Sorry , I was just using the term " leach" to generalize
the topology , I will be more specific.
I have used "Ideal" sources , and that does improve results.. but
I feel that is "false" in relation to the real circuit.
I did a thd20 on the apt clone , it does exibit more THD with 68/100pf inbalance. .005 vs .006 not bad .
for my leach test I went true to the original .. the "fighting"
was There (thd went way up), but the VAS'es did not "choke".
OS
Hi ostripper,
The question would be, do you think the increase would be audible? It may not stick out, but this increased distortion may give a sense of being "harsh" sounding, for lack of a better term.
My experience with measured THD is that if you can measure it, those distortion products can be heard by normal people.
-Chris
The question would be, do you think the increase would be audible? It may not stick out, but this increased distortion may give a sense of being "harsh" sounding, for lack of a better term.
My experience with measured THD is that if you can measure it, those distortion products can be heard by normal people.
-Chris
Bob is talking about a complementary VAS in which the impedance "seen" by the bases of the VAS transistors is large - on the order of tens of kOhms. This makes the Miller loop gain large, which in turn makes the output impedance of each half of the VAS small in the audio frequency range (because of the high local feedback). This leads to the VAS "fighting" problem. Neither the Leach amp nor the APT will ever approach this condition without some major changes, so these amps have nothing whatever to do with this subject.
ostripper said:
You do indeed and to twist and change topics too. You clearly stated your comparison sims in that other thread were at 1kHz, and now it is 20kHz!
anatech said:
LOL!
I was NOT talking about an EF buffered VAS! The topic was input bias current!
An EF buffer on the VAS does not reduce input bias current. I never even mentioned EF VAS buffers!
Re: Re: Re: VAS fighting
Hi Bob.
With 20k LTP loads is is not surprising that this circuit will have the VAS fighting issue.
However I have played with this circuit in the past and I can assure you that it does not work in real life with out a CMCL as the input stage gain is far too great (The VAS will not bias properly with the slightest Ic imbalance between the complementary long tail pairs.
Cheers,
Glen
Bob Cordell said:
Hi Bob.
With 20k LTP loads is is not surprising that this circuit will have the VAS fighting issue.
However I have played with this circuit in the past and I can assure you that it does not work in real life with out a CMCL as the input stage gain is far too great (The VAS will not bias properly with the slightest Ic imbalance between the complementary long tail pairs.
Cheers,
Glen
Hey Glen,
Have you tried CMCL with complementary input diff amps? I've seen posts describing its use in a CFB configuration, but never VFB. Just curious.
Have you tried CMCL with complementary input diff amps? I've seen posts describing its use in a CFB configuration, but never VFB. Just curious.
Thank you, that was easily "digestable" (very easy).
That leads me to the next simulation , the apt with emitter followers.. 😀
John , I don't like to be subjective , but human ears are.
I now have the ultimate "reference" (genesis stealth) to
compare topologies. I now have a genesis , apt , aska55(clone),
and a blameless in the same room (house).
rank...
1. the genesis .. class a rules.
2. apt and the aska clone (started out as a dx)
3 . blameless
4 . I had a leach for sub use , done to the letter.
I did not like the sound.
Hugh dean and I agree that the Distortion does not matter
to a point. The components do (h2/3/5/7). This knowledge
is not biasing my judgement as I am listening to these amps
daily as I work for long periods.
What I am hearing is absolutely being foretold by the FFT. 😎
OS
That leads me to the next simulation , the apt with emitter followers.. 😀
John , I don't like to be subjective , but human ears are.
I now have the ultimate "reference" (genesis stealth) to
compare topologies. I now have a genesis , apt , aska55(clone),
and a blameless in the same room (house).
rank...
1. the genesis .. class a rules.
2. apt and the aska clone (started out as a dx)
3 . blameless
4 . I had a leach for sub use , done to the letter.
I did not like the sound.
Hugh dean and I agree that the Distortion does not matter
to a point. The components do (h2/3/5/7). This knowledge
is not biasing my judgement as I am listening to these amps
daily as I work for long periods.
What I am hearing is absolutely being foretold by the FFT. 😎
OS
andy_c said:
To be fair any non-ideal symmetrical VAS WILL have an issue with common-mode signal currents (I'm not just talking about VAS fighting here due to Miller cap values), it's just a matter of degree.
Any asymmetry in the gain between symmetrical push-pull stages will cause common mode currents to flow, and this strictly has nothing to do with frequency compensation. Even an LTP, for example, with a current mirror load will suffer common mode effects if the current mirror does not perfectly complement the opposing leg signal current.
The effect exists in the Leach amp and The APT (but not much). My main point regrading these two amps was to point out that electrically, they are both symmetrical topologies that inherently cancel even harmonics.
Re: Re: Re: Re: VAS fighting
Absolutely true, Bob's circuit is one of those simulation traps. As you already know, in simulations the VAS biases correctly even without any LTP load. Dare to build it 🙂
G.Kleinschmidt said:
Absolutely true, Bob's circuit is one of those simulation traps. As you already know, in simulations the VAS biases correctly even without any LTP load. Dare to build it 🙂
Re: Re: Re: Re: Re: VAS fighting
The simulator will actually show you just how non-workable the circuit is, you just need to be aware of the issues and know what you are looking for.
