Years ago, I tested non-fedback triode amplifiers. I did frequency response and phase plots into an 8 Ohm non inductive resistor. The plots covered from 10Hz to 50kHz.
But I did something I had not seen in the literature. I overlaid the 8 Ohm load plots of frequency and phase, with Real Loudspeaker Load plots of frequency and phase.
(Yes, I know some do this with 'Simulated Loudspeaker' loads). A complex 'LCR woofer simulator' is not the same as an actual moving woofer in a cabinet. There is no harmonic distortion, and only predictable loading from a very good LCR simulator. Yes, an air core inductor is non distorting versus an inductor with a magnetic core.
I also did HD and IM tests into an 8 Ohm resistor. But what I did not do was IM tests into Real Loudspeaker Loads. Again, we have moving woofer, moving mid/tweet, crossover (some w/air core, some w/magnetic core). I am just saying that non linear speakers can make amplifiers less linear. Just using a 4 Ohm non inductive resistor on the 8 Ohm tap is not the same as a real loudspeaker.
But I did something I had not seen in the literature. I overlaid the 8 Ohm load plots of frequency and phase, with Real Loudspeaker Load plots of frequency and phase.
(Yes, I know some do this with 'Simulated Loudspeaker' loads). A complex 'LCR woofer simulator' is not the same as an actual moving woofer in a cabinet. There is no harmonic distortion, and only predictable loading from a very good LCR simulator. Yes, an air core inductor is non distorting versus an inductor with a magnetic core.
I also did HD and IM tests into an 8 Ohm resistor. But what I did not do was IM tests into Real Loudspeaker Loads. Again, we have moving woofer, moving mid/tweet, crossover (some w/air core, some w/magnetic core). I am just saying that non linear speakers can make amplifiers less linear. Just using a 4 Ohm non inductive resistor on the 8 Ohm tap is not the same as a real loudspeaker.
Smoking Amp:
Transformer distortion occurs with high drive levels with frequencies near the bass roll-off, as I said. the best way to detect/measure it is with the standard THD test, with the test tone taken down in frequency until distortion markedly increases, if it does. If it doesn't, this isn't the problem.
Grid blocking can be a problem, as others have said, during musical peaks. And if there is a short sharp transient, such as from record scratch, or tube glass discharge, which is audible but not particularly bothersome, grid blocking can turn it into terrible distortion, or even complete loss of signal lasting a fraction of a second or even longer in bad cases. All amplifiers designed with a claim to high quality or highly pleasant reproduction should be designed so that grid blocking does not occur. The best way to detect & measure it is with a pulse test. Your proposal will not detect it.
In push-pull output stages, even when nominally Class A, mean cathode current increases with signal. If there is a cathode bypass capacitor, there will be a grid shift, and the amplifier becomes a compromise - it is correctly biased at one signal level only, and may distort more at other levels, or it may be masked by distortions elsewhere. This problem is easily detected and managed with standard THD or intermod tests.
Amplifiers which are unstable under certain conditions often exhibit the symptom of high distortion that comes and goes. If your test detects this, it will be by pure luck. The best way to detect this is by testing with sinewaves or basic triangle waves, with the amplifier tested with a reasonable range of simulated speaker loads of R in series in C and L in series with C.
Transformer distortion occurs with high drive levels with frequencies near the bass roll-off, as I said. the best way to detect/measure it is with the standard THD test, with the test tone taken down in frequency until distortion markedly increases, if it does. If it doesn't, this isn't the problem.
Grid blocking can be a problem, as others have said, during musical peaks. And if there is a short sharp transient, such as from record scratch, or tube glass discharge, which is audible but not particularly bothersome, grid blocking can turn it into terrible distortion, or even complete loss of signal lasting a fraction of a second or even longer in bad cases. All amplifiers designed with a claim to high quality or highly pleasant reproduction should be designed so that grid blocking does not occur. The best way to detect & measure it is with a pulse test. Your proposal will not detect it.
