best location to place an analogue multipleoxor

[KlausST]

Hi,

Therefore Does it make sense to identify the negatives analogue ADC IOs to connect them to 0V_Sensor?

* If it has a dedicated negative input, it would be best.
* If there is a single "input reference" for differential mode, it would be just as well
* if there is the possibility for simultaneous sampling of two channels, it would be almost as good. Just subtract both conversion results as software.
* if neither of the above, then you might do an extra conversion of V_com and subtract it from the other conversion results, but the measurement is not that precise, because of different time sampling. But it still makes sense. At least wire the V_Com to the ADC (non muxed to an extra ADC input) then you are free to use it or not.

Furthermore the uC does not have an output pin vref/2

I did not expect that there is a VRef/2 pin.
But there should be a VRef pin. Use a simple voltage divider to get VRef/2.

Klaus

 

[flote21]

Hi!

After reading again the datsheet, the ADC module works mostly as the first option you say: " If it has a dedicated negative input, it would be best"

There are two internal ADCs with 24 analog multiplexed inputs . AIN0...AIN23. If you are working in differential mode, the first nine analogue inputs works like AINX_N and the rest works like AINX_P:
Differential mode setup:
AIN0...AIN7 => AINX_N
AIN8...AIN23 => AINX_P
Therefore there isn't a dedicated negative ADC pin (AIN_N) for every positive ADC pin (AIN_P). My plan is to route together all the AINX_N and connect them to the 0V_SENSOR. But in theory if just one AINX_N is connected to 0V_SENSOR, it should work too...

Finally the uC has a VDDCORE output, but it provides 1.2V and it is not enough to reach 1.25V (Vref/2). I will use an external zener to clamp the Vref/2. ¿Do you think it is good idea?

Thanks!

 

[KlausST]

Hi,

I just downlodad the datasheet.

Indeed - if one uses the PDF search function - it is not difficult to find.
You need to improve your datasheet reading skills ;-)

If you are working in differential mode, the first nine analogue inputs works like AINX_N and the rest works like AINX_P:

No. you may select one (out of 8) input as negative input. No need to use other inputs as negative inputs. You may still use them (AX0 .. AX7) as positive inputs.

My plan is to route together all the AINX_N and connect them to the 0V_SENSOR.

No. only one.

Finally the uC has a VDDCORE output, but it provides 1.2V and it is not enough to reach 1.25V (Vref/2).

No. Use VREF

****************
read about
* Voltage References System 19.8.8 VREF control
* VREFOE
* 45 ADC section, 45.1 Overview. .. tells about differential mode and the use of reference (internal, external)
* DIFFMODE
* MUXNEG (can always be left at your external Sensor V_COM)
* MUXPOS (scans one input after the other)
* 45.8.4 Input control
* 45.6.2.5 Reference configuration
* 54.10.4 Analog-to-Digital Converter (ADC) Characteristics

Klaus

 

[flote21]

Hi,

I just downlodad the datasheet.

Indeed - if one uses the PDF search function - it is not difficult to find.
You need to improve your datasheet reading skills ;-)

Klaus

hehe yes I need to read more uC datasheets. But I am finding this datasheet specially hard to understand. I have been always working with differential ADCs with dedicated P&N pins....

Anyway now is clear how the ADC works in differential mode. So Thank you very much for yout help.

However, I have using the ctrl+F shortkey to find information about the VREF and I am wondering about what the hell is telling here:



To what ADC input channel is routed the VREF? AIN0, AIN1, AIN10? Tere are also two independent ADCs with different I/O pins? Which ADC is talking about?

I was using the ctrl+F shortkey and I was not able to find a good clarification about this. I suppose that the VREF output is delivered by one of this pins:



But i need to confirm my idea...

Thanks in advance.



Thanks

 

[KlausST]

Hi,

as far as I understand (I have to admit it´s not that clear).

SUPC ---> VREF.SEL --> ADCx.REFCTRL(x = 0 or 1 for the ADCs independently) --> ADC. This is an ADC input (reference input) but not an ADC signal input
Also BANDGAP can be switched to an ADC input via ADCx.INPUTCTRL (here it is an ADC signal input).

Now it seems that the ADC Reference voltage can not be directly routed to an output.
One workaround id to use the DAC with the same reference as the ADC. Then set the DAC ouput to 50% and you will have VRef/2 at an DAC ouput pin. Not bad.
The DAC may cause some voltage-error at it´s ouput. (offset, noise, gain) .. but since you use it as negativ signal input the error should cancel out.

****
If you want it more accurate, then use an external reference voltage and feed it to REFA or REFB. Via REFCTRL you may use it as ADConversion reference.
To generate the external VRef/2 as sensor_COM you just need two equal resistors as voltage divider... as input to the buffer OPAMP.
****

To what ADC input channel is routed the VREF? AIN0, AIN1, AIN10?

It is routed internally to the ADCx.REFCTRL MUX as input (not to a pin).

Klaus

 

[flote21]

Hi!

I asked to the uC manufacturer and it says:

"A direct connection from the internal VREF has not been assigned to one of the microcontroller's pins because that functionality is already available via DACs.



Assigning the VREF to output pins via DACs has two advantages:
- Other values of reference voltages can be output, programming the DAC accordingly.
- The output pins of the DACs are suitable for analog outputs.



"

So I will use the DAD to generate the VREF voltage as the manufacturer says, but I will also design an alternative with the resistor divider just in case it does not work as i expected. Then I will be able to swich between two alternatives mounting or not mounting components.

Thanks everybody for your help!!!

Greetings

 

[flote21]

hi everybody

I know that this topic was a long time ago, but I am wondering about one issue is comming to my mind.

