Re: for once I agree with Peter...
You see, whatever good or bad performer the GC may be, I would be never building this type of cute, little amps, if it wasn't for LM3875 chip😉
http://www.diyaudio.com/forums/showthread.php?s=&postid=230711#post230711
DrG said:
You see, whatever good or bad performer the GC may be, I would be never building this type of cute, little amps, if it wasn't for LM3875 chip😉
http://www.diyaudio.com/forums/showthread.php?s=&postid=230711#post230711
Pjotr said:
If you're referring to what I'd mentioned previously, I wasn't talking about using a source/emitter follower. I was talking about using a current source tied to the 3875's output. Base current is irrelevant.
The datasheet says its internally current limited at 4 amps. 20 volts into an 8 ohm load will give you 25 watts and requires just 2.5 amps.
So I guess it depends what one defines as "reasonable output."
se
thermally challenged?
What are you saying here, Peter? Without these chips NOBODY would be building GC's!
I don't think Pjotr is thinking of a tie-down source, Steve. But looking at this scenario of yours:
For 25W/8R, 2.5App is correct, meaning your Iq source would need to be 1.25A and your speakers very forgiving. To maintain 25W/4R your source would need to sink 1.75A however. And for a perfect voltage amplifier you would need Iq=2.5A to guarantee 50W into 4R, thereby catering for most real-world speaker impedance curve variations.
Working from a 30v +ve supply your LM chip's upper half would dissipate 38, 53 or 75 Watts of heat continuously, requiring 2°C/W, 1.2°C/W or 0.46°C/W heatsinking (NS calculations on data sheet) - on the assumption that Ta is no more than 25°C and ignoring entirely the identical dissipation and similar heat-sinking requirements of your proposed current source!
Now you're approaching Nelson Pass-style heat sink acreage, not to mention that all of these proposed SE-chip dreams fall way beyond the SOAR curves on page 6 of the LM3875 data sheet. It looks like your chip should last... oh, around 10-100 milliseconds (joke). Point is you're gonna toast the little bugger. Luckily they're free, I hear...
So, why not keep everyone happy by (a) keeping the 3875 in-circuit (shriek!) and (b) whacking on a NFB-less class A source-follower... it's the path of least (thermal) resistance...hehehehe
What are you saying here, Peter? Without these chips NOBODY would be building GC's!
I don't think Pjotr is thinking of a tie-down source, Steve. But looking at this scenario of yours:
For 25W/8R, 2.5App is correct, meaning your Iq source would need to be 1.25A and your speakers very forgiving. To maintain 25W/4R your source would need to sink 1.75A however. And for a perfect voltage amplifier you would need Iq=2.5A to guarantee 50W into 4R, thereby catering for most real-world speaker impedance curve variations.
Working from a 30v +ve supply your LM chip's upper half would dissipate 38, 53 or 75 Watts of heat continuously, requiring 2°C/W, 1.2°C/W or 0.46°C/W heatsinking (NS calculations on data sheet) - on the assumption that Ta is no more than 25°C and ignoring entirely the identical dissipation and similar heat-sinking requirements of your proposed current source!
Now you're approaching Nelson Pass-style heat sink acreage, not to mention that all of these proposed SE-chip dreams fall way beyond the SOAR curves on page 6 of the LM3875 data sheet. It looks like your chip should last... oh, around 10-100 milliseconds (joke). Point is you're gonna toast the little bugger. Luckily they're free, I hear...
So, why not keep everyone happy by (a) keeping the 3875 in-circuit (shriek!) and (b) whacking on a NFB-less class A source-follower... it's the path of least (thermal) resistance...hehehehe
Re: thermally challenged?
What chips you have in mind?😉
If you mean LM3875, you just confirmed that they are special. If you mean other chips, I don't know what you are saying here.
DrG said:
What chips you have in mind?😉
If you mean LM3875, you just confirmed that they are special. If you mean other chips, I don't know what you are saying here.
I didn't confirm diddly... The only special aspects of these chips are (1) price (squat), and (2) the messianic following they enjoy.
Follow your own philosophy, Pete: "Say what you mean, mean what you think and think whatever you like..." Then you won't confuse people.
Follow your own philosophy, Pete: "Say what you mean, mean what you think and think whatever you like..." Then you won't confuse people.
But you just said that without those chips nobody would be building GCs. It's me who's confused here. What is so special about those chips that everybody's building amps with them? Certainly messianic following isn't the only reason (beside the price).
Re: thermally challenged?
I'm not sure what he was thinking. Which is why I prefaced my reply with "If you're refering to..."
