacing the 5v with another signal like Reset something like this:

3v3,GND,D01,D02,D03,D04,C0,RS,C1,3v3,GND,D05,D06,D07,D08 3v3,GND,D09,D10,D11,D12,A1,A2,A3,3v3,GND,D13,D14,D15,D16

piers 2019-02-26 00:59:22 UTC #12

Is 5V more useful than reset? I think for the cost of a series resistor and a pnp bjt to clamp the digital signal pins to 3v3 for the 2 digital lines either side of the 5v line would be a good option.

For the analogue inputs a unity gain op-amp powered from 3v3 with a series resistor will saturate at 3v3 (if you can get an op-amp that will tolerate 5v on its inputs). I suspect, however, the same active current limit will work for the stm32 analog inputs.

Perhaps a jumper could disconnect 5v from the 5v pin until the mixmod was plugged in? Or a jumper selector between reset and 5v for the top centre pin - with the jumper removed it would be 'safe' to plug in the mixmod.

One final thought ... if each of the 5 inputs has a series resistor to their pin and a diode to the emitter of a single pnp bjt, with its collector tied to ground and its base pulled to 3v3 by a pair of diodes of the same type in series ... this should clamp any input at 3v3*. In terms of power handling you'd have to decide whether one pin might be shorted to 5v or more than one - I'm not sure this active voltage clamp will work if there is a resistor on the collector (but if it does there's more scope to limit the series resistor on each input).

Note: * I'm assuming that the diodes have a similar or slightly higher voltage drop to the drop in the bjt when in active mode (bjt is not in saturation, so needs to be sized with this in mind).

Folknology 2019-04-05 14:28:28 UTC #13

All good points @piers but I am trying to reduce BOM.

But I am going with the jumper that needs soldering, my assumption here is if you do this, you know exactly what you are letting yourself into!

Here is the executive decision maximising Analog for MixMod (Mx) pinouts:

3v3,GND,D01,D02,D03,D04,A01,J5v,A02,3v3,GND,D05,D06,D07,D08
3v3,GND,D09,D10,D11,D12,A03,A04,A05,3v3,GND,D13,D14,D15,D16

J5v - Jumpered 5v, default is NC so user has to solder bridge jumper to get 57