In push-pull output stages, even when nominally Class A, mean cathode current increases with signal. If there is a cathode bypass capacitor, there will be a grid shift, and the amplifier becomes a compromise - it is correctly biased at one signal level only, and may distort more at other levels, or it may be masked by distortions elsewhere. This problem is easily detected and managed with standard THD or intermod tests.
Amplifiers which are unstable under certain conditions often exhibit the symptom of high distortion that comes and goes. If your test detects this, it will be by pure luck. The best way to detect this is by testing with sinewaves or basic triangle waves, with the amplifier tested with a reasonable range of simulated speaker loads of R in series in C and L in series with C.
Last edited:
I connect my non-feedback or low-feedback amps to a real loudspeaker, and then apply an impulse. Not a lot to see visually in the time domain. But then I look at the FFT and see if there are some peaks or nulls.
As I said, earlier, I have not used a real loudspeaker load to do IM testing of the amps. And this needs to be done with bursts and gated FFT.
If I can get the equipment to do this, I may build a suitable sound chamber. I can take the 'click' sound of an impulse, but I can not endure the sound of a two tone test.
As I said, earlier, I have not used a real loudspeaker load to do IM testing of the amps. And this needs to be done with bursts and gated FFT.
If I can get the equipment to do this, I may build a suitable sound chamber. I can take the 'click' sound of an impulse, but I can not endure the sound of a two tone test.
The ear plus brain's analysis power seems to be somewhat better than most T&M equipment's measurement capabilities.
Now what? Double blind?
I am reminded of a double blind test where the sound of the solid state amplifiers of the day were compared to an OTL vacuum tube amplifier's sound. There were trained and untrained listeners. The test results were that statistically, they were unable reliably identify the different amplifiers. But . . . The common characteristics of ALL these amplifiers was . . . Totem Pole Output stage, and Lots of Global Negative Feedback!
I am just saying, there may be something to that.
Now what? Double blind?
I am reminded of a double blind test where the sound of the solid state amplifiers of the day were compared to an OTL vacuum tube amplifier's sound. There were trained and untrained listeners. The test results were that statistically, they were unable reliably identify the different amplifiers. But . . . The common characteristics of ALL these amplifiers was . . . Totem Pole Output stage, and Lots of Global Negative Feedback!
I am just saying, there may be something to that.
Well, I for one will be following that thread on peeling the amplifier onion...
>people attack capacitors, just out of ignorance
I said why I proposed this thread, when I removed the speaker crossover from my listening experience by going to full range drivers. I postulated that the pleasant change was due to fewer capacitors in the overall signal path - and if so - why wouldnt a further reduction of these elements further improve the sound? That would be in the amplifier signal chain. So I asked the question.
I once tried a frequency response test on a tube amp using pink noise or white noise. The pink noise wanted a wide frequency band analysis to show a flat response, while the white noise wanted narrow band analysis to show flat response - using lots of band level averaging in both cases.
I recall the tube amp really didnt like being subjected to this kind of signal, sounding like the windings were physically rattling about in the output transformer. So I gave up the idea. That aside, it should be possible to inject an all frequencies simultaneously signal and get a FFT based frequency response plot in one shot. I dont know what was going on in the Dynaco MKIII's OPT...
Perhaps an all frequency noise signal could be AM modulated to show the effects of dynamics. You'd have to be able to take sample sets to FFT fast enough so that they would correspond to various levels of the AM modulated waveform. This would be different than statically setting various steady state levels, because the amplifier would be responding to the dynamics in the modulated signal case. Unsure if it could be made to work; maybe triggering the FFT at a specific level of the AM would allow one to see the frequency response difference between low levels and higher levels, each being averaged across multiple triggers taken at their specific amplitude.
Maybe it'd only be good for higher frequencies, as there wouldnt be enough time to capture the signal for an FFT down to 20Hz, with a realistic modulating frequency that could reveal effects of an amplifier's dynamic behavior - say, 5 - 10 Hz.
>people attack capacitors, just out of ignorance
I said why I proposed this thread, when I removed the speaker crossover from my listening experience by going to full range drivers. I postulated that the pleasant change was due to fewer capacitors in the overall signal path - and if so - why wouldnt a further reduction of these elements further improve the sound? That would be in the amplifier signal chain. So I asked the question.