Since it's single-ended, Iq would need to be 2.5 amps. And it just occurred to me that you'd have to drop it down to 2 amps because otherwise the top output device would need to go to a maximum of 5 amps which is more than the 4 amp internal current limit.
So instead of 25 watts, you're looking at 16 watts into 8 ohms. Still a viable amount of power for many. Not everyone owns power sponge 80-85dB loudspeakers.
And why would you need to guarantee twice the power into a 4 ohm load? We're not talking about a utilitarian Swiss Army knife here.
I was thinking a +/-20 volt supply so we're looking at a continuous dissipation more like 40 watts.
A 20 volt CE voltage and a 2 amp collector current looks doable to me.
Why would that necessarily keep everyone happy? The 3875 would still be operating in class AB. Also a source follower (or emitter follower or cathode follower) employs what amounts to 100% NFB so you're not escaping negtaive feedback with a follower. Though you could keep it out of the global loop.
se
DrG said:
I'm not sure what he was thinking. Which is why I prefaced my reply with "If you're refering to..."
Since it's single-ended, Iq would need to be 2.5 amps. And it just occurred to me that you'd have to drop it down to 2 amps because otherwise the top output device would need to go to a maximum of 5 amps which is more than the 4 amp internal current limit.
So instead of 25 watts, you're looking at 16 watts into 8 ohms. Still a viable amount of power for many. Not everyone owns power sponge 80-85dB loudspeakers.
And why would you need to guarantee twice the power into a 4 ohm load? We're not talking about a utilitarian Swiss Army knife here.
I was thinking a +/-20 volt supply so we're looking at a continuous dissipation more like 40 watts.
A 20 volt CE voltage and a 2 amp collector current looks doable to me.
Why would that necessarily keep everyone happy? The 3875 would still be operating in class AB. Also a source follower (or emitter follower or cathode follower) employs what amounts to 100% NFB so you're not escaping negtaive feedback with a follower. Though you could keep it out of the global loop.
se
Re: thermally challenged?
Would anyone ever make tube amps if there were no tubes? Would there ever be music if everyone was deaf? Would anyone ever drive cars if there were no cars? Would anyone ever be eating Chinese if there were no Chinese? Or Pizza without Italians, smoke without fire, reading without an alphabet, computers without monitors, carpet without fibres, CDs without CD players, experience without experiences...
You go figure!
Joe R.
DrG said:
Would anyone ever make tube amps if there were no tubes? Would there ever be music if everyone was deaf? Would anyone ever drive cars if there were no cars? Would anyone ever be eating Chinese if there were no Chinese? Or Pizza without Italians, smoke without fire, reading without an alphabet, computers without monitors, carpet without fibres, CDs without CD players, experience without experiences...
You go figure!
Joe R.
Re: point (to point?) conceded...
Fedde
I wonder how much RF is picked up by a 5cm antenna. I think that the inductance of 5cm wire at frequencies >100 Mhz is not negligible.DrG said:
Yes, I'd be interested to hear how the term "microdynamics" is defined...DrG said:
Fedde
ok, I'll bite
Um, what does inductance at over 100Mhz have do do with it?
Besides, it's an extra 5cm overall, not a 5cm piece, so if we're talking RF you're getting into some pretty hi freqs.
I also don't see how RF in the feedback loop can be any worse than RF in the input.
I too find this short signal path obsession to be a little overblown.
Chris
Um, what does inductance at over 100Mhz have do do with it?
Besides, it's an extra 5cm overall, not a 5cm piece, so if we're talking RF you're getting into some pretty hi freqs.
I also don't see how RF in the feedback loop can be any worse than RF in the input.
I too find this short signal path obsession to be a little overblown.
Chris
Well, the thing Kimura suggested in his article about opamps was that he experienced his chippies to be more stable using short feedback loops. Oscillation of the LM3875 is likely to occur around 200-400 Mhz. I would expect that if the wire of the feedback is long (say 5-10 cm), the extra inductance could shift some poles. Maybe an opamp expert can shed some light on this matter !?
Fedde
Fedde
Re: Re: thermally challenged?
I would not agree on this one. The was at least one composer who created great music, even if he was deaf.
I'm even more confused at this point with development of my single thought. So I will clarify it again. The chip amps were around for at least 20 years, yet nobody was building cute little (hi-end) amps with them, until somebody pointed LM chip (and recently few other), used specific application with it and made it news. If it wasn't for that (the great chip hype) I would still be building big class A monster amps and I would still be saying in my posts that chips are no good😉. But chips were always around, as far as I remember.
To me, the biggest appeal of the GC concept is possibility of creating interesting "structures" which are possible only when you deal with few parts, single chip and 2 PS caps. This allows for building "unique" designs (and really compact), which are not easy to attain with more conventional approaches. I'm not mentioning the quality of sound, but if it wasn't "that" good, whould I go through all that trouble and wasted my time?
