http://home.comcast.net/~mnjmiller/mosamp.gifCharles Hansen said:
I have stubstituted a 2SK1058 for the 2SK214 in this cicuit with pretty good results
Maybe with the Semelabs ?
Sorry guys, I have made a mistake.
I have made a factor of ten error - never post numbers done solely by mental arithmetic. After checking (and cross checking against some numbers from NIST), the thermal diffusivity of Al, at room temperature, is about 10^-4 metre squared per second, or more conveniently, 1 cm^2 per second.
So the time scale for 1 cm length is about 1s, and 10cm size heatsink fin would have a 100s time scale.
I guess the other important thing to point out is that there are prefactors that depend on the geometry of the problem. 3D heatspread from a small source into a nearly infinite mass is much faster than a 1D problem along the length of a rod.
This still means that temperature changes on musical time scales (seconds and faster) will not be isothermal on even small heatsinks.
I have made a factor of ten error - never post numbers done solely by mental arithmetic. After checking (and cross checking against some numbers from NIST), the thermal diffusivity of Al, at room temperature, is about 10^-4 metre squared per second, or more conveniently, 1 cm^2 per second.
So the time scale for 1 cm length is about 1s, and 10cm size heatsink fin would have a 100s time scale.
I guess the other important thing to point out is that there are prefactors that depend on the geometry of the problem. 3D heatspread from a small source into a nearly infinite mass is much faster than a 1D problem along the length of a rod.
This still means that temperature changes on musical time scales (seconds and faster) will not be isothermal on even small heatsinks.
This is a good point. We often see quite large heatsink (like 440mm x 130mm) with 40 wings or so, but quite thin base like 4 - 8mm. Though such a heatsink has quite low thermal resistance, it cannot be utilized effectively, due to restricted heat flow through the base. Provided there are 2 or 4 transistors mounted, some areas of such a heatsink would always stay "cold" and another areas "hot".
PB2 said:
Hi Pete,
I understand your question very well, but it is hard to answer, as my circuit is a different kettle of fish. Moreover, I wouldn't consider a system with a thermal sensor mounted on the heat sink as a golden standard. D. Self has written three articles about this subject in EW, 1996. Did you read them?
Anyhow, the influence of the feedback signal is neglectable, as it is not only very small, but added as a common mode voltage to the O/P stage. I have run a few THD sims at 20 Hz and 20kHz with and without bias feedback and the differences were insignificant. Also, I increased the loop BW by decreasing C3 to 100pF, which had no marked effects on the THD. Maybe this is only part of the story as I didn't look at IMD figures, for example. However, Marcel van de Gevel has built a comparable bias feedback system with a bandwidth far beyond the audio spectrum and he also reported good results.
I hope this answer will be adequate.
Cheers, Edmond.
PS1: I added the schematic to make this discussion understandable for other readers. More people should do this!
PS2:"Did you try .1 or even lower bandwidth?" No, why should I?
And here's a link to Marcel's schematic.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=101745&perpage=25&highlight=&pagenumber=47 post #1173
Attachments
estuart said:
Hi Edmond,
Thanks for the other links and your input.
I agree the thermal system is not the gold standard, however when the claim is made that the thermal system has lower distortion, my view is that a good experiment is to degrade the thermal performance to match the system with lower distortion, and then see if the two are then similar.
You say that the feedback was not the source of the distortion, and therefore it seems that we'd have to look elsewhere but then this feedback does not seem to be the fundamental problem.
No, I did not read D. Self's articles.
It seems that Re sensing has potential as a good solution.
There is a fundamental difference, it seems that the heat sink based sensor, obviously, measures temperature and then estimates the correct bias voltage to maintain the correct idle current. It functions "correctly" based on an estimate of the devices thermal characteristics even when the amp is amplifying a signal. Monitoring idle current on the other hand, operates correctly without a signal present, but usually not, depending on the design with a signal.
