It s waffles for one that don t understand how a differential
based amp settle the DC operating points..
Since there s a feedback at DC , any modifiation of
the output DC operating point made to the output,
will forcibly move the DC voltage aplied to the non
inverting input through the fedback network..
In reaction, the differential will shift its output to settle the
output stage to the same value as the non inverting input, i.e , 0,
and in the process, the current through the two legs of the
differential will be equally forcibly unbalanced, as it s now the
condition for the amp to have an output at 0V at iddle....
I can only advice you to read some litterature about these basics,
this will spare you many seemingly useless electronic tweakings.
regards,
wahab
Gentlemen, can we keep this discussion civilised please? Wahab, you are being very aggressive in expressing your criticism - but you never post anything valuable. If you are criticizing someone's ideas, make this a constructive criticism, and make sure it is well founded - otherwise please don't bother posting at all.
Well, i respond the same way i m talked to...
It is unfortunate that you didn t understand the
point i made 2 posts above, neither did THorsen.
One has to understand how a differential based amp work
before being capable to appreciate if a tweak is justified or
not..
There s much claim by there that are total non sense , in
mr Thorsen s own words..
Wahab, in engineering there are many “correct” principles for "optimum" operation of one stage or another, but that interact with all of the others, all operating at the same time
We have to put numbers on these to see how one or another effect that at extremes results in (dis)/improvements in a part of the amplifier’s performance balance out in the final product
Self has measurements in his "Class-XD" whitepaper found on the Cambridge Audio site,
http://www.cambridgeaudio.com/
also a chapter in his later edition Amplifier book
added output bias gives measurable distortion reduction - with Self's "optimum Class B bias, Blameless" amp - probably indicative for what could be seen with power chip amps - not huge: ~ 6 dB reduction at low levels - but an Improvement and really cheap
although trying the other rail could be good too due to the difference in the quasi-comp output for +,- current
for "unity gain" operation look up "noise gain compensation" - easiest with inverting gain, positive gain is going to require a low, stable source impedance at the + input from ~ 20 KHz - 2 MHz
We have to put numbers on these to see how one or another effect that at extremes results in (dis)/improvements in a part of the amplifier’s performance balance out in the final product
Self has measurements in his "Class-XD" whitepaper found on the Cambridge Audio site,
http://www.cambridgeaudio.com/
also a chapter in his later edition Amplifier book
added output bias gives measurable distortion reduction - with Self's "optimum Class B bias, Blameless" amp - probably indicative for what could be seen with power chip amps - not huge: ~ 6 dB reduction at low levels - but an Improvement and really cheap
although trying the other rail could be good too due to the difference in the quasi-comp output for +,- current
for "unity gain" operation look up "noise gain compensation" - easiest with inverting gain, positive gain is going to require a low, stable source impedance at the + input from ~ 20 KHz - 2 MHz
Last edited:
Dear All,
I tried to run the the chips at Class A as well, but with a cascode current source. (I saw a similar trick in some DIY headphone amp designs). I found small improvements in sound, especially (strange enough?) in the lower octaves. However, I went a bit to far and didn't had the availability of adequate heatsinks, so the thermal shut-down jumped in, and the chip got to hot. Now it is mentioned again here, it is interesting to experiment again with it. Lucky to see that this chips are well protected, and that is one of the big benefits of those chips. Some discrete designs will self-destruct themselves with such a experiments.
With kind regards,
Bas
I tried to run the the chips at Class A as well, but with a cascode current source. (I saw a similar trick in some DIY headphone amp designs). I found small improvements in sound, especially (strange enough?) in the lower octaves. However, I went a bit to far and didn't had the availability of adequate heatsinks, so the thermal shut-down jumped in, and the chip got to hot. Now it is mentioned again here, it is interesting to experiment again with it. Lucky to see that this chips are well protected, and that is one of the big benefits of those chips. Some discrete designs will self-destruct themselves with such a experiments.
With kind regards,
Bas
Hi,
Actually, I understand VERY well how this happens.
To wit, the input offset voltage, after accounting for the various parasitics and input bias currents, NFB networks etc. is actully around 10mV or more.
In many "gainclones" even with balanced DCR I measure more DC offset on the output than the total of the change in voltage at the output that a 0.1A extra bias current can cause.
