A better model...
JCX,
If you add a model of a microphone at the beginning of your audio circuit and a model of a speaker at the end of your audio circuit, don't you find that it matters not BJT or FET?
A very good microphone: 30hz to 50khz (>1db); noise: 8 uV rms; distortion: ~0.5% 1khz (hard to find specifications really)
Good speakers: 20hz to 30khz (3db); distortion: 0.05% 1khz
Good amplifier: DC to 530hz; noise < 1uV rms; distortion: <0.01% (20-20khz).
It seems like the electronic specifications exceed the transducer specifications by something like an order of magnitude. Seems like someone here would know, off hand, if this is the case or not.
JF
JCX,
If you add a model of a microphone at the beginning of your audio circuit and a model of a speaker at the end of your audio circuit, don't you find that it matters not BJT or FET?
A very good microphone: 30hz to 50khz (>1db); noise: 8 uV rms; distortion: ~0.5% 1khz (hard to find specifications really)
Good speakers: 20hz to 30khz (3db); distortion: 0.05% 1khz
Good amplifier: DC to 530hz; noise < 1uV rms; distortion: <0.01% (20-20khz).
It seems like the electronic specifications exceed the transducer specifications by something like an order of magnitude. Seems like someone here would know, off hand, if this is the case or not.
JF
ingrast said:
Actually my sim circuit topology is the classic diff pair - Miller integrator VAS and unity gain buffer – the open loop gain at the test frequency is very nearly gm_diff/(2*pi*f*C1) the output conductance/voltage feedback ratio terms in the input diff pair devices have many orders of magnitude less effect
The G1, C1 Miller integrator VAS input impedance is: 1/gm_G1 = 1 Ohm in this sim, easily determined “by inspection” as they say – G1 voltage dependent current source has infinite input impedance so any current from the diff pair must be balanced by C1 || R3 feedback branch current, since E1 also has infinite input impedance no current flows in E1 input, therefore the current in the local feedback branch exactly equals G1 current, the voltage at the input to G1 that gives the input balancing current is v = i/gm_G1, in this sim I've used G = 1 mho, rather higher than typically used in real amps but it keeps the math really simple in this example sim
andy_c said:
The infinite voltage gain of the ideal voltage controlled current source G1 hides the cost of added gain from reducing C1 feedback cap – any real VAS implementation will have finite input, output and shunt impedances that will limit the local voltage gain, by reducing the local feedback and requiring the same voltage swing for less input current the VAS has less error correcting loop gain to linearize its own transfer function – admittedly VAS linearity isn’t often a big concern but I think it worth pointing out that there is no such thing as free gain
Re: A better model...
I really doubt the number with 0.5% distortion !
If you take a look to my post at:
http://www.diyaudio.com/forums/showthread.php?postid=495793#post495793
You will find a distortionmeasurement of a speaker, done with a micro,
most of the distortions are far below 0.5%. If a micro would be that bad,
none of these measurements would make sense...
It was not a bad micro, but not the best on world.
Mike
johnferrier said:
I really doubt the number with 0.5% distortion !
If you take a look to my post at:
http://www.diyaudio.com/forums/showthread.php?postid=495793#post495793
You will find a distortionmeasurement of a speaker, done with a micro,
most of the distortions are far below 0.5%. If a micro would be that bad,
none of these measurements would make sense...
It was not a bad micro, but not the best on world.
Mike
Microphones are usually really low distortion at realistic input levels. Of course at 120-140 SPL they can measure some distortion, depending on their quality and sensitivity. This can be computed backward to estimate distortion at normal input levels, 90dB or so, by considering the elements class A in operation (no xover distortion) and noting the dominant harmonic (usually 2nd).
lumanauw said:
The surprising thing is that the NE 5534 make a much better null than the OPA 2604...
Do your home work and do the null test , you will understand a lot of things...
IF you don't want to do the home work see some values taken by a friend of mine...(Ricardo)
Post # 57 in
http://www.diyaudio.com/forums/showthread.php?s=&threadid=23030&perpage=10&pagenumber=6
Cheers
For me is clear the meaning of your test.
The NE5534 is the most linear device ever made in a sense that it makes the output equal to the input. The more you compare to other devices (I tested a lot of them), the more you realise that, and like your test shows and you don’t need expensive measurement equipment to get to that conclusion...
For me the basic function of an Amplifier is to Amplify without adding anything so, if you want your amplifier like to behave like a "musical instrument" adding al kinds of distortion go ahead and use other type of opamp. If you want to listen to your source without adding nothing you already now what to use because your test proved that.
Regards
Rick
The NE5534 is the most linear device ever made in a sense that it makes the output equal to the input. The more you compare to other devices (I tested a lot of them), the more you realise that, and like your test shows and you don’t need expensive measurement equipment to get to that conclusion...
