Hi,
In terms of components and assembly it would be no worse than a nested feedback amp. I am not neccesarily recommending it for your use.
You may find a LM3875 with a CCS or resistor to get a little bit of extra Class A more than sufficient.
That never really works.
Look at the existing crossover and eliminate the electrolytic capacitors on it in favour of film capacitors. Plus try the felt disks on the tweeter from my "Freakzoid Tweekaloid" article at enjoythemusic.com:
The Freakzoid Tweekaloid Strikes Again by Thorsten Loesch
The CD-52 uses an old Philips Bitstream Chipset. These not only produce slews of ultrasonic noise, they have on-chip I/V Op-Amp's that are pretty dire. I never spend much time with this chip, as despite claiming 20 Bit performance it measures and sounds way worse than the TDA1541A.
I suspect the circuit could be hacked to bypass the internal op-amp's of the DAC chip. I expect Nat Semi LM6181/82 (or AD811) as I/V converters would be a good choice, with maybe LM6171/72 (or similar) for the filtering.
Still, I'd personally be tempted to fit a Non-OS TDA1541 inside with a nice tube-output...
Ciao T
In terms of components and assembly it would be no worse than a nested feedback amp. I am not neccesarily recommending it for your use.
You may find a LM3875 with a CCS or resistor to get a little bit of extra Class A more than sufficient.
That never really works.
Look at the existing crossover and eliminate the electrolytic capacitors on it in favour of film capacitors. Plus try the felt disks on the tweeter from my "Freakzoid Tweekaloid" article at enjoythemusic.com:
The Freakzoid Tweekaloid Strikes Again by Thorsten Loesch
The CD-52 uses an old Philips Bitstream Chipset. These not only produce slews of ultrasonic noise, they have on-chip I/V Op-Amp's that are pretty dire. I never spend much time with this chip, as despite claiming 20 Bit performance it measures and sounds way worse than the TDA1541A.
I suspect the circuit could be hacked to bypass the internal op-amp's of the DAC chip. I expect Nat Semi LM6181/82 (or AD811) as I/V converters would be a good choice, with maybe LM6171/72 (or similar) for the filtering.
Still, I'd personally be tempted to fit a Non-OS TDA1541 inside with a nice tube-output...
Ciao T
Where should the CCS/resistor be mounted? You don't have access to bias the output stage.
Hi,
The usual way for Ap-Amp's?
Output to negative rail (actually, it is not always the negative rail).
This draws extra current through the positive output transistor...
You can try 100mA extra bias current if you like.
Ciao T
The usual way for Ap-Amp's?
Output to negative rail (actually, it is not always the negative rail).
This draws extra current through the positive output transistor...
You can try 100mA extra bias current if you like.
Ciao T
Can you clarify that for me please? Like I said before, I'm not terribly profficient with these things.ThorstenL said:
I understand Peter Daniel's design fairly well, so if you tell me what to change in it, this will certainly make much more sense to me:
DIY Chip Amplifier Kits, PCB's, Components and Information.
http://www.audiosector.com/images/lm3875_se_pcb.gif
Am I right in thinking that you suggest soldering a resistor between points labeled [OUT] and [V-] on that PCB? Sorry about such and awkward way of asking about this, I hope you know what I mean.
This will unbalance the circuit.
The negative feedback will shift the input differential
balance to cope with the added output offset..
If you think that this kind of childish trick
will improve the circuit, then you don t have any
clue about how a differential work...
Anyway, just stop misleading the public with such
ridiculous false solutions...
Hi,
Absolutely. For 100mA current the resistor should be:
V- / 0.1A = ??Ohm
So:
35V/0.1A = 350 Ohm
As NPV (Nearest Preferred Value) try 390 Ohm.
The power dissipated in the resistor is:
V- * 0.1A
So:
35V*0.1A = 3.5VA
It is good practice to not use resistors at more than 1/2 rated power, so you should use a resistor with at least twice this power rating. Note that the common Metal Clad resistors are rated with a heasink attached, they can handle much less power without that heatsink.
Of course, the same extra power is dissipated in the Poweramp Chip, so this may also need extra heatsinking.