I once tried a frequency response test on a tube amp using pink noise or white noise. The pink noise wanted a wide frequency band analysis to show a flat response, while the white noise wanted narrow band analysis to show flat response - using lots of band level averaging in both cases.
I recall the tube amp really didnt like being subjected to this kind of signal, sounding like the windings were physically rattling about in the output transformer. So I gave up the idea. That aside, it should be possible to inject an all frequencies simultaneously signal and get a FFT based frequency response plot in one shot. I dont know what was going on in the Dynaco MKIII's OPT...
Perhaps an all frequency noise signal could be AM modulated to show the effects of dynamics. You'd have to be able to take sample sets to FFT fast enough so that they would correspond to various levels of the AM modulated waveform. This would be different than statically setting various steady state levels, because the amplifier would be responding to the dynamics in the modulated signal case. Unsure if it could be made to work; maybe triggering the FFT at a specific level of the AM would allow one to see the frequency response difference between low levels and higher levels, each being averaged across multiple triggers taken at their specific amplitude.
Maybe it'd only be good for higher frequencies, as there wouldnt be enough time to capture the signal for an FFT down to 20Hz, with a realistic modulating frequency that could reveal effects of an amplifier's dynamic behavior - say, 5 - 10 Hz.
Last edited:
I mentioned the Impulse test earlier. It is a single sample pulse (1/44.1k in time) on a CD. The CD player filters it, and it has a flat spectra from very low frequency, all the way to 44.1k/2 (= 22.05kHz). The impulse is repeated at a regular rate. That defines the lowest frequency. That allows the FFT to make the complete test, without gating the spectrum. The impulse is good for a quick test, and for that I like it better than white or pink noise.
True white noise and true pink noises are non repetitive. That makes it more difficult to make any precision tests with just an FFT.
True white noise and true pink noises are non repetitive. That makes it more difficult to make any precision tests with just an FFT.
Last edited:
Use a pseudo-random pulse train filtered(analog of digital filter) so that it looks exactly like white noise to the amplifier under test. Because its digitally generated, it is precisely repeatable - you get the same results every time.
A very good (in that it is sensitive, results are proportional to the audible effects, and is a single test that detects ANY form of distortion), is the notched white noise test - easily done by digital filtering. In this test, the amplifier gets all frequencies as in white noise, but with a selected narrow band filtered out. At the output of the amplifier, anything detected by a spectrum analyser in the filtered out band is distortion products generated within the amplifier under test.
Of course, if you optimize the amplifier so that no products occur in the notched out band, what you get is an accurate amplifier - which is not what everyone wants.
The pulse test that you should use to detect amplifier distortions caused by grid blocking, grid shifts, and the like, is not what has been described in this thread. The proper test is a short burst of high amplitude preceded by and immediately followed by a low level tone. On some tube amplifiers what you see on an oscilloscope across the output is that the amplifier is paralysed for a short time after the burst of high level - the low level tone becomes very distorted or even completely missing during the brief recovery period.
It is possible to gate a THD meter so that it ignores the high level burst and measure the THD resulting in the low level tone.
The high level burst can be a single pulse or a few cycles of a sinewave. Each can best reveal particular amplifier design faults.
I'm glad you included the word "most". Because anything the ear can detect is detectable by test gear. Everything. You just have to use the right test gear.
Peter Walker of Quad fame decades ago came up with a good test. Essentially, it is this: You add a resistive mixer with one input connected to the output to the amplifier under test. Leave the speaker connected. Connect the other input of the mixer to a phase-inverted version of the amplifier input, with controls to match the level and the phase shifts inherent in the amplifier. This means the output of the mixer contains only the distortion products of the amplifier, and the amplified input signal is cancelled out. The signal used can be music to taste, sinewaves, pulses, white noise, whatever you like.
Feed the mixer output to a unity gain power amplifier and feed that to a loudspeaker in an acoustically isolated room.
This test revealed that with Quad amplifiers (tube and solid state) and with amplifiers from other manufacturers of professional studio standard, nobody could hear anything in the second speaker. But this is not the case with many retail amplifiers, and not the case with many (not all) early generation solid state amplifiers.
