Impedance Analyzer Build and Experiments

[jaxbird]

The noise is only really a problem with measurements in the very low range, in something like this inductor measurement it's got no effect:




But the noise really annoys me, it should not be there. 


Edit: No indicators, and the top log scale is not correct, but this chart is 1 Ohm to 500 Ohm scale.

[jaxbird]

The chip is a current drive. So the issue you are seeing maybe decoupling or earthing?

Good suggestion, I was thinking decoupling as well, but I already have 10uF tantalum + 100nF ceramic on it (as close as possible to the chip, soldered onto the pins), plus I don't think there shouldn't be a difference from using internal and external clock if it is caused by lack of decoupling. Also in the configuration I'm using, the chip is only driving a very benign load.

On the earthing, I am investigating, but I've not been able to measure any significant bounces anywhere, it is a fairly low consumption circuit after all, max something like 80mA.

[dannyf]

That's really nice. So it is connected to and controlled by a PC? Maybe after this, you can have Analog publish your work on their website.

[jaxbird]

That's really nice. So it is connected to and controlled by a PC? Maybe after this, you can have Analog publish your work on their website.

Thanks .. Yeah I've found it easiest to explore the functionality from a PC. Just simple USB <=> UART <=> mcu <=> I2C <=> AD5933.

Also I don't really need the CPLD anymore, by using a pic24 I can use a PWM to generate a clock up to ~6MHz, so one less device to program.

It's a fun project, but I doubt Analog would have any interest, I'm just playing around really Trying to push the AD5933 to the limits.

[jaxbird]

Actually the clock noise signal is present on the output (didn't notice it because I was dealing with a 2Vpp signal vs 10+mVpp noise), so at least that gives me some options to reduce it before it enters the current amp.

With some filtering applied (not visible, all smd stuff soldered on the copper side) measuring a pretty low impedance film capacitor:






Near 10 milliohm without any significant noise is pretty good I think, the capacitor resonance peak might be lower than that, but I don't think it would be realistic to aim for anything much below that with the setup I have

[jaxbird]

Been studying and playing around with the math involved to be able to plot an ESR curve. The key is really calculating the loss angle.

Here a capacitor measurement with ESR included: (it's a bit noisy, will have to clean that up)



I believe it's correct, but I cannot be 100% sure.

[jaxbird]

And a different capacitor (electrolytic):

[jaxbird]

And just for fun, the impedance of a small 2 way speaker: (Linear scale 20 Ohm)

[The Electrician]

Wow! All the way down to 10 Hz! 

Is 500 kHz the upper limit for the AD5933?

[jaxbird]

Wow! All the way down to 10 Hz! 

Is 500 kHz the upper limit for the AD5933?

With internal 16.776MHz clock it will do down to about 10kHz, below that it will no longer capture a full wave. But with a variable external clock signal it's no problem getting down to 10Hz .

It does take a slightly more elaborate algorithm to scan the lower frequencies, something like: set the clock to desired frequency (I use mcu PWM as clock for the lower frequencies) to capture at least 1 full wave for the 1024 point DFT, then scan 1 octave and repeat.

Downside is that it's pretty slow at low frequencies, as it needs to lock on a few waves first. But speed is not really high priority, beyond faster is preferred

The official, linear until, upper frequency is 100kHz as far as I remember, but it will do up to 500kHz with a roll off that starts around 200kHz. I can compensate for the roll off above 200kHz, with some loss of resolution.

I have considered trying e.g. a 20MHz external clock just to see if it will still run reliably 

[jaxbird]

Build up a simplified 2nd AD5933 to make sure the noise from external clock I'm experiencing, isn't just due to a faulty chip from all the prototyping short circuits etc.

But no, unfortunately it behaves exactly the same, very strange 




Tested with the same "gentle" clock signal I posted a capture of earlier, no over/under shoot, and lower in level than the AD5933 supply.

[jaxbird]

Also had a quick go at it with a 20MHz external clock, but, while the DDS parts works fine, the internal ADC does not want to play along, unfortunately.

Well, it was worth a try, and my clock signal wasn't the cleanest, but I don't think it will go much faster than the internal 16MHz.

Quick and dirty clock source:



Not too bad for 20MHz on a breadboard:

[sorin]

Which program did you use to visualize the output ?
Matlab, Labview ...

[jaxbird]

Which program did you use to visualize the output ?
Matlab, Labview ...

