Hi all.
I have no serious test equipment yet -with the exception of a LCR multimeter-; over the last years, I have made some line, RIAA and headphone preamps with relative success, but there is always a shade of doubt about the objective performance of the devices: they ¨sound¨ -some of them sound very good, some not so good-but not always is easy to discern if the differences in overall sound quality are related to the design per se, or are related to a suboptimal PCB layout, poor bypassing, misplaced components etc.
I wish to know if any of you has consistently detected audible -or noticeable by any other mean, i.e. excessive warming of a device- symptoms of subtle oscillation in a circuit.
Thank you very much in advance.
I have no serious test equipment yet -with the exception of a LCR multimeter-; over the last years, I have made some line, RIAA and headphone preamps with relative success, but there is always a shade of doubt about the objective performance of the devices: they ¨sound¨ -some of them sound very good, some not so good-but not always is easy to discern if the differences in overall sound quality are related to the design per se, or are related to a suboptimal PCB layout, poor bypassing, misplaced components etc.
I wish to know if any of you has consistently detected audible -or noticeable by any other mean, i.e. excessive warming of a device- symptoms of subtle oscillation in a circuit.
Thank you very much in advance.
Subtle oscillation - difficult without an oscilloscope. More severe, yes quite easy by heat and sonics.
When serious oscillation issues are present, random AM or even FM radio stations may be clearly heard in the output or in some nodes of the circuit. Less severe oscillation issues may be perceived as an increase in the noise floor and may be wrongly identified as 'humm', however, when that noise is amplified it tends to sound much like the 'buzz' from an AM radio not tuned to any station. The noise produced by op-amps suffering from mild oscillation at several tens of Mhz may sound just like white noise instead.
you do need a scope -- a decent analog scope isn't that expensive and is a worthwhile investment -- good probes help -- but one thing you can do in the absence of a scope is measure the current in your circuit -- if it is greater than you anticipate from the design you 've got a problem. you can also touch the IC's with your finger and if they are roasting -- well enough said.
take a look a the application notes on bypassing on Analog Devices website -- they are clear and to the point -- (just put bypass into their search engine) Texas Instruments has a good application note on "situating" opamps, ground planes etc.
take a look a the application notes on bypassing on Analog Devices website -- they are clear and to the point -- (just put bypass into their search engine) Texas Instruments has a good application note on "situating" opamps, ground planes etc.
I would guess in many cases
there will large DC-offset across output.
And this offset will not be very stable, constant.
This is easy to measure with a DC Voltage meter.
there will large DC-offset across output.
And this offset will not be very stable, constant.
This is easy to measure with a DC Voltage meter.
Why not build a high-speed active rectifier. AC-couple into it, so you can test various points in the circuit. Any DC offset at the output of the rectifier circuit will tell you if there's any hidden oscillation. Small-signal diodes are good to several MHz.
Cheers!
Cheers!
My experience is that you will NOT see any DC on the outputs from oscillation. When a power amp goes into HF oscillation - going RF as we say - it will usually still pass audio, but it does have a certain weak, washed out sound to it, and it is noisy in a hummy sort of way.
SOme non-scope methods would be:
Use the meter on AC current and monitor its draw from the mains. If that seems excessive, the unit might be cranking out the RF even at idle. The unit would run hot as well. A simple way is remove the mains fuse and connect the ammeter in its place.
The AM radio methiod sometimes works, but some 50kHz might not be enough to swamp the radio front end.
You can make a simple RF detector for your meter. The meter will read AC OK but not up to 50kHz, 100kHz, or even in to the MHz range. But a simple diode detector can. Even if there is 20v of signal on the output, it averages zero on DC, and at high freqs your AC meter can't read it. But if we sample the signal and rectifiy it, now it becomes a DC signal we CAN measure.
Basically from the test point, run a small diode to your meter. Then a small cap from the meter hot lead to its ground lead will smooth this DC. COnsider the voltages involved. If teh amplifier has 40 volt rails, the signal can be as high as 80v p-p, so the diode and cap should reflect this. Use a 1N4004 or something and a 100v cap. Doesn't much matter on the cap, a 0.1uf ought to do it. I suppose you could also add a series cap at the test probe tip to block actual DC, again value not critical, 0.1uf comes to mind. But at these freqs the cap offers little resistance.
