mikek:
Ok, most times this statement is true. But: i tried to build a chain with the less parts as possible, erery channel has except powersupplys now about 4 transistors. I don´t know how much transistors are used in my TDA 1305 DAC, but as i read in DIYOPAMP.PDF from NP, a typical Opamp could be build out of 4 transistors. You see, mostly there will be much more parts in the CDP than in the rest of the chain, but it isn´t necessarily that way. I´ll build another DAC next time, to get rid of some active devices at that place...
Ok, most times this statement is true. But: i tried to build a chain with the less parts as possible, erery channel has except powersupplys now about 4 transistors. I don´t know how much transistors are used in my TDA 1305 DAC, but as i read in DIYOPAMP.PDF from NP, a typical Opamp could be build out of 4 transistors. You see, mostly there will be much more parts in the CDP than in the rest of the chain, but it isn´t necessarily that way. I´ll build another DAC next time, to get rid of some active devices at that place...
example of pointless misuse of components
Why do I do this.......? The current source feed provides greater rejection of supply noise for the zener since it does not have a zero ohm impedance. This is a common circuit and can measurably increase the rejection of supply noise by the current source. This is not an"example of pointless misuse of components" any more than your post would be called a pointless misuse of this forum.
H.H.
Why do I do this.......? The current source feed provides greater rejection of supply noise for the zener since it does not have a zero ohm impedance. This is a common circuit and can measurably increase the rejection of supply noise by the current source. This is not an"example of pointless misuse of components" any more than your post would be called a pointless misuse of this forum.
H.H.
Hi, everyone!
The discussion about the specific circuit is totally of track. Maybe I will start a thread about audio gear philosophy.
I think Nelson Pass forgot to divide the DIY's into two groups: those who materialize (is there such a word in english? in swedish we have "materialsera") other ideas and those who materializes their own. Some ideas on the net aren't that good but I think Mr. Pass' ideas are cool, especially the lamp amp!
Since we are of track I would clearify my point of view:
Don't forget to listen to the music!
It's fun to test different gear but some differencies are small, hardly noticable and also not very important in order to enjoy the music.
My attitude is, it's fun to design but I don't need the performance. There are two halves of me, the designing process and the using of the gear.
The discussion about the specific circuit is totally of track. Maybe I will start a thread about audio gear philosophy.
I think Nelson Pass forgot to divide the DIY's into two groups: those who materialize (is there such a word in english? in swedish we have "materialsera") other ideas and those who materializes their own. Some ideas on the net aren't that good but I think Mr. Pass' ideas are cool, especially the lamp amp!
Since we are of track I would clearify my point of view:
Don't forget to listen to the music!
It's fun to test different gear but some differencies are small, hardly noticable and also not very important in order to enjoy the music.
My attitude is, it's fun to design but I don't need the performance. There are two halves of me, the designing process and the using of the gear.
"The feedback loop is the coolest thing about this amp. You can use a small film cap and still use low value feeback resistors so as not to run into input capacitance problems with the diff pair feedback input. I have used this several times and I love it." The Sumo Model 9 circa 1980 used this design too. Bongiorno used a 0.1µF cap with a 10M in parallel in the feedback loop. The FET inputs on the 9 had a stated input impedance of 10^12 ohms so the input off-set current was quite low even with the 10M resistor.
"I got very interested in these japanese design with a special stage which connects the diff input stage with the output. Could anyone explain a little bit more of the benefits of this weird connection? " The Lang amplifier in Audio Amateur 2/86 used this technique. The circuit board is still available
http://www.audioxpress.com/bksprods/pcbs/solidstate.htm
The 2/86 article cites a JAES reference that I remember reading, but can't remember the author/date off-hand. The JAES article has a lot of distortion graphs showing the improved performance of this topology.
