How to build an impedance measurement rig using an amplifier

vic1184

Registered
Thread Starter
Joined
Mar 1, 2021
Messages
5
Hello everyone. I've been using REW for quite some time and it's a very powerful software. I made my impedance measurement rig as described on the REW website and I used the headphone output and line input from my soundcard.
However, I'd like to use an amplifier to drive the speakers for more accurate results. The website says "If using a power amplifier the sense resistor can be much lower, 33 ohms or less, but the soundcard inputs should be connected via a resistive divider providing around 20dB of attenuation and ideally the inputs should also be protected by back-to-back zener diodes to clamp the input to less than 5V."

How do I pratically build that circuit? Does anyone have any schematics?
Thanks everyone!
 

jschwender

Member
Joined
Mar 16, 2021
Messages
140
Location
GERMANY
More  
Front Speakers
Nubert digital pro
Other Speakers or Equipment
Philips dss940
I use an external amplifier and a 29 and 4 ohm precision resistor with Ayrton-Perry-winding built in, and a 1/10 voltage divider built in. Works pretty well for impedance measurement between 10 mOhm to approximately 1 k Ohm.
49144

The amplifier has 10,0x amplification, and i can do measurements up 12 V, with a bandwith of 0.07 Hz…350 kHz. No clamping diodes or other protection, as my audio device can handle these voltages. If protection is needed, i would prefer a bidirectional TVS, not Zener. the output impedance of this amp is 5.7 mOhm + 1.55 µH. I bought a kit and assembled the pcb myself, this way i was able to modify and improve the amp to my needs. As the amp has distortion in the order of magnitude of 0.005% i can do linearity measurements on inductors too. The power supply is externally fed in with a multiwire cable 90 cm. I have different supplies for different requirements that i can just plug into, one with very low coupling capacity and low hf noise, made up by a cutting tape core transformer and capacitors, and another supply that is well stabilized made as switched power supply. The measurement resistors are built in because the connection of all the parts are critical in terms of resistance and inductance, this way i avoid clamps switches, plugs and long wires in the power path.
 

vic1184

Registered
Thread Starter
Joined
Mar 1, 2021
Messages
5
hello. thank your for your response. i have a way simpler and less accurate setup than yours, which consists in a AV Harman/kardon amplifier and a PC. I have a behringer umc-22 soundcard, which has stereo output and mono input, therefore I have to use the line-in on my integrated soundboard which actually is pretty decent noise-wise. I'm just looking for something that can drive speakers a little harder for measuring the impedance (and tuning) of bass reflex subwoofers and bigger, less sensitive speakers in general. how can I integrate everything for my system?
 

jschwender

Member
Joined
Mar 16, 2021
Messages
140
Location
GERMANY
More  
Front Speakers
Nubert digital pro
Other Speakers or Equipment
Philips dss940
If you have only one input (mono) impedance measurement will not work. It takes two and they should have best matching in their characteristics and need to utilize the same analog to digital converter to run synchronously. Having only one input is also for frequency response measurement a limitation, as you will not have a timing sync channel. If you want to measure speaker impedance at reference power level like 1W any simple amp will do. If you are looking to measure power compression, you will need to have a more powerful one. For integrating everything into one box, you may have seen this https://artalabs.hr/AppNotes/AN1-MeasuringBox-Ver2-Rev3Eng.pdf
 

vic1184

Registered
Thread Starter
Joined
Mar 1, 2021
Messages
5
Therefore I will first need to build that circuit, and using the integrated soundcard (won't be ideal but it's stereo) both for the output and the input I will have to do the "open circuit calibration" on REW with the speaker disconnected (or the switch in the CAL position) while entering the value for the sense resistor of 27 ohms, then just proceed as usual by doing a short circuit calibration by shorting the terminals on the "loudspeaker" side and then a reference calibration with a resistor of known value.
Did I get it right? Thanks for your help.
 

trobbins

Member
VIP Supporter
Joined
Nov 26, 2017
Messages
121
Location
Australia
Perhaps be mindful that the 'accuracy' of a measurement may have a few factors to consider. An accurate measurement of a passive part (eg. an inductor) likely comes down to your calibration and lead setup, and hence to the tolerance of your calibration reference resistor and how much you have to change the lead setup to test another part. Note that some parts like inductors with magnetic cores have an inductance value that depends on excitation voltage and so comparing measurements needs to also specify the excitation voltage. Some parts are not 'passive', such as the impedance of a speaker coil, which varies in time as the room acoustics set up standing waves, and in relation to the position of the coil in its magnetic structure.

An REW impedance measurement is a short duration (seconds to 10's of seconds) sweep of frequencies, and as such is not a steady-state measurement with respect to frequency. You may be able to alleviate that 'dynamic' frequency change by using longer FFT levels (eg. 4M) and by reducing the frequency sweep span, such as repeating the measurement but using a small span around an important frequency like a speaker resonance, and comparing the results.