So in this case we already have more unbalance from the normal offset than from my mod.
So, I repeat my point:
IN PRACTICE and with the LM3875 or LM3886 or LM4780 the imbalance in the differential input stage are very small, when using 0.1A extra bias to the negative rail.
Further, due to the design applied in these chips the effect of any imbalance on the output signal (distortion etc.) is very much minimised.
At the same time the application of the "unbalancing" has a very material positive impact on the distortion at low levels (much greater positive effect that the minimal imbalance).
You may bang on about imbalancing the differential pair, the point is that in the context and at the levels we are working with in my examples this imbalance is irrelevant and at best no worse than that from the normal offset sources.
It does not suffice to know the theory, one must also be able to evaluate the magnitude of the effects and contrast them to the impact they make.
I see you did not take up my offer to demonstrate that you actually know the theory sufficiently to present the necessary math to show just how much my modification upsets the apple cart.
For reference, D. Self shows that indeed the DC Balance is critical in cases where no emitter degeneration is used. Incidentally, his work also shoes that even with fairly minimal local degeneration (much less than the 1K used in the LM3875) will linearise the transconductance so much, that even several hundert mV DC imbalance on the inputs do not have any material effect.
Incidentally, for an un-degenerate differential stage D. Self suggests that limiting the imbalance to 1% is necessary. If you had done your homework you would realise that 0.1A extra current to the negative rail shift the current balance by less than that in the LM3875.
Anyway, you are still welcome to show your math. I have work to do.
Ciao T
Actually, I understand VERY well how this happens.
To wit, the input offset voltage, after accounting for the various parasitics and input bias currents, NFB networks etc. is actully around 10mV or more.
In many "gainclones" even with balanced DCR I measure more DC offset on the output than the total of the change in voltage at the output that a 0.1A extra bias current can cause.
So in this case we already have more unbalance from the normal offset than from my mod.
So, I repeat my point:
IN PRACTICE and with the LM3875 or LM3886 or LM4780 the imbalance in the differential input stage are very small, when using 0.1A extra bias to the negative rail.
Further, due to the design applied in these chips the effect of any imbalance on the output signal (distortion etc.) is very much minimised.
At the same time the application of the "unbalancing" has a very material positive impact on the distortion at low levels (much greater positive effect that the minimal imbalance).
You may bang on about imbalancing the differential pair, the point is that in the context and at the levels we are working with in my examples this imbalance is irrelevant and at best no worse than that from the normal offset sources.
It does not suffice to know the theory, one must also be able to evaluate the magnitude of the effects and contrast them to the impact they make.
I see you did not take up my offer to demonstrate that you actually know the theory sufficiently to present the necessary math to show just how much my modification upsets the apple cart.
For reference, D. Self shows that indeed the DC Balance is critical in cases where no emitter degeneration is used. Incidentally, his work also shoes that even with fairly minimal local degeneration (much less than the 1K used in the LM3875) will linearise the transconductance so much, that even several hundert mV DC imbalance on the inputs do not have any material effect.
Incidentally, for an un-degenerate differential stage D. Self suggests that limiting the imbalance to 1% is necessary. If you had done your homework you would realise that 0.1A extra current to the negative rail shift the current balance by less than that in the LM3875.
Anyway, you are still welcome to show your math. I have work to do.
Ciao T
Thank you JCX,
Very interesting link. See here for a direct download of the PDF.
http://www.cambridgeaudio.com/assets/documents/840Awhitepaper8-2-06web.pdf
With kind regards,
Bas
Very interesting link. See here for a direct download of the PDF.
http://www.cambridgeaudio.com/assets/documents/840Awhitepaper8-2-06web.pdf
With kind regards,
Bas
Hi,
EXACTLY.
And one has to understand the various effects of such techniques as local degeneration as well.
Note, I do not dispute that we will will see a very minor imbalance (though not necessary one that is much larger than what is present in normal operation and even Nat Semi's Datasheet Test circuits).
What I am pointing out is that the effect IN THE SPECIFIC CASE of the Nat Semi Overture Chips is very small in terms of quantity and it's impact is further reduced by various other design techniques implemented in the Nat Semi Chips.
Anyone can throw some halve-knowledge around. Those who understand what they talk about can give at least first order estimates of how large the impact of a given problem is.
