EEZ Bench Box 3 (BB3) and beyond...

[Kleinstein]

The input OP should be powered by the same voltage as the CMOS switch chip.  AFAIK the 74LVC1G 3157 is 5.5 max.
It should be OK to power the OP from 0 / 5 V.  The input is kind of floating, but via the resistors there is a link. In many cases one of the inputs would be essentially tied to ground.

If an OP with higher supply (e.g. +-5 V is used), one would need some extra current limiting resisor before the switches. The OP than would no longer need to support single supply operation.

[jan28]

No, you didn't, after all you started this interesting enhancement, and I'd like to add it into the next revision.
If LM293 is powered from dual rail then we need to limit "low" level with diode or something otherwise -V will appear on its output that I believe is not something that 2G08 would like to see on its input.

I was not completely clear:
I think only the opamp (IC4A) needs dual rail, the comparators can run from single rail to GND. The opamp (with resistors and caps) shifts the rprog- input to GND. Dual rail of the opamp allows a wider range of this shift.

Trying to figure out the limits of the MCP6001 I noticed the rail-to-rail definition is almost rail to rail: 25mV above ground is the minimum output. This is fine for the drawn jumper configuration because 0V output is around 53mV, but if we want 0mV on rprog = 0V output by switchting the jumpers from GND to GCOM 25mV is not enough to reach 0V output.

[Kleinstein]

A +-5 V supply for the OP could help for a larger range. The GND/ vs GCOM jumper would not make a big difference as the OPs ouput would still be the same, just the external votlage changes. Even the OP input would still see the same GCOM/2 for zero set point. There are plenty OPs for 10 V operation - starting from a 741, but also OPA202 if precision is wanted.

[prasimix]

Ok, so in the latest version of the schematic I only need to change the input op amp with someone that has a split rail power supply like the suggested OPA202?

[jan28]

Ok, so in the latest version of the schematic I only need to change the input op amp with someone that has a split rail power supply like the suggested OPA202?

Yes

[jan28]

Regarding the scaling of the rprog voltages, I found a very simple 'poor mans' solution that is very flexible and doesn't require addition components:

The user can just put equal resistors on the rprog+ (X1-1) and rprog- (X1-2) inputs. You can to this to achieve almost any scaling to make it less sensitive (not more sensitive).

This works because the difference amplifier setup is such that the output of the difference opamp is (rprog+  - rprog-) * R4/R5   
(R9 equals R5 and R11 equals R4 all the time to make this work). By adding additional equal resistors at the inputs (in series with the individual signals, not across/parallel) you effectively increase R5 in the formula.

Currently the rprog/output ratio is about 0.0625 V/V (increasing rprog with 62mV increases output with 1 V)
Adding 6k to both inputs makes the rprog/output ratio  0.1 V/V
Adding 150k to both inputs make the rprot/output ratio 1 V/V

The 0V offset due to GCOM gets amplified as well (to around 0.8V in case of 1V/V)

Just add this to the manual with formula and a table of some common example values (0.1, 0.5, 1 V/V...) and everybody has arbitrary scaling.
Personally I prefer the 0.1V/V above 0.0625 because it calculates a lot easier.

The formula if jumper is connected to GCOM:
Vout = 10k/(Ruser + 10k) * Vrprog * 16  or simplified: Vout = 160k/(Ruser + 10k) * Vprog

The formula if jumper is connected to GND:
Vout = 10k/(Ruser + 10k) * (Vrprog * 16) - (Ruser + 10k)/10k * 0.053

[Kleinstein]

Ok, so in the latest version of the schematic I only need to change the input op amp with someone that has a split rail power supply like the suggested OPA202?

With a supply to the OP that exceeds 0-5 V one should also have a resistor at the OPs ouput, to limit the current to the switch chips (mainly ground side) and a diode to GND, so that the comparators would see very much negative voltage. So it is not a big changes.

The not on reducing the input sensitivity is good - if a 10 V FS range is wanted just add external resistance (some 30 K each from a cride calculation).

