Well, you've calculated it. The bandwidth isn't just dependent of the feedback, it's also dependent of the transition frequency. The D44(5)H is described as audio amplifier and 'fast switch' but there's no transition frequency in the datasheet. Most of these transistors have a tf of <50khz or even lower. Your simulation looks like there was put in a very high (near infinite) tf, which means it will behave as a quasi-perfect transistor. So yes, simulation is all nice and well but only as accurate as your values you are using. That's why I'm a bit doubtful the df is that extremely high at these (for an audio amplifier) huge frequencies.
I do not doubt in any way it sounds good. And I like your baby version of it very much, though it might be a bit problematic to enter an airplane with it. 😀
Last edited:
Intersting, that H2/H3 thing, Maiko. So I just made an experiment with inserting a buffer between the VAS and the output devices and the H3 immediately got down to the same levels as the other harmonics. This buffer is only for experimenting since it ruins the HF and phase behavior.
Personally, I am quite sure that the H3 comes from the fact that the output devices are current driven. The hfe lowers as the signal gets stronger, that's a property of all power bjt's, so one can say that this H3 "sharpens the edges" slightly of the sinus curve.
In my opinion ( also Nelson's ) the H2, and H3 usually are harmless sonically. H2 gives some lightness and warmth and H3 makes the amp sparkle a bit and seem a bit more dynamical. My Baby Godzilla confirms this, I think.
Aha, you noticed my speaker cable! As a matter of fact it's made of some old Audio quest cables that I stripped from it's insulation. I'm using wooden "klädnypor" as it's called in Sweden to keep them separated. One every half foot or so. No dialectric absorbtion at least. This is mostly a benign surrendering to the cable myths. Placebo or not, it was indeed a cheap modd.
I talked to the Swedish cable manufacturer Jorma who makes very expensive cables. I met him in a local hifi store. He actually liked my little arrangement but said that such a cable could not be sold since the risk of short circuiting are too big.
And about simulation. I think simulators are astonishingly accurate. I rarely make real measurements. I have an oscilloscope with an FFT function but is is quite useless.
One last thing. My software, Circuitmaker, which is quite old, has a tendency to crash frequently. It's the spice engine that often reports things like " @6 : mo such vector ". And sometimes it crashes and takes down the whole application.
Are modern spice simulators more reliable? But I would prefer my old simulator - I have learned how to use it and I can work really fast with it.
Personally, I am quite sure that the H3 comes from the fact that the output devices are current driven. The hfe lowers as the signal gets stronger, that's a property of all power bjt's, so one can say that this H3 "sharpens the edges" slightly of the sinus curve.
In my opinion ( also Nelson's ) the H2, and H3 usually are harmless sonically. H2 gives some lightness and warmth and H3 makes the amp sparkle a bit and seem a bit more dynamical. My Baby Godzilla confirms this, I think.
Aha, you noticed my speaker cable! As a matter of fact it's made of some old Audio quest cables that I stripped from it's insulation. I'm using wooden "klädnypor" as it's called in Sweden to keep them separated. One every half foot or so. No dialectric absorbtion at least. This is mostly a benign surrendering to the cable myths. Placebo or not, it was indeed a cheap modd.
I talked to the Swedish cable manufacturer Jorma who makes very expensive cables. I met him in a local hifi store. He actually liked my little arrangement but said that such a cable could not be sold since the risk of short circuiting are too big.
And about simulation. I think simulators are astonishingly accurate. I rarely make real measurements. I have an oscilloscope with an FFT function but is is quite useless.
One last thing. My software, Circuitmaker, which is quite old, has a tendency to crash frequently. It's the spice engine that often reports things like " @6 : mo such vector ". And sometimes it crashes and takes down the whole application.
Are modern spice simulators more reliable? But I would prefer my old simulator - I have learned how to use it and I can work really fast with it.
Just to clarify things. The following picture shows the AC curves with "open" gain.
The coil L1 is there just to make the simulation work. The VAS load is now the base current of the output stage just is in real operation. R4 emulates the "normal" R5.
This shows that the gain is roughly 70 db ( that's misleading since the base of Q4 has around 50 ohms or so ) up to 200khz. That's really heavy. Actually unnecessary heavy.
So I think R5 may be changed to 330 ohm or more. That will make it unnecessary to have a special low voltage section and will make the amp more friendly to build.