If you can't hear distortion products by themselves, you certainly can't hear them when accompanied by the wanted signal.
A very good (in that it is sensitive, results are proportional to the audible effects, and is a single test that detects ANY form of distortion), is the notched white noise test - easily done by digital filtering. In this test, the amplifier gets all frequencies as in white noise, but with a selected narrow band filtered out. At the output of the amplifier, anything detected by a spectrum analyser in the filtered out band is distortion products generated within the amplifier under test.
Of course, if you optimize the amplifier so that no products occur in the notched out band, what you get is an accurate amplifier - which is not what everyone wants.
The pulse test that you should use to detect amplifier distortions caused by grid blocking, grid shifts, and the like, is not what has been described in this thread. The proper test is a short burst of high amplitude preceded by and immediately followed by a low level tone. On some tube amplifiers what you see on an oscilloscope across the output is that the amplifier is paralysed for a short time after the burst of high level - the low level tone becomes very distorted or even completely missing during the brief recovery period.
It is possible to gate a THD meter so that it ignores the high level burst and measure the THD resulting in the low level tone.
The high level burst can be a single pulse or a few cycles of a sinewave. Each can best reveal particular amplifier design faults.
I'm glad you included the word "most". Because anything the ear can detect is detectable by test gear. Everything. You just have to use the right test gear.
Peter Walker of Quad fame decades ago came up with a good test. Essentially, it is this: You add a resistive mixer with one input connected to the output to the amplifier under test. Leave the speaker connected. Connect the other input of the mixer to a phase-inverted version of the amplifier input, with controls to match the level and the phase shifts inherent in the amplifier. This means the output of the mixer contains only the distortion products of the amplifier, and the amplified input signal is cancelled out. The signal used can be music to taste, sinewaves, pulses, white noise, whatever you like.
Feed the mixer output to a unity gain power amplifier and feed that to a loudspeaker in an acoustically isolated room.
This test revealed that with Quad amplifiers (tube and solid state) and with amplifiers from other manufacturers of professional studio standard, nobody could hear anything in the second speaker. But this is not the case with many retail amplifiers, and not the case with many (not all) early generation solid state amplifiers.
If you can't hear distortion products by themselves, you certainly can't hear them when accompanied by the wanted signal.
Keit,
I Believe the inventors of the notched out ‘white noise’ measurement was invented by Bell Labs for their microwave links. There were multiple 4kHz bandwidth com channels stacked onto an FM modulator of a microwave link (0-4, 4-8, 8-12, etc.). One of the 4k channels was pulled out of the transmitter, and the filled in channel was measured at and by the receiver.
Later, for digitally modulated cell phone base stations, there was a measurement of the IM distortion signals in the immediate channels on either side of the transmitted channel (predominantly 3rd order distortion), and in the alternate channels that were just outside of the immediate channels (predominately 5th order IM).
I do not have the tools, but I came up with the idea of creating a digitally modulated test tone for audio. Example, a 2 kHz carrier frequency, modulated with QPSK, 16QAM, or 64QAM, at a 500Hz symbol rate. The spectrum would be from 1.75k to 2.25k. Any signals from the amplifier outside of that are IM distortion (and harmonic distortion, but that does not show up until 3.5 kHz).
You mention one cause of distortion . . . blocking.
There is another cause of distortion . . . sticking. When the signal clips, any negative feedback may also be clipped. It takes time for the negative feedback to recover.
There have been many demonstrations of null methods that allow you to listen for distortion when the original/desired signal has been removed.
One amp topology Quad used was feedforward. The major output power from the amp had distortion, a smaller more refined amp was fed the opposite of the distortion of the major amp. The two amp chains power was combined, canceling the distortion.
Please do not forget to read the literature on TIM (transient intermodulation distortion). Mattie Otala comes to mind, and I do not remember the name of the other pioneer from Finland.
