I agree - the overpriced stuff is hardly state of the art ... a lot of what goes on reminds me of someone with a car which is so heavy that the chassis drags on the ground - has no performance whatsoever - so they double the size of the engine, and add an extra turbocharger, as a "brilliant solution" ... I shake my head at most of this ...
I have noticed that in transistor thermal impedance charts like in OnSemi's BC327 datasheets, the curve often follows a 3db/oct curve like skin effect. Is thermal diffusion nonlinear like skin effect or can it be modeled well using a uniform RC line?
skin effect, lossy transmission lines, DA are usefully modeled as Linear Time Invariant - LTI systems don't produce any harmonics of the excitation frequencies
they are distributed/continuous though and show fractional powers in Laplace s domain descriptions instead of the integer powers in s seen in discrete linear systems
a discrete "pink noise" filter approximates a 1/s^(1/2) curve
they are distributed/continuous though and show fractional powers in Laplace s domain descriptions instead of the integer powers in s seen in discrete linear systems
a discrete "pink noise" filter approximates a 1/s^(1/2) curve
Last edited:
My opinion has been formed from what I have seen all these years of poking into audio related equipment.
Dissecting consecutive generations of TV sets and portable audio gadgets have shown me that the forward trend is there, years before it is adapted for higher power and repackaged into the larger enclosures for the audio fraternity.
DSP power has unleashed the powers from decades of systematic research on how our brains processes sound, on how we adapt to and perceive an aural environment.
If I have to assign a R&D activity to audio sector outside those lines, my vote would go to Nelson Pass for his new active semiconductor element.
George
George is right take ESS for example, they are making DAC's for phones now.
Heat flow is described by diffusion equations, the wiki is usually not bad for maths I find. Heat equation - Wikipedia, the free encyclopedia
You can build up a simple model where voltage = temperature, r= thermal resistivity, current = heat flow, and capacitance = thermal capacity. You need lots of finite elements like jcx's pink noise filter unless you simply want a final steady state answer, i.e. a transistor at X Watts and a heat sink at Y degrees C per Watt.
Last edited:
Right
And to shoot myself in the foot, I sat and listened (headphones) two days in the row for hours a bunch of CDs with string sonatas and concerti through an ancient SAE D102 (2x TDA1540P, x4 oversampling, 4xLM833) that I brought back into life.
I had to give it back though 🙂
George
Attachments
Last edited:
Just thought I'd offer the Bybees for the good of science 🙂 It seems a shame to have them collecting dust... Thanks for posting the Coto relay info. I have some of their 2911 series (collecting dust) that I've been wondering if I could use to switch cartridge gain and loading with. I might give them a try...
Mike
I think I read in some audio publication about measuring resistor distortion. Wonder if that will have any effect on any other products? 🙂
Exactly. Hi-end may improve, but not from their 'cutting edge R&D'- rather from basic, honest engineering by those that are sneered at by the hi-end crowd.
Jan
Some answers to direct and indirect questions previously placed here.
http://www.cotorelay.com/downloadables/Application%20Notes/Coto%20Technology%209852%20Best%20Practices.pdf
http://www.cotorelay.com/downloadables/Application%20Notes/Reed%20Relay%20Tech%20&%20App%20Section%20(all%20files)/Contact%20Resistance%20&%20Dynamics.pdf
http://www.cotorelay.com/downloadables/Application%20Notes/Reed%20Relay%20Tech%20&%20App%20Section%20(all%20files)/Reed%20Relay%20RF%20Parameter%20Measurements.pdf
Is this ‘basic, honest engineering’ Jan? 🙂
George
http://www.cotorelay.com/downloadables/Application%20Notes/Coto%20Technology%209852%20Best%20Practices.pdf
http://www.cotorelay.com/downloadables/Application%20Notes/Reed%20Relay%20Tech%20&%20App%20Section%20(all%20files)/Contact%20Resistance%20&%20Dynamics.pdf
http://www.cotorelay.com/downloadables/Application%20Notes/Reed%20Relay%20Tech%20&%20App%20Section%20(all%20files)/Reed%20Relay%20RF%20Parameter%20Measurements.pdf
Is this ‘basic, honest engineering’ Jan? 🙂
George
Thermal time constants are in the area of:
junction to chip attach:10 to 100 microseconds
to bottom of package through copper: 10 to 100 milliseconds
into heatsink 1 to 10 seconds.
to wit:
Apply a high power step dissipation to the chip. Within 100 microseconds, the chip top temperature will roughly stabilize with respect to it's bottom. At this time, the heatsink does not exist to the chip, but it's package bottom is starting to be seen. This regime is used to test the chip for anomalies in design, diffusion, and wirebonding.
within 100 milliseconds, the package heat capacity starts to fill, the die continues to rise in temp, but tracking directly the top of the package where the die is attached. This regime is used to test the chip to package attach integrity, looking for cold or hot voids.
A cold void is when no current is flowing from below, so that area of the die does not dissipate. Instead, the current will crowd the rest of the active element. A bipolar will use it's active gain to make matters worse, a diode will use it's negative tempco to make it worse. A mosfet will depend on where in it's current range it is being powered, either above or below it's zero tempco current.
