AudioXpress :
April 2007
May 2007
Do not miss these :
September 2007 (mail)
April 2009 (mail)
These articles show how it's easy to achieve excellent CCSs of a few mA.
Provided their PSRR is good enough, I think that in most circumstances as a long tail of a differential input stage or as a VAS load, a real CCS does not make a difference with what an ideal CCS should give.
However I lack of data on how CCS noise can impair the performance of amplifier circuits.
Last edited:
Hi Jan,
I have just read those articles, and, more importantly, the correspondance that followed.
I am afraid I disagree with Jung's conclusion in respect of, in particular, the amplified negative feedback current source, which he calls, rather inadequately, the "one Vbe" current source.
I know for a fact that the amplified negative feedback current source possesses a significantly higher output impedance than, for example, what Jung calls the "two diode" current source. This is also the view held by no less a luminary than the late Hans Camenzind on pg 77 of his book here:
http://www.designinganalogchips.com/_count/countdown.pl?designinganalogchips.pdf
Its noteworthy that Walt Jung was taken to task in this regard by Thomas Bohley in the correspondance section of AudioExpress of September 2007. Folk are incouraged to read his rebuttal.
Hi,
there are cases where an improvement of the CCS with regard to current constancy doesn´t improve the overall circuit´s performance.
Other parameters as e.g. noise may be of greater importance.
In that case it seems that the simplest ccs. the degenerated JFET is the top performer.
Such a circuit could be the interface between a current output DAC like the PCM179x or AD1955 to the IV-converter stage.
If Your converter is designed for an DAC-output current centrered around 0mA, a current source or sink may be required to null the DACs output offset current.
The noise of this ccs will directly affect the noise performance of the overall circuit. While the simple JFET ccs performs just well enough to not influence distortion figures, it outperforms any other more complicated CCS I´ve tried noisewise.
As so often in Audio ... simple is better ... at least in this application.
jauu
Calvin
there are cases where an improvement of the CCS with regard to current constancy doesn´t improve the overall circuit´s performance.
Other parameters as e.g. noise may be of greater importance.
In that case it seems that the simplest ccs. the degenerated JFET is the top performer.
Such a circuit could be the interface between a current output DAC like the PCM179x or AD1955 to the IV-converter stage.
If Your converter is designed for an DAC-output current centrered around 0mA, a current source or sink may be required to null the DACs output offset current.
The noise of this ccs will directly affect the noise performance of the overall circuit. While the simple JFET ccs performs just well enough to not influence distortion figures, it outperforms any other more complicated CCS I´ve tried noisewise.
As so often in Audio ... simple is better ... at least in this application.
jauu
Calvin
Hi,
well, the simplest seems to me to do a FFT and compare the noise ground.
Quite often the actual noise figure may be of lower interest than the qualitative value and the shape of noise amplitude response.
In that special case of a current sink nulling a DACs offset current, the output noise of the complete IV-stage will probabely mostly be due to the ccs-noise.
jauu
Calvin
well, the simplest seems to me to do a FFT and compare the noise ground.
Quite often the actual noise figure may be of lower interest than the qualitative value and the shape of noise amplitude response.
In that special case of a current sink nulling a DACs offset current, the output noise of the complete IV-stage will probabely mostly be due to the ccs-noise.
jauu
Calvin
Further in his response to Whitney (Audioxpress 09/07), Jung unwittingly demonstrates the incorrectness of relaying on the supply rejection of the current source to deduce its output impedance.
He realises that by splitting the reference resistor into two and connecting a capacitor from the mid point of the resistors to the supply rail, as recommended by Douglas Self, the supply rejection increases drastically.
This fact should have indicated to Jung that the inclusion of this capacitor in no way increases the output impedance of the current source even though supply rejection is increased, and, therefore, that using supply rail rejection to estimate output impedance for a current source gives completely erroneous results.
Additionally, Jung suggests that the amplified negative feedback current source, which he calls the "one Vbe current source", requires a small resistor in series with the base of the control transistor for stability.
This is completely untrue at least in audio amplifier applications as can be deduced by simulating the ANF current source's loop gain in SPICE. This should demonstrate that said ANF current source possesses completely innocuous LHP singularities giving a phase margin in excess of 90 degrees.
This is completely untrue at least in audio amplifier applications as can be deduced by simulating the ANF current source's loop gain in SPICE. This should demonstrate that said ANF current source possesses completely innocuous LHP singularities giving a phase margin in excess of 90 degrees.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.