With 330 ohm, the NFB will be around 500( 54db) up to 150khz.
With the proper output devices ( FJAF4210Y/4310Y ) with a beta of 150 and a high beta Q4, the NFB will be back to some 60db again.
One more thing. The wide NFB bandwidth is only there for making the amp effortless and low distorting at HF. A real amp should be limited to some 50khz or so.
The coil L1 is there just to make the simulation work. The VAS load is now the base current of the output stage just is in real operation. R4 emulates the "normal" R5.
This shows that the gain is roughly 70 db ( that's misleading since the base of Q4 has around 50 ohms or so ) up to 200khz. That's really heavy. Actually unnecessary heavy.
So I think R5 may be changed to 330 ohm or more. That will make it unnecessary to have a special low voltage section and will make the amp more friendly to build.
With 330 ohm, the NFB will be around 500( 54db) up to 150khz.
With the proper output devices ( FJAF4210Y/4310Y ) with a beta of 150 and a high beta Q4, the NFB will be back to some 60db again.
An externally hosted image should be here but it was not working when we last tested it.
One more thing. The wide NFB bandwidth is only there for making the amp effortless and low distorting at HF. A real amp should be limited to some 50khz or so.
Yes. They are. Unless you are entering wrong parameters. Or missing some. Like the tf.
And that here is the exact reason why you sometimes have to. Missing or wrong parameters throw off simulates by magnitudes. Unless you are simulating with the real values, it's all for the trashcan. A DF of over 4000 at 500kHz? Sorry, but if you can't measure it or deliver a simulation with complete parameters I call it the same as it sounds, too good to be true. Or in other words: utter BS. Sry.
You seem to want to continue your dream but no simulation in the world will give you proper results if you don't feed it correct and complete parameters.
> interesting if there were some transfer function for hfe
There is. Are!
1) There's the datasheet. Shows some marketing idealization of hFE vs current. The D44H8 sheet suggests "some" fall-off before you get up to the ~~3.5A that 30V supplies promise. Some droop below 100mA. And this is one of the flatter hFE curves I have seen. But always take datasheet "typical" with a large grain of salt.
2) There's A.S. Grove 1967 Physics and Technology of Semiconductor Devices, sections 7.2 and 7.3. Low current hFE is dominated by recombination. At high current the injected carrier concentration exceeds the base doping concentration, so it isn't the transistor you thought you made/bought. Modular Series on Solid State Devices: Volume III: The Bipolar Junction Transistor George W. Neudeck 1989 probably has more gruesome details, but is tough to chew.
3) And if you prefer to believe SPICE, plot hFE vs current over a wide range for some reasonable Vce. What you are really doing is making a picture of 2 or 3 parameters in your SPICE model. IK shapes the top end, and ratios of several other parameters seem to set the slope on the low side.
If the stages were designed with some "slugging" resistances so hFE mattered less, this might be minor. But you've stripped it down so the gain is dominated by final stage hFE and similar poorly-controlled parameters.
There is. Are!
1) There's the datasheet. Shows some marketing idealization of hFE vs current. The D44H8 sheet suggests "some" fall-off before you get up to the ~~3.5A that 30V supplies promise. Some droop below 100mA. And this is one of the flatter hFE curves I have seen. But always take datasheet "typical" with a large grain of salt.
2) There's A.S. Grove 1967 Physics and Technology of Semiconductor Devices, sections 7.2 and 7.3. Low current hFE is dominated by recombination. At high current the injected carrier concentration exceeds the base doping concentration, so it isn't the transistor you thought you made/bought. Modular Series on Solid State Devices: Volume III: The Bipolar Junction Transistor George W. Neudeck 1989 probably has more gruesome details, but is tough to chew.
3) And if you prefer to believe SPICE, plot hFE vs current over a wide range for some reasonable Vce. What you are really doing is making a picture of 2 or 3 parameters in your SPICE model. IK shapes the top end, and ratios of several other parameters seem to set the slope on the low side.
If the stages were designed with some "slugging" resistances so hFE mattered less, this might be minor. But you've stripped it down so the gain is dominated by final stage hFE and similar poorly-controlled parameters.
Attachments
TY, PRR! Just as I said, simulations can pretty much decieve, especally if the parameters aren't correct. Or the usage of the simulation.
OK, ICG.