So, you fail to quantify the magnitude of the problem you keep banging on about. I repeat again, please provide your numbers. This will:
A) Show the extent of the problem
B) Show that you understand the complete nature of the problem
As to "Well, i respond the same way i m talked to...", I believe you where the one to introduce the term "childish tweak". And that in the context of a tweak where you have repeatedly demonstrated an inability to quantify the issue you refer to.
Ciao T
EXACTLY.
And one has to understand the various effects of such techniques as local degeneration as well.
Note, I do not dispute that we will will see a very minor imbalance (though not necessary one that is much larger than what is present in normal operation and even Nat Semi's Datasheet Test circuits).
What I am pointing out is that the effect IN THE SPECIFIC CASE of the Nat Semi Overture Chips is very small in terms of quantity and it's impact is further reduced by various other design techniques implemented in the Nat Semi Chips.
Anyone can throw some halve-knowledge around. Those who understand what they talk about can give at least first order estimates of how large the impact of a given problem is.
So, you fail to quantify the magnitude of the problem you keep banging on about. I repeat again, please provide your numbers. This will:
A) Show the extent of the problem
B) Show that you understand the complete nature of the problem
As to "Well, i respond the same way i m talked to...", I believe you where the one to introduce the term "childish tweak". And that in the context of a tweak where you have repeatedly demonstrated an inability to quantify the issue you refer to.
Ciao T
Hi,
JCX beat me to it...
In case you missed it:
As for the actual practical implementation, sorry, that is a commercial product, so that shcematic stays under wraps I'm afraid.
Ciao T
JCX beat me to it...
In case you missed it:
As for the actual practical implementation, sorry, that is a commercial product, so that shcematic stays under wraps I'm afraid.
Ciao T
Hi Bas,
True, but when compared to the speaker load the kind of resistors we can use with normal chipamps are still quite high in impedance. So the difference going from there to a CCS is markedly less than with many other systems.
If you do a CCS you may as well make it dynamic, as per Mr. Self. He suggests that the extra current should be zero when the amp is maximum positive swing and at double the quiescent level when the amplifier is at maximum negative swing.
This seems to call for one of Nelson's dynamic current sources maybe?
As for pull to positive rail or negative rail, the results will be quite different in case of the LM3875, see attached principle schematic. Actually, that schematic is worth going over with a fine tooth comb. Many a proponent of discrete designs can see a lot of good ideas here that are almost never found in discrete designs...
Ciao T
True, but when compared to the speaker load the kind of resistors we can use with normal chipamps are still quite high in impedance. So the difference going from there to a CCS is markedly less than with many other systems.
If you do a CCS you may as well make it dynamic, as per Mr. Self. He suggests that the extra current should be zero when the amp is maximum positive swing and at double the quiescent level when the amplifier is at maximum negative swing.
This seems to call for one of Nelson's dynamic current sources maybe?
As for pull to positive rail or negative rail, the results will be quite different in case of the LM3875, see attached principle schematic. Actually, that schematic is worth going over with a fine tooth comb. Many a proponent of discrete designs can see a lot of good ideas here that are almost never found in discrete designs...
Ciao T
Attachments
Dear Thorsten,
But since it is Self's patent (pending) it doesn't feel good to copy him. Second how to vary the current source? I suspect in the Cambridge it is done with a PIC since the whole amplifier is PIC based. Analogue there are of course many ways to do it, but add skills and complexity.
With kind regards,
Bas
Hi,
Well, for private, personal use even patented technology is "fair game".
And the "pull down" current source for Op-Amp's has a long history, I am sure with sufficient time and effort there is probably enough prior art to challenge this patent, if one was so inclined.
As said, the dynamic current source by Nelson (Pass) could be a starting point.
With as linear a relation as that D. Self describes we just need a circuit that, referenced to the negative rail, outputs 0V if input voltage is twice the negative rail minus dropout voltage and outputs 1V with the negative rail voltage applied and 2V for no input voltage.
Looks like a single transistor (or maybe something temperature compensated like the Tl431) could do the job. I am sure, if I spend enough time I can do it with an LM317 and some resistors...
Shame the 317 has not enough voltage handling for our job.
Given Mr. Self's description I'd rather doubt it. It seems clearly directly linked to the output voltage of the amp.
I would suspect an inverting op-amp into a level shifter to the CCS, if you'd ask me.