I Believe the inventors of the notched out ‘white noise’ measurement was invented by Bell Labs for their microwave links. There were multiple 4kHz bandwidth com channels stacked onto an FM modulator of a microwave link (0-4, 4-8, 8-12, etc.). One of the 4k channels was pulled out of the transmitter, and the filled in channel was measured at and by the receiver.
Later, for digitally modulated cell phone base stations, there was a measurement of the IM distortion signals in the immediate channels on either side of the transmitted channel (predominantly 3rd order distortion), and in the alternate channels that were just outside of the immediate channels (predominately 5th order IM).
I do not have the tools, but I came up with the idea of creating a digitally modulated test tone for audio. Example, a 2 kHz carrier frequency, modulated with QPSK, 16QAM, or 64QAM, at a 500Hz symbol rate. The spectrum would be from 1.75k to 2.25k. Any signals from the amplifier outside of that are IM distortion (and harmonic distortion, but that does not show up until 3.5 kHz).
You mention one cause of distortion . . . blocking.
There is another cause of distortion . . . sticking. When the signal clips, any negative feedback may also be clipped. It takes time for the negative feedback to recover.
There have been many demonstrations of null methods that allow you to listen for distortion when the original/desired signal has been removed.
One amp topology Quad used was feedforward. The major output power from the amp had distortion, a smaller more refined amp was fed the opposite of the distortion of the major amp. The two amp chains power was combined, canceling the distortion.
Please do not forget to read the literature on TIM (transient intermodulation distortion). Mattie Otala comes to mind, and I do not remember the name of the other pioneer from Finland.
Last edited:
6A3sUMMER:
Not sure but I think you are right about the origin of the notched noise test. For quite a while Marconi sold a notch noise test set for use on comms channels.
TIM is unlikely to be a problem in tube amplifiers. Special test sets have been designed to measure TIM, but it is readily detectable with a standard 2-tone intermod test. It can be a serious problem in solid state amps if the designer doesn't know what he is doing.
TIM can be a problem in solid state amplifiers because they commonly use emitter follower (or source follower) output stages which have unity voltage gain. This means the VAS stage has to rapidly traverse from rail to rail or near to. Further, lag compensation on the VAS is used. Because of large amounts of neg feedback used to get good THD spec for marketing, it has to be fairly heavy. And lag compensation kills the VAS's ability to rapidly traverse its output.
It can get worse. Emitter followers are natural HF oscillators. Look at an emitter follower while visuallising parasitic capacitances and the reactive speaker load - its the configuration of a Colpitts/Clapp oscillator. Early solid state designers used the VAS lag compensation to fix that too - wrong thig to do. Also the capacitance of the speaker leads (and xover capacitors) can cause a phase shift and upset the neg feedback. Early designers used the VAS lag comp to fix that too. Wrong again. Both problems are best fixed by an LR circuit at the amplifier output.
Now, in a tube amplifier, the output stage has voltage gain, making life easier for the VAS. Lower amounts of neg feedback are normally used - indeed, with an output transformer you have to use less. So the brute force method of lag compensation isn't needed - all you need is the judicious use of a Zobel CR network or two, and lead compensation in the feedback.
So, quite unlike a solid state, TIM can be pretty much forgotten about in a tube amp, unless there is something weird in its design.
As far as "sticking" is concerned - I've been in this game for 60 years, designed many things from cheap lo-fi radios (even in cheap lo-fi, you tray your darndest to squeeze the best possible performance out the production cost the boss lets you have) to impeccable studio amplifiers, and I've never heard of it. Please explain it.
Not sure but I think you are right about the origin of the notched noise test. For quite a while Marconi sold a notch noise test set for use on comms channels.
TIM is unlikely to be a problem in tube amplifiers. Special test sets have been designed to measure TIM, but it is readily detectable with a standard 2-tone intermod test. It can be a serious problem in solid state amps if the designer doesn't know what he is doing.
TIM can be a problem in solid state amplifiers because they commonly use emitter follower (or source follower) output stages which have unity voltage gain. This means the VAS stage has to rapidly traverse from rail to rail or near to. Further, lag compensation on the VAS is used. Because of large amounts of neg feedback used to get good THD spec for marketing, it has to be fairly heavy. And lag compensation kills the VAS's ability to rapidly traverse its output.