Ciao T
Absolutely. For 100mA current the resistor should be:
V- / 0.1A = ??Ohm
So:
35V/0.1A = 350 Ohm
As NPV (Nearest Preferred Value) try 390 Ohm.
The power dissipated in the resistor is:
V- * 0.1A
So:
35V*0.1A = 3.5VA
It is good practice to not use resistors at more than 1/2 rated power, so you should use a resistor with at least twice this power rating. Note that the common Metal Clad resistors are rated with a heasink attached, they can handle much less power without that heatsink.
Of course, the same extra power is dissipated in the Poweramp Chip, so this may also need extra heatsinking.
Ciao T
Hi,
Intentionally.
BTW, how much error do we really impose?
The LM3875 Datasheet shows 0.45 Ohm Emitter resistors and I'd estimate around 20mA output stage quiescent current. Three diodes provide the actual biasing.
Due to the quasi-complementary nature of the Chip's output stage we have no emitter resistor in the emitter of the PNP pre-driver. so the actual voltage available in the output stage for the Bias of the whole stage is around 10mV.
If we draw 100mA we will have 45mV across this emitter resistor, which has now nearly pinched off the current in the PNP tansistor and NPN Power section, yet keeps this biased at a point where minimal signals will make this section conduct again.
So, we have actually shut down the current through the lower transistor completely and so at least for current up to 100mA (40mW) the output stage operates as single-ended class A follower.
Does this have measurable and audible consequences? Rest assured it will. You can try this and measure it on your AP2 or whatever you use,
Yes, I hope so. However this does not affect the change to the operation of the output stage biasing.
If you REALLY think the output stage of the LM3875 (and similar chips) is a differential circuit, please ask your tuition fees back from your university.
Hmmm, who is misleading people about how output stages work?
Please, try to first of all consider that those who promote a solution you do not immediately understand (BTW, the result of that resistor strap to -V os actually VERY easy to understand intuitively and in detail) MAY still have a point and at least try to understand what is really going on, before blasting off.
Kind regards Thorsten
Intentionally.
BTW, how much error do we really impose?
The LM3875 Datasheet shows 0.45 Ohm Emitter resistors and I'd estimate around 20mA output stage quiescent current. Three diodes provide the actual biasing.
Due to the quasi-complementary nature of the Chip's output stage we have no emitter resistor in the emitter of the PNP pre-driver. so the actual voltage available in the output stage for the Bias of the whole stage is around 10mV.
If we draw 100mA we will have 45mV across this emitter resistor, which has now nearly pinched off the current in the PNP tansistor and NPN Power section, yet keeps this biased at a point where minimal signals will make this section conduct again.
So, we have actually shut down the current through the lower transistor completely and so at least for current up to 100mA (40mW) the output stage operates as single-ended class A follower.
Does this have measurable and audible consequences? Rest assured it will. You can try this and measure it on your AP2 or whatever you use,
Yes, I hope so. However this does not affect the change to the operation of the output stage biasing.
If you REALLY think the output stage of the LM3875 (and similar chips) is a differential circuit, please ask your tuition fees back from your university.
Hmmm, who is misleading people about how output stages work?
Please, try to first of all consider that those who promote a solution you do not immediately understand (BTW, the result of that resistor strap to -V os actually VERY easy to understand intuitively and in detail) MAY still have a point and at least try to understand what is really going on, before blasting off.
Kind regards Thorsten
Don t stretch the words...
I m talking about a differential based amp, not about
the output stage which i pointed as a quasi complementary
since the very start of this thread..
That said, an unbalanced diferential will see increased
distorsion , even with only a few % unbalance in the
two transistors collector currents..
all da best,
My current GC runs off 18VAC, and it has some 24-25VDC on actual lines, if I remember correctly. So, if I wanted to try A-class bias with it, I will need a 240R, 6W or more resistor?ThorstenL said:
How important is the quality of resistors used here? Any specific brand recommendations? (of what's available in UK, preferably)
What are possible unwanted consequences of such addition? Other than increased heat dissipation, which is fine.
