Here is a very simple and easy THD meter project for beginners.
Of course, it cannot compare with commercial instruments, or even projects like Cordell's, but the investment in time, money and effort is on a completely different scale.
It only requires standard, consumer-grade parts (5% carbon resistors and 10% capacitors, common opamps), needs no component sorting or selection, and is completely adjustment-free.
It works on a single, non-regulated supply, and is easy to use, making it ideal for educational/didactic purposes.
It has three ranges: 100%, 10% and 1%. On the 1% range, the measurement floor and usable resolution is 0.01%.
The indicator is simply a digital multimeter in the 2V range, the full range voltage being 1V.
The circuit is based on a SVF (state variable filter), making it highly tolerant to components inaccuracies: they have no impact on the depth of the null, only on the exact frequencies.
Compared to a double T, it is a huge improvement, and it makes frequency variation easy with any low-cost stereo potentiometer.
With the values shown, the frequency range is 500~5000Hz, this can easily altered or augmented by changing/switching the integrator's capacitors.
The discrete precision rectifier renders operation possible from 30Hz to 30KHz (harmonics up to 150KHz).
The first version has a discrete residue amplifier.
The second schematic is annotated with DC quiescent voltages, and the third shows a fully integrated version.
Some pics of my prototype follow.
Of course, it cannot compare with commercial instruments, or even projects like Cordell's, but the investment in time, money and effort is on a completely different scale.
It only requires standard, consumer-grade parts (5% carbon resistors and 10% capacitors, common opamps), needs no component sorting or selection, and is completely adjustment-free.
It works on a single, non-regulated supply, and is easy to use, making it ideal for educational/didactic purposes.
It has three ranges: 100%, 10% and 1%. On the 1% range, the measurement floor and usable resolution is 0.01%.
The indicator is simply a digital multimeter in the 2V range, the full range voltage being 1V.
The circuit is based on a SVF (state variable filter), making it highly tolerant to components inaccuracies: they have no impact on the depth of the null, only on the exact frequencies.
Compared to a double T, it is a huge improvement, and it makes frequency variation easy with any low-cost stereo potentiometer.
With the values shown, the frequency range is 500~5000Hz, this can easily altered or augmented by changing/switching the integrator's capacitors.
The discrete precision rectifier renders operation possible from 30Hz to 30KHz (harmonics up to 150KHz).
The first version has a discrete residue amplifier.
The second schematic is annotated with DC quiescent voltages, and the third shows a fully integrated version.
Some pics of my prototype follow.
Attachments
Last edited:
Interesting. Add a couple of switchable attenuators and it would be dead handy for quick power amp testing.
The input potentiometer acts as a variable attenuator, the input level can range from less than 1.5V to 100V, that is sufficient for 99% of power amplifiers.
With R1 and C1, it could probably survive momentarily a direct connection of the input to the 230V mains.
Not something I'd recommend, but rugged enough to survive "normal" mishaps.
Some instructions for those unfamiliar with manual THD meters:
-Connect input to the A.U.T. and output to a DVM, range 2V, select "Set level" and adjust the level to read ~50% (0.5V)
-With the main tuning pot P2, search for the maximum amplitude. When done, adjust level to read exactly 1.000V.
-Select "100%" and adjust P2 for a minimum reading
-Switch to "10%" and adjust fine tune P3 for minimum
-If necessary do the same in 1% range
The output is linear, and 1V=100%, thus a reading of 110mV in the 10% range is equivalent to 1.1% distortion.
Note that the rectifier is not a true rms, and for high levels of multiple harmonics, the reading will be slightly underestimated, but at "normal" distortion levels, this effect is negligible.
For example, a pure triangle wave reads 11.2% instead of the actual ~12% it should, but this shouldn't be a problem as this instrument is not destined to be used in metrological applications
I'm not sure that your 22k input potentiometer survive on a 230V main direct connection... (it must dissipate ~2.4W).
Frex
Certainly not for a long time, in this sentence:
the important word is momentarily: the time to realize there is a funny smell around, and dive to pull the plug....
Hello,
You are right Elvee, but you probably need to unplug it very faster ! 😉
So, about your design, did you have done some Bode plots to show notch depth and width in it's frequency setting range ?
Often, graphs speak more than words.
Regards.
FRex
You are right Elvee, but you probably need to unplug it very faster ! 😉
So, about your design, did you have done some Bode plots to show notch depth and width in it's frequency setting range ?
Often, graphs speak more than words.
Regards.
FRex
Here you are:
First a global outlook with the frequency adjusted at 1KHz, marker at 2KHz (gain error = 0.4dB).