Of course you are right. But I am actually developing this concept now in "real time" and it would take much longer to actually build the circuit. Furthermore, I don't have any THD meter so I will have to stick to simulation.
Your criticism regards mainly the lack of model for the tf in the output devices. That's strange. I have always believed ( I'm not good at the physics and chemistry part ) that the TF mainly is due to various capacitances. If there are other mechanisms, what are they then? My spice model clearly includes those. The miller capacitances ( the important ones ) usually are well below 100pf. Both experience and simulation clearly shows that. Furthermore the D44H8 actually has a nominal TF of 50mhz as shown in the specs.
Assuming a pessimistic value of 150pf and two pairs are used the BW will be around 400khz.
When it comes to simulation accuracy, it seems to me that problem occurs when slight differences in the components stacks up and will create some unexpected behavior. But my circuit is really very simple, it should be of no challenge for any simulator.
So my option is either to shut up, and speak no more about my circuit, or continue my simulations. What do you say folk's, what do you want me to do?
ICG:
Then we have the thing about the behavior of hfe. The curves that you show are the usual ones. It's seems that the hfe varies quite a lot. I will now conduct an experiment with my simulator to figure out whether it simulates these curves reasonably good.
I will be back with the results.
Of course you are right. But I am actually developing this concept now in "real time" and it would take much longer to actually build the circuit. Furthermore, I don't have any THD meter so I will have to stick to simulation.
Your criticism regards mainly the lack of model for the tf in the output devices. That's strange. I have always believed ( I'm not good at the physics and chemistry part ) that the TF mainly is due to various capacitances. If there are other mechanisms, what are they then? My spice model clearly includes those. The miller capacitances ( the important ones ) usually are well below 100pf. Both experience and simulation clearly shows that. Furthermore the D44H8 actually has a nominal TF of 50mhz as shown in the specs.
Assuming a pessimistic value of 150pf and two pairs are used the BW will be around 400khz.
When it comes to simulation accuracy, it seems to me that problem occurs when slight differences in the components stacks up and will create some unexpected behavior. But my circuit is really very simple, it should be of no challenge for any simulator.
So my option is either to shut up, and speak no more about my circuit, or continue my simulations. What do you say folk's, what do you want me to do?
ICG:
Then we have the thing about the behavior of hfe. The curves that you show are the usual ones. It's seems that the hfe varies quite a lot. I will now conduct an experiment with my simulator to figure out whether it simulates these curves reasonably good.
I will be back with the results.
I'm back with results of my simulator.
From 50mA to 2.5A it reports an increase of 77 to 101.
PRR: Your submitted sheet must be wrong, the figures are unrealistically good. The real specs shows an increase of ( from 0.05 to 2.5A ) 80 to 125.
Lets say the simulator is wrong with 50%. That will probably generate THD figures that are not more than - generously - double as good as of the real circuit.
When using the FJAF4210Y/42310Y that has a larger beta, the figures probably will match my simulation.
So I think we must have some sort of gut feeling of what makes sense in this particular circuit.
From 50mA to 2.5A it reports an increase of 77 to 101.
PRR: Your submitted sheet must be wrong, the figures are unrealistically good. The real specs shows an increase of ( from 0.05 to 2.5A ) 80 to 125.
Lets say the simulator is wrong with 50%. That will probably generate THD figures that are not more than - generously - double as good as of the real circuit.
When using the FJAF4210Y/42310Y that has a larger beta, the figures probably will match my simulation.
So I think we must have some sort of gut feeling of what makes sense in this particular circuit.
The data sheets I've seen did not include any mention of the TF, I've looked at the ones of st.com and several other sources. However I did find it meanwhile in the ON Semiconductor data sheet. Okay, seems I didn't do my homework properly. Sorry for that.
The DH44H8 is listed with 50MHz, but the DH45H8 is noted with 40MHz. Strange, that's a significant drop. Well, let's see, what else does it say? Capacitance of the DH44H8 is listet with 90pF. The complementary DH45H8 is however stated with 160pF. 😱 Uh, that's interesting. And a lot worse than your pessimistic assumption. Well, I didn't know that but there is one of the reasons I didn't trust the simulation. And because of the asymmetry and much higher capacitance (among other things) I honestly have still more than reasonable doubts on the simulation, this causes the amplifier to show much more non-linearities at higher frequencies and especally the signal will not be symmetrical in both halve waves.