For me the basic function of an Amplifier is to Amplify without adding anything so, if you want your amplifier like to behave like a "musical instrument" adding al kinds of distortion go ahead and use other type of opamp. If you want to listen to your source without adding nothing you already now what to use because your test proved that.
Regards
Rick
Folks, 'null' testing is interesting and informative, but the residual, whatever it is, must be evaluated with a spectrum analysis, if you really want it to be meaningful. You can try it, of course, but usually LINEAR DISTORTION, do to phase shift, rather than nonlinear distortion, will dominate and limit the test.
I really wish that it did work well, BUT we (Walt Jung, Scott Wurcer, and I) found that even a single capacitor can have a measurable residual when directly compared to another capacitor of similar value, but slightly different material composition. How about an entire amp?
It is true that the apparent audible null is convincing, but why is it that the Hafler 280 is not the truly representative amplifier for the audio industry?
Actually the 280 is very similar in concept and execution to the first Levinson power amp, many of the Krell amps, and the Parasound amps. Yet, each and every amp sounds somewhat different. Why?
I really wish that it did work well, BUT we (Walt Jung, Scott Wurcer, and I) found that even a single capacitor can have a measurable residual when directly compared to another capacitor of similar value, but slightly different material composition. How about an entire amp?
It is true that the apparent audible null is convincing, but why is it that the Hafler 280 is not the truly representative amplifier for the audio industry?
Actually the 280 is very similar in concept and execution to the first Levinson power amp, many of the Krell amps, and the Parasound amps. Yet, each and every amp sounds somewhat different. Why?
Unfortunately the null test mixes non-linear errors and linear errors together. Even a small phase shift is the biggest source of errors. It is impossible to evaluate the results. I had been trying to use this method but left it. The best amp according to this method is that of frequency range DC - MHz region, regardless non-linearities, and that's not what we are trying to find.
For me phase shift is distortion... it's a distortion of time, and for me one of the greatest culprits of the solid state sound, because most comercial solid state amplifiers have huge phase shift...
Ok you can say that valve amplifiers sound soft and have phase shift as well but they have a fundamental reason that i leave to you guys to discover...
For me that an advantage... When you have a low null test, every type of distortion is low... can you identify any type of distortion in noise?
regards
Rick
jcx said:
jcx:
I had the purpose of working your sim last weekend but had to devote it to rebuilding my system after an unsucessful installation (of other software) trashed the disk
I get the point with your reasoning, but what I want to see - it is not clear to me but will work on it - is whether there is an important voltage swing at the fet's drain.
If that were the case, results will be highly dependent on the Vds / Id region where it is taking place because of pinchoff. This is not the case for bjt's, for they approach reasonably to ideal current sources frome the onset.
Will keep you informed whatever the outcome. Will try to include a cascode variation for comparison. Not really proficient for the time being with LTspice since I worked for years with Microcap.
Rodolfo
rickpt,
You are wrong. The phase shift of amplifiers is correlated to their amplitude characteristic. The amplifiers with norrower bandwidth (tube + transformer) have larger frequency dependent phase shifts.
This is quite unimportant, as human ear is unable to evaluate phase shift, it is only able to evalute phase difference between various sources. The phase shift of speakers is huge and exceeds this of amplifiers many many times.
According to the null method the best amp is that with unlimited frequency range. Do you thing that the DC - 10MHz amp must necessarily sound better than that of 10Hz - 35kHz?? If you think yes, then try the amp with unlimited bandwith and start to reduce it step by step. Then tell me your result.
You are wrong. The phase shift of amplifiers is correlated to their amplitude characteristic. The amplifiers with norrower bandwidth (tube + transformer) have larger frequency dependent phase shifts.
This is quite unimportant, as human ear is unable to evaluate phase shift, it is only able to evalute phase difference between various sources. The phase shift of speakers is huge and exceeds this of amplifiers many many times.
According to the null method the best amp is that with unlimited frequency range. Do you thing that the DC - 10MHz amp must necessarily sound better than that of 10Hz - 35kHz?? If you think yes, then try the amp with unlimited bandwith and start to reduce it step by step. Then tell me your result.
PMA said:
Gentlemen:
Please let me stress again a more up to date null test must be done off line processing datasets resulting from high quality input and output digitization.
This way not only AMPLITUDES are nulled, but all SYSTEMATIC linear deviations, so as to be left strictly with what really matters.
This nonwithstanding other references to successfull application of the null test as described elsewhere.
Rodolfo
but you foget one thing...because the transformer this creates a lag in the high frequecies... in the solid state case is quite the oposite... so it isnt only dependent on the amplitude like you say...you can have lag or advance of phase...
So you are wrong...
talk about your speakers, not mine...
For me an amplifier should have enough bandwith so that it dosent exibit phase shift in the audio range...
i dont use any type of filters on my amplifiers, so that should answer your question...
regards
Rick
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.