A very interesting article has appeared in EDN, about an ultra-low distortion oscillator:
https://www.edn.com/ultra-low-distortion-oscillator-part-2-the-real-deal/
Very nice and well researched (member @SandyTodorov1 proposed an oscillator using some of the principles described in the article), but one point caught my attention: the use of a LDR optoisolator for the amplitude control.
Quoting the author:
That was certainly not my experience. When used sparingly, in oscillators and THD meters, they work well enough, but their value needs to be diluted with parallel and series resistors, and they can easily degrade over time, ruining the performance of the instrument.
To be sure, I made a quick and dirty experiment:
I used a bridge configuration, because I thought it was required to reach a sufficient resolution. All the resistors are high-quality MF made by Ohmic, and U1 is the differential input of my ST1700 distortionmeter.
The voltage across the LDR is 1.4Vrms and I adjusted the current through the LED to be close to the equilibrium, leaving a small residue to allow the THD-meter to synchronise.
I tested various couplers, one from Clairex and another one bought from internet. I also brewed my own version, based on a professional metal/glass LDR mounted in front of a green LED.
The THD at the junction of R1 and R4 (suitably corrected) was comprised between 0.008% for my version to 0.1~0.15% for the others. Not exactly brilliant, and in fact I could have dispensed with the bridge and measured the THD directly
https://www.edn.com/ultra-low-distortion-oscillator-part-2-the-real-deal/
Very nice and well researched (member @SandyTodorov1 proposed an oscillator using some of the principles described in the article), but one point caught my attention: the use of a LDR optoisolator for the amplitude control.
Quoting the author:
Of course?
That was certainly not my experience. When used sparingly, in oscillators and THD meters, they work well enough, but their value needs to be diluted with parallel and series resistors, and they can easily degrade over time, ruining the performance of the instrument.
To be sure, I made a quick and dirty experiment:
I used a bridge configuration, because I thought it was required to reach a sufficient resolution. All the resistors are high-quality MF made by Ohmic, and U1 is the differential input of my ST1700 distortionmeter.
The voltage across the LDR is 1.4Vrms and I adjusted the current through the LED to be close to the equilibrium, leaving a small residue to allow the THD-meter to synchronise.
I tested various couplers, one from Clairex and another one bought from internet. I also brewed my own version, based on a professional metal/glass LDR mounted in front of a green LED.
The THD at the junction of R1 and R4 (suitably corrected) was comprised between 0.008% for my version to 0.1~0.15% for the others. Not exactly brilliant, and in fact I could have dispensed with the bridge and measured the THD directly
Thanks for the link in post 1 - and the Sandy oscillator is this :
https://www.diyaudio.com/community/threads/low-distortion-oscillator-gs-5534-8-ne5534.312692/ ?
https://www.diyaudio.com/community/threads/low-distortion-oscillator-gs-5534-8-ne5534.312692/ ?
LDR's (light dependent resistors) are useful as AC voltage dividers because they are very slow, so you don't need to integrate the rectified AC to use them on AC signals. They have a fairly fast response to light but recover (darkness) very slowly (seconds) so as the shunt part of a voltage divider, their response makes a good analog compressor. (A very long time ago) I combined them with a LED and a couple resistors around an amplifier to limit the output of the monitor amp so that the DJ couldn't blow up the monitor speakers. With just ~4 components on a ball hanging on the back of the amp, there was no effect except limiting the volume (no noticeable distortion).
But there is a design that uses one, and is examplary. http://www.janascard.cz/PDF/An ultra low distortion oscillator with THD below -140 dB.pdf - the trick is to make it only have a tiny effect on the loop gain, but enough to allow a slow feedback loop to hold the main oscillator loop gain at 1.000000 despite component drift and temperature changes.
So its distortion is diluted by a huge factor, and its a slow responding kind to further reduce non-linearity at 1kHz...
My first attempts were with LDR but the result turned out to be deplorable.
JFETs behave better, and reducing the voltage on them (10 times) and their differential connection solves the problems radically.
There is already an improved circuit with newer NE5534 op amps, but it is not yet released.
JFETs behave better, and reducing the voltage on them (10 times) and their differential connection solves the problems radically.
There is already an improved circuit with newer NE5534 op amps, but it is not yet released.
Dilution is the way to go if you want to reduce distortion whatever the control method, but high dilution means more stringent control methods, which has issues of its own, especially if you want to make the frequency variable.
I think I still have some old, sintered LDR's somewhere. They were produced by the Philips group (Valvo, Mullard, MBLE, RTC, etc) and since they are based on bulk conduction rather then 2D like omnipresent modern types, the current density should be much reduced, resulting (hopefully) in a lower non-linearity
I think I still have some old, sintered LDR's somewhere. They were produced by the Philips group (Valvo, Mullard, MBLE, RTC, etc) and since they are based on bulk conduction rather then 2D like omnipresent modern types, the current density should be much reduced, resulting (hopefully) in a lower non-linearity
You are right, but only when it is done without thinking.
When the JFET is used as a variable resistor when reducing the AC, the voltage across it decreases and the THD.
But by amplifying (x10) this small voltage THD does not increase.
And there is no longer a need for 10x the good uniformity of the RC circuits because we have 10x the voltage and its change on the JFET
When the JFET is used as a variable resistor when reducing the AC, the voltage across it decreases and the THD.
But by amplifying (x10) this small voltage THD does not increase.
And there is no longer a need for 10x the good uniformity of the RC circuits because we have 10x the voltage and its change on the JFET
If you use JFETs, add half the AC voltage to the gate control DC to make the distortion symmetric and double the linear range. This is usually just returning the DC filter cap to a low-Z divider instead of ground. I made compressor circuits this way that were faster than the audio, ie activated on the first half cycle. A friend used CMOS chips for the purpose with good results. It must have been "unbuffered" CMOS. (Many CMOS logic families have buffers and are not useful for analog purposes.)
https://www.ti.com/lit/an/scha004/scha004.pdf
https://www.ti.com/lit/an/scha004/scha004.pdf
Last edited:
Can you elaborate on this in the context of oscillator level control,
preferably with an example ? Thanks.
preferably with an example ? Thanks.
I used 74HC05 open drain inverters in a microphone preamplifier with switchable gain and a built-in limiter in 1997...2000, as part of a field memory recorder. You could switch the gain by switching on or off 74HC05 inverters and the limiter simply reduced the 74HC05's supply voltage.
Post 11 : you wrote about " .. use JFETs, add half the AC voltage to the gate control DC
to make the distortion symmetric and double the linear range .. " in the context of low
distortion sine wave oscillators, so perhaps this can be substantiated with a formula,
theory or practical example. I think we will not discuss crystal oscillators here.
to make the distortion symmetric and double the linear range .. " in the context of low
distortion sine wave oscillators, so perhaps this can be substantiated with a formula,
theory or practical example. I think we will not discuss crystal oscillators here.
From https://en.wikipedia.org/wiki/Jfet#Linear_ohmic_region
Adding Vds/2 to Vgs makes the coefficient constant.
I too used a JFET to regulate my sine-wave oscillator.
Ed
Adding Vds/2 to Vgs makes the coefficient constant.
I too used a JFET to regulate my sine-wave oscillator.
Ed
I have tested a Philips, sintered construction LDR.
The THD was atrocious, not even measurable with my sensitive bridge setup, but it was something between 5 and 10%.
Thus, counter-intuitively, the cheap, thin-film versions are much better regarding linearity
The THD was atrocious, not even measurable with my sensitive bridge setup, but it was something between 5 and 10%.
Thus, counter-intuitively, the cheap, thin-film versions are much better regarding linearity