mikek said:Actually...i have just discovered that the mjl1302 model has been recently updated by on-semi.....
sam9,
My thought here was that a good model of a similar device would be better than a bad model of the same device.
mikek,
One thing I noticed in simulation with the earlier On Semiconductor models of these devices was that with a class B complementary EF circuit with pure voltage source drive, the output inductance (of just the 2-transistor EF itself with no global feedback or any other additonal circuitry) became significant above 100 kHz. When simulating the performance with a load of 8 Ohms in parallel with 2 uF, there was a big-time resonance at this frequency, with lots of frequency response peaking and very sudden phase variation with frequency. Yet with the models for the Toshiba devices, the output inductance wasn't significant until above 3 MHz or so. This is a huge difference. Have you done any such experiments?
"
My thought here was that a good model of a similar device would be better than a bad model of the same device."
Now I just have to be careful that I don't start thinking that the model that shows best performance in the sim is the the best reflection of the real-world. I don't know about others but wishful thinking is a trap I can spring on myself sometimes.
My thought here was that a good model of a similar device would be better than a bad model of the same device."
Now I just have to be careful that I don't start thinking that the model that shows best performance in the sim is the the best reflection of the real-world. I don't know about others but wishful thinking is a trap I can spring on myself sometimes.
A modest proposal on models
If you are going to compare model parameters the first thing I would suggest is to have some understanding of what they actually represent:
http://newton.ex.ac.uk/teaching/CDHW/Electronics2/userguide/sec3.html#3.4.4
I see some very different values between the models including a couple of default values that are a couple of orders of magnitude different that actual values given in other models.
The Toshiba and ON semi parts should be fairly close to each other if they share the same base number. Parts with the same base part number are supposed to be second sources. This allows designers and purchasing to have several choices when building products in the real world and vendors to compete in the market place. My examination of the data sheets seem to indicate that they indeed do have very similar specs as one would expect for second source parts.
There are a very large number of poor Spice models out there and an understanding what the models mean is strongly advised. Comparing the model to the Data sheets and measured results is essential if you are to get any useful output from a Spice simulation. It is the difference between computer aided design and computer aided guessing. Of course as some wag (and I wish I remembered who so as to give him proper credit) "negative feedback should make component differences largely negligible anyway."
Fred
As always, please feel free to ignore my advice. There of several of us (and you guys know who you are!) that would miss the entertainment value of the usual free for all.
PS The Toshiba Power BJT models that came with the $3000 simulator I use are different from those that have been posted here. I dare not post them, as I would be accused, as usual, of attempting to increase the entertainment level at the distress of many on the forum.
If you are going to compare model parameters the first thing I would suggest is to have some understanding of what they actually represent:
http://newton.ex.ac.uk/teaching/CDHW/Electronics2/userguide/sec3.html#3.4.4
I see some very different values between the models including a couple of default values that are a couple of orders of magnitude different that actual values given in other models.
The Toshiba and ON semi parts should be fairly close to each other if they share the same base number. Parts with the same base part number are supposed to be second sources. This allows designers and purchasing to have several choices when building products in the real world and vendors to compete in the market place. My examination of the data sheets seem to indicate that they indeed do have very similar specs as one would expect for second source parts.
There are a very large number of poor Spice models out there and an understanding what the models mean is strongly advised. Comparing the model to the Data sheets and measured results is essential if you are to get any useful output from a Spice simulation. It is the difference between computer aided design and computer aided guessing. Of course as some wag (and I wish I remembered who so as to give him proper credit) "negative feedback should make component differences largely negligible anyway."
Fred
As always, please feel free to ignore my advice. There of several of us (and you guys know who you are!) that would miss the entertainment value of the usual free for all.
PS The Toshiba Power BJT models that came with the $3000 simulator I use are different from those that have been posted here. I dare not post them, as I would be accused, as usual, of attempting to increase the entertainment level at the distress of many on the forum.
