Now can we get some discussion - technical
I know by own experience, that not many have time to download PDFs
or follow links.
I seldome find it good to discuss circuits,
you try to describe in words. Very hard to do!
"One picture tells a thousand words"
So here I add 1000 words, to help this discussion.
I know we are many here, who can learn a lot,
if we start with tschrama's questions.
Well not the ones "who knows", but normal people ....
tschrama, he has the courage to ask, at the risk of stand out as beeing stupid.
I know a lot of "knowers" at this site,
that has not got
half that courage.
/halojoy - real stupid, to write things like this
tschrama said:
I know by own experience, that not many have time to download PDFs
or follow links.
I seldome find it good to discuss circuits,
you try to describe in words. Very hard to do!
"One picture tells a thousand words"
So here I add 1000 words, to help this discussion.
I know we are many here, who can learn a lot,
if we start with tschrama's questions.
Well not the ones "who knows", but normal people ....
tschrama, he has the courage to ask, at the risk of stand out as beeing stupid.
I know a lot of "knowers" at this site,
that has not got
half that courage. /halojoy - real stupid, to write things like this
Attachments
All these talks of 'oscillations' give me the shivers, bhhhhrrrrrrr!
How can an amp on the verge of oscillating sound any good?
Dave for once we can agree on something!
I read the hibernik guy page, seems like a lot of blanket, boisterous, balloney statements, with some spelling mistakes and not much substance other than the fact he can read the NS datasheet and he thinks that more is better. And this guy ain't no Nelson Pass either.
Anyways, has anyone actually built the canonical version, first schematic on the datasheet? What's wrong with that?
How can an amp on the verge of oscillating sound any good?
Dave for once we can agree on something!
I read the hibernik guy page, seems like a lot of blanket, boisterous, balloney statements, with some spelling mistakes and not much substance other than the fact he can read the NS datasheet and he thinks that more is better. And this guy ain't no Nelson Pass either.
Anyways, has anyone actually built the canonical version, first schematic on the datasheet? What's wrong with that?
Grataku wrote:
How do you think will most of the highly regarded discrete high-end power amps behave, when the user would fiddle around with the feedback network in the same way, as the average op-amp user does (and these IC amps are in fact high-power op-amps - BTW: most discrete SS power amps as well !!) ?
Regards
Charles
How do you think will most of the highly regarded discrete high-end power amps behave, when the user would fiddle around with the feedback network in the same way, as the average op-amp user does (and these IC amps are in fact high-power op-amps - BTW: most discrete SS power amps as well !!) ?
Regards
Charles
halo is right - I got more info to build a gainclone in 15 minutes by readying LM3875 datasheet in PDF than chasing and scrounging for 2 weeks after gainclone topics in the internet.
Mr Wang is even more right by quoting on “blameless” and Naim. I think it is very sophisticated circuit in almost ideal package (small) and should not sound any worse than those mentioned amps (just less power). But does it sound any better than some well regarded much simpler designs by Pass, Higara, J.L.Hood etc, I am yet to be convinced. Now comes the building, experimenting, evaluating part – Long live the DIY not FIB (follow it blindly).
argo
Mr Wang is even more right by quoting on “blameless” and Naim. I think it is very sophisticated circuit in almost ideal package (small) and should not sound any worse than those mentioned amps (just less power). But does it sound any better than some well regarded much simpler designs by Pass, Higara, J.L.Hood etc, I am yet to be convinced. Now comes the building, experimenting, evaluating part – Long live the DIY not FIB (follow it blindly).
argo
Hi,
Yup. All details, topology et al are covered by Selfs Articles in Wireless World and his Book. It is truesly "BOG STANDARD".
I thought this should be obvious.
Bipolar transistors operate by injecting a base current that modulates the resistance of the the collector emitter line. The proportion of this base current compared to the collector current is called "Beta" and is a measure of the transistors gain. However, as is well known, the Beta of a transistor varies with temperature and collector current.
Now most current amplifier designs tend to use the transistor in a circuit that is in essence voltage controlled. Thus our base current becomes in fact a parasitic current that is signal modulated. If we now drive the amplifier stage from a relatively high impedance (like a current source tail, current mirror differential pair) this current will cause distortion (the same applies BTW to the base inputs of the differential pair with respect to feedback loops and sources).
So this varying base current produces a "parasitic" and strongly signal dependent load on whatever preceeds the amplifying stage. It should be easy to see how this would introduce distortion.
By putting an Emitter follower ahead of the input to the given stage the base current modulation of emitter follower will be the base current modulation of the amplification stage divided by the Emitter Followers Beta, so in most cases the current modulation will be reduced by a factor of 100 or more (40db improvement).
