Actually you may not need nor want the deepest null possible. You need enough signal level for the analyzer to capture. You just need enough to make the analyzer operate in its more linear range. So a 40dB notch is fine... maybe to -60dB.
If just using the notch to do all the work, then the residual will be harmonics only if the notch is much deeper than the harmonic residue. In that case, as deep notch as possible or better than the distortion generated by signal under test.
However, a deep passive notch will attenuate the 2H and 3H to significant degree and needs to be known to correct the data. If possible, an active notch will have higher Q and not atten the harmonics you are trying to measure. Some added harmonics from opamp could add to the measured data but that wont be an issue unless you are going for the absolute lowest harmonic level measurement.
THx-RNMarsh
If just using the notch to do all the work, then the residual will be harmonics only if the notch is much deeper than the harmonic residue. In that case, as deep notch as possible or better than the distortion generated by signal under test.
However, a deep passive notch will attenuate the 2H and 3H to significant degree and needs to be known to correct the data. If possible, an active notch will have higher Q and not atten the harmonics you are trying to measure. Some added harmonics from opamp could add to the measured data but that wont be an issue unless you are going for the absolute lowest harmonic level measurement.
THx-RNMarsh
Last edited:
Here are two hacks I have developed to be able to tune the twin T:
The first is approximate, provides a limited notch depth and has a (very) limited adjustment range, but only requires one potentiometer.
In this example, a 100Ω pot is used, and the sim shows the rejection for various settings.
It is around -80dB on average. Note that better results could be obtained by tweaking the correction resistors R6 R7 more accurately, but it is time-consuming.
R2 fine-tunes the depth of the notch at all frequencies.
This topology is well-suited to hum rejectors, but is a little short for distortion analyzis.
The second circuit offers infinite rejection at all frequencies, provide a wide tuning range but requires a dual potentiometer.
The nice thing is that the two gang values are identical and need not be matched.
This means that a standard stereo potentiometer can be used.
The second pic shows the basic circuit: in this case, the two potentiometer sections are identical, at 220Ω.
The rejection stays identical at all frequencies.
In order to allow for mismatch between the tracks, the highest resistance one has to be positioned on the right (here 240Ω), and a parallel resistor is added to match the low resistance one (pic 3). In practice, this additional resistor is made up of a fixed resistor and a trimmer, and it will also compensate for all other residual mismatches in the capacitors and resistors.
Note that to benefit from the performances of this circuit, the components need to be highly stable: mica or ceramic COG for the caps, and high stability metal film or foil resistors.
Even PS or PP caps are unsuitable, because they have a low, but noticeable negative tempco.
A practical stereo pot will also degrade the performances, because the percentage of resistance vs. rotation won't be exactly identical between the two sections.
The first is approximate, provides a limited notch depth and has a (very) limited adjustment range, but only requires one potentiometer.
In this example, a 100Ω pot is used, and the sim shows the rejection for various settings.
It is around -80dB on average. Note that better results could be obtained by tweaking the correction resistors R6 R7 more accurately, but it is time-consuming.
R2 fine-tunes the depth of the notch at all frequencies.
This topology is well-suited to hum rejectors, but is a little short for distortion analyzis.
The second circuit offers infinite rejection at all frequencies, provide a wide tuning range but requires a dual potentiometer.
The nice thing is that the two gang values are identical and need not be matched.
This means that a standard stereo potentiometer can be used.
The second pic shows the basic circuit: in this case, the two potentiometer sections are identical, at 220Ω.
The rejection stays identical at all frequencies.
In order to allow for mismatch between the tracks, the highest resistance one has to be positioned on the right (here 240Ω), and a parallel resistor is added to match the low resistance one (pic 3). In practice, this additional resistor is made up of a fixed resistor and a trimmer, and it will also compensate for all other residual mismatches in the capacitors and resistors.
Note that to benefit from the performances of this circuit, the components need to be highly stable: mica or ceramic COG for the caps, and high stability metal film or foil resistors.
Even PS or PP caps are unsuitable, because they have a low, but noticeable negative tempco.
A practical stereo pot will also degrade the performances, because the percentage of resistance vs. rotation won't be exactly identical between the two sections.
Attachments
I want to see the distortions of the cheap oscillator.
At the moment the best notch is ~ -62dB and the remaining signal @ ~5.6mVac is still a clean looking sinewave.
The Thread/post that suggested this told us that the distortion is much better than specified and that it can be improved a lot by changing the opamps.
I want to know where it is as supplied before I attempt any changes.