According to the solution described previously, VREF/2 is generated by the DAC1 of the uC on this way:

KlauST said to use an OpAmp to feed the 0V_Sensor of all the current sensors (up to 38). I have decided to use the next part: LT1494CMS8#PBF. This part has a maximum output current = 20mA and it must feed the reference of 38 LEM HTA 300-S current sensors.

Could I have any fan out problem?

Thanks in advance.

 

[KlausST]

Hi,

I doubt 20mA is enough.

Klaus

 

[flote21]

Hi KlauST!

Thanks for your reply! I found this other part: LT1461 which is able to give an output current of 50mA. But I am not sure if it will be enough....

Which can be the min output current requirement to avoid fan out problems?

Thanks

 

[KlausST]

Hi,

The datasheet tells nothing abiut current or impedance.
You should do sorme tests with a single one.

Why do you use the 2k resistor? What is it for?

Klaus

 

[flote21]

Hi! I asked that question directly to the opamp manufacturer and he told me that the maximum output current for that part is 50mA You mean 2k2 resistor? With the 5k resistor is making a resistor divider. It allows to drive to the ADC of the uC a voltage inside of its input range Greetings

 

[KlausST]

Let's say every LEM outputs about 7V...this means 1mA each LEM.
You could use a HTA1000 .... and avoid all the voltage dividers?
For sure you have to decide whether this will increase overall error.

Klaus

 

[flote21]

Hi

The problem is that the hta300-s current sensor is not possible to change. It is an external project requirement.

Are you seeing any issue in the voltage divider? Do you think if I avoid the voltage divider I can use one opamp to feed all the 0v_sensor references?

I simulated the previous schematic and it is working properly. The vref/2 opamp have enough outout strength to feed the 0v_sensor. But I don't know how to do the same simulation with 38 hta300-s current sensors. Do you have any idea to assure a proper working of the design? Maybe a solution is to use one vref/2 opamp for every four current sensors...

Greetings

 

[KlausST]

No nee to do the test with all HTAs.

Just one, then measure the Opamp's output current (Vref/2).
The current from one to the other HTA may differ, but you get a clue.

Klaus

Now any noise in the Opamp input node gets amplified. A RC low pass filter may be an improvement.

I'd even add a relatively big capacitor at the output. If you want to do so:
* Opamp_output --> 100R (node A) --> 10uF --> GND
Don't use direct feedback, but:
* Opamp_output --> 10nF --> inverting_input
And
* node_A --> 10k --> inverting_input.

(The 100R ensures stability when driving 10uF. The 10k and 10nF ensures stability while maintaining DC accuracy.)

 

[FvM]

Regarding the original question, the HTA datasheet has no specification of 0V pin quiescent current. Obviously LEM assumes that it's connected to the power supply center rather than a generated reference voltage.

Feeding a widespread reference net by an OP output is basically a bad idea because it it's likely to create EMI susceptibility. If you rely on this circuit topology, I would consider an "infinite C load" OP like TI LM7321 with respective bypass capacitor.

The circuit can be however greatly simplified by using a higher positive reference for the voltage divider and tying the HTA 0V net to power supply common. This avoids feeding a driven analog net to the HTA sensors.

 

[KlausST]

Hi

The HTAs have their own power supply. The HTA's 0V is connected to this power supply 0V. Thus the Opamp does not need to drive any HTA_0V input current.
(I forgot about this ... some posts ago)
So the Opamp_output current could be fairly low.

But I agree with the wide net noise problem. Thus my "big capacitor" recommendation. There could be HF noise from the power supply as well as other induced noise.
In case this reference signal fluctuates too much (the Opamp can't regulate out) then one could continuously measure this signal (along with the other channels) as reference for software calculations.

Klaus

 

[flote21]

Hi guys!

As far I understood, the next setting should improve the noise reject performance of the vref/2 reference voltage:

And i should not be worried about the output strength of the opamp because the 0V_Sensor will be connected to the common pad of the traco dcdc converter and it will have enough power to supply supply the 38 hta 300-s current sensors. and therefore the next overall schematic should work properly:

However traco dcdc coverter is very expensive and I am thinking to use the same approach with two switching regulators (one of them configured in inverting mode):

GND signal on the sw regulator approach must be connected to the 0v_sensor right?

Greetings

 

[danadakk]

You can do a single chip approach with a PSOC -

This is single chip, many resources still left for other uses, see right hand window resources used/left

Differential possible as some parts have 2 SARS on them, if you need I will see if I can get one to route to 38
diff channels. One SAR is limited to 32 diff channels, thats where the other onchip one comes in. Again all on
one chip.

The chip has many more resources, PSOC calls them components. For example LUTs, OpAmps, DAC, DSP Filter,
PMW, cntrs, logic.....Attached a component catalog for a chip.

To handle the bipolar signal a pair of R's per channel, one from a Vref to pin, the other pin to input. This technique
can handle - signals.

IDE (PSOC Creator) and compiler free.

Laslty the part also has analog muxes and a 20 bit DelSig on it, if you need more accuracy.

Note part has onchip Vref, accurate to +/- .1%



Regards, Dana.

 

[KlausST]

Hi,

This is not the circuit that I described.
Can you post a link to the source?

Klaus

 

[danadakk]

Here is the R divider method to handle the - swing and offset it so ADC can handle single
polarity input. R's and Vref affect transfer function, I have a calculator that can compute
them for you if so interested. This approach has lower Zin that say a OpAmp follower, but
many sensors are already buffered into a V output so Zin not as important.

Red line is Vsignal, Green is result of offset.