One of the things the TF marks is the frequency the current gain drops to 1. So the actual usable frequency for a power amplifier is a lot lower, usually 2-3 digits, especally if the load contains phase shift and impedance fluctuations (which speakers DO).
Well, wouldn't you agree, that these are more than slight differences? As I said, simulations are fine and well but the more extreme something becomes the more a simulation is thrown off. And you have to admit, your Godzilla is extreme. 😉 You've simulated with near-ideal values and in such extraordinary regions you have to verify your simulation and back up your date with measurements.
I did not say that and I didn't mean it either. But sometimes a bit healthy scepticism is needed if it's too good to be true. At first I would suggest to use different, more symmetrical complementary transistors.
I didn't say that either, that's PRR's lines.
I just can't resist reporting my latest *measurements*. Aha, I have now actually made a THD analysis on my Baby Godzilla. Since the FFT function of my scope is crappy, I had to find a way to circumvent this.
First of all I simulated the behavior of the amp being fed by a 10Vpeak, 1khz signal into 3ohm. The THD is so low that my scope can't detect it. But the error signal at the base of the output devices will have a higher magnitude - my simulator reports 20mV of 3:rd order dist. That's -54db.
When I hooked the scope on that point, still the 3:rd order was below the noise floor, which is around -60db on my scope.
From this I deduce that the actual distortion will be at least the half of what my simulator asserts.
This makes sense since my FJAF-devices have a higher beta than the D44H8 devices.
Thank's ICG! Now I actually have something to back up my simulations on.
First of all I simulated the behavior of the amp being fed by a 10Vpeak, 1khz signal into 3ohm. The THD is so low that my scope can't detect it. But the error signal at the base of the output devices will have a higher magnitude - my simulator reports 20mV of 3:rd order dist. That's -54db.
When I hooked the scope on that point, still the 3:rd order was below the noise floor, which is around -60db on my scope.
From this I deduce that the actual distortion will be at least the half of what my simulator asserts.
This makes sense since my FJAF-devices have a higher beta than the D44H8 devices.
Thank's ICG! Now I actually have something to back up my simulations on.
Oh, I forgot to comment your latest reply, ICG.
90 +160 pf times two equals 500pf. That means a BW of 470khz. And that's far more than what's needed to get a DF of 4000 at 500khz.
One have to take things with a grain of salt and try to figure out what matters. When it came to that hfe transfer function I was quite unsure, but my last experiment have shown that my old Circuitmaker still is accurate enough.
I remember when I first had a chance to simulate circuits some 15 years ago. This was like a whole new world opened up. Theorys are necessary, but usually one calculates wrong and a simulator speeds up that iterated process of making/fixing errors.
But I regard your comments as a reaction of me being perceived sloppy and overly amateurish enthusiastic. Well I'm an amateur but I'm quite cautious.
But on the other hand, why am I presenting this concept at all? I suppose this forum is for sharing ideas and to inspire and help each other. So what I am doing is a bit like brainstorming. Constructive criticism is what's wanted.
90 +160 pf times two equals 500pf. That means a BW of 470khz. And that's far more than what's needed to get a DF of 4000 at 500khz.
One have to take things with a grain of salt and try to figure out what matters. When it came to that hfe transfer function I was quite unsure, but my last experiment have shown that my old Circuitmaker still is accurate enough.
I remember when I first had a chance to simulate circuits some 15 years ago. This was like a whole new world opened up. Theorys are necessary, but usually one calculates wrong and a simulator speeds up that iterated process of making/fixing errors.
But I regard your comments as a reaction of me being perceived sloppy and overly amateurish enthusiastic. Well I'm an amateur but I'm quite cautious.
But on the other hand, why am I presenting this concept at all? I suppose this forum is for sharing ideas and to inspire and help each other. So what I am doing is a bit like brainstorming. Constructive criticism is what's wanted.
Well, I must now take one step backwards. I just found that my measurements doesn't say anything at all.
So simulations can still be valid. Or not.
I measured at 1khz and at that frequency the NFB loop really isn't put under any stress.
To be able to measure THD, I imagine I'll have to build some sort of bandpass filter with a BW of perhaps one half octave or so that can be tuned up and down. In this manner each overtone may be measured without the scope getting saturated.