Also, how far can we go with biasing, before running into problems? What's the trade-off for increasing the A-class bias?
At iddle, one of the output transistor will run at higher
current, and the other will be switched off as
a consequence..
There s nothing that can be called class A in
this dreadful arrangement.
Class A require that the current should be increased
through both output transistors..
Hi,
I am not.
What you did was to produce some waffle about how the input stage behaves in response to modifications in the output stage.
Yet the modifications are applied to the output stage.
So, are you talking nonsense or are you trying to intentionally mislead those who may take your waffle as being based on some special understanding?
Hmm, lets see.
Thankfully the designers at Nat Semi's design center are quite competent designers. So they have provided substantial local degneration for the input pair and a quite clever variant of current mirror loading. What will be the unbalance if we vary the output by around 20mV with a "pull down" resistor or current source?
And more to the point, with so much degeneration in the input pair, will there be any appreciable increase in THD?
I still have a Manual and a Whitepaper to finish, so I don't have time to do the math myself. Would you volunteer to actually show the others just HOW BAD my "childish trick" is?
Ciao T
I am not.
What you did was to produce some waffle about how the input stage behaves in response to modifications in the output stage.
Yet the modifications are applied to the output stage.
So, are you talking nonsense or are you trying to intentionally mislead those who may take your waffle as being based on some special understanding?
Hmm, lets see.
Thankfully the designers at Nat Semi's design center are quite competent designers. So they have provided substantial local degneration for the input pair and a quite clever variant of current mirror loading. What will be the unbalance if we vary the output by around 20mV with a "pull down" resistor or current source?
And more to the point, with so much degeneration in the input pair, will there be any appreciable increase in THD?
I still have a Manual and a Whitepaper to finish, so I don't have time to do the math myself. Would you volunteer to actually show the others just HOW BAD my "childish trick" is?
Ciao T
Hi,
Actually, if you reference the internal schematic of the LM3875, you will see that inceasing the voltage on the single determining emitter resistor is not sufficient to fully shut off the second transistor, however it significantly reduces the current in the negative halve.
Well, if the signal level is kept at a sufficient level it clearly is class A and more to the point singe ended class A.
Actually, what happens with this "childish trick" is that we avoid what D. Self calls (inaccuratly IMNSHO) Gm doubling distortion, which is the result of having both output transistor halves operate in parallel at low levels.
You may also be aware that D. Self himself (pun intended) uses a similar "dreadful arrangement" and "childish trick" in his designs for Cambridge Audio Amplifiers. I hope you have made sure to let Mr Self know how you feel about his designs (dreadfull & childish) just as you have seen fit to do with me.
Hmm, what is the defintion of Class A again?
"Amplifying devices operating in Class A conduct over the whole of the input cycle such that the output signal is an exact scaled-up replica of the input with no clipping."
Electronic amplifier - Wikipedia, the free encyclopedia
I am afraid operating a single device at appropriate current qualifies as Class A.
And by running 100mA quiescent current in the positive output transistor section we have achieved the following:
1) Increase the power up to which the amplifier operates in Class A
2) Removed or substantially reduced the Gm doubling distortion
3) Introduced a very small current imbalance in a heavily degenerated input stage
So, I guess it's a little like Ronseal, innit? It does what it says on the Tin.
Further, for anyone running "Gainclones" it is easy to try. Solder in a single resistor per channel, turn on, listen (or measure, if you prefer).
Ciao T
Actually, if you reference the internal schematic of the LM3875, you will see that inceasing the voltage on the single determining emitter resistor is not sufficient to fully shut off the second transistor, however it significantly reduces the current in the negative halve.
Well, if the signal level is kept at a sufficient level it clearly is class A and more to the point singe ended class A.
Actually, what happens with this "childish trick" is that we avoid what D. Self calls (inaccuratly IMNSHO) Gm doubling distortion, which is the result of having both output transistor halves operate in parallel at low levels.
You may also be aware that D. Self himself (pun intended) uses a similar "dreadful arrangement" and "childish trick" in his designs for Cambridge Audio Amplifiers. I hope you have made sure to let Mr Self know how you feel about his designs (dreadfull & childish) just as you have seen fit to do with me.
