I have an aversion to CMOS logic tone generators. Goodness knows, I've built and experimented with enough of them. It's way off-topic but I suggest something that comes straight out of an audio tinkers junkbox, uses common audio BJTs and builds on a breadboard in less than hour - no fancy envelope stabilizing parts or anything to buy, for me at least. Spot-frequency Sine wave Generator - RED - Page92
My only regret is that a 5V USB power source isn't enough. Still, 0.11% THD @ 1kHz ain't bad for a simple circuit and will beat most logic derived generators.
I built a single 400Hz tone version and can scarcely hear any difference in comparison to 16 or 24bit software sources, which, if you're serious about high quality tone generation and application, is the way to go, needs no parts other than a suitable quality DAC and is often free! This site is simple, but not completely free. Go to the generator page and try a tone or two, you might find that your desktop system isn't so wonderful: Dynamic Test Tones
__________________
My only regret is that a 5V USB power source isn't enough. Still, 0.11% THD @ 1kHz ain't bad for a simple circuit and will beat most logic derived generators.
I built a single 400Hz tone version and can scarcely hear any difference in comparison to 16 or 24bit software sources, which, if you're serious about high quality tone generation and application, is the way to go, needs no parts other than a suitable quality DAC and is often free! This site is simple, but not completely free. Go to the generator page and try a tone or two, you might find that your desktop system isn't so wonderful: Dynamic Test Tones
__________________
Attachments
Ian, What seems to be lacking in these threads is simple ways of testing the end results. That circuit looks too simple to be stable. Usually a Wien Bridge needs a non linear resistance to work well in the feedback loop. Wiens can have a fixed arm R2C1, Usually ones made like this have different outputs at different frequencies,
Perhaps the best we can do is a CD made at -3dB of 0VU. It can be played on a CD player rather than the computer that burnt it to have simplest ground.
Using a hex inverter is for fun. The 4060 is 100% reliable in how it works so a good building block. To my eyes on an analogue scope it beats the majority of commercial signal generator. It's whole reason to be is noise free 50/50 square-waves most often for timing uses. Using a PP3 battery keeps noise to a minimum, Zinc batteries whilst not ideal noise wise are unlikely to exceed the CMOS noise.
Hex inverters are used to convert logic 1 to 0 or 0 to 1. I doubt if many were used exclusively for that. They have gain so must be intended for oscillators. A gain of 1 would serve the prime function. Hex inverters have a simi Schmidt action,if they didn't they would be unusable.
The real Schmidt inverter like CD40106B often work equally well as the non Schmidt if wanting square-waves, in theory they are better when that.
http://cpc.farnell.com/texas-instruments/cd40106be/ic-4000-cmos-40106-dip14-18v/dp/SC12812
Looking at square to sine filters Multiple Feedback Filters ( MFB ) and Sallen Key types the latter is more pleasing to the eye when doing layouts. The MFB might be better as it inverts at every stage. Usually by distortion curve cancellation MBF could give better results. This is more true if the prime frequency gain is 1. MFB also can use slower op amps. Sometimes even the 10 MHz NE5532 is too slow.
One ideal I might use is op amp one as buffer and virtual ground. Seeing as 2Vrms would be plenty of output it's worth a thought. More likely I will use 2 x PP3 9V batteries as they will simplify the idea.
Perhaps the best we can do is a CD made at -3dB of 0VU. It can be played on a CD player rather than the computer that burnt it to have simplest ground.
Using a hex inverter is for fun. The 4060 is 100% reliable in how it works so a good building block. To my eyes on an analogue scope it beats the majority of commercial signal generator. It's whole reason to be is noise free 50/50 square-waves most often for timing uses. Using a PP3 battery keeps noise to a minimum, Zinc batteries whilst not ideal noise wise are unlikely to exceed the CMOS noise.
