Low-distortion Audio-range Oscillator

Jump to Latest
M

mikebeth

Member
Joined 2016
Hello all.
Just finished laying out my PCB for an active Twin-Tee notch filter and wondering if anyone else is interested in the project currently.
Would not be adverse to constructive comments about the schematic or layout. Also, willing to share boards (at cost) with any others in the group.
Thanks, Mike
 

Attachments

  • Twin-tee_bal.pdf
    53.7 KB · Views: 253
  • PCB front.pdf
    22.8 KB · Views: 170
  • PCB back.pdf
    41.7 KB · Views: 139
B

Bob Cordell

Member
Joined 2006
cordellaudio.com
This looks like a very nice project. I do have one question. You have 3 pots for trimming the resistor branches. What is your procedure for adjusting them for the deepest null? When I have built twin-T's I have just brute force matched all of the resistors, starting with resistors that are about 2% high and then shunting them with a much higher resistance as needed to hit the target. I did the same with the capacitors. There is one further trick to tuning. If you just allow for tweaking the shunt resistor leg +/- a couple of percent, the notch can be optimized even further, since under this condition the deviation is very small, but effective.

Cheers,
Bob
 
fubar3

fubar3

Member
Joined 2011
dennismiller55 said:
I am looking to add a low distortion audio sig gen to my bench. I have been watching eBay for good deals. Any worth bidding soon rise above my budget.

Anyway, I starting looking into a DIY Signal Generator. My research turned this up. Low-distortion Audio-range Oscillator

Anyone ever build this? Sounds good but is it? There is no build tips or PCB layout. Just a diagram and parts list.

This author has a book of all his circuit designs but comes at a steep price. However, my local library has a copy and I will pick it up sometime this week.
Click to expand...
I find the Sine oscillator very interesting. I am starting with the simple design here .. https://www.elektormagazine.com/index.php/files/attachment/326
which is based on the JKL 1869 lamp. I have already KiCad-ed a bare PCB for experiment.

The Wien bridge lamp causes a bit of distortion but other management ideas also have disadvantages.

The redcircuit design has ganged pots which are imperfect and has a lot of parts to complicate troubleshooting. Maybe you would be happy with just 1khz which seems to be the standard on REW threads.
 
M

mikebeth

Member
Joined 2016
IVX said:
mikebeth, the input buffer is missed, hence, different sources will affect the filter's Q, Freq. Also, I'd like to reduce the Notch impedance 100x times to keep it capable measure THD+N as well.
Click to expand...
An astute observation. I had considered using one of the op-amp sides as a unity input buffer, and I may change over to that instead of having two separate output amps. How would you reduce the notch impedance 100X? Change the R & C values respectively?
 
M

mikebeth

Member
Joined 2016
Bob Cordell said:
This looks like a very nice project. I do have one question. You have 3 pots for trimming the resistor branches. What is your procedure for adjusting them for the deepest null? When I have built twin-T's I have just brute force matched all of the resistors, starting with resistors that are about 2% high and then shunting them with a much higher resistance as needed to hit the target. I did the same with the capacitors. There is one further trick to tuning. If you just allow for tweaking the shunt resistor leg +/- a couple of percent, the notch can be optimized even further, since under this condition the deviation is very small, but effective.

Cheers,
Bob
Click to expand...
Hi Bob.
For the breadboard (with the C's handpicked) I used three 100ohm pots and found the null at 1kHz for Victor's oscillator (once it fully warmed up) to be fairly easy to achieve a 90dB null by "twiddling" the three single turn pots with the Q factor chosen by the R9/R10 ratio.

This project is the precursor to doing the version 4 of my 50Wx2 amplifier based heavily on your latest book. This version will be a change to a 3-stage output predominantly, along with a few other refinements.

Cheers, Mike
 
M

mikebeth

Member
Joined 2016
IVX said:
yep, just too high an impedance will produce a lot of noise. I use up to 200nF caps to see down to -135db of THD+N. However, I have never seen so low THD+N on my bench(Matlab simulation only) due to the flaw of my sine source(DAC+LPF), only -132db.
Click to expand...
So you would suggest bumping the C's up to 220 and 22nF and cutting the R legs by 1/10? Will look into that. Thanks
 
D

deciaml

Member
Joined 2022
vicnic said:
According my own measurements all the 1kHz boards, that was built within last year or little more, have the harmonics lower than -150dB at full output 2,8V RMS. Typical harmonic levels are around -155dB, but some boards may go to -160dB. Also I observed the artifact when the boards often may have significantly lower 3H, when the opamp output is under the load around the 1 kohm.
From the first time, all my measurements was done via the passive twin T notch. Now I am using 1kHz twin T built from the selected polystyrenes from RCL components and the metal film resistors of course. As I previously wrote, every capacitor must be tested before using for the ULD measurement purposes,, because the some examples may distort relatively high. Only statistics, logic and step by step implementation can help in this way.
Typical problem for my customers is to get proper ULD measurements, because they do not use or do not have passive twin T filters with the tested performance.
Little about the other things. This box perfectly fits for the one oscillator board:
1pcs White Electrical Instruments Aluminum Box 100*65*35mm DIY | eBay
Needs only to drill the face panel holes.
The regulator for 35VDC also can be easy found:
LM317 DC Linear Voltage Regulator Step down Power Supply Module 5V 6V 9V 12V 24V | eBay
Easiest and cheapest way is to use small 2W transformer with 15VAC output and rectifier-voltage doubler for to power the LM317 regulator. The adapter type supply is preferred, when the transformer is far from the oscillator board.
Click to expand...
Hello vicnic ,I 'm in China, how can i get your board? I dont know how to contact you > may i have your email?
 
