Re: Re: Re: 100mV Line Level
Pavel,
Thanks for your informative answer!
Is that generator developed by your firend perhaps something you could present, with allowance from your friend of course, I would be interested in it!
You can also e-mail me direct in order to keep it outside of the threads mainissue for the moment.
Michael
PMA said:
Pavel,
Thanks for your informative answer!
Is that generator developed by your firend perhaps something you could present, with allowance from your friend of course, I would be interested in it!
You can also e-mail me direct in order to keep it outside of the threads mainissue for the moment.
Michael
Re: Re: Re: CFP vs CE
Oops. I posted when I was in a hurry. When I said "I hate seeing stuff like that" I meant the peak in transconductance above 10 MHz for the CFP. Such things can be pretty hard to get rid of. I haven't looked at Walt's PDF yet but am looking forward to reading it. But the way I wrote it, it looks like it's come kind of commentary on what Walt did, and I didn't mean it that way at all.
Ultima Thule said:10 MHz for the CFP I guess you meant, ok?!
Sorry for asking, I am just not familiar with LTSpice, I used PSpice for 3 years ago the last time and nowdays don't have access to that SW, LTSpice would take too long time, ..at least my curiosity can't wait! Lazy..?
Oops. I posted when I was in a hurry. When I said "I hate seeing stuff like that" I meant the peak in transconductance above 10 MHz for the CFP. Such things can be pretty hard to get rid of. I haven't looked at Walt's PDF yet but am looking forward to reading it. But the way I wrote it, it looks like it's come kind of commentary on what Walt did, and I didn't mean it that way at all.
Mike Gergen said:
Unfortunately a frequent conceptual error.
The differential input signal (i.e. the difference between the + and - input pins) is what actually gets amplified to yield the output.
You can look at it as the desired signal plus a predistorted correction that gets (fairly) cancelled in turn by the nonperfect open loop transfer function.
The higher the open loop gain, the smaller both the input differential desired signal and the correction factor (or error if you want). In a customary aproximation, it may be considered zero, so the designation "virtual ground".
Again, the differential input should be seen as a small predistorted input to be amplified by the open loop transfer characteristic. Look at the circuit. The only thing that stands between +/- inputs and output is the naked device.
Rodolfo
I've done the series buffer opamp array. And I do find that NE5532 gives minimal difference to the original sound. Why I do that? Because I cannot hear clearly the difference if I only use 1 opamp. (This is the DIYer way to compare opamp sound without measurement equipment)
About NE5532 is good or bad, maybe it is given back to the user. If the user needs opamp that will be put in alot of number in series (Like in Mixer manufacturer) maybe he will say that NE5532 is the best opamp.
But if the one who use opamp is just a hobbyist with single opamp for preamp, he can choose whatever opamp that gives good sound (not knowing that it is the "coloured" sound, or very clean/non distorted sound).
About NE5532 is good or bad, maybe it is given back to the user. If the user needs opamp that will be put in alot of number in series (Like in Mixer manufacturer) maybe he will say that NE5532 is the best opamp.
But if the one who use opamp is just a hobbyist with single opamp for preamp, he can choose whatever opamp that gives good sound (not knowing that it is the "coloured" sound, or very clean/non distorted sound).
lumanauw said:
The Diyer way ...but also a much clever way!
The NE 5532 is the best , for people searching for accuracy and not a "nice sound".
I wonder how dull and lifeless the recordings will be , if the chain of op amps that the signal pass trough in the recording studios were not NE5532.
Some call this "voicing" the sound to one 's preference...but that's not accuracy.
Congratulations Lumanauw...you see the light!!
Cheers
bjt - fet input comparison
Taking from the excellent work by jcx, I tweaked the simulation a little more in order to address some issues raised at least in my oppinion.
The original simulation makes clever use of some tricks in order to isolate what the actual device behaviour is and as such it works fine.
Yet I was left with a nagging feeling the configuration does not match adequately with actual operating conditions in the field.
The strong feedback constructed from ideal controlled current and
voltage sources could in principle be emulated by a local FB loop (in the case of an actual amplifier) but could not be really applicable to a complete device (i.e. a global loop).
With this in mind, I modified both the Miller integrator resistor from 5 Megs to 10k and rised the feedback divider from the original 13 / 1 to 1300 /1. In doing so I lowered significantly the applied feedback.
The other variation is to include 210 ohm emitter resistors for the bjt pair - following jcx advise - bringuing the equivalent transconductance to aproximately the same as the one corresponding to the fet pair.
Taking from the excellent work by jcx, I tweaked the simulation a little more in order to address some issues raised at least in my oppinion.
The original simulation makes clever use of some tricks in order to isolate what the actual device behaviour is and as such it works fine.
Yet I was left with a nagging feeling the configuration does not match adequately with actual operating conditions in the field.
The strong feedback constructed from ideal controlled current and
voltage sources could in principle be emulated by a local FB loop (in the case of an actual amplifier) but could not be really applicable to a complete device (i.e. a global loop).
With this in mind, I modified both the Miller integrator resistor from 5 Megs to 10k and rised the feedback divider from the original 13 / 1 to 1300 /1. In doing so I lowered significantly the applied feedback.
The other variation is to include 210 ohm emitter resistors for the bjt pair - following jcx advise - bringuing the equivalent transconductance to aproximately the same as the one corresponding to the fet pair.
bjt - fet waveforms
A further observation is both Q1 and J2 are loaded by high impedance current sources, and I am not convinced this situation is fair or representative, a cascode load for the fet in principle should be a distinct advantage for several reasons. The bjt by nature behaves much better as a current source and is therefore less affected.