I use a 1k sense resistor and a soundcard headphone output and nicely get accurate results down to the milliohm and uH level, and I have access to a 10 ohm 0.05% reference resistor for calibration.
 

sm52

Member
Joined
Mar 14, 2019
Messages
863
Note that some parts like inductors with magnetic cores have an inductance value that depends on excitation voltage
Do you have observations of how much the inductance changes in this case?
 

sm52

Member
Joined
Mar 14, 2019
Messages
863
Maybe this phenomenon is typical for large inductors, the inductance of which is 30 H or more? But for crossover inductors, the inductance of which is 2-10 mH, this is not the case?
 

trobbins

Member
VIP Supporter
Joined
Nov 26, 2017
Messages
121
Location
Australia
It is related to the core material and construction and how that affects operation on the BH curve, including if dc bias is present. Certainly the situation for steel laminated transformers and chokes and cross-over inductors. If the core has a very linear BH curve then not such an issue, including where an air-gap linearizes BH curve. An air-core inductor shows no such affect. There could be a lot of different core materials used, including a whole range of ferrite and iron-dust and amorphous etc.
 

jschwender

Member
Joined
Mar 16, 2021
Messages
140
Location
GERMANY
More  
Front Speakers
Nubert digital pro
Other Speakers or Equipment
Philips dss940
Perhaps be mindful that the 'accuracy' of a measurement may have a few factors to consider. An accurate measurement of a passive part (eg. an inductor) likely comes down to your calibration and lead setup, and hence to the tolerance of your calibration reference resistor and how much you have to change the lead setup to test another part. Note that some parts like inductors with magnetic cores have an inductance value that depends on excitation voltage and so comparing measurements needs to also specify the excitation voltage. Some parts are not 'passive', such as the impedance of a speaker coil, which varies in time as the room acoustics set up standing waves, and in relation to the position of the coil in its magnetic structure.

An REW impedance measurement is a short duration (seconds to 10's of seconds) sweep of frequencies, and as such is not a steady-state measurement with respect to frequency. You may be able to alleviate that 'dynamic' frequency change by using longer FFT levels (eg. 4M) and by reducing the frequency sweep span, such as repeating the measurement but using a small span around an important frequency like a speaker resonance, and comparing the results.

I use a 1k sense resistor and a soundcard headphone output and nicely get accurate results down to the milliohm and uH level, and I have access to a 10 ohm 0.05% reference resistor for calibration.
That is amazing, it means that your setup must have 120 dB usable dynamic range (1000 Ohm to milliohm). So your milliohm results have only very limited accuracy, right? The systematic error is minimum if the object resistance has the same value as the reference resistor, below or above the error is larger. Accuracy also largely depends on temprerature, so a measurement with phones output has certainly very little effect to object temp. This is a problem if i measure with higher power: i can see the rise in resstance for example within a 2 second sweep. And if you have a core material other than air, it has a hysteresis, which is nonlinear, and causes systematic error that is not in the measurement model, right?
 

trobbins

Member
VIP Supporter
Joined
Nov 26, 2017
Messages
121
Location
Australia
I was able to confirm accuracy of circa 0.1uH for a reference 10uH aircore inductor, and was able to measure inductances of around 1uH in a relative accuracy sense (to confirm a lowering inductance value as turns were taken off a commercial inductor to use in the feedback loop of a valve amp to improve high frequency stability as I was after a nominal 1uH level).

On a recent 0.1% precision decade resistance box I measured about 7-8 milliohm for the 0 ohm pass through using a HP3497A with 1 milliohm resolution, and an REW impedance sweep showed 10 milliohm at the DC end (I was doing sweeps of the decade box to show up impedance change with frequency). Doing a repeat short circuit test (using the same setup as the short circuit calibration) the impedance plot typically shows down at the 0-1 milliohm level. I haven't really tried to do such low resistance REW tests so far due to need, although I have a few current shunts and other known milliohm 1% resistances that I've used to confirm operation of two double Kelvin bridges and an ESR meter based on diyaudio Jay_Diddy_B's meter that provides 0.1mohm resolution.

I typically do an impedance measurement at the max known (low distortion) output level of my soundcard in order to maximise S/N ratio. Getting good impedance plots can be fickle for some parts like inductors and I often have to do repeat tests until I get a 'clean' plot.

Measuring resistance, or using a resistor in measurements (whether it be a test load for a 50W amp or a sense resistor in an impedance jig) certainly can require some care. There have been a few interesting diyaudio threads on distortion introduced by the test load resistor for power amp measurements. What sort of % power is dissipated in your sense resistor relative to its power rating?