Hi Fred,
The reference I use for device modeling is Massobrio and Antognetti "Semiconductor Device Modeling with SPICE" second edition. It's a really excellent book that gets into all the gory mathematical detail. I consider it an essential SPICE reference (besides "The SPICE Book" by Vladimirescu).
http://www.amazon.com/exec/obidos/t...=lpr_g_1/102-8997260-6384162?v=glance&s=books
I still have the models you're referring to that you posted in an earlier conversation we had on this same subject. I decided to use MOSFETs so this issue has been on the back burner for a while. I will compare the simulated ft values with the data sheet for your models as well. I previously wasn't aware of the simple relationship between ft and output inductance for an emitter follower given here http://www.reed-electronics.com/ednmag/archives/1996/042596/09df3.htm But now that I have this info it should be a simple matter to do the ft simulation at various collector currents and compare with the data sheet values. This should tell which model is closest (if indeed any are close).
I like to think of this stuff in terms of Hamming's motto, "The purpose of computing is insight, not numbers".
The reference I use for device modeling is Massobrio and Antognetti "Semiconductor Device Modeling with SPICE" second edition. It's a really excellent book that gets into all the gory mathematical detail. I consider it an essential SPICE reference (besides "The SPICE Book" by Vladimirescu).
http://www.amazon.com/exec/obidos/t...=lpr_g_1/102-8997260-6384162?v=glance&s=books
I still have the models you're referring to that you posted in an earlier conversation we had on this same subject. I decided to use MOSFETs so this issue has been on the back burner for a while. I will compare the simulated ft values with the data sheet for your models as well. I previously wasn't aware of the simple relationship between ft and output inductance for an emitter follower given here http://www.reed-electronics.com/ednmag/archives/1996/042596/09df3.htm But now that I have this info it should be a simple matter to do the ft simulation at various collector currents and compare with the data sheet values. This should tell which model is closest (if indeed any are close).
I like to think of this stuff in terms of Hamming's motto, "The purpose of computing is insight, not numbers".
Re: A modest proposal on models
I don't think anyone has given any indication that they 'have no understanding of what the parameters actually represent'.....
They do not.
We're gratefull for your munificence i am sure....Oh maestro.
Fred Dieckmann said:
I don't think anyone has given any indication that they 'have no understanding of what the parameters actually represent'.....
Fred Dieckmann said:
They do not.
Fred Dieckmann said:
We're gratefull for your munificence i am sure....Oh maestro.
Mike K:
>I have just discovered that the mjl1302 model has been recently updated by on-semi.....<
OK...
>VAF=19.8199<
A forward Early voltage of 19.8199 for the MODPEX model, while the data from Andy C's model shows 100? Perhaps it is better to hand-plot the junctions of those Ic curves...
When you also compare VAR (reverse Early voltage); 2.77936 vs. 100. Again, totally different.
OTOH, all of this sounds distressingly familiar.
regards, jonathan carr
>I have just discovered that the mjl1302 model has been recently updated by on-semi.....<
OK...
>VAF=19.8199<
A forward Early voltage of 19.8199 for the MODPEX model, while the data from Andy C's model shows 100? Perhaps it is better to hand-plot the junctions of those Ic curves...
When you also compare VAR (reverse Early voltage); 2.77936 vs. 100. Again, totally different.
OTOH, all of this sounds distressingly familiar.
regards, jonathan carr
...on reflection...i reckon you have a point there..They do not.
I think everybody will consider that a pretty simplistic statement without some actual comparisons of the parameters you find to be so different.
I made the post to raise the point that you need to look at the models closely and have a basic understanding of how realistic the models are. Knowing which parameters to vary is useful to know what transistors with different Hfe's will do in the circuit. Idss on FETs vary enough that you will not even get the right DC bias values from the simulation. I don't think this is very different than the points Mr. Carr and Andy are making. Borrowing from Mr. Carr to describe yet another of your caustic diatribes "all of this sounds distressingly familiar". Maybe that has something to do with why Mr. Self does not return your Emails...........
I think everybody will consider that a pretty simplistic statement without some actual comparisons of the parameters you find to be so different.
I made the post to raise the point that you need to look at the models closely and have a basic understanding of how realistic the models are. Knowing which parameters to vary is useful to know what transistors with different Hfe's will do in the circuit. Idss on FETs vary enough that you will not even get the right DC bias values from the simulation. I don't think this is very different than the points Mr. Carr and Andy are making. Borrowing from Mr. Carr to describe yet another of your caustic diatribes "all of this sounds distressingly familiar". Maybe that has something to do with why Mr. Self does not return your Emails...........