[prasimix]

New proposal: Op amp model changed and its supply to split rails, D3 and R1 added for switches protection.

[prasimix]

IC5 pin4 line has to be something else: e.g. I_SET_OUT

[Kleinstein]

The OPA202 can deliver quite some current (35 mA typical) in case of a negatvie signal. This could be a significant load and cause problems, e.g. if the external input is open and just picking up hum.  As far as I understand the circuit connected to the set signals does not care very much about the signal impendance. So one could add another resistor (e.g. 1-2 K range) between die OP and D3 to limit the currrent.

[prasimix]

... or just to move D3 after R1?

[Kleinstein]

Die Switch chip likely alteady includes diodes to the negative supply, so behind R1 there is no real need for an extra diode. The tricky part is that with current through a substrate diode, the normal function is no longer guarantied and the normal set point my be effected. The signal is not that high in impedance, so it may still work - just have to test the swtich chip in real life.

[prasimix]

Ok, I will make a small PCB to test this circuit so I will leave room to add a resistor on both sides of the diode D3.

[prasimix]

The new MUX14D module is ready to be added to BB3. In fact it is a companion to the MIO168 module that can bring more signals to the ADC inputs on one of the connected AFEs.

The module has two 7:1 multiplexers that can be combined into one 14:1. It is made with reed relays so that the switching speed is not high. The inputs are 2-wire instead of single ended which consumes twice as many pins on the connectors. It may be worthwhile to make a variant with single-ended inputs for more inputs per module. The outputs from the multiplexer also go to the external terminal, but can also be connected internally to the MIO (or some other module) via ADIB connectors (currently not mounted on the prototype in the picture below).

I also glued a "cold junction" sensor to the bottom of the lower terminal connector.



Schematics in attachment.

[AlanS]

A thing of beauty!

[fuzzoli]

We are all going to need a bigger box!

All lame lines from 45 year old shark attack movies aside, will there be any options for some kind of auxiliary / expansion box for all of these wonderful boards?

[prasimix]

One idea is to make an expansion chassis that could accommodate up to 7 non-power modules in the same form factor as the BB3 (i.e. those that do not need a Mean Well 48 Vdc converter). Communication with BB3 or PC could be via USB or Ethernet.
However, there are a lot of things to define: whether the AUX-PS and MCU modules will remain separate, or perhaps merge them into one? Should AUX-PS/MCU will be connected behind the new 7-slot backplane or should it be a module that goes into a dedicated (8th slot), will the FPGA be added from the beginning or will we just stay on STM32, etc.

[prasimix]

Another prototype made by @goran.mahovlic of MCU-FPGA (ULX3S based) module is ready for testing. We now have a functional pass-thru FPGA of all DIB lines which will give us more flexibility in defining their functions and allow higher transfer speeds.

[prasimix]

We started doing the first tests with AFE3 for the MIO168 module. Unlike AFE1 where all four analog inputs share the same GND, here we have two fully isolated channels for which TI isolators with DC/DC converters were used.
The first input (AIN1) is voltage only, the AMC3330 is used and has two ranges: ±50 V and ±450 V so it will allow measurement of AC mains up to 320 Vrms. The second input (AIN2) is current only, AMC3301 is used and also has two ranges: ±1 A and ±10 A. This in combination with the remaining two inputs which are combined voltage/current as in the case of AFE1, and with simultaneous A/D conversion I expect that there could be a usable solution for measuring input power, output power and the efficiency of AC/DC and DC/DC converters, battery chargers, etc.



I have to say that the mentioned isolators look very promising. Here we come to one interesting detail: the problem of measuring large currents (e.g. over 4 A) where the TCR of the current shunt is very pronounced. I have already discussed this with @Kleinstein and complained that the ROHM of 0R010 (GMR100HTBFA10L0) used with a TCR of only ±20 ppm/K does not seem to be good enough. After all it’s just an SMT component in the 2512 packaging so that shouldn’t come as a surprise. I tested it on AFE1 where I have it on AIN3 and AIN4 inputs. I use the same with AFE3 on the AIN2 and AIN4 inputs with the difference that behind the one on AIN2 is the already mentioned AMC3301.