The selected input signal level and operating point I find is more nearly matched to what is to be expected for an input stage.
This plot shows input voltage to the diff pairs and collector / drain current. Input signall voltage is from the same source, and feedback voltage plots superimpose neatly as can be seen.
Input voltages are almost identical and output currents are also alike, meaning equivalent transconductance for both pairs if fairly well matched.
A further observation is both Q1 and J2 are loaded by high impedance current sources, and I am not convinced this situation is fair or representative, a cascode load for the fet in principle should be a distinct advantage for several reasons. The bjt by nature behaves much better as a current source and is therefore less affected.
The selected input signal level and operating point I find is more nearly matched to what is to be expected for an input stage.
This plot shows input voltage to the diff pairs and collector / drain current. Input signall voltage is from the same source, and feedback voltage plots superimpose neatly as can be seen.
Input voltages are almost identical and output currents are also alike, meaning equivalent transconductance for both pairs if fairly well matched.
Attachments
fet - bjt FFT
In this configuration, the output signal spectrum shows the fet pair to be slightly worse than the bjt, with a well defined second harmonic and associated IM products.
No attempt has been made to optimize each pair in its best performance regime, much less to select devices representative of what is best suited in each realm - bjt / fet - be it discrete devices or monolitic examples for IC application.
Rodolfo
In this configuration, the output signal spectrum shows the fet pair to be slightly worse than the bjt, with a well defined second harmonic and associated IM products.
No attempt has been made to optimize each pair in its best performance regime, much less to select devices representative of what is best suited in each realm - bjt / fet - be it discrete devices or monolitic examples for IC application.
Rodolfo
Attachments
Good work, Rodolfo and AndyC.
I might suggest that it is the higher order odd harmonics that are of most concern. You should match your devices, just because we normally would use matched FET's or bipolars in quality designs, and the resultant 2'nd harmonic from any mismatch potentially obscures the resolution of higher order harmonics. For the record, we can happily listen to a great deal of added third harmonic distortion as well as 2'nd. This has been shown by listening to decades of analog tape recording that have typically 1-10% third harmonic at 0 Vu and higher levels. 0.1% harmonic distortion is virtually un-noticable, but don't think that almost any amount of 7th harmonic is! Also, you might consider noise, when you add resistive degeneration to bipolar transistors. Fets will be lower noise, and have almost no bias current to deal with as well!
I might suggest that it is the higher order odd harmonics that are of most concern. You should match your devices, just because we normally would use matched FET's or bipolars in quality designs, and the resultant 2'nd harmonic from any mismatch potentially obscures the resolution of higher order harmonics. For the record, we can happily listen to a great deal of added third harmonic distortion as well as 2'nd. This has been shown by listening to decades of analog tape recording that have typically 1-10% third harmonic at 0 Vu and higher levels. 0.1% harmonic distortion is virtually un-noticable, but don't think that almost any amount of 7th harmonic is! Also, you might consider noise, when you add resistive degeneration to bipolar transistors. Fets will be lower noise, and have almost no bias current to deal with as well!
Unmatched Transistors...
JC,
so you say that unmatched transistors in the diffpair, or why not also the outputstage with 2 or more pairs, WILL add more distortion...
How dramatically is this?
However if we just focus on the input diff stage, I myself wondered this just for couple of days how an unmatched diffpair would distort compared to a matched pair.
I actually red Upupa Epop's comment in the issue seen in post #68 and #70 in another thread which got me thinking IF we use unmatched pair then each transistor will work under diffrent biasing condition and thereby parameters will be slightly offseted compared to each other.
But also, can DC servo fix the "problem"?
Michael
john curl said:
JC,
so you say that unmatched transistors in the diffpair, or why not also the outputstage with 2 or more pairs, WILL add more distortion...
How dramatically is this?
However if we just focus on the input diff stage, I myself wondered this just for couple of days how an unmatched diffpair would distort compared to a matched pair.
I actually red Upupa Epop's comment in the issue seen in post #68 and #70 in another thread which got me thinking IF we use unmatched pair then each transistor will work under diffrent biasing condition and thereby parameters will be slightly offseted compared to each other.
But also, can DC servo fix the "problem"?
Michael
Michael, I had build many the same amps with these SJ/SK. They was matched to groups in tolerance 1 mA of Ids, not better. In listening comparisions was any differeces of group with higher end of this group or with lower end ( all was with " BL " marking ), so I thing although maybe by measuring you can find differences, by listening not ( I am talking now about listening with the " same " amps, not about difference types of amps. DC servo helps with offset and DC drift, but sound depend on basic connection of amp. By many listening experiences in many types of amps, I may to say, that on all cases ( types ) was better listening results with fet's than with bjt's
.
.
Hi, Mr.JC,
Your design do use FET for differential, but for VAS it is still bipolar. There also RE degeneration in this VAS.
How important is the resistor-noise in VAS degenerated bipolar? VAS seems always have this RE degeneration, especially in symmetrical designs, mainly to set bias in VAS.
And I also find that if using bipolar differential, RE degeneration is a must, even small values like 22ohm
Your design do use FET for differential, but for VAS it is still bipolar. There also RE degeneration in this VAS.
How important is the resistor-noise in VAS degenerated bipolar? VAS seems always have this RE degeneration, especially in symmetrical designs, mainly to set bias in VAS.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.