But this is true for my baby Godzilla. The grown up one has such hefty NFB factor, and that will assure low THD.
For me this THD thing is not so very important. It's the complexity of the distortion that matters. Simple circuits have low complexity in the distortion pattern. But now, I have entered the philosophical realm. Or perhaps psycho acoustical.
Normally I use to build amps with:
1. Mosfet
2. Class A
3. Low NFB
4. Simple circuits.
All the usual audiophile hobbyhorses. This time I actually violated all those ideas except one - simplicity.
So simulations can still be valid. Or not.
I measured at 1khz and at that frequency the NFB loop really isn't put under any stress.
To be able to measure THD, I imagine I'll have to build some sort of bandpass filter with a BW of perhaps one half octave or so that can be tuned up and down. In this manner each overtone may be measured without the scope getting saturated.
But this is true for my baby Godzilla. The grown up one has such hefty NFB factor, and that will assure low THD.
For me this THD thing is not so very important. It's the complexity of the distortion that matters. Simple circuits have low complexity in the distortion pattern. But now, I have entered the philosophical realm. Or perhaps psycho acoustical.
Normally I use to build amps with:
1. Mosfet
2. Class A
3. Low NFB
4. Simple circuits.
All the usual audiophile hobbyhorses. This time I actually violated all those ideas except one - simplicity.
Svitjod,
Your amps are unconventional and many do not understand; small in parts count, complicated in topology. They are a little left field for many readers here, and understandably.
Most like a fully designed, built, and debugged on this forum. You can understand that, they want to build something that they know will work right off. Your R&D approach is a little inaccessible for most.
Keep going. Build a few versions of your amp, refine it. Then present it for comment here. There are very good amps here that are popular; have a good look at the very simple quasi amp, it's very easy to build, lots of pcbs to use, and sounds wonderful. The other is CFH7 from XRK971, an excellent amp!
There are quite a few amps with very low THD that sound terrible..... you need to examine the harmonics to find out the truth and fiddle with the fb ratios.
Merry Christmas!!
Hugh
Your amps are unconventional and many do not understand; small in parts count, complicated in topology. They are a little left field for many readers here, and understandably.
Most like a fully designed, built, and debugged on this forum. You can understand that, they want to build something that they know will work right off. Your R&D approach is a little inaccessible for most.
Keep going. Build a few versions of your amp, refine it. Then present it for comment here. There are very good amps here that are popular; have a good look at the very simple quasi amp, it's very easy to build, lots of pcbs to use, and sounds wonderful. The other is CFH7 from XRK971, an excellent amp!
There are quite a few amps with very low THD that sound terrible..... you need to examine the harmonics to find out the truth and fiddle with the fb ratios.
Merry Christmas!!
Hugh
Last edited:
ah yes, that's probably a very possible reason behind the added H3 svitjod, the hfe, I was in a very selective thinking mode.
The hfe curves seen in the attached picture by prr shows clearly how hfe starts dropping around 1 A should give a hint, but as you mentioned H2 and H3 are quite benign to our ears.
I remember from the past contemplating a concept amplifier using BJT's in the output and went through a few different output transistor datasheets from different vendors, and in particular I looked for a flat hfe, I just remember I found some of the popular Sanken transistor to be quite good, but in general all the BJT's seems a bit limited in their character so there's a bit of a challenge when using few silicon devices, but that's one charming part squeezing out max from KIS topologies.
Regarding analyzing amplifier distortion, have you seen this article?
CordellAudio.com - Build a High Performance THD Analyzer
However, in general I think it would be a good (a much better!) idea to use a sound card with your computer, some of the popular software's are these:
Latest News. Audio Rightmark (I believe this one is recommended to start with)
https://www.trueaudio.com/
ARTA Software (the developer Ivo Mateljan "iMat" is nowadays here on DiyA, the thread is: http://www.diyaudio.com/forums/multi-way/76977-arta.html)
Using oscilloscope analyzing THD is quite limited, and DSO types don't use more than something like 8 or 10 bit video ADC which is I believe to some degree is due to our eyes can't really read out information from the monitor as accurate as our ears decodes sound.
You can always try for a starter with the sound card you already have in your computer, though the vast majority aren't very good, so a separate card is a good investment, perhaps an external connected via USB as computers provides a very noisy environment inside the case.