"I got very interested in these japanese design with a special stage which connects the diff input stage with the output. Could anyone explain a little bit more of the benefits of this weird connection? " The Lang amplifier in Audio Amateur 2/86 used this technique. The circuit board is still available
http://www.audioxpress.com/bksprods/pcbs/solidstate.htm
The 2/86 article cites a JAES reference that I remember reading, but can't remember the author/date off-hand. The JAES article has a lot of distortion graphs showing the improved performance of this topology.
peranders:
You wrote a posting i can agree with fully. Very good comment.
Start this pholosophy thread.
We should keep in mind that Mr Pass has not only ideas which are really worth to use time for understand them, her also seems to be a very tolerant human beeing.
Another point is, Mr Haller writes in a style that is formal provocative, but the content of his postings are often worth reading. We just should overlook the style, maybe the style of speech is correlated with his location, wasn´t his President from the same location, from Texas too?
Theyer socialisation maybe is very different from a european one, but i think he is a "friendly cowboy", he reads the "********" we greenhorns write, and uses his time to write comments. Thats ok.
(PS, if we talk wine, in mr. passes country Mr. Robert Mondavi produces a quite good but expensive Cabernet Sauvignon, my favorite is a same quality Lagrein from Südtirol / Italy made by Joseph Brigl for half the money)
You wrote a posting i can agree with fully. Very good comment.
Start this pholosophy thread.
We should keep in mind that Mr Pass has not only ideas which are really worth to use time for understand them, her also seems to be a very tolerant human beeing.
Another point is, Mr Haller writes in a style that is formal provocative, but the content of his postings are often worth reading. We just should overlook the style, maybe the style of speech is correlated with his location, wasn´t his President from the same location, from Texas too?
Theyer socialisation maybe is very different from a european one, but i think he is a "friendly cowboy", he reads the "********" we greenhorns write, and uses his time to write comments. Thats ok.
(PS, if we talk wine, in mr. passes country Mr. Robert Mondavi produces a quite good but expensive Cabernet Sauvignon, my favorite is a same quality Lagrein from Südtirol / Italy made by Joseph Brigl for half the money)
pointless misuse....
Hi harryhaller,
on the contrary, the current source whose reference is biased by yet another current diode is suboptimal with respect to supply rail rejection, (which includes rf on the rails), precisely because its rejection immunity decreases with increasing frequency, due to the parasitic inter-electrode capacitances of the jfet current diode.
a much better way is to use two resistors, R1,R2,(see attached), in place of the current diode, and decouple their intersection to the supply rail.......supply immunity now increases with frequency and at less cost....a time constant, (T=C*R2), of the order of two seconds is sufficient.
connecting the capacitor directly across the zener is again sub-optimal, as a comensurately larger component would be required for the same time constant.
on a separate issue....many audiophiles are carefull not to define precisely what is meant in practical terms by the 'simpler is better' concept....the overall lack of consistency boggles the mind......for instance, cascodes seem to fall in and out of favour like autumn leaves, (see recent threads on subject
).
Hi harryhaller,
on the contrary, the current source whose reference is biased by yet another current diode is suboptimal with respect to supply rail rejection, (which includes rf on the rails), precisely because its rejection immunity decreases with increasing frequency, due to the parasitic inter-electrode capacitances of the jfet current diode.
a much better way is to use two resistors, R1,R2,(see attached), in place of the current diode, and decouple their intersection to the supply rail.......supply immunity now increases with frequency and at less cost....a time constant, (T=C*R2), of the order of two seconds is sufficient.
connecting the capacitor directly across the zener is again sub-optimal, as a comensurately larger component would be required for the same time constant.
on a separate issue....many audiophiles are carefull not to define precisely what is meant in practical terms by the 'simpler is better' concept....the overall lack of consistency boggles the mind......for instance, cascodes seem to fall in and out of favour like autumn leaves, (see recent threads on subject
).
hi djk
i think you'll find the connection you've reffered to is infact due to peter blomley....article in wireless world..
i should'nt be too impressed with the japanese disign though....the critical balance of collector current in the double-ended diff. stages, essential for cancellation of even order artifacts), is disrupted by p-n, and jfet diodes which seem to syphon off an imprecisely quantified amount of current from the diff. stages.
