How does NI PCI 4474 DSA detect an overload to input channel?

[Hyun Oh]

I have read "NI dynamic signal acquisition user manual." It is written that NI PCI 4474 only support digital domain overload detection.  

Here's my question. What is the digital domain overload detection? If an analog signal is already converted to digital signal, there may not be signals exceeds the limit. Or the DSA may be already damaged during ADC process, if there exist overload from the input signal. Thank you.

Hyun

[gaving]

---

Hyun Oh wrote: 

Here's my question. What is the digital domain overload detection?

Hyun

---

"Digital" overload detection simply means that if the bits returned are too big we detect this as overload.  The ADC is returning 24-bit data, we apply calibration factors and look to see if what is being returned is bigger than our largest possible 24-bit word (of course, similar for largest negative).  If it is, we return an overload.  It gives you the ability to know if your input signal is too large (either positive or negative) even if the board does not have another ADC to detect this condition.

-gaving

[Hyun Oh]

Thank you for your reply.Some of your comments was not clear to me.

As you said, let's assume the ADC offers 24 bit data. Since NI PCI 4474 has 24 bit resolution, the "largest possible word" would be the largest word of 24 bit. There may not be any overload although the word can be as big as the the largest word of 24 bit.

By the way, it does not make sense that the ADC accepts the value that exceeds the overload in digital domain. If the ADC already has a capability to endure the value exceeds the limit, why do we need overload detection? Or the limit in analog domain should be different from the limit in digital domain.

I am not an expert for DSA and do not know the details of DSA. I can be totally wrong. However, as for me, digital domain overload detection does not make sense. I think we use overload detection in order to protect electrical circiut from damaging by removing overloaded inputs. 

Can you help me to understand this issue?

Hyun

[gaving]

This type of detection doesn't protect anything without user intervention.  It is only a read property that the user can read and make decisions on.

The ADC returns 24-bit data, but the chip that processes it on the board can use more than 24 bits, so, for instance, if you have 0xFFFFFF and you have a calibration factor (let's assume a linear fit slope) that is not 1.00000 in slope or not 0.00000 in offset you could overflow the math inside the chip (which could be 32-bits, 64-bits, whatever). This is how we would "overflow" the 24-bit data.

It's only a math thing.

Again, while the detection doesn't do anything unless the user chooses to do something, it is still useful.  I have seen it used by customers to check all their connections before running an acquisition.  Their program brings up a panel that shows booleans for all the channels and check if any of them are overflowing (possible bad wiring or sensor).  Then they can decide to fix the problem with that particular channel before their acquisition runs.

...is this helping?

-gaving

[Hyun Oh]

I agree with your comment that the detection doesnot protect anything without user intervention.

I would like to ask you more about the 24 bit data from ADC. My basic understanding about ADC is that ADCs will convert analog data "within a range" to digital data "within a range". For example, +10V input to an ADC with a range of +-10V can be converted to 0xFFFFFF as an output. If an analog input is +11V, the output digital value will be saturated to 0xFFFFFF or the ADC process will be screwed up.

Do you mean that the 0xFFFFFF is just max value of output from ADC and the chip would have more information about the analog input? If this is true, why just give analog domain overload detection instead of digital domain overload detection?

Hyun

[gaving · Accepted solution]

Yes, the chip has more information about the board and needs to apply calibration factors, which is how it can overflow.

What we term as "analog domain overload detection" requires more hardware (i.e. a comparator circuit) to watch the input signal.  Since the board in question does not have a comparator to monitor the input signal, and seeing as how we still want to give the user some idea about whether or not their signal is too large, we just use the math inside the chip. It's better than nothing.

Oh, and I should mention that protection is handled by a completely different circuit.  The spec should say how much voltage you can apply to the board before protection kicks in.

-gaving

[Hyun Oh]

Thank you so much. Now, I could understand that digital domain overload detection is nothing but another way of DSA overload detection.

By the way, can I ask what is your affiliation to NI? I guess you have lots of experience in DSA.

Hyun