modeling 4461 card

[MisterDAQ]

Hi all,

Recently I try to make some measures with the 4461 pci card. At the beginning, I use it for for audio. Nevertheless, in audio, we use caps, and working this caps of different types, I would want to measure their ESR using this card.

I try to model the circuitry using pspice. Many things have happened

- First I use the circuitry block diagram shown in the ni 4461 user manual

- bad idea, because I discover that there exists a capacitance between the bnc plug and the chassis ground. Why is it not present in the circuitry diagram ? This information is given in the section "input impedance": 1Meg//217pF between positive input and chassis ground and 1Meg//229pF between - and the chassis. Ok, strange asymmetry in these values (why ???)

- Testing the BNC plug, we can easily see that, this cap exists on the output channels too. Nothing about this in the docs ... STRANGE

- Reading the "Specifications Change Note to Users " I discover that the output impedance value is 2k5Ohms and not ~50kOhm like in the block diagram... why  ? Dont know. Maybe, the impedance changes when the DAC is working.

This is really annoying. There is everything in this card to make a proper work. Why maintaining such mysteries ? Trying to measure my esr by estimating the phase shift, I have error of 5 or 10 degrees depending on the model I choose in pspice... quite dangerous

The are not a lot of components between the dac and the adc. I believed that it was possible to model the behavior of circuitry.  

Message Edité par MisterDAQ le 08-06-2007 08:54 PM

[ClementG]

MisterDAQ,

I'm not surprised that not all of the components of the card are list in the circuitry block diagram of the user manual. As stated in the manual, the block diagram is aimed to describe the theory of operation, not to list all components of the card.

Moreover, I don't think any company that sells products will give every details which would allow anyone to make or copy its product.

Regards,

Message Edité par ClementG le 08-08-2007 09:14 AM

[MisterDAQ]

yes but, it is quite easy to dismantle the card and see all the used components. 

There a lot of components, more than the one described in the doc. It is easy to see that there is a cap between the out channel like in the input channel. In fact, the front end circuitry is very similar on the two types of channels. But, they only try to describe the caps present in the input channel in the doc and they say nothing about the output channel... why ??? in the last manual of 2006-2007, corrections have appeared relative to the equivalent resistor between the output channel and the ground, going from 100k to 4.5k... quite different.

I do not think this is a good strategy to hide everything to be more simple. Nevertheless explainging everything is a tough job. So, maybe, there is no solution to this problem.

Message Edité par MisterDAQ le 08-08-2007 05:45 AM

[EBL]

The block diagram is probably over-simplified. It doesn't indicate the parasitic capacitance of the many components in the input section, nor does it indicate the 150 pF filter capacitors to ground at the input connections, which account for the majority of the capacitance. (By the way, the reason for the asymmetry between the capacitances to ground from the + and - inputs is probably that the outer shell of the BNC has a non-trivial amount of capacitance to ground, as would also any BNC cable attached to it.) This capacitance obviously affects the frequency response of the board whenever the source has a significant amount of impedance. To maintain a frequency response flat to within 1 dB all the way to 100 kHz, the source impedance should be less than 4000 ohms.

The filters are there to prevent high-frequency signals from entering the board and aliasing inside the A/D converters. They're also there to prevent the board from radiating RF power from the BNC cables. Remember that NI products are required to pass CE and FCC requirements for electromagnetic radiation.

The output section has the identical filter structure at the connectors, for the same reasons as for the input, and so there is also at least 150 pF to ground from the + and - connections. Again, the block diagram is probably over-simplified. However, these capacitors are probably not as consequential as the capacitors at the input.

The block diagram is definitely misleading as to the common-mode output impedance. While there really are 100 k resistors to ground from both the + and - connections, the common-mode output impedance is dominated by the need to control degeneration in the balanced transformerless output stage. Without this degeneration-control, the output impedance would indeed be high, but so would the common-mode output DC offset, leaving the board with an almost uncontrolled and certainly unacceptable amount of DC common-mode voltage. So as a result of this, the board has a much lower common-mode output impedance (and much lower common-mode offset), which comes to around 2500 ohms and is listed in the most recent specifications as 2400 ohms.

Hope this helps,
Ed L.

[MisterDAQ]

I agree with you. Nice information. thx

Nevertheless, it is quite difficult to work with it. Around the DAC, there are two 4800ohms resistors connecting each channel to the AO ground. This implies that (if I understand well) that any resistors connected to the BNC will be in parallel with these two resistors 4800+4800=9600. So if I want to my resistor should be <<< than this value. Moreover,  the source impedance must be greater than 600 ohms in order to stay under the 15mA limitation of the card. So... quite difficult we can only work with 1k resistors...

Mainly, but I can only say that know, we can only use these card in differential configuration, I mean with the same components on the + and - chanel.

Nevertheless its THD is great, and ni is good for this achievement. This card can not be excellent for everything.

Maybe I should have another card, to measure for example ESR or things like that, where distortion is not the problem. What do you think of that ?

Message Edité par MisterDAQ le 08-08-2007 07:34 PM

[EBL]

Actually, I measure impedance, including ESR, all the time using the 4461 and a simple fixture (an impedance bridge, of sorts). Most recently I've been using it to look at the dielectric properties of different circuit board materials. To my knowledge, it's the best impedance-measuring tool we have in house up to 100 kHz.