It can get worse. Emitter followers are natural HF oscillators. Look at an emitter follower while visuallising parasitic capacitances and the reactive speaker load - its the configuration of a Colpitts/Clapp oscillator. Early solid state designers used the VAS lag compensation to fix that too - wrong thig to do. Also the capacitance of the speaker leads (and xover capacitors) can cause a phase shift and upset the neg feedback. Early designers used the VAS lag comp to fix that too. Wrong again. Both problems are best fixed by an LR circuit at the amplifier output.
Now, in a tube amplifier, the output stage has voltage gain, making life easier for the VAS. Lower amounts of neg feedback are normally used - indeed, with an output transformer you have to use less. So the brute force method of lag compensation isn't needed - all you need is the judicious use of a Zobel CR network or two, and lead compensation in the feedback.
So, quite unlike a solid state, TIM can be pretty much forgotten about in a tube amp, unless there is something weird in its design.
As far as "sticking" is concerned - I've been in this game for 60 years, designed many things from cheap lo-fi radios (even in cheap lo-fi, you tray your darndest to squeeze the best possible performance out the production cost the boss lets you have) to impeccable studio amplifiers, and I've never heard of it. Please explain it.
Last edited:
An amp that is overdriven by a square wave is probably the simplest test for sticking. If it happens, you can easily see it in the time domain.
Perhaps someone has a photo of it. I have not used negative feedback for years, except for UL.
In 1971 I heard about the Bell Labs missing channel test. So easy to do, just pull one of the transmitter channel modules out. Then see how much signal you get at the receiver end of things, should be zero, but it is not. I say 1971, because I believe the test was already decades old at that time.
I bet Marconi got the idea for a T&M product from the Bell Labs test. But Bell did not have to purchase anything, other companies had to.
It is good you are familiar with oscillators. Here is another: Some wonder why parallel tubes can oscillate. The wires that join them are inductors. The oscillator is called a Buttler oscillator.
One engineer needed to increase the current from an IC Op amp. He used an NPN and PNP totem pole emitter follower, and wrapped the output into the feedback loop. He tied the bases together. The collectors went to their respective + and - supplies that were bypassed by local caps at the collectors. The emitters were also tied together. There were no resistors in the follower. The input to the op amp was a DAC (but was operated at slow rates as an adjustment voltage, not a signal). As the voltage across the NPN and PNP changed, and the current through them changed, a point was reached where the parts effective cutoff frequency, ft was approached, the wiring and capacitance caused a resonance that made the transistor oscillate. You could only find the problem if you stepped the DAC through its number range. A couple of base resistors fixed the problem. Sound familiar (grid stoppers).
Perhaps someone has a photo of it. I have not used negative feedback for years, except for UL.
In 1971 I heard about the Bell Labs missing channel test. So easy to do, just pull one of the transmitter channel modules out. Then see how much signal you get at the receiver end of things, should be zero, but it is not. I say 1971, because I believe the test was already decades old at that time.
I bet Marconi got the idea for a T&M product from the Bell Labs test. But Bell did not have to purchase anything, other companies had to.
It is good you are familiar with oscillators. Here is another: Some wonder why parallel tubes can oscillate. The wires that join them are inductors. The oscillator is called a Buttler oscillator.
One engineer needed to increase the current from an IC Op amp. He used an NPN and PNP totem pole emitter follower, and wrapped the output into the feedback loop. He tied the bases together. The collectors went to their respective + and - supplies that were bypassed by local caps at the collectors. The emitters were also tied together. There were no resistors in the follower. The input to the op amp was a DAC (but was operated at slow rates as an adjustment voltage, not a signal). As the voltage across the NPN and PNP changed, and the current through them changed, a point was reached where the parts effective cutoff frequency, ft was approached, the wiring and capacitance caused a resonance that made the transistor oscillate. You could only find the problem if you stepped the DAC through its number range. A couple of base resistors fixed the problem. Sound familiar (grid stoppers).
