Hi!
I googled a bit, but can't seem to find a solid answer. How to check for oscillations in a solid state discrete design? So far I got this:
Happy to hear your point of view!
Circuit in question:
![20230716_213018[1].jpg](https://www.diyaudio.com/community/data/attachments/1101/1101692-94d897ce11fa165d58b25598de637975.jpg?hash=lNiXzhH6Fl "20230716_213018[1].jpg")
Schematic:
I googled a bit, but can't seem to find a solid answer. How to check for oscillations in a solid state discrete design? So far I got this:
- Overheating components can indicate oscillation
- Using oscilloscope to look for oscillations probing the circuit (sadly my only scope is ADALM2000...)
- Increased distortion (beyond expected values) when measuring the circuit
Happy to hear your point of view!
Circuit in question:
Schematic:
I would always go for the oscilloscope.
Overheating components and high distortion are possible symptoms of oscillation - but the scope will show it.
Without knowing the details of ADALM2000 (I just gave it a brief look) it seems that it would be capable of showing oscillations - even if only being a broad band over the screen.
Word of caution: the Analog datasheet for the ADALM2000 hints at a maximum input of +/- 20V and your design may generate a little more than this - given the 24V rails.
So I would recommend using a "devide by ten" probe with the oscillocope.
Cheers
Martin
Overheating components and high distortion are possible symptoms of oscillation - but the scope will show it.
Without knowing the details of ADALM2000 (I just gave it a brief look) it seems that it would be capable of showing oscillations - even if only being a broad band over the screen.
Word of caution: the Analog datasheet for the ADALM2000 hints at a maximum input of +/- 20V and your design may generate a little more than this - given the 24V rails.
So I would recommend using a "devide by ten" probe with the oscillocope.
Cheers
Martin
The current draw can jump in value as the signal amplitude is varied smoothly, indicating the onset or end of oscillations. 25MHz is going to catch many problems so that 'scope is worth using, but you many need an attenuator to protect its front end as it doesn't look like it can take 10x probes.
Oscillation can often be heard as a low level fizzling noise superimposed on the signal - try a lowish frequency sine wave as your test signal so such noise is easier to hear. The Zobel network resistor may heat up noticably too.
Oscillation can often be heard as a low level fizzling noise superimposed on the signal - try a lowish frequency sine wave as your test signal so such noise is easier to hear. The Zobel network resistor may heat up noticably too.
Input a 1KHz square wave and look for sine waves on square wave edges or overshoot.
I think I would have gone for a Vbe multiplier rather than a couple of LEDs.
I think I would have gone for a Vbe multiplier rather than a couple of LEDs.
Weird bias points are also a typical sign of oscillations, especially if they change when your finger comes close to a component (which you shouldn't try when the voltage is too high or when you wear a highly conductive ring, obviously).
The most reliable way to measure it is with an oscilloscope with a large bandwidth or with an RF spectrum analyser. I never saw the 200 MHz...300 MHz oscillations you sometimes get in discrete circuits with BC550s and BC560s when I still had a 20 MHz oscilloscope, but my 150 MHz oscilloscope shows them clearly (albeit with the wrong amplitude and waveform, as 150 is less than 200 to 300). Connecting an oscilloscope probe's ground lead to its tip makes a simple loop antenna that you can move around above the circuit to find where the oscillatory signal is the strongest.
The most reliable way to measure it is with an oscilloscope with a large bandwidth or with an RF spectrum analyser. I never saw the 200 MHz...300 MHz oscillations you sometimes get in discrete circuits with BC550s and BC560s when I still had a 20 MHz oscilloscope, but my 150 MHz oscilloscope shows them clearly (albeit with the wrong amplitude and waveform, as 150 is less than 200 to 300). Connecting an oscilloscope probe's ground lead to its tip makes a simple loop antenna that you can move around above the circuit to find where the oscillatory signal is the strongest.
If want to limit your support of oligarchic suppliers, I can detect ultrasonic oscillations with a $35 analog voltmeter. One with 20 vac, 50 vac, 200 vac scales. 5000 ohms/volt is fine. You need to block one input (I block the black one) with a capacitor to keep the needle from reading on DC values. Use alligator clip leads. A .047 uf cap will pass audio and ultrasonics. 390 pf cap will only pass ultrasonics. High steady voltage out with no audio in the speaker is the tip off to me. Make cap voltage exceed any voltage your circuit can produce.
That way if you step on the probe, you are not out >$50. Also there are no electrolytic caps in an analog VOM to kill the product in 5, 10 or 15 years. I bought my simpson 266xlpm meter in 1986.
Forget DVM. The inexpensive ones ($35) with a 200 vac scale produce random numbers on music. The expensive RMS ones ($160) read music okay, but have a top frequency response of 7000 hz. The digital meters with a pointer are no more useful for ultrasonic detection than the ones with numerals only. You need a coil pulling on a magnetic pointer to read radio.
That way if you step on the probe, you are not out >$50. Also there are no electrolytic caps in an analog VOM to kill the product in 5, 10 or 15 years. I bought my simpson 266xlpm meter in 1986.
Forget DVM. The inexpensive ones ($35) with a 200 vac scale produce random numbers on music. The expensive RMS ones ($160) read music okay, but have a top frequency response of 7000 hz. The digital meters with a pointer are no more useful for ultrasonic detection than the ones with numerals only. You need a coil pulling on a magnetic pointer to read radio.