Again, this is covered in detail and with measurements in Selfs work. I rarely agree with Mr. Self, but I do agree that it is desirable to linearise a given circuit as much as possible BEFORE applying negative feedback, it not only measures better, it also sounds better.
BTW, all those guy's who are now smug about using FET's in their designs which do not have this peculiarity found with BJT's can wipe that smuggrin off their faces. FET's have other strongly signal modulated parasitics (in this case capacitance) with their own problems. Most FET Amplifiers have a strongly signal dependant open loop bandwith which can make life interesiting after feedback is applied and is (IMNSHO) the cause of the often cited "Mosfet Fog" sound. It can of course be fixed easily by adding an Emitter follower on the Gate of the Mosfet...
See above. A Class A emitter follower (or source follower or cathode follower) implemented well (I'd suggest cascoding and using a current source as load) will have very low distortion of it's own. So no, in this case the emitter followers buffering each stage will materially decrease the open loop distortion and thus increase the open loop linearity of thge Amplifier.
Local degeneration. This reduces the gain in the LTP (again, Self actually does cover this) and thus improves linearity and bandwidth.
It is clearly set for what Self calls "blameless Class B". What this means is that it is tuned to a minimum of crossover distortion. However crossover distortion still exists and usually increasing the quiescent current to higher than the "blameless class B" operation sounds subjectively better, even if measured distortion (THD) is higher.
Then you have not looked very far AT ALL.
If you strip out the emitter follower buffering of the LTP and VAS stage and take out the local degeneration in the LTP you have pretty much the absolute standard quasi complementary amplifier from the late 1960's and early 1970's from the RCA Book. Naim still builds all their amplifiers this way BTW...
Is that what is bothering you? Diodes usually provide poor bias stability. They only really work well on the same chip. The VBE multiplier tends to "throttle" the bias current when the amp gets hot avoiding thermal runaway (at the expense of sound quality), it makes amplifiers more reliable...
I have seen them in quite a few schematics of commercial amplifiers actually. Also, in the 1990's D. Self extensively documented this use and it's benefits. I also seem to remember Ed Cherry and others to have commented in similar tracks.
Because they materially improve things.
If I was making a discrete amplifier I would NOT use this topology at all. I would not use SEPP, I would not use the LTP/VAS/Follower arrangement. I would do something entierly different.
But that is hardly the point here. As far as conventional structure Amplifiers are concerned the LM38XX series is a sensible implementation that ensures good performance within the limits of the design remit. Rather than mucking about with similar discrete circuits which will invariably be less ideal than the chip amp do something differnet and shift the paradigm.
Sayonara
Yup. All details, topology et al are covered by Selfs Articles in Wireless World and his Book. It is truesly "BOG STANDARD".
I thought this should be obvious.
Bipolar transistors operate by injecting a base current that modulates the resistance of the the collector emitter line. The proportion of this base current compared to the collector current is called "Beta" and is a measure of the transistors gain. However, as is well known, the Beta of a transistor varies with temperature and collector current.
Now most current amplifier designs tend to use the transistor in a circuit that is in essence voltage controlled. Thus our base current becomes in fact a parasitic current that is signal modulated. If we now drive the amplifier stage from a relatively high impedance (like a current source tail, current mirror differential pair) this current will cause distortion (the same applies BTW to the base inputs of the differential pair with respect to feedback loops and sources).
So this varying base current produces a "parasitic" and strongly signal dependent load on whatever preceeds the amplifying stage. It should be easy to see how this would introduce distortion.
By putting an Emitter follower ahead of the input to the given stage the base current modulation of emitter follower will be the base current modulation of the amplification stage divided by the Emitter Followers Beta, so in most cases the current modulation will be reduced by a factor of 100 or more (40db improvement).
Again, this is covered in detail and with measurements in Selfs work. I rarely agree with Mr. Self, but I do agree that it is desirable to linearise a given circuit as much as possible BEFORE applying negative feedback, it not only measures better, it also sounds better.
BTW, all those guy's who are now smug about using FET's in their designs which do not have this peculiarity found with BJT's can wipe that smuggrin off their faces. FET's have other strongly signal modulated parasitics (in this case capacitance) with their own problems. Most FET Amplifiers have a strongly signal dependant open loop bandwith which can make life interesiting after feedback is applied and is (IMNSHO) the cause of the often cited "Mosfet Fog" sound. It can of course be fixed easily by adding an Emitter follower on the Gate of the Mosfet...
See above. A Class A emitter follower (or source follower or cathode follower) implemented well (I'd suggest cascoding and using a current source as load) will have very low distortion of it's own. So no, in this case the emitter followers buffering each stage will materially decrease the open loop distortion and thus increase the open loop linearity of thge Amplifier.
Local degeneration. This reduces the gain in the LTP (again, Self actually does cover this) and thus improves linearity and bandwidth.