Moreover the topology develops a double zero, rapidly followed by a treble(!), pole in its transfer function, all within an octave of the transition frequency.....a recipe for gross instability....these appear not to have been compensated for in the designs?
i think you'll find the connection you've reffered to is infact due to peter blomley....article in wireless world..
i should'nt be too impressed with the japanese disign though....the critical balance of collector current in the double-ended diff. stages, essential for cancellation of even order artifacts), is disrupted by p-n, and jfet diodes which seem to syphon off an imprecisely quantified amount of current from the diff. stages.
Moreover the topology develops a double zero, rapidly followed by a treble(!), pole in its transfer function, all within an octave of the transition frequency.....a recipe for gross instability....these appear not to have been compensated for in the designs?
for example of pointless misuse of components
So you are trading one part for three parts, I suppose thats progress. The RC decoupling will work but with the following considerations:
No improvement over resistor alone until after corner frequency defined by RC time constant and only 6dB per octave after that. The current diode (or jfet with resistor to set current) works to DC.
The current souce will not start conducting till the cap charges up.
If one were to use this circuit in the front end of the Aleph amp circuit for example, The DC offset when the amp was powered up can do serious damage to the speaker in some cases. I wasted a good pair of tweeters in a BIAMP set up this way. Circuit design has to be considered in the context of the whole circuit you are design with awareness of the tradeoffs and gains.
H.H.
So you are trading one part for three parts, I suppose thats progress. The RC decoupling will work but with the following considerations:
No improvement over resistor alone until after corner frequency defined by RC time constant and only 6dB per octave after that. The current diode (or jfet with resistor to set current) works to DC.
The current souce will not start conducting till the cap charges up.
If one were to use this circuit in the front end of the Aleph amp circuit for example, The DC offset when the amp was powered up can do serious damage to the speaker in some cases. I wasted a good pair of tweeters in a BIAMP set up this way. Circuit design has to be considered in the context of the whole circuit you are design with awareness of the tradeoffs and gains.
H.H.
Re: pointless misuse....
I agree about the junction capacitance of the jfet. One way to handle this would be to include a series resistor, which would also lower the power dissipation in the jfet.
I don't think I concur about the paragraph I quoted. Connecting the capacitor across the zener would have two benefits: it would suppress some of the noise generated by the zener and render the reference voltage more immune to load changes (base current of current source transistor). In terms of HF rail noise rejection, I don't see any disadvantage. After all, we'd be concerned about noise that the + rail carries with respect to the - rail that happens to be the reference node for the current source. So the time constant for HF rejection is determined by the impedance to the + rail times the capacitor size. Whether there is some DC load to the capacitor or not is insignificant. Besides, using a split resistor, the DC current would still be the same. So we could save one resistor. This is actually a case of a simpler and better design
Eric
mikek said:
I agree about the junction capacitance of the jfet. One way to handle this would be to include a series resistor, which would also lower the power dissipation in the jfet.
I don't think I concur about the paragraph I quoted. Connecting the capacitor across the zener would have two benefits: it would suppress some of the noise generated by the zener and render the reference voltage more immune to load changes (base current of current source transistor). In terms of HF rail noise rejection, I don't see any disadvantage. After all, we'd be concerned about noise that the + rail carries with respect to the - rail that happens to be the reference node for the current source. So the time constant for HF rejection is determined by the impedance to the + rail times the capacitor size. Whether there is some DC load to the capacitor or not is insignificant. Besides, using a split resistor, the DC current would still be the same. So we could save one resistor. This is actually a case of a simpler and better design
Eric
Lest some wisecrack point this out to me: in a constant current source the collector current and hence also the base current is constant. What I meant to say was that voltage changes on the collector have less of a chance to feed through via the miller capacitance if the base is well buffered by a large capacitor.