Joe Rasmussen said:
I would not agree on this one. The was at least one composer who created great music, even if he was deaf.
I'm even more confused at this point with development of my single thought. So I will clarify it again. The chip amps were around for at least 20 years, yet nobody was building cute little (hi-end) amps with them, until somebody pointed LM chip (and recently few other), used specific application with it and made it news. If it wasn't for that (the great chip hype) I would still be building big class A monster amps and I would still be saying in my posts that chips are no good😉. But chips were always around, as far as I remember.
To me, the biggest appeal of the GC concept is possibility of creating interesting "structures" which are possible only when you deal with few parts, single chip and 2 PS caps. This allows for building "unique" designs (and really compact), which are not easy to attain with more conventional approaches. I'm not mentioning the quality of sound, but if it wasn't "that" good, whould I go through all that trouble and wasted my time?
Would anyone be tempted to enter into a repetitious philosphical soliloquy withaout a repetitious philosphical soliloquy? Apparently...YES!
I agree fedde, that could be catastrophic! Not to mention what speaker cables might do with random RF... But this is an audio forum, not a SETI page, so maybe it really doesn't matter eh, whaddaya think? Hearing any little green men on you GC radio?
Within the audio band, a MOSFET follower pair would draw less bias current than the LM3875's internal Iq... which leaves it in class A bias. Go crunch the numbers... And yes, I'm fully aware of the operational structure and inherent 100% (current) FB of any single-device follower. Which is irrelevant to the point being made. The point was to get realistic, symmetric class A output from an LM3875 chip without resorting to a tie-down and without lengthening the NFB path. Perhaps you can suggest a better option, Steve...
If these are your reasons Peter, then excluding the "compactness" motive you probably are indeed wasting your time. What part of working with these chips offers scope for 'interesting' or 'unique' 'structures'...??? GC lore (ex buffers, reg PSU's etc) leaves you with inverting or non-inverting op-amp connections and all the minimal necessary caps. That's it! Nothing new there... hasn't been for 30 years in fact. And every GC mod was invented or applied in the discrete realm first.
Now tell me "cheap, easy, compact and I like the sound" and that's fine, I'll believe you but until you've discovered the hundreds of additional mods/upgrades/whatever possible in discrete design you won't realize the limitations of working with chips. Every chip mod can be implemented in some way to a discrete amp, but not even 10% (guess) of possible discrete mods could be done with chips.
DrG said:
True enough. I didn't stop to think it'd be driving a much higher impedance.
But it's still complimentary and my point was that you could turn the 3875's class AB, complimentary output stage into a single-ended class A output stage.
That's fine. It's yet another option for people to choose from, which was all that I was offering. Which is why I was wondering why you were insisting that it must be able to double power into 4 ohms and why you think this approach would keep everyone happy as if everyone prefers the same things?
Better for whom? I've already suggested an option, one which is inherently no better or no worse than the option you suggest. Which is "better" will depend on the tastes and preferences of the individuals who would be choosing between the two. For some one option may be better and for others, the other option may be better.
se
Of course the part which requires only 5 components to built the complete amp (including chip), the requirement for only a modestly sized heatsink and enough satisfying sound. No discreet SS circuit is that appealing😉DrG said:
anything but that...
Now let's not go there, Steve. I've never questioned personal choice, only some of the proffered reasons whereupon such choices are made. If you prefer then read "different" or "challenging"...
The exchanges revolved around exploring ways to have a class A-output GC. Therefore "better" was simply asking you to propose a different, possibly "better" solution to the question of whether this is possible and how such a class A GC might be configured. Given that a tie-down CCS would limit output power severely... as you pointed out, some 16W/8R which would become all of 8W/4R... a little on the low side unless you have 96+dB/W speakers or prefer not to listen too loudly.
Now let's not go there, Steve. I've never questioned personal choice, only some of the proffered reasons whereupon such choices are made. If you prefer then read "different" or "challenging"...
The exchanges revolved around exploring ways to have a class A-output GC. Therefore "better" was simply asking you to propose a different, possibly "better" solution to the question of whether this is possible and how such a class A GC might be configured. Given that a tie-down CCS would limit output power severely... as you pointed out, some 16W/8R which would become all of 8W/4R... a little on the low side unless you have 96+dB/W speakers or prefer not to listen too loudly.
Re: anything but that...
Fair 'nuff.
But I thought my original suggestion was a different and possibly "better" solution. Just as your suggestion was a different and possibly "better" solution. So why should I go on proposing more solutions? It was simply something that occurred to me while I was replying to SY's post and something I don't recall having seen implemented at this scale so I just tossed it out onto the table.