It is clearly set for what Self calls "blameless Class B". What this means is that it is tuned to a minimum of crossover distortion. However crossover distortion still exists and usually increasing the quiescent current to higher than the "blameless class B" operation sounds subjectively better, even if measured distortion (THD) is higher.
Then you have not looked very far AT ALL.
If you strip out the emitter follower buffering of the LTP and VAS stage and take out the local degeneration in the LTP you have pretty much the absolute standard quasi complementary amplifier from the late 1960's and early 1970's from the RCA Book. Naim still builds all their amplifiers this way BTW...
Is that what is bothering you? Diodes usually provide poor bias stability. They only really work well on the same chip. The VBE multiplier tends to "throttle" the bias current when the amp gets hot avoiding thermal runaway (at the expense of sound quality), it makes amplifiers more reliable...
I have seen them in quite a few schematics of commercial amplifiers actually. Also, in the 1990's D. Self extensively documented this use and it's benefits. I also seem to remember Ed Cherry and others to have commented in similar tracks.
Because they materially improve things.
If I was making a discrete amplifier I would NOT use this topology at all. I would not use SEPP, I would not use the LTP/VAS/Follower arrangement. I would do something entierly different.
But that is hardly the point here. As far as conventional structure Amplifiers are concerned the LM38XX series is a sensible implementation that ensures good performance within the limits of the design remit. Rather than mucking about with similar discrete circuits which will invariably be less ideal than the chip amp do something differnet and shift the paradigm.
Sayonara
Hi Kuei Yang Wang
I look at the input stage a little different than you do.
First the fact: Apart from emitter followers, a BJT works the most linear if it is used as a CURRENT CONTROLLED CURRENT SOURCE -period !!
In my opinion the input emitter followers are there to avoid high input bias currents and therefore broaden the range of resistor values that can be used for the feedback network.
Because the bases of the LTP are now driven by a voltage source it's emitter resistors have to be increased to the 1k instead of the commonly used low values (between 56 and 120 Ohms) to retain a reasonably large linear input range.
Regards
Charles
I look at the input stage a little different than you do.
First the fact: Apart from emitter followers, a BJT works the most linear if it is used as a CURRENT CONTROLLED CURRENT SOURCE -period !!
In my opinion the input emitter followers are there to avoid high input bias currents and therefore broaden the range of resistor values that can be used for the feedback network.
Because the bases of the LTP are now driven by a voltage source it's emitter resistors have to be increased to the 1k instead of the commonly used low values (between 56 and 120 Ohms) to retain a reasonably large linear input range.
Regards
Charles
Hi,
[QUOTE
Can we pray for similar exposition on the topic: inverting versus non-inverting (op)amps?
[/QUOTE]
Oh, that's easy. If employed "inverting" the voltage on the positive and negative input are in effect 0V (reality is more complex though). This means that the common mode input voltage for the amplifier circuit ais also 0V.
Operate the Amplifier non-inverting and the long tailed pair (LTP) is subjected to a common mode voltage (inverting input follows non-inverting input), namely the signal. Due to a number of varying effects a LTP with a common mode signal is not terribly happy (no matter if BJT, FET or Valve) and will cause increased distortion or more precisely reduced open loop linearity. Again this is all covered most excellently by Mr. Self.
This BTW holds for each, all and sundary long tailed pairs. Hence my general dislike (sonically as much as technially) for the use of LTP based Circuits and in solid state my liking for circuits based on the "Diamond Transistor" structure. Here the common mode signal in non-inverting mode is not a problem. In most cases common Audio Op-Amp's and Power Op-Amps sound better inverting (taking to re-invert the polarity by swapping the speaker wires), subjectively speaking. Measured distortion also is usually better.
Sayonara
[QUOTE
Can we pray for similar exposition on the topic: inverting versus non-inverting (op)amps?
[/QUOTE]
Oh, that's easy. If employed "inverting" the voltage on the positive and negative input are in effect 0V (reality is more complex though). This means that the common mode input voltage for the amplifier circuit ais also 0V.
Operate the Amplifier non-inverting and the long tailed pair (LTP) is subjected to a common mode voltage (inverting input follows non-inverting input), namely the signal. Due to a number of varying effects a LTP with a common mode signal is not terribly happy (no matter if BJT, FET or Valve) and will cause increased distortion or more precisely reduced open loop linearity. Again this is all covered most excellently by Mr. Self.
This BTW holds for each, all and sundary long tailed pairs. Hence my general dislike (sonically as much as technially) for the use of LTP based Circuits and in solid state my liking for circuits based on the "Diamond Transistor" structure. Here the common mode signal in non-inverting mode is not a problem. In most cases common Audio Op-Amp's and Power Op-Amps sound better inverting (taking to re-invert the polarity by swapping the speaker wires), subjectively speaking. Measured distortion also is usually better.