You can also make them with the diode in shunt, doesn't matter.
50 years ago when I was in amateur radio, "RF probes" were a common accessory for one's meter. I haven't seen one in a long time, and my scope makes it unecessary here, but they probably still make them commercially.
In fact, Google "RF probe" and the first few hits there show exactly how to make this simple tool.
SOme non-scope methods would be:
Use the meter on AC current and monitor its draw from the mains. If that seems excessive, the unit might be cranking out the RF even at idle. The unit would run hot as well. A simple way is remove the mains fuse and connect the ammeter in its place.
The AM radio methiod sometimes works, but some 50kHz might not be enough to swamp the radio front end.
You can make a simple RF detector for your meter. The meter will read AC OK but not up to 50kHz, 100kHz, or even in to the MHz range. But a simple diode detector can. Even if there is 20v of signal on the output, it averages zero on DC, and at high freqs your AC meter can't read it. But if we sample the signal and rectifiy it, now it becomes a DC signal we CAN measure.
Basically from the test point, run a small diode to your meter. Then a small cap from the meter hot lead to its ground lead will smooth this DC. COnsider the voltages involved. If teh amplifier has 40 volt rails, the signal can be as high as 80v p-p, so the diode and cap should reflect this. Use a 1N4004 or something and a 100v cap. Doesn't much matter on the cap, a 0.1uf ought to do it. I suppose you could also add a series cap at the test probe tip to block actual DC, again value not critical, 0.1uf comes to mind. But at these freqs the cap offers little resistance.
You can also make them with the diode in shunt, doesn't matter.
50 years ago when I was in amateur radio, "RF probes" were a common accessory for one's meter. I haven't seen one in a long time, and my scope makes it unecessary here, but they probably still make them commercially.
In fact, Google "RF probe" and the first few hits there show exactly how to make this simple tool.
Enzo said:
Enzo
Thanks for info, that there is often no DC-offset at output
so this is no good method to detect oscillation.
My guess was not so good.
To be clear, there may or may not be DC offset, just as there may or may not be oscillation. But they are not usually linked.
Oscillation may cause offsets only when the circuit suffers from highly asymmetric slew rates. A simple diode rectifier may not detect mild oscillation at all because the voltage amplitude may be well below one diode drop.
Also, the most usual and most troublesome oscillation issues in power amplifiers appear only at certain output current and voltage levels, so it's required to feed some signal in the audio band to the circuit and play with its amplitude while monitoring the RF that is coming from the output through a high pass filter. Obviously, several load impedances have also to be tested. Most oscillation issues appear only near zero crossing or near clipping, but sometimes oscillation may appear only between these points.
The RC output zobel may be a very useful built in HP filter when the capacitor is connected to the output and the resistor to ground. In these circumstances, all the RF will be present across the zobel resistor without any bass signal disturbing the measurement.
Also, the most usual and most troublesome oscillation issues in power amplifiers appear only at certain output current and voltage levels, so it's required to feed some signal in the audio band to the circuit and play with its amplitude while monitoring the RF that is coming from the output through a high pass filter. Obviously, several load impedances have also to be tested. Most oscillation issues appear only near zero crossing or near clipping, but sometimes oscillation may appear only between these points.
The RC output zobel may be a very useful built in HP filter when the capacitor is connected to the output and the resistor to ground. In these circumstances, all the RF will be present across the zobel resistor without any bass signal disturbing the measurement.
Eva said:
Right - that's why an active rectifier (what I meant - one without any volttage drop) will work better. But Eva's right - often oscillations show up only when the circuit's actively driven; i.e. using a square wave to look for jaggedness and ring in the response. That means some kind of cancellation would be needed to remove the test signal, leaving the residual oscillations, if any.
Easier to buy an old analog oscilloscope! 🙂
Cheers!
Active rectifiers are usually op-amp based, so they are not likely to work properly with signals above 1Mhz.
detection of oscillations
Thank you very much for all the tips; they are really interesting for me, because I`m doing my first steps in DIY -with a cuote of good luck, happily-, but I have no electronics background, nor experience in testing.
It appears that a good move will be the acquisition of an oscilloscope, and some reading and learning. It will be fun.
