LCR Meter using Sound Card - revisited

I just set up one of my soundcards to do LCR measurements. I took an old radial lead cap from bench marked 1100 uF, 35Vdc, and I mean old. I purposely measured using too large a series resistor, 10k, which became immediately apparent from the 'ratty' results. Toned it back down to 220 ohm, still too large, but the results were 'cleaner'.

Did not verify the methodology a lot, easier to ask this knowledgeable group about the results. The results were somewhat as expected.

Cap measured, approx 890uF, with Resr approx 90 milliohms. with age and never having voltage previously applied, the low value seemed appropriate (maybe) the Resr was disappointly high, but as expected. too

But what I saw regarding characteristics vs freq I don't know are true, or are these artifacts of what I did.

The plot of C vs log(f) produced a rather linear straight line descending down as freq went up. The plot of log(Resr-offset) vs log(f) also had a similar straight line. with offset slightly less that 84 milliohms

QUESTION: Are these relationships real, or artifacts of technqiue and/or accuracy?

For example, the Q of this cap at 20 Hz is around 25, so it didn't seem reasonable that the soundcard couldn't pick up the phase well enough to discern the Resr there, but Resr had increased at the lower frequency to around 100-200 milliohms [forgot which now!]

Also, C droppoed to around 830uF going up infreq and looked like it was asymptotically approaching 900+uF going down to '0' freq. But again, the Xc is very small compared to the Resr as you go up, so the values could be in question.

My point is, are the values about right or should I comb through the measurement technique?

PS: Earlier I said you needed a good VTVM to measure voltage and resistance. WRONG! Using soundcard for measuring C you only need a pure resistance of known value. You don't need to know ANYTHING else. assuming your soundcard's frequency is accurate :) Should be able to get better than 0.1% accuracy, too.

Reply to
RobertMacy
Loading thread data ...

Did you consider induction ?

Jamie

Reply to
Maynard A. Philbrook Jr.

...

Probably the capacitor is intended for power supply filtering (120 Hz), and you are seeing inductive impedance. Mostly, such capacitors are spiral wrapped (though stacked-foil is available for higher frequency applications).

Your measurement, if it includes phase shift as well as 'C value' should deviate from the lagging-by-90-degrees phase shift of a capacitor, for the higher frequency values, if this is the case.

Reply to
whit3rd

Hi, I missed the original posting of this thread, so I missed the reference to the software used. I tried a similar approach from "ELEKTOR"

on my PC. I think the idea of using a soundcard for impedance measuremenrt is VERY clever and would like to use it. Can anybody send me the original posting or a link to it?

Frank

Am 23.01.2014 15:26, schrieb RobertMacy:

Reply to
"Müller, Frank-Stefan (MRT)"

Frank, feel free to contact me offline to discuss in detail. You're in Germany! I use the streaming audio interface from Steinberg, called ASIO. Well done package. I then wrote C/C++ code to exercise that interface and control my soundcard directly, make it do EXACTLY what I need, and then provide back the resulting values.

For other posters I was planning on measuring 'equivalent' inductance in future testing expecting less than 100nH. Usually, inductance rears its ugly head as a rise in impedance as you go up in frequency, but just didn't seem to show up here. So ignored it partially to 'simplify' the resulting model,which just made last night.

Over freq range of 20 to 1000 Hz, using 220 ohm resistor:

The best curve fit to the data for this antique part using a series Resr and C:

Resr = 80 milliohm + 5/f, I'm NOT kidding! where did THIS come from? It's like my measuring technique got drug into it. C = 1e6*(888.34 - 34.8*log10(f/20)), again could be due to the accuracy going away as you go up in frequency.

For those who don't undertand how to use their sound card to measure an unknown impedance: [Note: the system has higher range, and rejects AC mains better when used as a 'balanced' system; it's far easier to work with if you set up the card in 'unbalanced' mode. requires only a single resistor, not two, 'absolutely' equal resistors. Smaller range, just easier to work with.]