Hmm, what is the defintion of Class A again?
"Amplifying devices operating in Class A conduct over the whole of the input cycle such that the output signal is an exact scaled-up replica of the input with no clipping."
Electronic amplifier - Wikipedia, the free encyclopedia
I am afraid operating a single device at appropriate current qualifies as Class A.
And by running 100mA quiescent current in the positive output transistor section we have achieved the following:
1) Increase the power up to which the amplifier operates in Class A
2) Removed or substantially reduced the Gm doubling distortion
3) Introduced a very small current imbalance in a heavily degenerated input stage
So, I guess it's a little like Ronseal, innit? It does what it says on the Tin.
Further, for anyone running "Gainclones" it is easy to try. Solder in a single resistor per channel, turn on, listen (or measure, if you prefer).
Ciao T
Hi,
Yes, at least for a quick trial.
Quality is important, insofar that the resistor should withstand the power dissipated well... Otherwise I'd not loose too much sleep.
You may not like the sonic results. I seem to remember Peter Daniels complained that this "Class A Biasing" (Maybe we better go with D. Self's definition - "Class XD"?) made his gainclones sound too smooth.
Okay, in principle you are only limited by chip temperature and current limits. If you can devise an effective cooling scheme you could probably run this scheme in single ended Class A for full power.
If you use normal arrangements of cooling etc. you very quickly run into the limits of this technique. For example 100mA at 25V mean an extra dissipation of 2.5W for an increase in "Class A" operation for an 8 ohm Load from 16 mW to 40mW.
If you go up to (say) 300mA you now dissipate 7.5W continous, which means quite significant size heatsinks, as you still also need to provide for the dynamic dissipation with higher signal levels. On the plus side, with 300mA you now have around 300mW in "Class A".
I suspect this is one of the things where one may have to run several test series with both blind listening tests and measurements. I personally prefer to run things quite safe, even much less than 100mA can give measurable and audible improvements.
So, best try. Luckily enough, the LM38XX chips are pretty hard to fry.
Ciao T
Yes, at least for a quick trial.
Quality is important, insofar that the resistor should withstand the power dissipated well... Otherwise I'd not loose too much sleep.
You may not like the sonic results. I seem to remember Peter Daniels complained that this "Class A Biasing" (Maybe we better go with D. Self's definition - "Class XD"?) made his gainclones sound too smooth.
Okay, in principle you are only limited by chip temperature and current limits. If you can devise an effective cooling scheme you could probably run this scheme in single ended Class A for full power.
If you use normal arrangements of cooling etc. you very quickly run into the limits of this technique. For example 100mA at 25V mean an extra dissipation of 2.5W for an increase in "Class A" operation for an 8 ohm Load from 16 mW to 40mW.
If you go up to (say) 300mA you now dissipate 7.5W continous, which means quite significant size heatsinks, as you still also need to provide for the dynamic dissipation with higher signal levels. On the plus side, with 300mA you now have around 300mW in "Class A".
I suspect this is one of the things where one may have to run several test series with both blind listening tests and measurements. I personally prefer to run things quite safe, even much less than 100mA can give measurable and audible improvements.
So, best try. Luckily enough, the LM38XX chips are pretty hard to fry.
Ciao T
I guess that's the reason why such relatively simple modification is not included as a standard in your kit... Is it really so hard to substantially improve upon the original Gaincard design?... Have you tried any of the nested topologies mentioned in this thread?Peter Daniel said:
I think I might give it a try anyway, like Thorsten said, just to see how it works out; it's not an expensive modification, and I can always reverse it if I don't like it. I'll resolder all the connections while I'm at it, as I've noticed that one channel was losing sound occasionally when I was moving my equipment rack around.
Last edited:
Peter,
My own tests (including blind listening) came to the opposite conclusion, however, this was of course with different parts, circuits and speakers than your test...
Ciao T
My own tests (including blind listening) came to the opposite conclusion, however, this was of course with different parts, circuits and speakers than your test...
Ciao T
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.