Besides that, the method described by Enzo is very interesting as experimentation, and a good way to excercise creative approaches to electronics.
AM radio is an interesting tip too, as well as the different sonic signatures of different kind of oscillations.
I believe that a mix of methods will be useful in casual situations; the serious approach afforded by the oscilloscope will be the icing of the cake.
Thanks to all of you again.
Thank you very much for all the tips; they are really interesting for me, because I`m doing my first steps in DIY -with a cuote of good luck, happily-, but I have no electronics background, nor experience in testing.
It appears that a good move will be the acquisition of an oscilloscope, and some reading and learning. It will be fun.
Besides that, the method described by Enzo is very interesting as experimentation, and a good way to excercise creative approaches to electronics.
AM radio is an interesting tip too, as well as the different sonic signatures of different kind of oscillations.
I believe that a mix of methods will be useful in casual situations; the serious approach afforded by the oscilloscope will be the icing of the cake.
Thanks to all of you again.
Greetings from Norfolk
Just a thought - a 1N4000 series diode will loose performance somewhee around 1 MHZ I think - I would suggest that a signal diode be used (1N914 is OK to about 75 V rev. voltage) and will rectify better at around 1 MHz.
Also suggest that the smoothing capacitor be a ceramic one - not a 'wound plastic film' type - this may have a significant inductance in series with it due to its construction.
Good luck
Richard
Just a thought - a 1N4000 series diode will loose performance somewhee around 1 MHZ I think - I would suggest that a signal diode be used (1N914 is OK to about 75 V rev. voltage) and will rectify better at around 1 MHz.
Also suggest that the smoothing capacitor be a ceramic one - not a 'wound plastic film' type - this may have a significant inductance in series with it due to its construction.
Good luck
Richard
Eva said:
Hi Eva, of course you're right, but I was thinking of one of the hi-speed FET input op-amps, or an alternative active circuit to track the bias on the diode.
Hi gandalph - agreed, in914 or 1n4148, or in34...
Cheers!
Yes I suppose we should differentiate between instability and parasitic oscillation. I usually run into amps that will sit there cranking out the RF. I don't nearly as often see triggered oscillations except when the circuit is otherwise really screwed up too. I also deal with high powered stuff, so I worry about voltages. I suppose the diode Vrev is not an issue in a 30 watt amp.
As to not reaching the diode junction voltage, again I assumed that the amp was cranking out the HF. SO losing a half a volt from the signal is no big loss.
As to not reaching the diode junction voltage, again I assumed that the amp was cranking out the HF. SO losing a half a volt from the signal is no big loss.
For everybody that can get access to an oscilloscope:
Connect one end of a 100pf capacitor to the output of the amplifier, the other end to a 1K resistor, and the other end of the resistor to ground. This will form a nice 1Mhz high-pass filter. Connect one probe to the junction betwen the C and the R, and the other probe to the output of the amplifier. Set the trigger source to the low-pass filtered probe.
This setup allows to find out exactly when and where an amplifier oscillates. I've just used it to find and fix a 14Mhz oscillation with only 100mV amplitude that otherwise was very hard to analyse because it only appeared very briefly near zero crossing.
When everything is right, the filtered probe should show almost no signal at all (a few milivolts), even when the the amplifier is playing at full volume. Clipping may introduce some HF content on the output, but this is not a major issue.
Note that oscillation does introduce serious distortion because it impairs the error-correction properties of the global feedback loop while it happens.
Connect one end of a 100pf capacitor to the output of the amplifier, the other end to a 1K resistor, and the other end of the resistor to ground. This will form a nice 1Mhz high-pass filter. Connect one probe to the junction betwen the C and the R, and the other probe to the output of the amplifier. Set the trigger source to the low-pass filtered probe.
This setup allows to find out exactly when and where an amplifier oscillates. I've just used it to find and fix a 14Mhz oscillation with only 100mV amplitude that otherwise was very hard to analyse because it only appeared very briefly near zero crossing.
When everything is right, the filtered probe should show almost no signal at all (a few milivolts), even when the the amplifier is playing at full volume. Clipping may introduce some HF content on the output, but this is not a major issue.
Note that oscillation does introduce serious distortion because it impairs the error-correction properties of the global feedback loop while it happens.
- Status
- Not open for further replies.