Find a great broadband resistor of KNOWN value, 0.01%, or 0.05% if you got it, or measure a 1% one and note its R

Then use the output of one channel to drive, provide the signal source Now using the two input channels perform the following: connect L & R to the output to calibrate the two channels connect L to the drive and then resistor to R to calibrate the input impedance of R, knowing the value of R connect the unknown component across R's input

The first calibrates L channel and R channel to be identical voltmeters. The second calculates the input impedance of the R channel so you can remove that value later. [for my soundcard a close fitting model was around 10.094k with a parallel capacitance of around 0.002384uF in series with 132.2 ohm resistor (fit within 0.04% window over the range of interest), which closely matched the manual's desc. of 10k in parallel with 0.0025uF The third step provides the data to then be manipulated to find what impedance was added. Simple.

You can store the values in a fileand reload for continually running Step

3, or Initialize once and continually run Step 3. I recommend running Step 1 and then Step 2 each time you have turned on your system, let it warm up [about an hour], and then want to take measurements. The accuarcy is high, you can still see drift in the ppm after letting sit there for a couple hours. But is ok, because if you touch your sense resistor, its tempco will swamp that out. I even see the change caused by holding my fingers within a 1/4 inch of the setup during measurements. Don't even get me started on the effect of using your fingers to hold the cap under test! I digress.

But I wonder if somewhere during all this process. I missed something that injects a frequency dependent term erroneously in there. Any help appreciated.

For example, removing the frequency sensitive term from Resr and assuming

80 milliohms, and moving that term over to the C, makes C become sensitive to frequncy instead of log(frequency), kind of.

QUESTION: Does anybody KNOW what values I should be obtaining?

Reply to
RobertMacy

On a sunny day (Fri, 24 Jan 2014 07:27:04 -0700) it happened RobertMacy wrote in :

Circuit diagram would help...

Reply to
Jan Panteltje

Thought word description was enough.

I can't post a picture onto a 'service' website. Often can't even retrieve pics from them.

so here is the ascii stuff: Rsens is added known resistor, DL is drive FROM soundcard, L & R are the input channels TO the soundcard.

Step 1, Calibrate L & R channels

DL --+--+ | | L R

Step 2. Find Input Impedance of R soundcard

Rsens DL --+--/\/\/\---+ | | L R

Step 3. Measure capacitor

Rsens DL --+--/\/\/\---+----+ | | | | | === Cmeas | | | L R GND

The equivalent input impedance of the R channel is

----+-------------+ | | | | | | | === 0.0025uF | | / | \ | / 10k / \ \ / / 130 | \ | / | | | | | | | | GND GND

and the model for C

||

-------||-----/\/\/\------- ||

C Resr

hope this helps, but sure thought the verbal description was sufficient.

Reply to
RobertMacy

On a sunny day (Fri, 24 Jan 2014 08:39:25 -0700) it happened RobertMacy wrote in :

I understand this so far, you measure the current in the series RC at some frequency, and from that calculate losses in C?

My test is always differential scope probe on each side of C, if you see fast edges then C is kaput ;-) Can be done in circuit, live. I do not usually give a ... what the exact 'ESR' is, or whatever, it can take many forms, I have seen electrolytics go inductive if the foil on the edge was not making good contact, what not. The important point is that a C should be a C only, and at the value written on it.

This aloso works for a C out of circuit:

scope 1V calibration square wave output ---->------->-- ch 1 input | --- --- C under test (elco) | ///

If you see fast edges than C is kaput...

For smaller Ls and Cs I build this LC meter, very good actually:

formatting link

Reply to
Jan Panteltje

Yes, including FULL characterization over the whole spectrum.

C, Resr, R dielctric loss, L

Neat circuit and construction, thanks for sharing!

It's just in my old age, I'm getting lazy about building hardware! Plus, the soundcard can get you beyond 0.005% relative accuracy comparing parts and with a good R can get to 0.01%, even finding the Capacitance value, or inductor value.

range of C from 100,000 uF down to less than 100 pF To do with one machine is handy.

Reply to
RobertMacy

On a sunny day (Fri, 24 Jan 2014 11:37:26 -0700) it happened RobertMacy wrote in :

Yes that is a huge range, I do not think I have 100,000 uF in the box...

Reply to
Jan Panteltje

ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.