Last edited:
Thank you all for helping out! I'll try my best to sniff it out with the ADALM2000. Would you say something like the Hantek DPO6202C (or similarly priced scope) would be enough to make sure the circuit is oscillation free? It's my first solid state design and I want it to be stable before I try it longer than a few minutes.
BTW as you can see in the photo I opted for a rather large 220p Miller compensation capacitor. I would like to transition into better tailored TMC solution in the future.
BTW as you can see in the photo I opted for a rather large 220p Miller compensation capacitor. I would like to transition into better tailored TMC solution in the future.
In any case, I would recommend using a supply with some sort of (crude) current limiter (real current limiter, series resistor, lightbulb) for the first tests.
@indianajo 's response is interesting. When you have a small coupling capacitor and some fast, basic rectifier circuit, you can make oscillations anywhere up to hundreds of megahertz visible with a meter that doesn't have such a large bandwidth.
For checking frequency compensation, a 20 MHz scope should normally be more than enough. Higher frequency oscillations are usually due to local interactions between a transistor and its wiring (the reason why base stoppers are sometimes needed).
@indianajo 's response is interesting. When you have a small coupling capacitor and some fast, basic rectifier circuit, you can make oscillations anywhere up to hundreds of megahertz visible with a meter that doesn't have such a large bandwidth.
For checking frequency compensation, a 20 MHz scope should normally be more than enough. Higher frequency oscillations are usually due to local interactions between a transistor and its wiring (the reason why base stoppers are sometimes needed).
You can try one of these cheaply, tested for RF over 1MHz.
https://www.circuitlib.com/index.ph...ffer-field-strength-meter/category_pathway-30
https://www.circuitlib.com/index.ph...ffer-field-strength-meter/category_pathway-30
I have a regulated power supply with current limiter. The circuit behaves as expected. The currents and voltages are in line with simulations, I have played a bit of music through it and it was fine.
@indianajo I don't have that on hand, but it's not something a quick shopping can't fix.
The oscilloscope is essential to look for oscillation. Any peaking in the frequency response indicates instability. Especially look at the output when turning the amplifier on/off. Transients can provide the necessary initial conditions for the oscillation to start at those times. Square wave testing can also be helpful. Use a more demanding load, 4R, as this will put the amplifier under more stress and oscillations are more likely to show. Then you can also add some output capacitance and see how it performs with these.
Sometimes connecting the scope will kill the oscillation.
Hence the RF sniffer, which does not interact with the circuit.
Hence the RF sniffer, which does not interact with the circuit.
Small update. I'll have a 100MHz scope, I'll build that sniffer proposed by @rayma and I'll try the method proposed by @indianajo.
I'll keep you updated, but no ETA for now. Cheers to everybody!
I'll keep you updated, but no ETA for now. Cheers to everybody!
A scope is handy for identifying transitory oscillations that only occur at the peak of the waveform (quite a common thing in my experience), and in particular can readily identify which peak is involved and how it changes with amplitude/load/bias. Knowing the frequency of oscillation is useful for identifying the relevant fix, be it motor-boating through to VHF oscillation of an output device.
A "castlellated" sine wave could also be used to "tickle" an amp to help find her difficult spots. It is a small amplitude square wave riding on top of a large amplitude sine wave. 
The sine wave would drive the amp to just shy of clipping, and that's often where the stability margin is less ample.
This would be a more gentle way of poking and probing than with a full amplitude square wave, as it doesn't stress the amp as much.
In the attached waveform, the sine is at about 1.6KHz, the square is at 16x that frequency and they are phase locked, locked so that a clear and stable waveform can be obtained on an oscilloscope.
I documented this signal generator project in This Thread
The sine wave would drive the amp to just shy of clipping, and that's often where the stability margin is less ample.
This would be a more gentle way of poking and probing than with a full amplitude square wave, as it doesn't stress the amp as much.
In the attached waveform, the sine is at about 1.6KHz, the square is at 16x that frequency and they are phase locked, locked so that a clear and stable waveform can be obtained on an oscilloscope.
I documented this signal generator project in This Thread
I usually have an AM radio in the workshop. Some oscillations hetrodyne with the radio!
You can current limit with LM317 and LM337s in the Power supply lines
You can current limit with LM317 and LM337s in the Power supply lines
Like jackinnj said:
I have my $3 spectrum analyzer (AM plastic battery loop stick radio) tune to a blank station and walk it around the product. If the signal is modulated it can be heard, if not modulated the noise goes quiet. I have been using this for 50 +years and never failed me.
I also use my TEK 2465B & HP 8594E to finalize the search.
Duke
I have my $3 spectrum analyzer (AM plastic battery loop stick radio) tune to a blank station and walk it around the product. If the signal is modulated it can be heard, if not modulated the noise goes quiet. I have been using this for 50 +years and never failed me.
I also use my TEK 2465B & HP 8594E to finalize the search.
Duke
Thanks for this tip, Jo! It just happens that I grabbed a nice condition Micronta analog meter about six months ago as I'd always wanted one for various tasks (such as testing potentiometers on cars). It's basically a cheaper copy of the Simpson style meter. Fun to play with... Plus I only have an old analog 20MHz scope, so this is another tool in the armoury. Cheers!
I work in SPICE.
But when circuit is not stable then the DC offset at output is not very stable.
To fix this I add more compensation. Increase Ccomp.
JFE2140 JFET was new to me. High quality device!