6A3sUMMER:
You said sticking of the negative feedback. Given that neg feedback is invariably via passive components, I don't think there is such a thing.
There is a sticking phenomena in solid state amplifiers, which can be exposed by your square-wave test if one looks carefully, but it's better detected by a large amplitude pulse immediately followed by a low level 10-12 kHz sinewave - the sinewave may not come thru the amp until some time after the pulse.
What happens is that overdrive of the input stage due to clipping at the output causes saturation in the VAS transistor, or even in the emitter follower output transistors. Saturation i.e., turned hard on) causes bipolar transistor to accumulate an excess of current carriers in the base region. When the signal changes to turn the transistor off, it doesn't until these excess carriers diffuse out.
It can cause a surprisingly long duration turn-off lag in power transistors. In 1966 Mullard and Philips released circuits as application notes for "high quality" 10W stereo amplifiers featuring their then new low cost complementary transistors AD161/162. These circuits were quickly used by a multitude of stereo manufactuers, with the AD161/162 or with other maker's transistors.
These amplifiers sounded pretty good MOST of the time, by the standard of the day. But they had a serious flaw - as there was no low cost NPN in the Philips range to use as a driver current source, they used bootstrapping to make the VAS load resistor voltage constant. This meant that the VAS could drive the upper output transistor well into saturation - and there was "sticking".
The cure was extremely simple, and published by Mullard a short time (a few months as I recall) later - a diode between the power rail and the VAS collector, clamping the VAS output and preventing output transistor saturation. But too late - thousands of stereos by dozens of manufacturers already in volume production. I worked for a small outfit, and I got my design corrected.
You said sticking of the negative feedback. Given that neg feedback is invariably via passive components, I don't think there is such a thing.
There is a sticking phenomena in solid state amplifiers, which can be exposed by your square-wave test if one looks carefully, but it's better detected by a large amplitude pulse immediately followed by a low level 10-12 kHz sinewave - the sinewave may not come thru the amp until some time after the pulse.
What happens is that overdrive of the input stage due to clipping at the output causes saturation in the VAS transistor, or even in the emitter follower output transistors. Saturation i.e., turned hard on) causes bipolar transistor to accumulate an excess of current carriers in the base region. When the signal changes to turn the transistor off, it doesn't until these excess carriers diffuse out.
It can cause a surprisingly long duration turn-off lag in power transistors. In 1966 Mullard and Philips released circuits as application notes for "high quality" 10W stereo amplifiers featuring their then new low cost complementary transistors AD161/162. These circuits were quickly used by a multitude of stereo manufactuers, with the AD161/162 or with other maker's transistors.
These amplifiers sounded pretty good MOST of the time, by the standard of the day. But they had a serious flaw - as there was no low cost NPN in the Philips range to use as a driver current source, they used bootstrapping to make the VAS load resistor voltage constant. This meant that the VAS could drive the upper output transistor well into saturation - and there was "sticking".
The cure was extremely simple, and published by Mullard a short time (a few months as I recall) later - a diode between the power rail and the VAS collector, clamping the VAS output and preventing output transistor saturation. But too late - thousands of stereos by dozens of manufacturers already in volume production. I worked for a small outfit, and I got my design corrected.
Last edited:
On further thought, there is another sort of apparently sticking phenomena that one sometimes sees in solid state (do you get the impression I don't like solid state by my listing of all sorts of distortion sources not applicable to tubes? Actually, I'm ok with solid state, but it is far more important you know what you doing than with tubes) that can happen with tube amplifiers, though not terribly likely.
If the amp undergoes slew rate limiting, there isn't just the problem of TID. Slew rate limiting imposes a phase lag. This means an amplifier with global neg feedback that is marginally stable, and works well most of the time, may briefly oscillate for a few cycles after a large amplitude pulse. If an amp is oscillating (may well be supersonically and inaudible) its too darn busy to pass signal.
If the amp undergoes slew rate limiting, there isn't just the problem of TID. Slew rate limiting imposes a phase lag. This means an amplifier with global neg feedback that is marginally stable, and works well most of the time, may briefly oscillate for a few cycles after a large amplitude pulse. If an amp is oscillating (may well be supersonically and inaudible) its too darn busy to pass signal.
