RF amplifiers have less efficient local feedback due to group delay
being variable not only with frequency but also with the transistor s
moving parameters in function of temperature.
Negative feedback efficency is inherently dependant of the output s
phase value.
Idealy , it is 180° , but if phase delay increase , say to 270° ,
NFB efficency will be drasticaly reduce and linearity will decrease.
In audio apps nothing of the sort as the dynamic non linearity
evolution is way slower than the NFB loop , so the 10khz signal
has nothing to be modulated by/with.
Besides , one has to look closely at the phenomenon , as it was
too much rapidly branded as being of negative influence.
Let s admit than a pulse rise the device temperature.
Although with a significant delay , the transistor see its beta
increasing as well as the zero crossing standing current also
increased , assumed an eventual thermal tracking is dependant
of the main heatspreader.
Would such parameters variations be negative in respect
of the amp linearity ? I would find it unlikely.
being variable not only with frequency but also with the transistor s
moving parameters in function of temperature.
Negative feedback efficency is inherently dependant of the output s
phase value.
Idealy , it is 180° , but if phase delay increase , say to 270° ,
NFB efficency will be drasticaly reduce and linearity will decrease.
In audio apps nothing of the sort as the dynamic non linearity
evolution is way slower than the NFB loop , so the 10khz signal
has nothing to be modulated by/with.
Besides , one has to look closely at the phenomenon , as it was
too much rapidly branded as being of negative influence.
Let s admit than a pulse rise the device temperature.
Although with a significant delay , the transistor see its beta
increasing as well as the zero crossing standing current also
increased , assumed an eventual thermal tracking is dependant
of the main heatspreader.
Would such parameters variations be negative in respect
of the amp linearity ? I would find it unlikely.
So based on some typical datasheets along with experiences and other references I dont see why there wouldn't be temperature modulation with signal.
I would expect to see some rectified droop type variation of the die temperature at 10Hz.
Not commenting on the audibility or if its a dominate source when such signals are present.
Might it be easier to just single pulse an amp while looking at the input to the output stage recovery?
Thanks
-Antonio
I would expect to see some rectified droop type variation of the die temperature at 10Hz.
Not commenting on the audibility or if its a dominate source when such signals are present.
Might it be easier to just single pulse an amp while looking at the input to the output stage recovery?
Thanks
-Antonio
No,no! Thermal feedback, something like Vbe multiplier, adds some additional details to the interesting picture!
It would be great if we could test separately influences of Kirk, Early, thermal effects, base width, Ft modulation, beta dependence on current and temperature, Vcbo, capacitances, and so on, one by one on the surgeon bench. Then put a vacuum triode there and check do they exist at all.
I am downloading gEDA now, let's see how convenient it is, after I build it for Scientific Linux (Fedora-like)
Sy,
I completely grasp it. And if you have IR equipment with a sufficient resolution and speed (some USB IR Cameras do) you can see thermal modulation in Plastic Transistors on heat sinks without opening them up.
BTW, the presence of thermal memory in typical Op-Amp circuits was confirmed previously by non other than Scott Wurcer (#769), though in his case (and in real circuits) the thermal modulation (or memory) mainly applied to Input and VAS transistors...
I find you continued denial of the actual existence of these effects, as well you making up experimental proof that never existed somewhat baffling, but I am sure you have your reasons...
Ciao T
I completely grasp it. And if you have IR equipment with a sufficient resolution and speed (some USB IR Cameras do) you can see thermal modulation in Plastic Transistors on heat sinks without opening them up.
BTW, the presence of thermal memory in typical Op-Amp circuits was confirmed previously by non other than Scott Wurcer (#769), though in his case (and in real circuits) the thermal modulation (or memory) mainly applied to Input and VAS transistors...
I find you continued denial of the actual existence of these effects, as well you making up experimental proof that never existed somewhat baffling, but I am sure you have your reasons...
Ciao T
Hi,
Yes, in other words, music.
There are differences when running devices hot (e.g. hot transistors have higher beta, which reduces loading distortion in the circuit) so a circuit running at elevated temperature will likely measure different than one run cold. There are as remarked also other factors that are not obviously electronic in nature.
If we take these out of the equation however, the difference in thermal load is likely to introduce very similar levels of parameter modulation no matter at what absolute temperature you start, the relative load change and relative temperature change will be the same.
It is however non-trivial to separate out all these (and other) thermal effects and as remarked, in terms of measured distortion they are often swamped out by many factors...
Ciao T
Yes, in other words, music.
There are differences when running devices hot (e.g. hot transistors have higher beta, which reduces loading distortion in the circuit) so a circuit running at elevated temperature will likely measure different than one run cold. There are as remarked also other factors that are not obviously electronic in nature.
If we take these out of the equation however, the difference in thermal load is likely to introduce very similar levels of parameter modulation no matter at what absolute temperature you start, the relative load change and relative temperature change will be the same.
It is however non-trivial to separate out all these (and other) thermal effects and as remarked, in terms of measured distortion they are often swamped out by many factors...
Ciao T
Hi,
Appypollylogy.
There was 'sposed to be a link with this:
by non other than Scott Wurcer (#769)
Ciao T
Appypollylogy.
There was 'sposed to be a link with this:
by non other than Scott Wurcer (#769)
Ciao T
Last edited:
Pano,
Yes. The key is in this (italics for emphasis added by me):
In this case the point is that the settling time, which is specified incidentally in hundreds of nanoseconds for the part being discussed, being disturbed by thermal effects and Scott notes that these thermal effects together with other issues will compromise the performance of modern DAC's...