I have corresponded with self briefly.....re his book, I find it superb at what it does, but it is a bit pompous in respect of 'this is the only way to do it, and all other ways are crap' having listened to some excellent amps that are not a la self, sound theory is not the only way to sound 'sound' if you get my drift.
personally, I think that a lot of the techniques that go to make a perfect amp. tend to kill the sound.
Someone close to self alleges slone could have plagarised his book....they ARE very similar in essence.
personally, I think that a lot of the techniques that go to make a perfect amp. tend to kill the sound.
Someone close to self alleges slone could have plagarised his book....they ARE very similar in essence.
Slone cites Self as a reference, so I don't think that strictly speaking it is "plagerism" Also having read both, I find Slone's restatment of Self is sometimes more clear than the original. Certainly the graphics in the Slone's book are superior --whoever was responsible for the illustrations as they appear in Self should be strung up by the thumbs!
Slone's book is organized a bit better in that Self seems to be basicly reprints of magazine articles with result that it is sometimes difficult to ferret out a specific item if you don't remember where it is.
You don't have to agree with Self's overall emphasis to find great value in the work he has done. It's good that someone has made the effort to push on all the "butons" that relate to THD+N. Even if your prime goal is something other than minimized THD+N, knowledge regarding how your design decisions impact the THD+N figure is valuable. There are always trade-offs where to get better results in one aspect, you have to accept worse in other areas.
A simple example" in a a project I have under way for a modest powered amp one problem is that my first version exhibited significant "rail-sticking" with the probable result that it would sound horrendus when clipping or near clipping (an important consideration when modest power levels are involved). The solution turned out to be replacing the beta-enhanced VAS with someting less agressive. Self's work was helpful in understanding the consequences of this as well alternatives to this VAS configuration that still provide some of the benefits. The results are by no means a "blameless" amp, but it benefits from some of the ideas.
What I would like now is for someone to attack other areas of amp performance, such a behavior at or near clipping with similar systematic detail that Self did for THD+N.
Slone's book is organized a bit better in that Self seems to be basicly reprints of magazine articles with result that it is sometimes difficult to ferret out a specific item if you don't remember where it is.
You don't have to agree with Self's overall emphasis to find great value in the work he has done. It's good that someone has made the effort to push on all the "butons" that relate to THD+N. Even if your prime goal is something other than minimized THD+N, knowledge regarding how your design decisions impact the THD+N figure is valuable. There are always trade-offs where to get better results in one aspect, you have to accept worse in other areas.
A simple example" in a a project I have under way for a modest powered amp one problem is that my first version exhibited significant "rail-sticking" with the probable result that it would sound horrendus when clipping or near clipping (an important consideration when modest power levels are involved). The solution turned out to be replacing the beta-enhanced VAS with someting less agressive. Self's work was helpful in understanding the consequences of this as well alternatives to this VAS configuration that still provide some of the benefits. The results are by no means a "blameless" amp, but it benefits from some of the ideas.
What I would like now is for someone to attack other areas of amp performance, such a behavior at or near clipping with similar systematic detail that Self did for THD+N.
don't get me wrong, I think self is great, slone, too...helped me understand an awful lot. It helps if you understand it as to be simply a book on reducing distortion to its min.
I would love to see self extend his book, and offered him some suggestions, like discussing symmetrical amps, opamp inputs, generally extending it to include loads of other stuff...he was not keen on the idea...thought it would take an age
would be nice to see someone else write a dedicated book on audio amps to cover all the questions we want answers to.
It amazes me how the japs come up with so many myriads of designs, all different.
I would love to see self extend his book, and offered him some suggestions, like discussing symmetrical amps, opamp inputs, generally extending it to include loads of other stuff...he was not keen on the idea...thought it would take an age
would be nice to see someone else write a dedicated book on audio amps to cover all the questions we want answers to.
It amazes me how the japs come up with so many myriads of designs, all different.
I guess I was responding to quick to a percieved point of view that tends to dismiss outright anyone who even measures distortion. My mistake.
On a related not, I've been wanting to check out a book by Ben Duncan on amps, but it's not in any local libraries and availble only in hardcover at about US$60 ! Is anyone familiar with this and is there anything new or worthwhile in it?
On a related not, I've been wanting to check out a book by Ben Duncan on amps, but it's not in any local libraries and availble only in hardcover at about US$60 ! Is anyone familiar with this and is there anything new or worthwhile in it?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.