I am myself am contemplating a few different sound card options, a few external have been mentioned under the ARTA thread such as Focusrite and Steinberg(Yamaha) but I have to do more research into it.
ps. if and when you have breadboarded something, take a picture and put it up for display, that does always attract people.TOP
The hfe curves seen in the attached picture by prr shows clearly how hfe starts dropping around 1 A should give a hint, but as you mentioned H2 and H3 are quite benign to our ears.
I remember from the past contemplating a concept amplifier using BJT's in the output and went through a few different output transistor datasheets from different vendors, and in particular I looked for a flat hfe, I just remember I found some of the popular Sanken transistor to be quite good, but in general all the BJT's seems a bit limited in their character so there's a bit of a challenge when using few silicon devices, but that's one charming part squeezing out max from KIS topologies.
Regarding analyzing amplifier distortion, have you seen this article?
CordellAudio.com - Build a High Performance THD Analyzer
However, in general I think it would be a good (a much better!) idea to use a sound card with your computer, some of the popular software's are these:
Latest News. Audio Rightmark (I believe this one is recommended to start with)
https://www.trueaudio.com/
ARTA Software (the developer Ivo Mateljan "iMat" is nowadays here on DiyA, the thread is: http://www.diyaudio.com/forums/multi-way/76977-arta.html)
Using oscilloscope analyzing THD is quite limited, and DSO types don't use more than something like 8 or 10 bit video ADC which is I believe to some degree is due to our eyes can't really read out information from the monitor as accurate as our ears decodes sound.
You can always try for a starter with the sound card you already have in your computer, though the vast majority aren't very good, so a separate card is a good investment, perhaps an external connected via USB as computers provides a very noisy environment inside the case.
I am myself am contemplating a few different sound card options, a few external have been mentioned under the ARTA thread such as Focusrite and Steinberg(Yamaha) but I have to do more research into it.
ps. if and when you have breadboarded something, take a picture and put it up for display, that does always attract people.TOP
Yes, AKSA. You're a manufacturer of amplifiers, you should know.
I like to construct amplifiers ( when a new idea comes to my mind ) and this diyaudio thing, for me, is just about sharing ideas. Of course I would be glad ( and also boost my ego ) if someone appreciates my constructions.
So I deliberately presented this amp in a sketchy way so people may make comments and react on it as the circuit develops.
But i will now drop this thing. If anyone of you wants to build one, you may take contact with me personally or visit my simple site ( - Embuddy ) and check out my other constructions.
I don't have any schematics published of my Godzilla, but if you want, I can send you schematics on my Baby Godzilla which is completed. I also have pcb layouts and also some boards lying in my drawer that I can send you.
Maiko, I actually have a 24 bit external soundcard. I tried to use it a couple of weeks ago in order to measure THD but I was a bit impatient and I didn't get the thing to work properly. With such a low THD, it's not easy to measure it - it tends to be below the noise floor.
But I'm not really into this THD thing. The only dist that really matters is the cross over dist and that one is usually very low and hard to measure( even though our ears are very sensitive to it ).
This Godzilla project is basically a way for me to compete a bit with the "blameless" guys. A simplistic amp can be made with very low THD.
And I have actually recently inserted a special buffer that turns it to a Turbo Godzilla". The THD is well below -100db up to 10khz and the phase margin still seems to be magnificent - with any capacitive load. It never drops below 120 degrees up to 10mhz.
Simulated, that is, ICG 😀.
But that's more to play with figures. It's not sure at all that it sounds particularly good as AKSA suggested. But my Baby Godzilla sounds very nice, I think.
I like to construct amplifiers ( when a new idea comes to my mind ) and this diyaudio thing, for me, is just about sharing ideas. Of course I would be glad ( and also boost my ego ) if someone appreciates my constructions.
So I deliberately presented this amp in a sketchy way so people may make comments and react on it as the circuit develops.
But i will now drop this thing. If anyone of you wants to build one, you may take contact with me personally or visit my simple site ( - Embuddy ) and check out my other constructions.
I don't have any schematics published of my Godzilla, but if you want, I can send you schematics on my Baby Godzilla which is completed. I also have pcb layouts and also some boards lying in my drawer that I can send you.
Maiko, I actually have a 24 bit external soundcard. I tried to use it a couple of weeks ago in order to measure THD but I was a bit impatient and I didn't get the thing to work properly. With such a low THD, it's not easy to measure it - it tends to be below the noise floor.
