Hi syn08,
You know they must design to "their sound". The advertising sold this distortion and their client base bought into it. They really can not afford to prove the product. Besides, they are overcharging for poor design and making a lot of money doing it.
You know marketing. Tell me what would happen if those companies suddenly started building clean amps with no "character" anymore?
-Chris
No.
You know they must design to "their sound". The advertising sold this distortion and their client base bought into it. They really can not afford to prove the product. Besides, they are overcharging for poor design and making a lot of money doing it.
You know marketing. Tell me what would happen if those companies suddenly started building clean amps with no "character" anymore?
-Chris
Re: Re: Re: Re: Re: Re: Re: hec != hoax ?
Hi Bob,
I don't think it's very appropriate to discuss the pros and cons of the various simulators in this thread. Nevertheless I can tell you that I did have tried LTSpice and I don't like it at all. It's ugly and primitive. Here is one example: to view the distortion residual, in Micro-Cap just use the function RES(V(xxx)), while LTSpice lacks such a function and you have to make it by yourself.
BTW, it is unlikely that discrepancies have anything to do with the simulators, rather the used models.
Cheers, Edmond.
Bob Cordell said:
Hi Bob,
I don't think it's very appropriate to discuss the pros and cons of the various simulators in this thread. Nevertheless I can tell you that I did have tried LTSpice and I don't like it at all. It's ugly and primitive. Here is one example: to view the distortion residual, in Micro-Cap just use the function RES(V(xxx)), while LTSpice lacks such a function and you have to make it by yourself.
BTW, it is unlikely that discrepancies have anything to do with the simulators, rather the used models.
Cheers, Edmond.
anatech said:Hi syn08,
No.
You know they must design to "their sound". The advertising sold this distortion and their client base bought into it. They really can not afford to prove the product. Besides, they are overcharging for poor design and making a lot of money doing it.
You know marketing. Tell me what would happen if those companies suddenly started building clean amps with no "character" anymore?
-Chris
Cheers, Edmond.
Hi, Edmond,
I'm confused. Is it the same calculation if the connection point is base of VAS or base of (-) differential input? (Like Bob's http://www.diyaudio.com/forums/attachment.php?s=&postid=1368243&stamp=1196892957)
Re: Re: Re: Re: Re: Re: Re: Re: hec != hoax ?
You're probably right about this discussion belonging in the SPICE thread, but I think it is certainly an important one.
I certainly don't see the LTSPICE as ugly and primitive.
Being able to show the distortion residual directly is a nice feature, though. But features from simulator to simulator will always vary a bit. The lack of such a convenient feature is no reason to call the simulator ugly and primitive. If this is the best complaint you can come up with, you have not convinced me at all.
Cheers,
Bob
Edmond Stuart said:
You're probably right about this discussion belonging in the SPICE thread, but I think it is certainly an important one.
I certainly don't see the LTSPICE as ugly and primitive.
Being able to show the distortion residual directly is a nice feature, though. But features from simulator to simulator will always vary a bit. The lack of such a convenient feature is no reason to call the simulator ugly and primitive. If this is the best complaint you can come up with, you have not convinced me at all.
Cheers,
Bob
Re: Re: Re: hec != hoax ?
Hi Bob,
Actually, I wasn't in the mood to recollect all my complaints about LTSpice. The missing RES feature was just the last thing that stroke me. Of course their are more things I dislike (layout, accuracy, human interface, data handling, clumsy FFT implementation, etc.).
BTW, it was you who told me that the EKV models ran terrible slowly. How did you solved this problem?
But the best advice I can give is download the trial version of Micro-Cap and judge by yourself.
Cheers, Edmond.
Bob Cordell said:
Hi Bob,
Actually, I wasn't in the mood to recollect all my complaints about LTSpice. The missing RES feature was just the last thing that stroke me. Of course their are more things I dislike (layout, accuracy, human interface, data handling, clumsy FFT implementation, etc.).
BTW, it was you who told me that the EKV models ran terrible slowly. How did you solved this problem?
But the best advice I can give is download the trial version of Micro-Cap and judge by yourself.
Cheers, Edmond.
Re: Re: Re: Re: Re: Re: Re: hec != hoax ?
Hi Ovidiu,
Indeed, there is a discrepancy, but that doesn't necessarily mean that my simulation is wrong.
Perhaps you have overlooked my remark:
BUT more importantly, when I decreased the TMC resistor R36 from 1K to 330 Ohm, that means that Ft of the TMC network equals the Fc of your HEC NFB loop, the distortion is just the same, 6ppm.