Ciao T
Yes. The key is in this (italics for emphasis added by me):
In this case the point is that the settling time, which is specified incidentally in hundreds of nanoseconds for the part being discussed, being disturbed by thermal effects and Scott notes that these thermal effects together with other issues will compromise the performance of modern DAC's...
Ciao T
According to this graph from wahab's datasheet, if the device starts to dissipate 30W of heat, it takes only 10mS for the junction temperature to rise by 10 degrees.
So in a class B (or AB) amp reproducing a 50Hz square wave and dissipating 30W alternately in the two output devices, I would expect the junction temperatures to be fluctuating by about 10 degrees at 50Hz in a roughly sawtooth fashion. This is apart from the slow rise in temperature as the heatsink warms up.
Debate about the significance could drag on indefinitely, but can we at least agree that this kind of temperature fluctuation with signal is possible?
So in a class B (or AB) amp reproducing a 50Hz square wave and dissipating 30W alternately in the two output devices, I would expect the junction temperatures to be fluctuating by about 10 degrees at 50Hz in a roughly sawtooth fashion. This is apart from the slow rise in temperature as the heatsink warms up.
Debate about the significance could drag on indefinitely, but can we at least agree that this kind of temperature fluctuation with signal is possible?
Attachments
I don't know if you are referring to this, but when monitoring bias on optimally biased
class-ab amplifiers, i noticed the following.
When feeding the amp with a large-signal stimulus (approaching clipping on the output attached to a dummyload),
immediately after removing the stimulus the measured bias could easily be twice the optimum value.
Under these (severe) conditions imo you can hardly say that the amp is “optimally biased”.
In a conventional arrangement (vbe-multiplier on main heatsink)
the vbe-multiplier does not accurately track the temperature of the die of the outputs.
Well, yes, but I was wondering more about a certain chemist.
Needs evidence, since the expected rise in distortion with lowered frequency doesn't seem to be observed.
Note that we're talking about power amplifiers with discrete output devices having considerable thermal mass- bringing in tiny opamps is a red herring.
How can you observe a rise in distortion below the LF rolloff of the amp? This effect, if present, would be triggered by the signal envelope so that is where we should be looking.
Stuart, you are beginning to sound like a politician: "There is no evidence for blah-blah". Instead of simply asserting the absence of evidence why not work with us to agree how the evidence might be found if the effect is real?
OK, here is a suggestion (I don't have the equipment to do this myself):
Feed in an audio signal consisting of two signals at, say, 1kHz and 1.001kHz. Measure the third harmonic and the output device chip temperature. Look for a correlation as the frequency difference (1Hz) is varied. The idea is that temperature changes will affect crossover distortion, and that will show itself as third-order distortion which changes with signal envelope.
Stuart, you are beginning to sound like a politician: "There is no evidence for blah-blah". Instead of simply asserting the absence of evidence why not work with us to agree how the evidence might be found if the effect is real?
OK, here is a suggestion (I don't have the equipment to do this myself):
Feed in an audio signal consisting of two signals at, say, 1kHz and 1.001kHz. Measure the third harmonic and the output device chip temperature. Look for a correlation as the frequency difference (1Hz) is varied. The idea is that temperature changes will affect crossover distortion, and that will show itself as third-order distortion which changes with signal envelope.
The time constants being mentioned (e.g., by godfrey) would cause distortion well above the LF cutoff of most amps.
Your suggested experiment is quite interesting and I would love to see the results of it.
Hi,
It may be expedient to first review the extant body of work on the subject.
Jan Didden at his linearaudio.nl site has a number of excellent articles on a number of controversial audio topics, I presume of course with full agreement of the copyright holders.
They do include a number on thermal distortion in solid state amplifiers inlcuding an AES Preprint on Thermal distortion induced by temperature changes in transistor output stages (the very existence of such an AES paper argues that there are signal related changes in output transistor temperature)..
Ciao T
It may be expedient to first review the extant body of work on the subject.
Jan Didden at his linearaudio.nl site has a number of excellent articles on a number of controversial audio topics, I presume of course with full agreement of the copyright holders.
They do include a number on thermal distortion in solid state amplifiers inlcuding an AES Preprint on Thermal distortion induced by temperature changes in transistor output stages (the very existence of such an AES paper argues that there are signal related changes in output transistor temperature)..
Ciao T
It may be worthwhile for you to review exactly what we're talking about so as to keep focused on the question at hand rather than bringing in other issues related to thermal effects.
Since you mentioned Jan's site, it might be worthwhile for you to get a copy of his Baxandall/Self reprint book, with previously unpublished commentary by Baxandall. The thermal distortion issue that we're actually discussing here is discussed extensively. Self points out that the thermal mass of a silicon chip used in power transistors causes enough thermal inertia to eliminate thermal modulation as a significant distortion mechanism.
Since you mentioned Jan's site, it might be worthwhile for you to get a copy of his Baxandall/Self reprint book, with previously unpublished commentary by Baxandall. The thermal distortion issue that we're actually discussing here is discussed extensively. Self points out that the thermal mass of a silicon chip used in power transistors causes enough thermal inertia to eliminate thermal modulation as a significant distortion mechanism.
- Status
- Not open for further replies.