Wrt BH curve non-linearity, I recall that some ferrite types in gapped coresets had quite linear BH curves and only showed up significant non-linearity as saturation region was entered. Gapping helps a lot. The only time I've attempted to look out for transformer core related distortion was when testing Williamson valve amp related output transformers, and yes that indicated that REW distortion results were often 'in the weeds' and showing up as partially transparent.
 

sm52

Member
Joined
Mar 14, 2019
Messages
863
If the core in the inductor is not ferrite, but a set of metal plates with high magnetic permeability, the inductance of such an inductor will depend on the magnitude of the current? For example, the inductance is declared 7 mH. The current can vary as in a conventional 60-120W power amplifier.
 

jschwender

Member
Joined
Mar 16, 2021
Messages
140
Location
GERMANY
More  
Front Speakers
Nubert digital pro
Other Speakers or Equipment
Philips dss940
Measuring resistance, or using a resistor in measurements (whether it be a test load for a 50W amp or a sense resistor in an impedance jig) certainly can require some care. There have been a few interesting diyaudio threads on distortion introduced by the test load resistor for power amp measurements. What sort of % power is dissipated in your sense resistor relative to its power rating?

Wrt BH curve non-linearity, I recall that some ferrite types in gapped coresets had quite linear BH curves and only showed up significant non-linearity as saturation region was entered. Gapping helps a lot. The only time I've attempted to look out for transformer core related distortion was when testing Williamson valve amp related output transformers, and yes that indicated that REW distortion results were often 'in the weeds' and showing up as partially transparent.
Do you mean the resitor itself causes thermal distortion? The resistor is a 50W with ayrton-perry winding and 20 ppm temperature coefficient, that is at least much lower than copper in objects. In theory power load could be 100%, but i never did that, usually i apply signal like 6 Veff on the amp output, which for low impedance objects that would be less than 20%.
Right, gapping reduces saturation as well as hysteresis. I have a setup to measure saturation of inductors up to 100 A, this is useful in order to make sure a impedance measurement is in the more or less linear range. But hysteresis is something that appears in the result within the resistance part, but the computational model in REW does not take this behavior into account. So you get a result, but you have no idea about the accuracy.
 

jschwender

Member
Joined
Mar 16, 2021
Messages
140
Location
GERMANY
More  
Front Speakers
Nubert digital pro
Other Speakers or Equipment
Philips dss940
If the core in the inductor is not ferrite, but a set of metal plates with high magnetic permeability, the inductance of such an inductor will depend on the magnitude of the current? For example, the inductance is declared 7 mH. The current can vary as in a conventional 60-120W power amplifier.
Yes, it does. More or less. Metallic alloy cores and some ferrite core materials mostly have a wide linear range, that goes pretty sharp into saturation. Powder cores have more a soft transition to saturation, or in other words they are mostly non-linear. If you have a core it is important to know the saturation limit, as application should stay away from this limit. The form of the core also has influence on the saturation curve, as the flux density may vary with location spot.
 

sm52

Member
Joined
Mar 14, 2019
Messages
863
Thank you. From what I have noticed when measuring inductors with a three decimal place accuracy instrument, it is that the inductance increases slightly with each successive measurement.
 

trobbins

Member
VIP Supporter
Joined
Nov 26, 2017
Messages
121
Location
Australia
Some discussion and a link in diyaudio on what was being reported as load resistor related distortion, and test setup : https://www.diyaudio.com/community/...urements-with-rew.338511/page-43#post-6758592

I view inductance as a measure of incremental inductance, although the BH loop being exercised gets larger and larger as AC only excitation gets larger. The start and finish of any such transient excitation may not be on the steady-state excitation BH loop, especially if the start is from BH origin. As the BH loop is not a linear straight line then I agree that a sinewave excitation voltage would result in a sinewave current with distortion. It's likely that REW only assesses the fundamental voltage signal available to it to calculate a model result, not any harmonic levels that may relate to BH curve characteristics. Any start/finish characteristic (if it is being caught by REW) would be at the start and finish of the sweep, and any concern could be alleviated by only assessing within the sweep and use a long sweep time and a short frequency span.

It's likely that some/many other methods of inductance measurement rely on Vrms or Vav measurement (not V-H1) and so there may be some deviance in measurements between different test methods. As I see it, stating any inductance value of an inductor that is not air-core (or with a large air gap) should really also state the test conditions to allow others to relate that to their application or to compare to other test methods - not that any one method is more accurate than any other (apart from the absolute values being measured in the test method). Not that REW easily allows (unless the choke has a separate winding or an inserted magnet for biasing the BH curve) but the other way to characterise inductance is to make measurements of incremental inductance at different static bias points from the BH origin out to the saturation region (an important aspect of performance for some applications like mains frequency power supplies using choke filtering for DC generation).
 
Top Bottom