The attached image illustrates the fixture I use. With it and the 4461, I'm able to measure impedances from about 1 milliohm to about 1 Gohm over the 100 Hz to 100 kHz frequency range simultaneously, limited mainly by the stability of the ratio of the gains of the two input channels. AO0 generates a multi-tone signal, which is fed into the fixture. AI0 measures the voltage at one side of the 1 kohm test resistor, and AI1 measures the voltage at the Device Under Test. I look at the vector ratio of the two inputs to determine complex impedance over the frequency range.

A calibration routine eliminates all parasitics and unknowns in the 4461, the fixture, and the cabling. During the calibration I connect a short, an open, and a well-known 1 k resistor at the DUT connection, and the routine stores the correction constants in a cal file.

If you're interested in the code, I'll clean it up and post a copy of it here.

Ed L.

[MisterDAQ]

<< AO0 generates a multi-tone signal, which is fed into the fixture. AI0 measures the voltage at one side of the 1 kohm test resistor, and AI1 measures the voltage at the Device Under Test. I look at the vector ratio of the two inputs to determine complex impedance over the frequency range.>>

Good job and great idea. I think to this idea this afternoon, trying to use the two AI channel, but when I test it... do no succeed.

I do not understand why there is a 200uF cap the 6.8 resistor... Im sure they are usefull, but for what ?

The 1 ko resistor is the impedance you choose in order to compare with the impedance of the DUT  ? It could be 100 or 5ko ?

Im very interested in your fixture and your application. My motivation has been multiplied by 10, and I'm sure now that I can continue to work with this card.

<<<A calibration routine eliminates all parasitics and unknowns in the 4461, the fixture, and the cabling. During the calibration I connect a short, an open, and a well-known 1 k resistor at the DUT connection, and the routine stores the correction constants in a cal file.
If you're interested in the code, I'll clean it up and post a copy of it here.>>

I do not understand. I believed that when the 4461 was calibrating, a simple relay disconnet the inner machinery of the card from the BNC plug. Apparently what you say make me think that this is not the case at all.  Im very interested in this code and in your method, but im relatively quite new to the calibration procedure of NI.

In any case, thx for the great help

[EBL]

>>I do not understand why there is a 200uF cap the 6.8 resistor... Im sure they are usefull, but for what ?<<

I often measure power supply impedance, where there could be a DC voltage of 20 V or more. Since the 4461 has no AC-coupling in its output stage, I need the capacitor to keep DC voltage out of the 4461 driver stage. The 6.8 k resistor is there to provide a ground reference on the DUT side of the capacitor. Its value is not critical and it probably should be much larger, but when I originally built the fixture (many years ago), it's what I had sitting around in the lab. Perhaps 100 k would be more appropriate. If you're never going to measure a circuit with voltage on it, you don't need either of these components.

>>The 1 ko resistor is the impedance you choose in order to compare with the impedance of the DUT  ? It could be 100 or 5ko ?<<

Yes, the 1 k is the "comparison", or bridge resistor. Essentially it's the characteristic impedance of the bridge, so 1 k is the impedance that the bridge measures best. But since the 4461 has so much dynamic range, the system can still measure impedances 1000000 times higher or 1000000 times lower. You could make your bridge impedance higher or lower, but as you pointed out, the 4461 AO0 can only drive 15 mA or so (it can typically go up to about 25 mA), so you can't use much less than 1 k. One good reason to use 1 k is that you can easily find precision 0.1% or 0.01% 1 k resistors, which you'll want for the bridge resistor and you'll absolutely need for the calibration resistor (connected as a DUT during calibration).

>>I do not understand. I believed that when the 4461 was calibrating, a simple relay disconnet the inner machinery of the card from the BNC plug. Apparently what you say make me think that this is not the case at all.  Im very interested in this code and in your method, but im relatively quite new to the calibration procedure of NI.<<

Sorry for the confusion. When I say "calibration", I'm referring to my own calibration VI, which calibrates the entire system, including fixture and cabling. That's independent of and in addition to the NI calibration for the board itself.

I'll work on cleaning up my code into a presentable state and try to post it here within a few days. It occurs to me that another good reason to use 1 k for your bridge is that I may have inadvertently hard-wired that value into the code.

Cheers,
Ed

[MisterDAQ]

thx you a lot for all your information.

I go on holidays for ten days and after that I will come back
to this very interesting problem !!!

sys EBL

[MisterDAQ]

<<<I'm referring to my own calibration VI, which calibrates the entire system, including fixture and cabling. That's independent of and in addition to the NI calibration for the board itself.>>>

I tried to study this option. But its seems that the inputs of the card have at least 200pF // 1M impedance. So when we work at 50kHz, for example, the impedance falls to the value of the capacitance. It seems to be too complicate to evaluate the exact impedance of the input and the ouput. Nevertheless your fixture entails to forget the problem of the output chain because we have an input that stores the generated signal. Using the two 4461 inputs, we can consider that they have the same impedance (for a fixed frequency). But we do not know it... I do not succeed in finding how you can calibrate your calculus in order to bypass all these problems. So the challenge is always opened for me.

Always on holydays 🙂 sys

Message Edité par MisterDAQ le 08-13-2007 08:11 AM