Hex inverters are used to convert logic 1 to 0 or 0 to 1. I doubt if many were used exclusively for that. They have gain so must be intended for oscillators. A gain of 1 would serve the prime function. Hex inverters have a simi Schmidt action,if they didn't they would be unusable.
The real Schmidt inverter like CD40106B often work equally well as the non Schmidt if wanting square-waves, in theory they are better when that.
http://cpc.farnell.com/texas-instruments/cd40106be/ic-4000-cmos-40106-dip14-18v/dp/SC12812
Looking at square to sine filters Multiple Feedback Filters ( MFB ) and Sallen Key types the latter is more pleasing to the eye when doing layouts. The MFB might be better as it inverts at every stage. Usually by distortion curve cancellation MBF could give better results. This is more true if the prime frequency gain is 1. MFB also can use slower op amps. Sometimes even the 10 MHz NE5532 is too slow.
One ideal I might use is op amp one as buffer and virtual ground. Seeing as 2Vrms would be plenty of output it's worth a thought. More likely I will use 2 x PP3 9V batteries as they will simplify the idea.
With a "twin" (cascaded/series) MFB you could cover the complete square-to-sine "conversion". As it is inverting, the biasing at the opamp "+" input could become trivial layout-wise, esp. if you use a quad like TL074. A MFB has another advantage; you can use the MFB input resistive divider to "pull" the preceding opamp into Class-A. I mean, you do want to tickle a purebred Class-A amp, so doing this could be one step in the right directon, haha.
I was told that the only way to get a true 50/50 square is to use a divider. I mean how can you be 100% sure that a Schmidt triggers at 100% equal offsets both swings?
I was told that the only way to get a true 50/50 square is to use a divider. I mean how can you be 100% sure that a Schmidt triggers at 100% equal offsets both swings?
As said Nigel, I built and used the Red circuit myself in a comparison with an on-line tone source of 16 and and 24 bit tones. I have a reasonable 24 bit DAC coupled to the PC and then to a desktop audio system based on Tannoy reveal monitors. I test distortion with a Quantasylum QA400 sound card and software which you'll find commented on and supported by Richard Marsh and other members here.
The generator is a spot frequency type, so frequency and levels are preset for any or each discrete tone. That's nowhere near the stability problem presented by general purpose analog instruments with full coverage from 20Hz -20,000kHz. On my tiny matrix board assembly, I measured 0.17% THD at 400Hz and 0.12% at 1kHz 0dBu - very close to the spec.
The point of it is that it's analog, simple, quick and does a good job of providing a clean tone where required. If a full range of tones or continuous coverage and sweeps are needed, the hardware issues like quality pots don't make this easy or cheap for either approach.
On the CMOS digital generation trail, you might look into the Natsemi switched capacitor filters like 4th order configurable MF100. They're probably rare now but such devices can track a digital waveform perfectly and with a little more filtering of the output, can shape a raw 8bit ladder generated sinewave down to the 0.1% distortion area.
The generator is a spot frequency type, so frequency and levels are preset for any or each discrete tone. That's nowhere near the stability problem presented by general purpose analog instruments with full coverage from 20Hz -20,000kHz. On my tiny matrix board assembly, I measured 0.17% THD at 400Hz and 0.12% at 1kHz 0dBu - very close to the spec.
The point of it is that it's analog, simple, quick and does a good job of providing a clean tone where required. If a full range of tones or continuous coverage and sweeps are needed, the hardware issues like quality pots don't make this easy or cheap for either approach.
On the CMOS digital generation trail, you might look into the Natsemi switched capacitor filters like 4th order configurable MF100. They're probably rare now but such devices can track a digital waveform perfectly and with a little more filtering of the output, can shape a raw 8bit ladder generated sinewave down to the 0.1% distortion area.
I hit an interesting problem with this idea. It will be virtually impossible to be 20 dB lower in distortion than the JLH. This is the standard usually required. 0.08% = 20Log 0.0008 = -62dB. If we take 0.03% suggested for my BC327/337-40 it's -70 dB via simulation. That needs a sine-wave -90dB pure or 0.003%. The best I will get I calculate is -75dB. What this allows is to say working well enough. I have real doubts any 1960's test gear was better than that.