Sayonara
Swapping the speaker wires is easy but the volume pot issue brought up by tschrama still bogs me.
Even if lower value pot (10k for an example) is used, gain and therefore also feedback ratio is affected. As feedback changes so does distortion and we get different sound (distortion) for different volumes. That means, basically LM series power opamps in inverting mode must have preceding buffer stage (which will have its own preceding volume control) and we can not do away with just simple volume pot + blocking capacitor. Or am I confusing something here?
argo
tschrama said:
Even if lower value pot (10k for an example) is used, gain and therefore also feedback ratio is affected. As feedback changes so does distortion and we get different sound (distortion) for different volumes. That means, basically LM series power opamps in inverting mode must have preceding buffer stage (which will have its own preceding volume control) and we can not do away with just simple volume pot + blocking capacitor. Or am I confusing something here?
argo
if your gainclone sounds only "average" to you, there's 2 possibilities i can think of. 1.) you didn't build it right, or 2.) your speakers are a poor match for them.
it's been said a hundred times but the circuit layout and implementation of this little chip means everything. this includes the quality of the power supply components you use - diodes, caps, the transformer itself. the chip also seems to favor warmer-sounding parts such as "old-school" carbon resistors - the wrong metal film can make it sound cold and sterile. if your gainclone is heating up a lot then there's probably an oscillation problem as well, which not only kills power but the sound quality as well. even if it's not oscillating full power, there can still be low-level oscillations that can be buggers. getting the thing stable requires a lot of care.
the gainclone is NOT a universal amplifier by any means, you really need the right speaker. i've yet to find hard and fast rules as to what speakers work well, though a reasonably high and flat load impedance seems to be one obvious one. i recently reviewed a commercial amp which is essentially a gainclone. it sounded phenomenally musical with a bookshelf 2-way and a small transmission line 2-way, both very benign loads. it sounded like **** with my Merlin TSMs, which are actually easy loads. another commercial gainclone was reviewed by my editor and he confirmed that it sounded extraordinary with some, mundane with others.
that all said, when done right, not oscillating, and connected to the right speaker, it can sound VERY good indeed. it strains when you play it loudly but at moderate volume levels it can really bring out nuances and shadings in a highly enjoyable and natural manner... many "conventional" amplifiers will sound blurred and flat by comparison.
it's been said a hundred times but the circuit layout and implementation of this little chip means everything. this includes the quality of the power supply components you use - diodes, caps, the transformer itself. the chip also seems to favor warmer-sounding parts such as "old-school" carbon resistors - the wrong metal film can make it sound cold and sterile. if your gainclone is heating up a lot then there's probably an oscillation problem as well, which not only kills power but the sound quality as well. even if it's not oscillating full power, there can still be low-level oscillations that can be buggers. getting the thing stable requires a lot of care.
the gainclone is NOT a universal amplifier by any means, you really need the right speaker. i've yet to find hard and fast rules as to what speakers work well, though a reasonably high and flat load impedance seems to be one obvious one. i recently reviewed a commercial amp which is essentially a gainclone. it sounded phenomenally musical with a bookshelf 2-way and a small transmission line 2-way, both very benign loads. it sounded like **** with my Merlin TSMs, which are actually easy loads. another commercial gainclone was reviewed by my editor and he confirmed that it sounded extraordinary with some, mundane with others.
that all said, when done right, not oscillating, and connected to the right speaker, it can sound VERY good indeed. it strains when you play it loudly but at moderate volume levels it can really bring out nuances and shadings in a highly enjoyable and natural manner... many "conventional" amplifiers will sound blurred and flat by comparison.
argo said:
I'll admit that this baffled me for a bit, when I tried to figure out how to convert the linear pot into a stepped attenuator with equal 2dB steps.
To understand the feedback in an inverting op-amp, you must consider that the inverting input is a virtual ground point in the circuit. The way gain works is that the op-amp ensures that the potential of the inverting input is ground. The only way it can do this is to provide an equal and opposite current from the output to sum with the current from the input. So, you must calculate how much current is presented to the inverting input by the 10k resistor. The voltage divider (really current splitter) shunts the rest of the current away to ground, never to be seen by the op-amp.
As a result, the gain of the op-amp is always -R(feedback)/R(in). In the gainclone circuit, this is -220k/10k, or -22, always.
So, while the total gain of the circuit can approach zero, the gain of the op-amp is always -22, ensuring stability. The small price to pay for this is that the input impedence changes with the overall circuit gain.
Jeremy
oh, i didn't know you could still get carbon comp (not carbon film) resistors. not all carbon comp resistors are equal though, i'm not sure how the Xicons sound. they probably have steel endcaps/leads which usually sound not so good. i know some people really like the way Allen-Bradleys sound, maybe the Xicons are worth a shot.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.