As this 6ppm figure of HEC is well in accordance with real life measurements and also in accordance with my view on TMC, I don't see any reason to express your doubts about my simulations.
Cheers, Edmond.
syn08 said:
Hi Ovidiu,
Indeed, there is a discrepancy, but that doesn't necessarily mean that my simulation is wrong.
Perhaps you have overlooked my remark:
BUT more importantly, when I decreased the TMC resistor R36 from 1K to 330 Ohm, that means that Ft of the TMC network equals the Fc of your HEC NFB loop, the distortion is just the same, 6ppm.
As this 6ppm figure of HEC is well in accordance with real life measurements and also in accordance with my view on TMC, I don't see any reason to express your doubts about my simulations.
Cheers, Edmond.
Re: Re: Re: Re: Re: Re: Re: Re: hec != hoax ?
Did I say anything about your simulations being wrong? It could be Bob's or yours, it doesn't really matter. Fact is, there is a 50% difference that has to be covered and explained - if you like consistency.
I haven't expressed any doubts about your simulations however, based on a one time shot at that 6ppm, I wouldn't consider that there's always a good fit between your numbers and the real world.
To summarize:
1. Make sure that all simulation platform deliver results within a few percents. It could be the models responsible for discrepancies, but knowing a little about how the Spice core is working (funny, not much changed since the original Berkley Spice implementation) it could be other reasons as well.
2. Check the simulation results against the real world. This could be done based on reusable "building blocks". This is where simulation models should be fine tuned.
But then again, you don't have to do anything. Lacking any analytic approach, the above steps is what would make me happy and more trustful in TMC, DTMC (BTW, why not TTMC?) and other. Which is not necessary what would make you happy.
Do you know what made me believe in NDFL? Certainly not your simulations only, but mostly a document created by Dr Richard Mitchell at University of Reading in the UK, as part of his "Computer Controlled Feedback Systems" course notes, that I was able to download at http://www.personal.rdg.ac.uk/~shsmchlr/pof/ (unfortunately the NDFL part was meantime pulled). That document speaks a language that I can compile: analytic approach, transfer functions, diagrams, concepts, etc...
Edmond Stuart said:
Did I say anything about your simulations being wrong? It could be Bob's or yours, it doesn't really matter. Fact is, there is a 50% difference that has to be covered and explained - if you like consistency.
I haven't expressed any doubts about your simulations however, based on a one time shot at that 6ppm, I wouldn't consider that there's always a good fit between your numbers and the real world.
To summarize:
1. Make sure that all simulation platform deliver results within a few percents. It could be the models responsible for discrepancies, but knowing a little about how the Spice core is working (funny, not much changed since the original Berkley Spice implementation) it could be other reasons as well.
2. Check the simulation results against the real world. This could be done based on reusable "building blocks". This is where simulation models should be fine tuned.
But then again, you don't have to do anything. Lacking any analytic approach, the above steps is what would make me happy and more trustful in TMC, DTMC (BTW, why not TTMC?) and other. Which is not necessary what would make you happy.
Do you know what made me believe in NDFL? Certainly not your simulations only, but mostly a document created by Dr Richard Mitchell at University of Reading in the UK, as part of his "Computer Controlled Feedback Systems" course notes, that I was able to download at http://www.personal.rdg.ac.uk/~shsmchlr/pof/ (unfortunately the NDFL part was meantime pulled). That document speaks a language that I can compile: analytic approach, transfer functions, diagrams, concepts, etc...
Attachments
Re: Re: Re: Re: Re: Non-HEC-amp
Hi Edmond.
The dynamic impedance of those zener diodes is pretty low. Combined with the low input capacitance of the VAS bjt's you're using, I doubt that it is the reason for the lack of oscillation.
It would probably be just as oscillation free in simulation if you replaced the zeners with ideal voltage sources.
The LTP's and the VAS are the only sources of voltage gain in my amp. BTW, you already have a schematic of my amp, which is why I only posted a low res version this time, just to show the VAS topology again.
Also, what models are you using for the MOSFET devices? I'd like to do some comparison sims with them if they are LTspice compatible.
Cheers,
Glen
PS
I'm not dating Rachel Weisz anymore. She dumped me after catching me with these two........
Edmond Stuart said:
Hi Glen,
As you have probably seen, In my latest design I've used 3.3V zeners. The sim doesn't reveal any trace of instability. Perhaps the internal resistance of the zener acts as a base stopper.