But I'm not really into this THD thing. The only dist that really matters is the cross over dist and that one is usually very low and hard to measure( even though our ears are very sensitive to it ).
This Godzilla project is basically a way for me to compete a bit with the "blameless" guys. A simplistic amp can be made with very low THD.
And I have actually recently inserted a special buffer that turns it to a Turbo Godzilla". The THD is well below -100db up to 10khz and the phase margin still seems to be magnificent - with any capacitive load. It never drops below 120 degrees up to 10mhz.
Simulated, that is, ICG 😀.
But that's more to play with figures. It's not sure at all that it sounds particularly good as AKSA suggested. But my Baby Godzilla sounds very nice, I think.
> PRR: Your submitted sheet must be wrong
I snipped that image from the ST sheet.
I now see that the Fairchild sheet gives a different curve.
I do remember saying, though, that "SPICE has an in-built assumption (unless specifically declared) on low-I fall-off which seems to be pessimistic for many modern device processes (recombination is not as bad as the old days)."
hFE *WILL* fall-off at low and high currents. But what is "low" or "high" depends a lot on process parameters. Generally, on a more complete sheet, there may be Min hFE specs for several currents, but a Max hFE is usually specified for some medium current. This cleverly avoids specifying just where hFE falls off, and leaves leeway for "better" processes, or not, at the maker's convenience.
I snipped that image from the ST sheet.
I now see that the Fairchild sheet gives a different curve.
I do remember saying, though, that "SPICE has an in-built assumption (unless specifically declared) on low-I fall-off which seems to be pessimistic for many modern device processes (recombination is not as bad as the old days)."
hFE *WILL* fall-off at low and high currents. But what is "low" or "high" depends a lot on process parameters. Generally, on a more complete sheet, there may be Min hFE specs for several currents, but a Max hFE is usually specified for some medium current. This cleverly avoids specifying just where hFE falls off, and leaves leeway for "better" processes, or not, at the maker's convenience.
Thank's PRR for digging up this information. I think not many knows the mathematical behavior of the hfe since most constructions simply relies on approximations.
Whats interesting in this very case is the un-linearity of it since that is what creates the distortion. I found out that my simulator indeed estimated a proper "curvation" even though it was bended "the other way around". So regarding the amount of cross over dist, my simulator probably gives a worse result than real life.
Whats interesting in this very case is the un-linearity of it since that is what creates the distortion. I found out that my simulator indeed estimated a proper "curvation" even though it was bended "the other way around". So regarding the amount of cross over dist, my simulator probably gives a worse result than real life.
For those who are interested: ( about simulation vs real life )
I had to find out that issue we discussed about driving the BJT output stages high ohmic. I rigged my amp in such a way that I could measure the THD reasonably good. With no NFB the figures are high enough to be measured with the FFT function of my scope.
Simulations had indicated that THD is much higher when driven high ohmic.
Well, the real life actually had the decency to be fairly similar to my spice simulations. About +-30% or so. The only thing that was worse in real life was the H2, due to the fact that the hfe differs between NPN and PNP and my sim didn't take that into account.
OK, the high THD when current driven should perhaps put one off a bit. But the amount of NFB actually takes away the annoying dist just enough. And measurements show that the high order components are not represented much higher when driven high ohmic.
Then I think the simplicity of the NFB path by itself leads to a nicer subjective impression since the distortion has lower complexity, even if it has a slightly higher level.
I had to find out that issue we discussed about driving the BJT output stages high ohmic. I rigged my amp in such a way that I could measure the THD reasonably good. With no NFB the figures are high enough to be measured with the FFT function of my scope.
Simulations had indicated that THD is much higher when driven high ohmic.
Well, the real life actually had the decency to be fairly similar to my spice simulations. About +-30% or so. The only thing that was worse in real life was the H2, due to the fact that the hfe differs between NPN and PNP and my sim didn't take that into account.
OK, the high THD when current driven should perhaps put one off a bit. But the amount of NFB actually takes away the annoying dist just enough. And measurements show that the high order components are not represented much higher when driven high ohmic.
Then I think the simplicity of the NFB path by itself leads to a nicer subjective impression since the distortion has lower complexity, even if it has a slightly higher level.
- Status
- Not open for further replies.