Then a zoom at the two frequency extremes.
And finally, the impact of random component deviations on the depth of the null.
Surprisingly, the fundamental rejection performance of the physical prototype is significantly better than predicted by the sim.
Perhaps caused by parasitic capacitance of the potentiometers, I don't know exactly
First a global outlook with the frequency adjusted at 1KHz, marker at 2KHz (gain error = 0.4dB).
Then a zoom at the two frequency extremes.
And finally, the impact of random component deviations on the depth of the null.
Surprisingly, the fundamental rejection performance of the physical prototype is significantly better than predicted by the sim.
Perhaps caused by parasitic capacitance of the potentiometers, I don't know exactly
Attachments
Hi,
Interesting, but when i had speak about Bode plots, it was real world measurements (done with sound card for example), not simulations.
Frex
Interesting, but when i had speak about Bode plots, it was real world measurements (done with sound card for example), not simulations.
Frex
I made some more detailed measurements.
I didn't make complete plots, since it is of little interest: the general form is that of any other second order system.
Two details have an importance, mainly the fundamental leakage and accessorily the attenuation of harmonics.
As I suspected, the actual performance is better than predicted by the sim, at least for mid-range frequencies: at 1KHz, I measured 0.0065% leakage.
This degraded to 0.14% at 300Hz and 0.36% at 3KHz (I used 2.2nF tuning capacitors to center the geometric mean on 1KHz).
These figures are somewhat surprising: one would expect a degradation at the frequency extremes, but not of that magnitude.
In fact, all has to do with the parasitic capacitances: at 1KHz, the wiring to the frequency pots creates a slight capacitive overcompensation, which is about right to compensate the phase shift of the TLO84.
But at the frequency extremes, it works in the wrong direction: at the maximum frequency, there is no compenstion at all when it would be most needed, and at the minimum, the compensation is maximal, ie too large, and in addition, the available loop gain is decreased by the potentimetric arrangement of the frequency controls.
Better performances at low frequency would be achievable by connecting the pots as variable resistors, but I opted for the potentiometric configuration to improve the stability and reduce the effect of wiper resistance.
With high quality wirewound types, the resistive method would be preferable.
It should be possible to properly compensate the filter for the whole frequency range by adding small capacitors (several pF) across R4 and R6, and between the wiper and cold terminals of the pots.
I'll test it one of these days.
One (important) thing I forgot to mention last time: the tuning capacitors must be of a low loss type to be able to achieve the advertised null depths.
Polystyrene is best, that's what I used, second best is polypropylene (foil).
Mica or good COG ceramics are a third option.
No other commonly available dielectric is suitable: even polycarbonate would degrade performances by an order of magnitude.
I didn't make complete plots, since it is of little interest: the general form is that of any other second order system.
Two details have an importance, mainly the fundamental leakage and accessorily the attenuation of harmonics.
As I suspected, the actual performance is better than predicted by the sim, at least for mid-range frequencies: at 1KHz, I measured 0.0065% leakage.
This degraded to 0.14% at 300Hz and 0.36% at 3KHz (I used 2.2nF tuning capacitors to center the geometric mean on 1KHz).
These figures are somewhat surprising: one would expect a degradation at the frequency extremes, but not of that magnitude.
In fact, all has to do with the parasitic capacitances: at 1KHz, the wiring to the frequency pots creates a slight capacitive overcompensation, which is about right to compensate the phase shift of the TLO84.
But at the frequency extremes, it works in the wrong direction: at the maximum frequency, there is no compenstion at all when it would be most needed, and at the minimum, the compensation is maximal, ie too large, and in addition, the available loop gain is decreased by the potentimetric arrangement of the frequency controls.
Better performances at low frequency would be achievable by connecting the pots as variable resistors, but I opted for the potentiometric configuration to improve the stability and reduce the effect of wiper resistance.
With high quality wirewound types, the resistive method would be preferable.
It should be possible to properly compensate the filter for the whole frequency range by adding small capacitors (several pF) across R4 and R6, and between the wiper and cold terminals of the pots.
I'll test it one of these days.
One (important) thing I forgot to mention last time: the tuning capacitors must be of a low loss type to be able to achieve the advertised null depths.
Polystyrene is best, that's what I used, second best is polypropylene (foil).
Mica or good COG ceramics are a third option.
No other commonly available dielectric is suitable: even polycarbonate would degrade performances by an order of magnitude.
You can use this one: it is based on the one from ESP, and its performances approximately match those of the THD-meter
Attachments
- Status
- Not open for further replies.