Last edited:
And from a different angle... if you need a super low distortion sine wave of any given frequency to test your filter then you can create one in seconds with Audacity. Or a collection. 1000Hz, 1001 Hz and 1002Hz etc. Just burn it to a CDR and you have a better and lower distortion standard then many oscillators.
Andrew... why not record your oscillator output onto the computer and then look at the distortion spectrum. Audacity again.
I tried using a sound card and the test file posted by a Member. What ever the test file did, it stopped the sound card outputting any signal, not even as a music card.
I tried to uninstall the sound card software and re-install it. still not working.
I even went as far as removing the sound card and uninstalling. Buying second sound card and installing the software. But still no music.
What ever that downloaded file did, I can't undo it.
I'm not buying any more sound cards. They cost too much to just lie on the shelf.
I tried to uninstall the sound card software and re-install it. still not working.
I even went as far as removing the sound card and uninstalling. Buying second sound card and installing the software. But still no music.
What ever that downloaded file did, I can't undo it.
I'm not buying any more sound cards. They cost too much to just lie on the shelf.
Sounds odd. No idea what could have happened there tbh.
That's one for another thread but I would have thought it should be fixable. There should be some clues in the sound card settings and in 'device manager'.
That's one for another thread but I would have thought it should be fixable. There should be some clues in the sound card settings and in 'device manager'.
There are better alternatives to the Twin Tee Notch
Twin Tee Notch is painful to tune. You can get an ideal null using the Bainter Filter, and it can be easily tuned.
www.changpuak.ch/electronics/downloads/Active-Notch-Filter-(Bainter).pdf
More details and practical implementations are shown in Doug Self's book on crossover design...see page 258, Figure 9.6.
Twin Tee Notch is painful to tune. You can get an ideal null using the Bainter Filter, and it can be easily tuned.
www.changpuak.ch/electronics/downloads/Active-Notch-Filter-(Bainter).pdf
More details and practical implementations are shown in Doug Self's book on crossover design...see page 258, Figure 9.6.
It is hard to realise a deep notch (better than 60dB) with passive parts. Here is a variable notch of about -60dB with a 600 Ohm loading the passive filter.
I added an opamp to it for higher Q. I can switch between passive and active. If gain in opamp, then notch can be deeper. In my use with an analyzer it was fine as shown without atten 2H and 3H. (disregard offset for each plot).
The variable freq notch is very helpful.
THx-RNMarsh
I added an opamp to it for higher Q. I can switch between passive and active. If gain in opamp, then notch can be deeper. In my use with an analyzer it was fine as shown without atten 2H and 3H. (disregard offset for each plot).
The variable freq notch is very helpful.
THx-RNMarsh
Last edited:
I printed out a spreadsheet version of this about 15years ago.
Far quicker to look up a table than to input data into some software one needs to find, load and wait till it starts.
I got sent to fig 9.6 to see the Bainter Filter.
The previous page shows Bridged-differentiator fig 9.5d and it's the same as the Filter you posted. But D.Self does not show the relationship between the resistances in the various arms. He shows the values for an 80 to 180Hz notch. But he has swapped R1 and R2. The higher value is on the input side rather than on the opamp side.
comparing the two sch it looks like the centre frequency is given (approximately) by:
Fnotch ~ 4 / {Pi*Rupper*C}
Last edited:
Similar, the one I found gave all combinations within a specified precision nice if there are holes in your parts drawer. We only have complete runs with SMT now, the stock room no longer has refills for leaded resistors. This also helps if the answer involves resistors decades apart since the possible error is still virtually +-1%. If you say for instance .02% there can often be dozens of choices
Last edited:
Success !
Thank you all for all the various ideas.
Got lots of progress yesterday when I ran out of adjustment range on the three rheostats but achieved -64dB with the passive notch.
Today I added another parallel resistor to each of the three resistor legs to lower the values and now the range covers the minimum with spare.
I ended up using the dual log vol pot as the adjuster on the two 6800r arms. The log law gives a nice even range of adjustment. (appologies to Simon for ignoring his parasitic capacitance advice).
The resistor equality between the two resistance tracks is pretty bad, especially at the lower end. But when used in parallel to the "added" resistor, the error in resistance between the two arms never exceeds 1r1 for a total that ranges from ~7220 to ~7260.
The 3400r arm used a 10T 50k for the paralleled adjuster. This is far too fiddly, it needs too much of a turn to see any effect. But it's the only 50k I had.
Now that it is in tune I have -78dB to -79dB notch and the scope hardly shows any signal at all. I need quite a bit of gain to see the residual.
For the few minutes of operation I am not aware of any drift.
