Which is discussed in APAD 6th edtion on page 148.
It is then demonstrated that cascoding fixes this.
Maybe you missed that bit.
I would suggest that this is not proven. A simple current-source usually works very well. Of course it depends on what you mean by 'near-ideal'.
A technique for stabilising RF power amplifiers against parasitic oscillation.
See:
Neutralization of RF Amplifiers
I have never heard of its use in audio amplifiers.
In saying that VAF is the voltage at which hfe doubles, I assume you are referring to Vcb, so the "reference" hfe is at Vcb=0V. Is that correct? I had not heard that description of VAF before, either. I think that you are also right about the nonlinearity in Early effect being fairly small once one is reasonably far away from quasi saturation. Even if Earlt effect is modeled as linear, the key is that it models current gain as being a function of Vce, and when a transistor parameter varies with signal, we have the potential for distortion. So even though SPICE models Early effect as linear, the simulation should catch the majority of distortion caused by Early effect.
In looking at input current signal and its distortion, a significant component can be in the common mode, due to Early effect or other effects in the tail current source. With perfectly balanced circuitry and a perfect current mirror, this will largely come out in the wash EXCEPT for input current. If the tail current has some signal current in it, and that signal current is distorted, that distortion will appear as distorted input current.
It is worth noting that nonlinearity in the tail current will appear in the amplifier output even with zero source and feedback network impedances if the current mirror is not perfect (i.e., if it has finite common-mode rejection). Amplifiers that do not use a current mirror will be especially vulnerable to distortion in the IPS tail current. The IPS tail current source wants to be really good in any high-performance design.
In simulation, I discovered significantly distorted signal input current, to the tune of about 2.5% THD-1 at 1kHz in the Blameless design that Self put forth in discussing input current distortion. Through several simulation experiments, I discovered that this distortion was mainly coming from distorted signal current in the LTP tail current (NOT input-referred distorted current from the forward path). Further experimentation showed that the distorted tail current signal was a result of the bias line that was shared between the VAS feedback current source and the IPS tail current source (NOT from Early effect in the tail current source transistor). This was discovered by creating the tail current source from a completely independent feedback current source, and seeing that the distortion in the tail current and input current went down by a factor of about 5.
The sharing of the bias among the VAS and tail current sources is a very bad idea, asking for trouble. It appears to be the major contributor to the nonlinear input current that I have seen thus far in the simulations I have done. That problem is fixed by spending merely that one transistor and resistor that was foolishly saved by sharing the bias line.
Cheers,
Bob
The late Charles Hansen mentioned neutralization techniques as providing clues for developing an approach to mitigating oscillation in his output stages.
mlloyd1
mlloyd1
Hi Harry,
thank you for the effort and the kind words. But I think it is better to not to contribute here.
Thanks again for the D44/D45 models that you made available some time ago.
Apart from the technical issues I have learnt in this thread that OTL amps may be mentally dangerous if abused.
Kind regards,
Matthias
Last edited:
It can do that.... but not its main use.
It works well for any amplifier... audio included. I told John Curl how to do it decades ago for diff pair ips. JC tried it and it worked for him also.
It extends BW by cancelling the IPS stage C's. And, thereby also reduces distortion from that device's C.
THx-RNMarsh
Last edited:
I don't think that that makes a difference. There is nothing connected to the node R1 -> bQ1. So each and every electron through R1 goes into the base.
Then how can I(R1) != I(bQ1)? Schematic post 8488, right?
BTW Nice to see you here again Harry. Trust you are doing well!
Jan
Isn't neutralization a form of positive feedback? Even Harry Black wrote about it about a century ago ;-)
Scott used it in his AD797 iirc.
Jan
take diff ips (bjt) ---- connect a value of c to collector of one and to base of other. do it that way to both bjt.
Roughly, if the C are same as ips c's, they cancel.
You can use a backwards diode or better, another transistor as diode of same type as ips diff. and get better dynamic cancellation using diode C.
I just found you a schematic of my description. Best though to use c of same transistor connected as diode. [for CN1 and CN2]
![US07256646-20070814-D00000[2].png](https://www.diyaudio.com/community/data/attachments/591/591700-ebb2dfc3a0525dfda444850b2264479d.jpg "US07256646-20070814-D00000[2].png")
THx RNMarsh
Roughly, if the C are same as ips c's, they cancel.
You can use a backwards diode or better, another transistor as diode of same type as ips diff. and get better dynamic cancellation using diode C.
I just found you a schematic of my description. Best though to use c of same transistor connected as diode. [for CN1 and CN2]
THx RNMarsh
Last edited:
In high school geometry it's called "similar triangles".
Triangle AOX is similar to triangle ABY since (i) both are right triangles, and (ii) both have the same acute angle XAO (== acute angle YAB).
Since the base of triangle ABY is exactly twice the base of triangle AOX, the height of triangle ABY is also exactly twice the height of triangle AOX.
At the same base current, when VCE=Va the collector current has doubled, from X to Y. Thus the ratio Ic/Ib has doubled.
It's geometry.
_
Attachments
Hi
Is this something that you have measured or is it done by simulation?
If it's done by simulation I think that it's important to understand that when it comes to VAF some BJT "LTSpice" Models are very wrong.
BTW.: I dont think that "Vaf is the voltage at which Hfe doubles" is a correct statement.
All the best
Reodor
It also fixes it by doing the tail current source as an independent feedback current source, which is the way it should have been done in the first place. By obtaining the bias for the tail current source transistor from the VAS feedback current source, you lost all of the Early effect immunity that would have been achieved simply by using a separate feedback current source for the LTP.
I should have been more clear about what I meant by ideal current source. It should be one with a very high output impedance. I did not mean that it needed to be extremely accurate or accurately temperature compensated.
Cheers,
Bob
I had to go back over the post and remove all my errors, that one slipped by. the 57nApp would be mostly fundamental, and the 5% guesstimated nonlinearity applies to it, so only a small fraction of it would be harmonics.
I make mistakes!
People don't always catch them though.
Last edited:
I mentioned in the post that I compared the simulation models to actual measured data, which is where I got the figures from. The measured data is noisy, but it does give an idea of the worst case nonlinearity of Early effect.
Last edited:
You are right - many transistor models out there (LTspice or not) are very wrong when it comes to the VAF value. Measuring VAF with DC measurements must be done very carefully to avoid heating the transistor junction and thereby increasing beta. A better approach may be to set up a jig that measures dynamic output resistance of the transistor when connected in CE mode and properly biased.
Cheers,
Bob
Yes, there is. Nodes either side of R1 are labelled "r" and "b" respectively. These nodes are then connected to C1 and C3 respectively.
Not too bad thanks but don't get any time to do any audio research
Too busy with other things at home and work. Like this (open-access (free access to all) IEEE Transactions on Power Electronics paper).
in http://www.diyaudio.com/forums/soli...lls-power-amplifier-book-849.html#post5297500 #8490
r (left side of C1) is connected to r (right side of C2)
b (base of Q1) is connected to b (left of C3)
and from there to R2, R6 and U1
Last edited:
Yes, I like to do that, it keeps the drawing clear and clean (but than... adds it's own mechanism of confusion
).