Here is a design that looks very interesting. The lamp takes distortion down to - 96dB or better. Compared with a Bubba oscillator -40 dB is what it gives. And they often don't work. Same op amp count. My State Variable Filter Oscillator using diode limiter was about -65dB. The lamp pushes it down 30 dB as it always is a sine-wave regardless of where in the circuit. 2 x 1N4148 diodes gives a squared wave ( between sine and square ).
https://www.elektormagazine.com/files/attachment/326
https://www.hscott.net/bubba.pdf
Here is a design that looks very interesting. The lamp takes distortion down to - 96dB or better. Compared with a Bubba oscillator -40 dB is what it gives. And they often don't work. Same op amp count. My State Variable Filter Oscillator using diode limiter was about -65dB. The lamp pushes it down 30 dB as it always is a sine-wave regardless of where in the circuit. 2 x 1N4148 diodes gives a squared wave ( between sine and square ).
https://www.elektormagazine.com/files/attachment/326
https://www.hscott.net/bubba.pdf
This adaptation of Rosens well-known design will get you down to 0.003% and better if you want a wide range DIY oscillator. Note the opto-coupled stabilization here. Very nice compared to my simpler build of it that uses either the unobtanium RA53, diode/ resistor network or lamp stabilization Finding and paying for a decent dual gang pot. is often the limitation though.
Low-distortion Audio-range Oscillator - RED - Page82
Low-distortion Audio-range Oscillator - RED - Page82
Attachments
Last edited:
Here's the original WW article of Rosen's Phase-shift oscillator from 1982, not that long after JLH had his version of the Wien Bridge oscillator design published there. It promises THD of typically 0.0005% but I think that would have required best quality components and laboratory grade test equipment back then. I had to use wirewound pots and the help of a local test lab. to get near that figure. And the cost? .... well, I'd rather forget about it.
http://www.keith-snook.info/wireles.../Phase-shifting oscillator - Roger Rosens.pdf
http://www.keith-snook.info/wireles.../Phase-shifting oscillator - Roger Rosens.pdf
Picoscope seem to be offering something good enough to ask the questions. I have been told a good soundcard with Freeware gets to the same point. Any thoughts?
I think I can make an oscillator for less than £10 to do low distortion sine-waves. Could the whole thing cost less than £50 and be as close as we need to an Audio Precission test set. The Picoscope isn't cheap. That is to measure -80dB ( 0.01% distortion ) up to 25 kHz with some chance the results are truely the amplifer and not quirks of the additon of the test device and the test gear. An example of that was wow and flutter tests on turntables. Depending on where the test disk was placed the errors added or subtracted as the disk was made on a lathe with it's own errors. 0.04% and 0.08% could be the result, the true reading 0.06%.
Do read the analysis of the Kenwood amp. Looks to me the JLH is better if saying at all levels, no surprise really. The Kenwood is really nice all the same and not unlike the JLH in style.
Using an FFT spectrum analyzer to test audio amplifiers
16 bit high resolution oscilloscope | Overview
This last is worth a read. Not least it sets out to measure capacitor distortion. The Elektor SVF is very simple which is the way I want to go . It is very close to being perfect. I will have to ask friends to see who can measure it for me. Then I will have some benchmarks. This is a tricky project for me as I am packed up ready to move house.
https://linearaudio.nl/sites/linearaudio.net/files/Bateman EW 07 2002 test oscillator.pdf
I think I can make an oscillator for less than £10 to do low distortion sine-waves. Could the whole thing cost less than £50 and be as close as we need to an Audio Precission test set. The Picoscope isn't cheap. That is to measure -80dB ( 0.01% distortion ) up to 25 kHz with some chance the results are truely the amplifer and not quirks of the additon of the test device and the test gear. An example of that was wow and flutter tests on turntables. Depending on where the test disk was placed the errors added or subtracted as the disk was made on a lathe with it's own errors. 0.04% and 0.08% could be the result, the true reading 0.06%.