It was easy to tune in. I set all three rheostats to maximum. Then turn down the dual and immediately the fundamental falls. Kept going till I reach a minimum. Then swapped to lowering the third arm resistance. The Null got lower to leave a tiny signal. Back to the dual and I had to turn it up very slightly to acheive a nosiy line that the scope can't resolve properly. The scope is set to 2mV/div.
The input reads 7.85Vac the outputs reads 0.8mVac to 0.9mVac, but based on what the scope is trying to show I think much of what the DMM is reading is noise and interference and distortion. It's all open lying on the bench.
My next step is to convert to an active Twin T notch filter to help preserve the 2nd and 3rd harmonics.
The active uses two opamps both wired as unity gain followers.
Can the first one be set up with a bit of gain? maybe 10times (+20dB) to give more signal for the scope probe to read.
Or must a third opamp be added to multiply the output from the notch?
Where does one tap in to take out the signal in the active twin T? At the passive notch or at the output of the first opamp? Found the answer, from the output of the first opamp.
Thank you all for all the various ideas.
Got lots of progress yesterday when I ran out of adjustment range on the three rheostats but achieved -64dB with the passive notch.
Today I added another parallel resistor to each of the three resistor legs to lower the values and now the range covers the minimum with spare.
I ended up using the dual log vol pot as the adjuster on the two 6800r arms. The log law gives a nice even range of adjustment. (appologies to Simon for ignoring his parasitic capacitance advice).
The resistor equality between the two resistance tracks is pretty bad, especially at the lower end. But when used in parallel to the "added" resistor, the error in resistance between the two arms never exceeds 1r1 for a total that ranges from ~7220 to ~7260.
The 3400r arm used a 10T 50k for the paralleled adjuster. This is far too fiddly, it needs too much of a turn to see any effect. But it's the only 50k I had.
Now that it is in tune I have -78dB to -79dB notch and the scope hardly shows any signal at all. I need quite a bit of gain to see the residual.
For the few minutes of operation I am not aware of any drift.
It was easy to tune in. I set all three rheostats to maximum. Then turn down the dual and immediately the fundamental falls. Kept going till I reach a minimum. Then swapped to lowering the third arm resistance. The Null got lower to leave a tiny signal. Back to the dual and I had to turn it up very slightly to acheive a nosiy line that the scope can't resolve properly. The scope is set to 2mV/div.
The input reads 7.85Vac the outputs reads 0.8mVac to 0.9mVac, but based on what the scope is trying to show I think much of what the DMM is reading is noise and interference and distortion. It's all open lying on the bench.
My next step is to convert to an active Twin T notch filter to help preserve the 2nd and 3rd harmonics.
The active uses two opamps both wired as unity gain followers.
Can the first one be set up with a bit of gain? maybe 10times (+20dB) to give more signal for the scope probe to read.
Or must a third opamp be added to multiply the output from the notch?
Where does one tap in to take out the signal in the active twin T? At the passive notch or at the output of the first opamp? Found the answer, from the output of the first opamp.
Last edited:
It's been running for over an hour.
It did drift.
The notch had dropped to about -70dB.
I retuned it and measured the warmed up frequency: 1001Hz. But I did not measure the frequency at the beginning.
It did drift.
The notch had dropped to about -70dB.
I retuned it and measured the warmed up frequency: 1001Hz. But I did not measure the frequency at the beginning.
The active twin-T just needs one opamp, which must be wired with slightly less gain than unity. You could use other opamps as buffers at input and output, I suppose. Or you could use two buffers with some attenuation between them. You may need some gain adjustment, as an imperfect twin-T (i.e. any real twin-T) can deliver almost random phase near the null so you might get an oscillator instead of a notch. Anyway, the aim is to drive the 'ground' connection of the twin-T with a very slightly attenuated version of its output - essentially a bootstrap. Apologies if you already knew all this.AndrewT said:
I'm curious: what resistance did you end up with? How close was it to my estimate?
The added resistor now measures 462r and 460r
A bit below what my first soldered version could get down to.
This is the only active style Twin T I have seen.
It's what David Moore used as well.
Another question.
The common node for the notch is in the middle.
It that where the GND symbols should connect?
That makes the second opamp with -IN pin, and Out pin and the bottom of the divider ladder all going to the same common node, where the input and output returns also meet up.
A bit below what my first soldered version could get down to.
This is the only active style Twin T I have seen.
It's what David Moore used as well.
Another question.
The common node for the notch is in the middle.
It that where the GND symbols should connect?
That makes the second opamp with -IN pin, and Out pin and the bottom of the divider ladder all going to the same common node, where the input and output returns also meet up.
Attachments
Last edited:
- Status
- Not open for further replies.