Thanks again Edmond,
Pete B.
hitsware said:I have stubstituted a 2SK1058 for the 2SK214 in this cicuit with pretty good results
There are three disadvantages of this substitution, some small, some large:
a) The bigger part is more expensive.
b) The bigger part has higher interelectrode capacitances, and hence a lower bandwidth.
c) The small part has a zero tempco point at around 15 - 20 mA, which is probably close to the actual idle current. The large part has a zero tempco point at around 100 - 120 mA, which will create a lot of DC drift in this application.
The N-channel Hitachi parts are fine. It's the P-channel parts that are not so great.
autobias of mosfets
You're welcome.
Hi Pete,
I'm not sure what you mean by: "when the claim is made that the thermal system has lower distortion" ? A system with a temperature sensor? If so, compared to what other system and who made this claim? Please, could you be more specific?
True, I did say that, but that was only according to a simulation with a constant junction/channel temperature. One run without RE sensing and a fixed bias voltage, and a second run with RE sensing enabled, in which case the THD was unchanged, This THD BTW, was largely created by the VAS and O/P stage, so not by the bias feed back system.
If you are interested, I'll try to get copies in electronic form.
Absolutely, what else could I say?
True, it's just an estimate. Such system is inaccurate because of thermal "attenuation" and "delay". The problem is that a quiet passage of music, proceeded by a prolonged high power burst, will result in (temporary) under bias (and more distortion).
For this reason, I think that a faster bias FB loop is in advantage. I can't prove this, as measuring such effect is rather difficult.
The feed back of my system is still active (and correctly working ) in the presence of a signal, unless it exceeds (temporarily) a certain level. In case of the latter, the integrator "remembers" the correct bias voltage.
It should be noticed that my scheme only works with vertical MOSFETs, no BJT's, neither the (f-word) P-channels from IR.
Whether it works with lateral MOSFETs, I don't know. I didn't try, as these devices are already thermally stable.
Thank you for your interest.
Cheers, Edmond.
PB2 said:
You're welcome.
PB2 said:
Hi Pete,
I'm not sure what you mean by: "when the claim is made that the thermal system has lower distortion" ? A system with a temperature sensor? If so, compared to what other system and who made this claim? Please, could you be more specific?
PB2 said:
True, I did say that, but that was only according to a simulation with a constant junction/channel temperature. One run without RE sensing and a fixed bias voltage, and a second run with RE sensing enabled, in which case the THD was unchanged, This THD BTW, was largely created by the VAS and O/P stage, so not by the bias feed back system.
PB2 said:
If you are interested, I'll try to get copies in electronic form.
PB2 said:
Absolutely, what else could I say?
PB2 said:
True, it's just an estimate. Such system is inaccurate because of thermal "attenuation" and "delay". The problem is that a quiet passage of music, proceeded by a prolonged high power burst, will result in (temporary) under bias (and more distortion).
For this reason, I think that a faster bias FB loop is in advantage. I can't prove this, as measuring such effect is rather difficult.
PB2 said:
The feed back of my system is still active (and correctly working ) in the presence of a signal, unless it exceeds (temporarily) a certain level. In case of the latter, the integrator "remembers" the correct bias voltage.
It should be noticed that my scheme only works with vertical MOSFETs, no BJT's, neither the (f-word) P-channels from IR.
Whether it works with lateral MOSFETs, I don't know. I didn't try, as these devices are already thermally stable.
PB2 said:
Thank you for your interest.
Cheers, Edmond.
>a) The bigger part is more expensive.
God save us from that.
>b) The bigger part has higher interelectrode capacitances, and hence a lower bandwidth.
To some extent.
>c) The small part has a zero tempco point at around 15 - 20 mA, which is probably close to the actual idle current. The large part has a zero tempco point at around 100 - 120 mA, which will create a lot of DC drift in this application.
The large source resistor (1K in the example)
negates that concern.
I'm saying if that one (for some reason)
wanted a single xsistor front end for a
complementary pair of outputs ......
using one of the output devices as the front end
(or better yet a N and P(one for each channel)
It would lower # of parts involved.........
God save us from that.
>b) The bigger part has higher interelectrode capacitances, and hence a lower bandwidth.
To some extent.
>c) The small part has a zero tempco point at around 15 - 20 mA, which is probably close to the actual idle current. The large part has a zero tempco point at around 100 - 120 mA, which will create a lot of DC drift in this application.
