I am attemting to design a low distorsion THD analyzer. The ones I found online were either too complicated or too simplistic.
So, I would like to take on designing one. It wouldn't be very good quality at first, but would eventually get there.
For the signal source, I was tempted to use a simple signal generator (either external or use a XR2206 for instance).
The problem is to have a signal as pure as possible. I was thinking of 2 approaches:
1- Have a band-pass filter that would reject all the harmonics and only keep the fundamental from the source signal.
2- Have a circuit that would measure the THD from the source signal and set the zero-THD at that level (assuming that I could simply substract the THD from the source from the THD from the device under test).
And maybe a mix of the 2.
Are those approach bad ideas, from your knowledge? If yes, what would be a better approach.
The other goal of this project is also learning, that is why I won't buy one.
So, I would like to take on designing one. It wouldn't be very good quality at first, but would eventually get there.
For the signal source, I was tempted to use a simple signal generator (either external or use a XR2206 for instance).
The problem is to have a signal as pure as possible. I was thinking of 2 approaches:
1- Have a band-pass filter that would reject all the harmonics and only keep the fundamental from the source signal.
2- Have a circuit that would measure the THD from the source signal and set the zero-THD at that level (assuming that I could simply substract the THD from the source from the THD from the device under test).
And maybe a mix of the 2.
Are those approach bad ideas, from your knowledge? If yes, what would be a better approach.
The other goal of this project is also learning, that is why I won't buy one.
hi michouthefirst,
I like the Idea of both like so you can calibrate it as things do change age,temp,ect.
I have wanted to get a THD analyzer but as you said to simple or too deep and some steep in the pockets ones too,LOL.
Have you a circuit in mind or you starting it from scratch?
I will build one!
NS
I like the Idea of both like so you can calibrate it as things do change age,temp,ect.
I have wanted to get a THD analyzer but as you said to simple or too deep and some steep in the pockets ones too,LOL.
Have you a circuit in mind or you starting it from scratch?
I will build one!
NS
motherboard sound chipsets are typically good to better than -80 dB
<$200 soundcards get below -100dB without added notch filters
adding the notch filter, gain in a few op amps and R,C and you should be able to get -140 dB resolution if you fliter the soundcard's sine steeply enough, use today's better op amps, take care in passives choice...
other indirect 2-tone IMD techniques can extend soundcard resolution for measuring nonlinearities even further
<$200 soundcards get below -100dB without added notch filters
adding the notch filter, gain in a few op amps and R,C and you should be able to get -140 dB resolution if you fliter the soundcard's sine steeply enough, use today's better op amps, take care in passives choice...
other indirect 2-tone IMD techniques can extend soundcard resolution for measuring nonlinearities even further
@bear: I tried using Bob Cordell's design, but it is very hard to understand for a newcomer like me.
I need to go step by step in order to learn how a THD meter works.
@noSmoking: I don't have any design yet, just trying to figure out how I could achieve a passable design then move on to a better one until I get a solid design. I will try a few things on my own to see how it goes.
@df96: Without taking too much of your time, could you expand on this?
I need to go step by step in order to learn how a THD meter works.
@noSmoking: I don't have any design yet, just trying to figure out how I could achieve a passable design then move on to a better one until I get a solid design. I will try a few things on my own to see how it goes.
@df96: Without taking too much of your time, could you expand on this?
The original articles for the Cordell design cover the approach very well. IMO, you should re-read them. I'm not sure what good a poor THD analyzer is, since almost everything you might want to measure (in the solid state world) will be better than the analyzer. You can't subtract distortion numbers, so you need the basic performance to be pretty good. Watch out for the 2206. It was a nice chip in its day, but what you can get today is not at all the same and my experience was that they're supply sensitive and don't meet the original voltage control specs. The state variable oscillator in the Cordell design works well. If you make or get the boards, it isn't hard to build, except for the switching. I'd go with the PC solution or maybe pick up an old HP 33x to play with.
You might be better off starting with a simple very low THD sine wave oscillator as in the first page of linear technologies LT1007 datashet rather than using a XR2206 for a signal source,
http://cds.linear.com/docs/en/datasheet/100737fbs.pdf
I am sure any good quality opamp will work just fine.
The better the opamp the lower the THD will be too as there are some that are rated at better than .00005% THD.
Once you get a working model going then you can test the output of a XR2206 and is just how bad it really is!!
jer
http://cds.linear.com/docs/en/datasheet/100737fbs.pdf
I am sure any good quality opamp will work just fine.
The better the opamp the lower the THD will be too as there are some that are rated at better than .00005% THD.
Once you get a working model going then you can test the output of a XR2206 and is just how bad it really is!!
jer
Assume you measured 10.0 units of stuff as your base, with an accuracy of 1%. That means you have around 2 significant figures. Then you measured 12.0 on the device under test, also 1% accuracy. Subtract to get a measurement of 2.0. However, this only has about 6% accuracy. Why?michouthefirst said:
The original base was somewhere in the region 9.9 to 10.1. The second figure was 11.88 to 12.12. Subtract to get 1.78 to 2.22, which is 2.0 +- 6%.
As you look at the details involved you'll probably find good reasons why those performing well are "too complicated".
Besides the constraint mentioned by DF96 (the uncertainity quickly increases as the distortion of the DUT approaches/goes beyond that of the source/analyzer) there are several practical difficulties involved. First, for proper subtraction you need to measure not only the amplitude, but also the phase (relative to the fundamental) of the harmonics. Most standard FFT software tools don't support this. Second, both the source and the analyzer contribute to the distortion residual, and the distortion residual of each is dependent on the fundamental frequency, fundamental level and load/source impedance conditions. So unless the DUT has unity gain and the same input/output impedance as the analyzer/generator, a short-circuit test is not necessarily a good estimate of the effective distortion contribution of the measurement gear under actual operating conditions. Third, the distortion contribution of the source and analyzer may be a function of time (both short-term variations and long term drift) and environmental operating conditions (e.g. temperature and humidity). So a firm distortion residual is not easily establish.
As noted before, designing both source and analyzer such that their independent distortion contribution is at least 10 dB (preferably 20 dB) below that of the DUT will usually be the more feasible path.
Samuel
BTW, you might want to check this post for an easily implemented low-pass filter to clean up a source.
Samuel
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.