Ciao T
Well, for private, personal use even patented technology is "fair game".
And the "pull down" current source for Op-Amp's has a long history, I am sure with sufficient time and effort there is probably enough prior art to challenge this patent, if one was so inclined.
As said, the dynamic current source by Nelson (Pass) could be a starting point.
With as linear a relation as that D. Self describes we just need a circuit that, referenced to the negative rail, outputs 0V if input voltage is twice the negative rail minus dropout voltage and outputs 1V with the negative rail voltage applied and 2V for no input voltage.
Looks like a single transistor (or maybe something temperature compensated like the Tl431) could do the job. I am sure, if I spend enough time I can do it with an LM317 and some resistors...
Shame the 317 has not enough voltage handling for our job.Given Mr. Self's description I'd rather doubt it. It seems clearly directly linked to the output voltage of the amp.
I would suspect an inverting op-amp into a level shifter to the CCS, if you'd ask me.
Ciao T
we're able to hear that.
For some reason most of the rest of the world cannot or won't believe.
Take a look at the DEQX HDP3. It takes digital or analog input, each one from either an XLR or RCA input (adding up to four inputs in total), does the crossover, speaker correction, room correction, EQ, DAC, and volume control thing, and sends the analog output signal through six transformer balanced XLRs. Or electronically balanced outputs, or single ended RCAs, or digital outputs, for that matter. Quite a few options, and lots of possibilities.
I have one at the core of my system, followed by six channels of Hypex Class-D amplification, and I cannot see any reason for swapping it out.
www.deqx.com
For the next iteration of my system, I will need fourteen channels of amplification, so high quality chipamps seem quite interesting.
Just to let everyone know that I haven't abandoned the thread - I'm stocking up on things like silver content solder, signal wires, heatsinks etc. I will move on to building stuff soon enough
While I was looking through some audio shops, I noticed that some of them sell chokes in many flavours, intended to be used as "filters" for cables. Can someone please explain to me if they are worth buying - what can be gained and what can be lost as a result of employing them? Also, which cable sections are they suitable for? Power only?
Overview
Crossover board (most parts are visible on this photo - ones that aren't, are in the pictures below)
Cap 1
Cap 2
Cap 3
Coil 1
Film capacitor?
Resistor 1
While I was looking through some audio shops, I noticed that some of them sell chokes in many flavours, intended to be used as "filters" for cables. Can someone please explain to me if they are worth buying - what can be gained and what can be lost as a result of employing them? Also, which cable sections are they suitable for? Power only?
Thanks for the link asbjbo, DEQX HDP3 seems like a great piece of kit - but it's a bit outside my budget constraints for the time being...asbjbo said:
I've had a look at the crossover, but there seems to be only one electrolytic capacitor in it. The other parts, except two resistors, have some meaningless markings on them. I would appreciate if someone could tell me what they actually mean and what would be a reasonable upgrade, if applicable. Here are the pictures (v.high resolution - will take a while to load):ThorstenL said:
Overview
Crossover board (most parts are visible on this photo - ones that aren't, are in the pictures below)
Cap 1
Cap 2
Cap 3
Coil 1
Film capacitor?
Resistor 1
Looks like a polyswitch to me, not the value here, but this type of device.
http://cpc.farnell.com/tyco-electronics-raychem/lvr016k/polyswitch-0-16a/dp/FF02037
so it's not something to change.
http://cpc.farnell.com/tyco-electronics-raychem/lvr016k/polyswitch-0-16a/dp/FF02037
so it's not something to change.
If they are protection devices, can I remove them altogether? I understand their role would be somewhat similar to a fuse? I don't intend to drive my speakers to insane power levels, and neither do I intend to use them with untested equipment, so might as well get rid of them as they cause unnecessary resistance in the signal path?
Thanks for confirming that the yellow devices are actually capacitors.
I'm quite positive that the 47uF electrolytic cap could do with replacing as it's very old - any suggestions as to what to use? Would two film capacitors in parallel (eg. 22uF+25uF) be better than a single electrolytic?
Thanks for confirming that the yellow devices are actually capacitors.
I'm quite positive that the 47uF electrolytic cap could do with replacing as it's very old - any suggestions as to what to use? Would two film capacitors in parallel (eg. 22uF+25uF) be better than a single electrolytic?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.