howdy capslock...
my mistake....i should have indicated that the bias network in the diagram is connected between ground and the -ve rail......ripple and rf from -ve rail bypasses the zener and R2 to ground through R1.
including a resistor in series with the jfet source increases its output impedance somewhat...however this is ultimately nullified by the reduction of output impedance with increasing frequency due to parasitic capacitances in the jfet.
as for dc offsets harry haller,, i reckon you need dc offset protection anyway for a safe amplifier...as you can never gaurantee that your power supply capacitors, or indeed circuit parasitics will charge up symmetrically....
further the jfet working to dc is unnecesary, as ripple components and rf become significant beyond 100hz, (in eu), or 120hz in us.
A single pole, 2second time-constant filter is sufficient, as attenuation at 100hz will be greater than 90db!
much is often made about filtering zeners for noise...sadly the equivalent output impedance of a typical zener is well below 20R, requiring a huge electrolytic to suppress the noise....
A better solution is to use a low-volt device, (5V~10V), running at relatively high current, (~10mA). Further reading: onsemi's zener diode databook dl150
www.onsemi.com
do'nt have to use zeners though...could go for led...or chain of p-n diodes......etc, to avoid noise issue altogether......insist on decoupled dual resistors though whatever the solution...
my mistake....i should have indicated that the bias network in the diagram is connected between ground and the -ve rail......ripple and rf from -ve rail bypasses the zener and R2 to ground through R1.
including a resistor in series with the jfet source increases its output impedance somewhat...however this is ultimately nullified by the reduction of output impedance with increasing frequency due to parasitic capacitances in the jfet.
as for dc offsets harry haller,, i reckon you need dc offset protection anyway for a safe amplifier...as you can never gaurantee that your power supply capacitors, or indeed circuit parasitics will charge up symmetrically....
further the jfet working to dc is unnecesary, as ripple components and rf become significant beyond 100hz, (in eu), or 120hz in us.
A single pole, 2second time-constant filter is sufficient, as attenuation at 100hz will be greater than 90db!
much is often made about filtering zeners for noise...sadly the equivalent output impedance of a typical zener is well below 20R, requiring a huge electrolytic to suppress the noise....
A better solution is to use a low-volt device, (5V~10V), running at relatively high current, (~10mA). Further reading: onsemi's zener diode databook dl150
www.onsemi.com
do'nt have to use zeners though...could go for led...or chain of p-n diodes......etc, to avoid noise issue altogether......insist on decoupled dual resistors though whatever the solution...
Re: howdy capslock...
Now you've lost me! First of all R1 is not a very efficient shunt to ground. And why would you be concerned about noise on the -ve rail? Both the zener and the emitter are referenced to this rail, so the current will be free of this noise. If you are concerned about capacitive coupling to the collector, your filtering scheme will not help this. What you need to do is include a cascode with the base decoupled to ground.
Eric
mikek said:
Now you've lost me! First of all R1 is not a very efficient shunt to ground. And why would you be concerned about noise on the -ve rail? Both the zener and the emitter are referenced to this rail, so the current will be free of this noise. If you are concerned about capacitive coupling to the collector, your filtering scheme will not help this. What you need to do is include a cascode with the base decoupled to ground.
Eric
hi eric...
...the single pole, high-pass filter comprised of C and R1 across the zeners internal resistance in series with R2, improves the regulation of the voltage drop across the zener diode by diminishing power supply ripple in the current established by R1 & R2...........cannot make it plainer than this i am afraid....
...the single pole, high-pass filter comprised of C and R1 across the zeners internal resistance in series with R2, improves the regulation of the voltage drop across the zener diode by diminishing power supply ripple in the current established by R1 & R2...........cannot make it plainer than this i am afraid....
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.