A hot void is when current in the silicon can traverse horizontally over the void so the chip dissipates without a direct heat path where the void is. Bipolars and diodes will thermally runaway in this hotspot.
At 10 seconds, you are testing the package to heatsink capability.
If you model it as three stages of RC using those time constants, you can get an idea of what is happening. The time constants for any specific device will depend on the chip thickness, it's length and width, the die attach metal and thickness, the baseplate material and thickness, how the heatsinking is accomplished.
The transient pulse graphs in the IR book show this effect, although their presentation is a tad confusing to most. At very high speed, they use the silicon thermal mass, at lower speeds they use the package base mass, then the graph eventually settles down to steady state thermal response, which is the spec'd number in the datasheet normalized.(IIRC)
jn
Thanks.
Is this to confirm that a discrete amplifier’s bias compensation scheme based on temp sensing the intimate exterior of a power device is lagging at least by 1sec in response, while within a power op amp chip this time lag is ~ x1000 shorter?
George
Hi Wayne and Nelson. Yes, you got a great review in TAS. You will give some other designers (including me) a real run in comparative sonic quality.
Of course, that is the point. The CTC Blowtorch was a challenge for us to do everything that we could to make the best preamp possible, and we added nothing extra that would have given us more convenience, but might change the sound quality. In my last years, it will be good enough for me, but all bets are off for amps and phono stages, digital, etc. They can almost always be improved.
Of course, we can't share details with each other, but I presume that you also find the complementary jfets, especially from Toshiba, are the best in the business for many applications, just like Charlie Hansen and me. We will have to see what falls out through the reviewers, mainly, but I am glad that I am not in direct competition with you at this time.
On another subject, does anyone else out there own a Stanford Research SR-1? I am having trouble getting the most out of it. Love it though.
Of course, that is the point. The CTC Blowtorch was a challenge for us to do everything that we could to make the best preamp possible, and we added nothing extra that would have given us more convenience, but might change the sound quality. In my last years, it will be good enough for me, but all bets are off for amps and phono stages, digital, etc. They can almost always be improved.
Of course, we can't share details with each other, but I presume that you also find the complementary jfets, especially from Toshiba, are the best in the business for many applications, just like Charlie Hansen and me. We will have to see what falls out through the reviewers, mainly, but I am glad that I am not in direct competition with you at this time.
On another subject, does anyone else out there own a Stanford Research SR-1? I am having trouble getting the most out of it. Love it though.
Last edited:
JCX, Jneutron said skin/proximity effect caused a 2nd harmonic, and this seems to conflict with your statement.
One reason I asked is that for the graphs on OnSemi's BJT datasheets, it's SOOOO much easier to use the URC spice element to model thermal impedance. Surely I'm not the only one who's noticed how hard it is to do manually with RC's? Of course this probably won't extend as well to weird thermal curves supplied by Vishay with their FET thermal models.
Is it that a thermal impulse spreading in a uniform material spreads as square root of time or something like this, and this is why TO92 thermal impedance tends to have a large 3db/oct region? And if true, that the higher slope in the 100uS region would be due to having a flat die, where one dimension of thermal expansion was effectively removed? And also, that a slope of less than 3db/oct would be caused by an interface from a less thermally conductive material to a more thermally conductive material?
One reason I asked is that for the graphs on OnSemi's BJT datasheets, it's SOOOO much easier to use the URC spice element to model thermal impedance. Surely I'm not the only one who's noticed how hard it is to do manually with RC's? Of course this probably won't extend as well to weird thermal curves supplied by Vishay with their FET thermal models.
Is it that a thermal impulse spreading in a uniform material spreads as square root of time or something like this, and this is why TO92 thermal impedance tends to have a large 3db/oct region? And if true, that the higher slope in the 100uS region would be due to having a flat die, where one dimension of thermal expansion was effectively removed? And also, that a slope of less than 3db/oct would be caused by an interface from a less thermally conductive material to a more thermally conductive material?
I need to see his equations for an example physical arrangement and the calculation of the magnitude of the nonlinearity - he hasn't shown that or linked to a good example yet
lacking that I have great confidence in the fact that EM, Maxwell's equations are linear - in vacuum up to ridiculous energy density
in materials the situation is a little more complex, but for good insulators (not piezo, ferro electrics), metallic conductors with only para/diamagnetism (not ferromagnetic) EM is very linear - copper, polypropylene and air for an instance in many cables
there is Lorentz Law Ordinary Mangetoresistance Magnetoresistance - Wikipedia, the free encyclopedia - seen with the Corbino disk - which is not at all a large effect even with the geometry optimized to see it when again using nominally linear, nonmagetic metallic conductors
and in skin effect in normal wires the symmetry even eliminates that concern
Parseval's Theorem allows for linear systems to show frequency dependent loss without nonlinearity
I can't find any discussion of skin effect nonlinearity in DSL cable analysis - where miles of wire, much put in place decades before for ordinary voice now carries 100s of kHz with a need for extreme low nonlinearity to prevent IMD from wiping out S/N in at least 256 symbol QAM - you would think they would care, talk about it if it can be easily seen, measured
Last edited:
- Status
- Not open for further replies.