I use C# and do all the drawing for now, but it shouldn't be too hard to either export or write a direct interface for Labview etc.

[fonak]

I've been playing around with designing and building an impedance analyzer based on the AD5933.

It's a nice little chip. Capable of generating a sinewave from low Hz (some clock scaling needed) up to about 500kHz output sweep. And on the input, using a 12 bit ADC with up to 1 MSPS, it will do a 1024 point DFT and provide the complex imaginary and real values for each step in a sweep.

It is a bit complicated to get the best performance out of it.

And I'm mostly interested in low range impedance (below 100 Ohm). The chip is not really designed for that. But adding a buffer on the output making it capable of sourcing some more current should help.

My current setup is something like this:




And the current prototype:




(Not ideal to breadboard, but it works, only sacrificing some resolution really)

I will need to have some kind of reference measurement to compensate for phase and impedance changes caused by the buffer, I think the best approach will be a few calibration resistors that can be invoked on the input.

Impedance can be calculated as:

Z = SQRT(I * I + R * R)

And phase:

Phase = ATAN(I/R)

Should be possible to calculate capacitance and inductance as well once the resonance frequency is found.

Not sure if I can calculate ESR and any other interesting parameters, suggestions much welcome

Hi all

This is a very interesting project. In atachment please find formulas to calculate other parameters.
sorry for my bad english

[jaxbird]

Hi all

This is a very interesting project. In atachment please find formulas to calculate other parameters.
sorry for my bad english

Thanks, that is a very nice description and collection of key calculations 

I have been doing a bit of work on a 2nd iteration prototype, mainly added some filtering to compensate for the roll off above 100kHz plus the ability to measure voltage across DUT using MCU ADC and some programmable lowpass filters and gain.

Sill waiting for a few parts, so not built it up yet.

[jaxbird]

Another detail I have found is that the AD5933 does not behave so nice when switching frequency range:




It does not have much effect when measuring a capacitor, but with an inductor it can cause a jump in the impedance.

[jaxbird]

Designed a new prototype. It includes kelvin connections, some frequency response compensation, and a few more features.

Ready to engrave on CNC:

[jaxbird]

And the 2nd proto pcb after 20 minutes of engraving, drilling and milling:



Happy I now have my CNC being 99%+ wysiwyg with my custom software.

[jaxbird]

And testing the nearly fully assembled 2nd proto pcb:

[jaxbird]

Testing went well until I had a minor setback oops moment, heard a pop and sensed that smell you just don't want to smell when working on electronics, I'm sure everyone here is intimately familiar with it 

This is where the smell came from:



Of course it had to be the most inconvenient component to replace 

Not sure exactly what happened, but I was flipping over the pcb to get some measurements, I think some wires on my messy table shorted it out, I did have my psu limited to 200mA though, so not entirely sure.

Anyway, it seemed to be working fine before the "incident" so I'll just replace whatever blew and go from there.

[tonyarkles]

And the 2nd proto pcb after 20 minutes of engraving, drilling and milling:

As a curiosity, what are you using for bits? Those look like very nice isolation routes.

[jaxbird]

As a curiosity, what are you using for bits? Those look like very nice isolation routes.

Hi Tony,

I use cheap Chinese 1/8" carbide 30-40 degree v-bits. With a new sharp bit I can do really fine details down to 0.5mm pitch/ 8-10 mil traces or better, but it's a bitch to solder, and I tend to use my bits beyond dull

I find the most important factor is a very solid CNC and minimal runout of the spindle for best results.

[jaxbird]

Got it up and running again after the unfortunate oops moment 

The damage went beyond the blown 74hc4053 and included the whole negative rail (2 x 7660) plus the main AD5933 

Not sure how that happened, but I guess the large amount of decoupling did it during an unluck short.

Looking on the bright side, the opamps and buffer are fine

So back to testing.

[Odysseus]

I spent a semester working on a similar project for an independent research project.  I took a very different approach, preferring to produce the IQ values as DC  voltages using mixers, subsequently read by a precision delta-sigma ADC.  I found a lot of fixable issues with my design once the dust settled, but I haven't gotten around to addressing them in a new design.  It also relied on the use of bench-top lab equipment, so it wasn't a stand alone device, either.

Anyway, here's a link to all the documentation I produced for that project.  Maybe it will be of use to you or others. Good bedtime reading at least.

https://drive.google.com/folderview?id=0BzbfuLG7ZVrvcDhmajVXejZkN1k