The large source resistor (1K in the example)
negates that concern.
I'm saying if that one (for some reason)
wanted a single xsistor front end for a
complementary pair of outputs ......
using one of the output devices as the front end
(or better yet a N and P(one for each channel)
It would lower # of parts involved.........
Re: autobias of mosfets
Hi Edmond,
I probably misunderstood you when you responded to this:
John Curl wrote:
"It has been tried by LT, but the chip failed to make low high frequency distortion at 20KHz, so necessary for THX. I wasted months of time."
with this response:
"If you had used a simulator (as I did), you had only wasted one hour."
I take it now, based on your article that you're autobias design has acceptable distortion performance, and I misread you previously. I also didn't notice that this was previously discussed in this very thread, I should have searched first.
Pete B.
estuart said:
Hi Edmond,
I probably misunderstood you when you responded to this:
John Curl wrote:
"It has been tried by LT, but the chip failed to make low high frequency distortion at 20KHz, so necessary for THX. I wasted months of time."
with this response:
"If you had used a simulator (as I did), you had only wasted one hour."
I take it now, based on your article that you're autobias design has acceptable distortion performance, and I misread you previously. I also didn't notice that this was previously discussed in this very thread, I should have searched first.
Pete B.
Re: Re: autobias of mosfets
Hi Pete,
I have already sent them to: basconsultants@aol.com
Is this address the correct one?
Cheers, Edmond.
PB2 said:
Hi Pete,
I have already sent them to: basconsultants@aol.com
Is this address the correct one?
Cheers, Edmond.
Re: Re: autobias of mosfets
Hi Pete,
Never mind. As for distortion, I would like to go one step further. The autobias doesn't add any distortion, that is, under steady state conditions. How it behaves under thermal transient conditions in real life, I don't know (as I said before, difficult to measure), but I do know that at 20Hz and skipping the first half cycle, the simulated THD wasn't affected by the bias circuit.
Cheers, Edmond.
PS: To get correct results from a simulation, one needs the BSIM3.3 (or higher) models of the vertical MOSFETs, as lower level models doesn't cover the weak inversion.
PB2 said:
Hi Pete,
Never mind. As for distortion, I would like to go one step further. The autobias doesn't add any distortion, that is, under steady state conditions. How it behaves under thermal transient conditions in real life, I don't know (as I said before, difficult to measure), but I do know that at 20Hz and skipping the first half cycle, the simulated THD wasn't affected by the bias circuit.
Cheers, Edmond.
PS: To get correct results from a simulation, one needs the BSIM3.3 (or higher) models of the vertical MOSFETs, as lower level models doesn't cover the weak inversion.
Estuart, perhaps the simulation did not show you what we found in reality. We found that we could not use the chip and keep the high frequency THD low. I contacted LT and they admitted to the problem. This was about 15 years ago. We could NOT use the chip directly, or modify our circuits sufficiently, to use the chip, and meet THX specs.
This is REALITY, not some simulation.
This is REALITY, not some simulation.
john curl said:
Hi John,
Reality in the test lab is certainly what counts at the end of the day. No argument there. I wonder what simulation model of the LT1166 Estuart was using. I'm not aware of one being available.
It would be fun and instructive to play with a sim of that part if one was available.
If I recall correctly, the LT1166 was designed for use with MOSFET output power transistors (although I would not rule out its use with BJTs in the right arrangement). Was your experience with the LT1166 with MOSFETs or with BJTs?
Thanks,
Bob
Hi John,
Perhaps, what you fond wasn't exactly the same as I found, but our end conclusions are just the same, as I also found that the chip (LT1166) contributed too much to distortion at high frequencies. So, I'm glad to learn from you that your realities agree with with my simulations.
BTW and off-topic, last weekend I saw a beautiful coil, and while thinking of you -yes, I know you don't like output coils- I couldn't resist the temptation to make a picture of it. In the event you're designing a 100kW amp and need an inaudible coil, look at this, as it is rock solid mounted on the ground plane by means of concrete.
Cheers, Edmond.