IVX

IVX

Member
Joined 2003
e1dashz.wixsite.com/index
Aha, I use 100nF and 100nF//100nF for 1k and 10nF 10nF//10nF for the 10k. In that case, we 100.0% need a serious buffer to drive the notch perfectly.

To be honest, you also need the differential stage(again with the lowest impedance possible) before the buffer, and each of In+ and In- also needs buffers to get high input impedance. Without such a differential=balanced input I wouldn't go to measure <-130..-140db harmonics.
2022-06-11_16-50-25.jpg


And the simulated or rather, semi-simulated result of the THD+N -105.2db-30db=135.2db


2022-07-05_11-44-47.jpg
 
Last edited:
  • Like
Reactions: 1 user
B

Bob Cordell

Member
Joined 2006
cordellaudio.com
Samuel Groner said:
In the last months, I've worked on a bibliography covering a wide range of electrical oscillator: bibliography_electrical_oscillators.pdf

While mostly concerned with RF oscillators, the document also includes a list of references discussing low distortion RC oscillators.
Click to expand...
This is a nice piece of work! I also have always been fascinated with oscillators, starting with an EICO 377 kit when I was a teenager. A major moment in my journey was when back in the very early 70's I discovered HP's elegant oscillator with a JFET AGC element that had 50% drain signal feedback to the gate to greatly reduce distortion. Inspired by Bruce Hofer's convention paper (then at Tektronix before co-founding Audio Precision), I designed my first state variable oscillator for my THD analyzer back around 1980 or so.

Cheers,
Bob
 
M

mikebeth

Member
Joined 2016
IVX said:
Aha, I use 100nF and 100nF//100nF for 1k and 10nF 10nF//10nF for the 10k. In that case, we 100.0% need a serious buffer to drive the notch perfectly.

To be honest, you also need the differential stage(again with the lowest impedance possible) before the buffer, and each of In+ and In- also needs buffers to get high input impedance. Without such a differential=balanced input I wouldn't go to measure <-130..-140db harmonics.


And the simulated or rather, semi-simulated result of the THD+N -105.2db-30db=135.2db
Click to expand...
I will have to evaluate the difference switching to that more elaborate front end would have on performance for my system. It will be easy enough to add to the existing breadboard work.
 
diyralf

diyralf

Member
Joined 2016
IVX said:
Aha, I use 100nF and 100nF//100nF for 1k and 10nF 10nF//10nF for the 10k. In that case, we 100.0% need a serious buffer to drive the notch perfectly.

To be honest, you also need the differential stage(again with the lowest impedance possible) before the buffer, and each of In+ and In- also needs buffers to get high input impedance. Without such a differential=balanced input I wouldn't go to measure <-130..-140db harmonics.
View attachment 1088360

And the simulated or rather, semi-simulated result of the THD+N -105.2db-30db=135.2db
Click to expand...

The resistors are a bit too small for me. Don't you provoke a few harmonics?
 
B

Bob Cordell

Member
Joined 2006
cordellaudio.com
Bob Cordell said:
This is a nice piece of work! I also have always been fascinated with oscillators, starting with an EICO 377 kit when I was a teenager. A major moment in my journey was when back in the very early 70's I discovered HP's elegant oscillator with a JFET AGC element that had 50% drain signal feedback to the gate to greatly reduce distortion. Inspired by Bruce Hofer's convention paper (then at Tektronix before co-founding Audio Precision), I designed my first state variable oscillator for my THD analyzer back around 1980 or so.

Cheers,
Bob
Click to expand...
I should mention that the low-distortion state variable oscillator that I designed for my THD analyzer can be found on my website at cordellaudio.com. It is described in the THD analyzer project writeup on the site. It uses a JFET for the agc with a full-wave rectifier and an agc loop filter with time constants that are switched to be optimal for each band. The agc loop also includes an integrator. The JFET used is a 2N4091 and can be hard to get these days. It can be replaced by a 2N4391 available from Linear Systems. It has high pinchoff voltage and low on resistance, making it well-suited to this application. The original design used 5534 op amps, but today the LM4562 would be a good choice. Frequency is selected by switches for frequency band and for 1 out of 11 frequencies in each band. THD+N is well below 0.001% at 20 kHz. There is also a frequency range that goes up to 200 kHz.

Cheers,
Bob
 
You must log in or register to reply here.
Top Bottom