Just insert a voltage source of several mV in series with the emitter of one of the LTP transistors to simulate an offset error between the complementary halves, then double check the quiescent VAS current 🙂
Cheers,
Glen
syn08 said:
The simulator will actually show you just how non-workable the circuit is, you just need to be aware of the issues and know what you are looking for.
Just insert a voltage source of several mV in series with the emitter of one of the LTP transistors to simulate an offset error between the complementary halves, then double check the quiescent VAS current 🙂
Cheers,
Glen
Wow 😀 this actually did not turn out too bad. I tried
emitter followers on both my simulations.
On the complementary input diff amp I saw the "effect"
the VAS "died" 😎 .
I am very happy with the APT with followers "uncrippled"
BUT with the inbalance THD went way up.(but it did not die)
OS
emitter followers on both my simulations.
On the complementary input diff amp I saw the "effect"
the VAS "died" 😎 .
I am very happy with the APT with followers "uncrippled"
BUT with the inbalance THD went way up.(but it did not die)
OS
Re: Re: Re: Re: Re: Re: VAS fighting
I just simmed Bobs circuit. An input offset error of +/-15mV (between the complementary LTP's) is enough to cause the quiescent VAS current to vary ~ +/- 10mA from the nominal 10mA.
I guess the issue here could be described as input offset fight. If there is an opposing input offset error between the long tail input pairs, the global nfb can never steer both LTP's into balance (in attempting to balance one the other will be thrown out).
For optimum performance you really do not want the LTP imbalance to cause a variation of more than +/-10% to the calculated, or ideal VAS standing current. In the case of Bobs amplifier, this would mean trimming the input offset error between the complementary LTP's to 1.5mV or better.
The most basic fix is to limit the LTP gain so that the resultant Ic imbalance and the deviation in quiescent VAS current is maintained within reasonable limits (and this will also fix the VAS fighting issue 😀 )
Cheers,
Glen
G.Kleinschmidt said:
I just simmed Bobs circuit. An input offset error of +/-15mV (between the complementary LTP's) is enough to cause the quiescent VAS current to vary ~ +/- 10mA from the nominal 10mA.
I guess the issue here could be described as input offset fight. If there is an opposing input offset error between the long tail input pairs, the global nfb can never steer both LTP's into balance (in attempting to balance one the other will be thrown out).
For optimum performance you really do not want the LTP imbalance to cause a variation of more than +/-10% to the calculated, or ideal VAS standing current. In the case of Bobs amplifier, this would mean trimming the input offset error between the complementary LTP's to 1.5mV or better.
The most basic fix is to limit the LTP gain so that the resultant Ic imbalance and the deviation in quiescent VAS current is maintained within reasonable limits (and this will also fix the VAS fighting issue 😀 )
Cheers,
Glen
Attachments
Why one has to bother with "classical" VAS approach, when more simple and stable folded cascode "VAS" solve so many problems?
Please do not give me the argument of not enough gain ... If the LTP is well designed, there are not much distortions to reduce through GNFB 😉
Please do not give me the argument of not enough gain ... If the LTP is well designed, there are not much distortions to reduce through GNFB 😉
Re: Re: ratio
In either case, that means a weak rationale behind the ratio. 😉
Cheers,
Edmond.
Bob Cordell said:
In either case, that means a weak rationale behind the ratio. 😉
Cheers,
Edmond.
Audio is not RATIONAL, it is subjective! It is what it is, and you know when you get it right, or at least, closer to being 'right'. The test equipment, simulations, etc, are TOOLS, not the audio experience, itself.
Most successful audio designs that rise above mid-fi, in subjective appreciation, have something about them that works better than most other audio designs. To be sure, there are 1000's of different audio designs that have been produced by serious engineers, and put into the marketplace. How many make the top 100 of all time? How many are still sought after by collectors and music listeners, who are willing to pay many times more for the item than it originally cost, and still are thankful for getting it?
Some here, including many of the most recent contributors here, will scoff, and say that it is 'marketing', 'pretty finish', 'name recognition' , that makes a successful design, forgetting that the competition is fierce, and most audio designs just come and go, after a short sojourn in the marketplace, to be heard about no more.
It is this difference, that the BLOWTORCH thread has tried to teach people how to build into their designs, so that they too can be successful in the audio world.
Most successful audio designs that rise above mid-fi, in subjective appreciation, have something about them that works better than most other audio designs. To be sure, there are 1000's of different audio designs that have been produced by serious engineers, and put into the marketplace. How many make the top 100 of all time? How many are still sought after by collectors and music listeners, who are willing to pay many times more for the item than it originally cost, and still are thankful for getting it?
Some here, including many of the most recent contributors here, will scoff, and say that it is 'marketing', 'pretty finish', 'name recognition' , that makes a successful design, forgetting that the competition is fierce, and most audio designs just come and go, after a short sojourn in the marketplace, to be heard about no more.
It is this difference, that the BLOWTORCH thread has tried to teach people how to build into their designs, so that they too can be successful in the audio world.
famous names
I had once a famous McIntosh MC2300: http://mcintoshlaboratory.tripod.com/aa/mc2300.htm
That beast (58kg!) looks very impressive, but it sounded mediocre and it measured mediocre.
I had once a famous McIntosh MC2300: http://mcintoshlaboratory.tripod.com/aa/mc2300.htm
That beast (58kg!) looks very impressive, but it sounded mediocre and it measured mediocre.
- Status
- Not open for further replies.