With those witches it was what was inside (their heads) that was alarming. In wires it also can be an iron core that is clad (saves money . . .) I also have an alps potmeter, the high end Black expensive part, where the lugs are made of clad iron. You bet I am following that in my own witch hunt. [My sound gradually deteriorated as I bought and installed some new components. Now I imagine I am fixing things.] It is well known there is contact resistance and a dielectric effect for dissimular connector metal parts like in an RCA plug. I have an old Revox amp with rhodium plated female plugs - should be connected with ditto plated plugs. Not gld plated ones. Recently I got a set of old Philips parts, Rhodium plated wires, a knightmare to solder after 50 years, my friend her father ran a Ph tube factory
About transformer coupled devices, what about the hysteresis, saturation and other effects of transformers?
I understand permalloy is almost the only one having a good behavior with a very high permeability leading to excellent low (<5Hz is common) and high (>100KHz is often seen) transformers that can be overdriven (>+10dB) but is is mostly non-DC, at most push-pull.
Is amorphous core (metal glas) the second best?
Hysteresis causes a power loss. Simple cures - either use a bigger core and/or bigger airgap, or just a bigger output power stage.
Saturation causes distortion of bass frequencies , and must be managed by the techniques I have already explained (more than once)
Other transformer effects affect frequency response and power loss.
Transformers are like pleasure yachts - the more money you spend, the bigger they get and the better the performance - but there is a law of diminishing returns. Unlike yachts, the fewer the transformers you have the better.
I think if you overload the amp the NF will try and drive the output stages even harder to correct the issue causing the grid to go even more positive. This will make the blocking worse.
As for transformers verses Yachts, transformers don't need continuous maintenance and money spending on them to keep them working.
As for transformers verses Yachts, transformers don't need continuous maintenance and money spending on them to keep them working.
As I said before, a competent designer of any tube amplifier that is claimed to be a quality sound amplifier does things so that grid blocking is minimal or does not occur. It's not hard. Those of us who did it professionally did it as routine.
Its a matter of ensuring that the preceding driver swing is not much greater than the required peak-to-peak drive of the grid, using appropriate values of grid resistor, etc. Use clamp diodes on the preceding anode if you need to, most of us did not.
In any amplifier that is claimed to provide quality sound, grid blocking should not significantly occur, regardless of whether it has global negative feedback or not, though you are correct in that feedback will make a bad amplifier worse in this respect.
Australian Neville Thiele was a very competent design engineer in the valve era. It always amuses me that his writings on everything except speakers has been ignored by the hi-fi cognoscenti, apparently because only his speaker paper was reprinted in a certain American magazine. We in Australia knew about it years before.
In one of his papers on amplifier design, he related how he eliminated audible distortion from grid blocking in a new amplifier (a clone of the Quad amplifier using low cost TV triode pentodes) by just reducing the value of a grid resistor. He demonstrated the difference to his totally cloth-eared but financially keen boss, who then thought the sun shone out of Thiele's fundamental orifice.
Last edited:
Keit, you say there is a law of diminishing returns in OT...
Ready to apply this to tubes,
with transformers my experience speaks for the contrary.
I experienced the most dramatic sound changes with premium output transformers in the 600+
I would just advice to not fall into the extremes such as this thread suggests. Sure a good transformer coupled design with wide bandwidth could play better than simple capacitor couplings....
the reason behind that is that the tubes are being choke loaded, gain is maximized, thd is at its lowest of the tube operation with transformers. The cost is very big and using overpowered drivers and input tubes should give similar results.
Ready to apply this to tubes,
with transformers my experience speaks for the contrary.
I experienced the most dramatic sound changes with premium output transformers in the 600+
I would just advice to not fall into the extremes such as this thread suggests. Sure a good transformer coupled design with wide bandwidth could play better than simple capacitor couplings....
the reason behind that is that the tubes are being choke loaded, gain is maximized, thd is at its lowest of the tube operation with transformers. The cost is very big and using overpowered drivers and input tubes should give similar results.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.