I guess that your design has no trouble with a 3k load resistor because there is somewhere else an extra gain stage, right?
Say to Rachel that she should stop moaning about spectacles. Even under an electron microscope I couldn't decipher the resistor values.
Cheers, Edmond.
Hi Edmond.
The dynamic impedance of those zener diodes is pretty low. Combined with the low input capacitance of the VAS bjt's you're using, I doubt that it is the reason for the lack of oscillation.
It would probably be just as oscillation free in simulation if you replaced the zeners with ideal voltage sources.
The LTP's and the VAS are the only sources of voltage gain in my amp. BTW, you already have a schematic of my amp, which is why I only posted a low res version this time, just to show the VAS topology again.
Also, what models are you using for the MOSFET devices? I'd like to do some comparison sims with them if they are LTspice compatible.
Cheers,
Glen
PS
I'm not dating Rachel Weisz anymore. She dumped me after catching me with these two........
Attachments
Re: Re: Re: Re: Re: Re: Non-HEC-amp
Hi Glen,
The difficulty is that the Hamksford cascode is sometimes unstable. It appears that I don't need base stoppers in my non-HEC amp (at least, according to my sim), but I did have troubles with it in a different amp. That's why I am cautious with this kind of cascode.
As for the models (EKV), I'll send them to you.
Cheers, Edmond.
PS: As for Rachel, just be glad. Jealous women are the source to all evil and misery.
G.Kleinschmidt said:
Hi Glen,
The difficulty is that the Hamksford cascode is sometimes unstable. It appears that I don't need base stoppers in my non-HEC amp (at least, according to my sim), but I did have troubles with it in a different amp. That's why I am cautious with this kind of cascode.
As for the models (EKV), I'll send them to you.
Cheers, Edmond.
PS: As for Rachel, just be glad. Jealous women are the source to all evil and misery.
Re: Re: Re: Re: Re: Re: Re: Re: hec != hoax ?
Hi Bob,
Although it's true that this arrangement eliminates the Early effect in the input transistors, but now the cascode trannies are subjected to Vce modulation and, consequently, exposed to the Early effect.
Admittedly, it helps a little bit, but not to the extent of the Hawksford cascode.
BTW, using jfets in the input stage as in your original design is a different story.
Cheers, Edmond.
Bob Cordell said:
Hi Bob,
Although it's true that this arrangement eliminates the Early effect in the input transistors, but now the cascode trannies are subjected to Vce modulation and, consequently, exposed to the Early effect.
Admittedly, it helps a little bit, but not to the extent of the Hawksford cascode.
BTW, using jfets in the input stage as in your original design is a different story.
Cheers, Edmond.
I can understand your absorption with the tubes. Solid state designers can learn a lot from tude designers. Often, tube designs really play music well even though they have relatively high THD. Must be the fault of the listener...
I did mean closed loop gain. I was unable to reconcile the figures you gave for feedback factor without assuming the CL gains were different. I may need a diagram.
Re: Re: Re: Re: Re: Re: Re: Re: Re: hec != hoax ?
Hi Ovidiu,
Lacking any analytic approach? I bag you pardon.
An analytic approach is not limited to only what you think is 'analytic'. There also exist something called logic.
To me it is crystal clear that TMC and DTMC does work and I don't need any math or whatsoever to understand why. I can even prove, still without any math, that TMC reduces the distortion to the same extent as HEC.
BTW, believing that something cannot work, just because it is sooooh simple, is a nice example of lacking any analytic insight.
Cheers, Edmond.
PS: why not TTMC? NO, because it can be read as triple-TMC.
syn08 said:
Hi Ovidiu,
Lacking any analytic approach? I bag you pardon.
An analytic approach is not limited to only what you think is 'analytic'. There also exist something called logic.
To me it is crystal clear that TMC and DTMC does work and I don't need any math or whatsoever to understand why. I can even prove, still without any math, that TMC reduces the distortion to the same extent as HEC.
BTW, believing that something cannot work, just because it is sooooh simple, is a nice example of lacking any analytic insight.
Cheers, Edmond.
PS: why not TTMC? NO, because it can be read as triple-TMC.
TTMC was obviously a semantic joke. Following the same "logic" you are mentioning, if TMC is good, and TTMC is better then why wouldn't a triple TMC be even better? )
But I see you became quite passionate on the topic so I will refrain from any further comments until I will have more of a 3D insight (paper, simulation, experimental).
)But I see you became quite passionate on the topic so I will refrain from any further comments until I will have more of a 3D insight (paper, simulation, experimental).