Do read the analysis of the Kenwood amp. Looks to me the JLH is better if saying at all levels, no surprise really. The Kenwood is really nice all the same and not unlike the JLH in style.
Using an FFT spectrum analyzer to test audio amplifiers
16 bit high resolution oscilloscope | Overview
This last is worth a read. Not least it sets out to measure capacitor distortion. The Elektor SVF is very simple which is the way I want to go . It is very close to being perfect. I will have to ask friends to see who can measure it for me. Then I will have some benchmarks. This is a tricky project for me as I am packed up ready to move house.
https://linearaudio.nl/sites/linearaudio.net/files/Bateman EW 07 2002 test oscillator.pdf
Indeed, and you don't need difficult hardware like a low distortion oscillator. I have posted that you can download high quality continuous tones, sweeps, chirps or anything that comes into our heads if we only search it with the browser. I keep tone generator and analyser programs on 2 PCS so that the test signal can be monitored and altered whilst the analysis is running.
Mooly's suggested program - Download | Audacity(R)
(it's massive but our needs are simple and not too difficult to get started)
Mooly's suggested program - Download | Audacity(R)
(it's massive but our needs are simple and not too difficult to get started)
I just did a little layout for the oscillator using 2 x NE5532. I will risk the stability as the higher current is good to drive a lamp. The Elektor design claims claims the lamp isn't for gain. Instead it gives phase changing results. This might say how a TL074 give -108dB at 1 kHz. I only have a 28V 40 mA lamp. Hope it will work before getting his recomendation. It's a very simple circuit.
That's very much the question. He says about NE5532 in passing with 9dB lower distortion. The real example is TL074 which suits fine. TL074 is the most stable op amp I ever tried. I have strapped all of it's input and outputs together ( all in's to in's all out's to out's ) and made it work! That's because it has a resistive output and JFET input. Not many op amps get close to allowing that.
JLH pnp
Hello everybody,
I have some AD149 germanium pnp transistors, and would like to try and use them in the JLH circuit.
The most obvious solution would be, I guess, to mirror the whole circuit and change all npn transistors to pnp and vice versa.
But is it also possible to just change the output devices to pnp and keep the npn driver? I tried to drive a couple of pnp with an npn in ltspice, and got it working, but only with unity voltage gain. Is this inevitable, or is there a way to make it work?
Hello everybody,
I have some AD149 germanium pnp transistors, and would like to try and use them in the JLH circuit.
The most obvious solution would be, I guess, to mirror the whole circuit and change all npn transistors to pnp and vice versa.
But is it also possible to just change the output devices to pnp and keep the npn driver? I tried to drive a couple of pnp with an npn in ltspice, and got it working, but only with unity voltage gain. Is this inevitable, or is there a way to make it work?
try it and see.
At last I found an hour to build the SVF Oscillator from Elektor. It seems to do exactly what it should. It exceeds my ability to test it. If I could I would tweak the 27K to get the lowest distortion. It worked first time and with the guesses I tried. Output ranges from 0.8V to 3.65V with better distortion at about 2.27V. It is true the distortion improves as it heats up. This allowed me to quickly find the ideal voltage whilst the small residual vanished. The capacitors are the very cheapest green Mylar types. The author states the working is not the same as other lamp stabilised circuits. I feel I noticed this. The lamp is the cheapest of a common type. I think it is a light aircraft type. The hum is just that it should be in a metal case. I guess 1 nF for 10 kHz would be ideal. The caps were about 10.1nF, resistors 16K 1%. Very easy to build. I used a 50 x 25mm stripboard. Zinc 9V batteries and a double pole switch. Notice the letters on the chip. This was to me the better layout.