It seems to me that the only reason you asked me to propose something different is because my original suggestion wasn't an approach you'd care to take.
So? There's a significant number of people for whom that 16 into 8 and 8 into 4 would do just fine. In fact there's a surprising number of people who are doing just fine with less than 1 watt. And some of these people might find that a chip amp with a single-ended output stage sounds better than a chip amp with a complimentary output stage.
se
DrG said:
Fair 'nuff.
But I thought my original suggestion was a different and possibly "better" solution. Just as your suggestion was a different and possibly "better" solution. So why should I go on proposing more solutions? It was simply something that occurred to me while I was replying to SY's post and something I don't recall having seen implemented at this scale so I just tossed it out onto the table.
It seems to me that the only reason you asked me to propose something different is because my original suggestion wasn't an approach you'd care to take.
So? There's a significant number of people for whom that 16 into 8 and 8 into 4 would do just fine. In fact there's a surprising number of people who are doing just fine with less than 1 watt. And some of these people might find that a chip amp with a single-ended output stage sounds better than a chip amp with a complimentary output stage.
se
sacrilege?
You are correct, that is not the approach I would care to take. And the reasons are (1) that there are real limitations to output power and (2) that I question the LM's ability to sustain a high, constant Iq. You are also correct in pointing out that 8W (or even <1W) suffices for some. But all things being equal, most people would like a little more. And so I was trying to arrive at a solution more acceptable to a larger subset of the GC audience, not to mention a larger subset of loudspeakers.
How about a "compromise" - a smaller tie-down CCS (say 250-400mA) into a class A FET follower, with or without global NFB. What do you think of such a concept?
In fact, I've been toying with putting on paper a sort of "ultimate-banzai-clone", embodying most major GC mods discussed in various threads, and implementing both class A approaches, as above, both yours and mine.
How about a tube LTP, tube buffered (ie 4xtubes total) and cap-coupled to the LM which is differentially driven, with inverted FB and CCS tied-down output, DC into class A buffer... I wonder if the minimalists are going to scream "sacrilege"...
You are correct, that is not the approach I would care to take. And the reasons are (1) that there are real limitations to output power and (2) that I question the LM's ability to sustain a high, constant Iq. You are also correct in pointing out that 8W (or even <1W) suffices for some. But all things being equal, most people would like a little more. And so I was trying to arrive at a solution more acceptable to a larger subset of the GC audience, not to mention a larger subset of loudspeakers.
How about a "compromise" - a smaller tie-down CCS (say 250-400mA) into a class A FET follower, with or without global NFB. What do you think of such a concept?
In fact, I've been toying with putting on paper a sort of "ultimate-banzai-clone", embodying most major GC mods discussed in various threads, and implementing both class A approaches, as above, both yours and mine.
How about a tube LTP, tube buffered (ie 4xtubes total) and cap-coupled to the LM which is differentially driven, with inverted FB and CCS tied-down output, DC into class A buffer... I wonder if the minimalists are going to scream "sacrilege"...
Hi,
The remarks concerning “short wiring” your GC here is looking in the wrong direction IMHO. It has not much to do with “delays” and HF stability in the over 100 MHz region. Anyone who has build a discrete amp knows, or should know, that board layout is very important for good results. This applies especially traces and/or wires that carry large current. What plays havoc here is inductive coupling between traces or wires and error voltages generated by ground return currents. Keeping the wires short minimises coupling between wires and minimises error voltages.
To avoid confusion about a SEC in my previous post. I was intending a set up like the attached schematic. Thus including the buffer fet into the FB loop. I see no good reason to fry the LM chip itself. But going into that direction the good old LM391 is probably a better solution for the op amp. It has at least more freedom to stabilise the circuit. Unfortunately it is obsolete now.
Cheers
The remarks concerning “short wiring” your GC here is looking in the wrong direction IMHO. It has not much to do with “delays” and HF stability in the over 100 MHz region. Anyone who has build a discrete amp knows, or should know, that board layout is very important for good results. This applies especially traces and/or wires that carry large current. What plays havoc here is inductive coupling between traces or wires and error voltages generated by ground return currents. Keeping the wires short minimises coupling between wires and minimises error voltages.
To avoid confusion about a SEC in my previous post. I was intending a set up like the attached schematic. Thus including the buffer fet into the FB loop. I see no good reason to fry the LM chip itself. But going into that direction the good old LM391 is probably a better solution for the op amp. It has at least more freedom to stabilise the circuit. Unfortunately it is obsolete now.
Cheers
Attachments
- Status
- Not open for further replies.