A thing that is interesting and I don't know how to interpret at the moment: when measuring a current of 5 A at the AIN4 input in a period of 15 minutes, I have a drift of +6 mA. However when measuring the same current on AIN2 (with AMC3301) in the same period and using the same current shunt I have a 0 mA drift!

The first thing that comes to my mind is that the current shunt (R11) on AIN2 is better cooled because there is more copper on the PCB. However, this is not true. The traces to the shunt (R32) on the AIN4 are longer, so there is more copper to keep it cooler. Perhaps this drift should not be attributed to the current shunt at all but to some other component. The schematics (which does not include the latest changes but is relevant to the prototype I am testing is attached).

[exe]

A thing that is interesting and I don't know how to interpret at the moment: when measuring a current of 5 A at the AIN4 input in a period of 15 minutes, I have a drift of +6 mA. However when measuring the same current on AIN2 (with AMC3301) in the same period and using the same current shunt I have a 0 mA drift!

Is it possible to just measure the voltage across the shunt? 6mA over 5A is 0.12% which should be easily measurable. This way we eliminate the shunt. I quickly checked the schematic, but can't say anything specific. Can it be a board contamination with flux and some leakage currents?

Is the problem only present at 12A range? I'd check on smaller ranges if they drift too. I'd also use a hotair gun to warm up the board a little bit and see how that affects readings.

[Kleinstein]

Some 0.1% drift for the current reading would need 50 K temperature rise to get his from 20 ppm/K TC of the shunt.
This is a bit much for the low value shunt.  The fuses may be the larger sources for heat than the shunt.

Beside the TC, also the way the voltage sense is exactly connected could cause an error. It may effectively include some copper with the shunt. So it could be about the layout details at the shunt.

Some of the clearances on the board are quite small, e.g. the Ain3 input.  So one should not consider the input suitable for direct mains use. More like a class 1 safety limit.

Die a 5 A current reading an not much extra series resistance in the sense line leakage currents on the board are likely not an issue.

[prasimix]

AIN3 is not intended for more then ±240 Vdc, but ok more clearances wouldn't hurt. I have to check what is possible.

On the output of AIN2 high range shunt I have tiny filter, don't know if that helps in getting better measurement, but that shouldn't be connected with drift caused by rise of temperature (if we assume that is the major reason for drifting).

The exactly the same fuses are used on both AIN2 and AIN4 channels, so don't know how they could participate in such huge drift.

I'll try to see if 0.12% error is measurable directly on shunt as @exe suggested, also try to play a bit with hot air gun.

PCB is also everything but clean . Lots of flux is all over it. I'll try to get rid off that mess. One more thing, although the same circuit is used for non-isolated input on AFE1 and AFE3 prototypes, results on AFE3 (6 mA drift) is more then three times better. I had 20 mA drift for 5 A on AFE1 in much shorter time.

[prasimix]

I did the following simple experiment. I soldered ROHM Rsense to the PCB, measured the temperature before letting 5 A through it. After a few minutes the temperature stabilized at about 40 oC. So it turns out that the drift should be around 400 ppm which is 30 times better than 0.12% at the AIN4 input.

[exe]

As I understand, some inputs use isolated amplifiers. Those can drift badly. I checked datasheets, they have 40-45ppm/C worst-case gain drift. So, I'd say, they can noticeably contribute to the total error. So, 20mA drift on those channels can be a bit different in nature.

I still hope it's just flux leftovers . Most fluxes are slightly conductive, even if claimed "no clean". This is especially a problem in elevated temperatures. So, even if flux passes tests at 25C, it doesn't mean it will behave the same under 50-70C, or in humidity, or when it's electrically stressed (according to one flux chemist on youtube).

[prasimix]

The thing is, it behaves better on AIN2 which has an izolator than on AIN4 without it. One possible explanation is that the TCRs of Rsense and Isolators are of different sign and cancel each other out?