Long time ago and long time before we have heard of a Gaincard, I already used the LM3875's and LM3886's for an active speaker system. These integrated power amps outperformed the rest of the available power IC's IMO, including the powerdriver IC's LM391 from NS and the µPC1225 from NEC (those IC’s who need just a pair of external discrete power transistors).
All credits for the really good performance of the LM38xx power IC’s belong to the design team of NS indeed.
A year ago, I first heard a 47-Labs Gaincard and was really impressed that such a small block of aluminium could squeeze out such a lot of music. For a real Gaincard you have to use also the rather thin teflon isolated speaker cables. It is part of the lore, not my idea.
But yes you have to give Mr. Kimura the credits for bringing these single chip amps to the audiophile world. I wonder if so many people would enjoy building low cost gain cards without Mr. Kimura. IMHO Gaincards are a big joke in a sense that these cheap amps blow away many commercial expensive audio gear if it comes to enjoying recorded music.
😉
All credits for the really good performance of the LM38xx power IC’s belong to the design team of NS indeed.
A year ago, I first heard a 47-Labs Gaincard and was really impressed that such a small block of aluminium could squeeze out such a lot of music. For a real Gaincard you have to use also the rather thin teflon isolated speaker cables. It is part of the lore, not my idea.
But yes you have to give Mr. Kimura the credits for bringing these single chip amps to the audiophile world. I wonder if so many people would enjoy building low cost gain cards without Mr. Kimura. IMHO Gaincards are a big joke in a sense that these cheap amps blow away many commercial expensive audio gear if it comes to enjoying recorded music.
😉
Just thot i'd repost this JoeList (ie tube amp freaks of the 1st degree) message that was my 1st exposure to the idea of the ChipAmp. From 22-Nov-99:
This was before all the hype ...
dave
This was before all the hype ...
dave
MBK said:
True. Fullrange has a certain aesthetic appeal to it but I think it tends toward the "too simple" side of the scale. For me at least. I'm hoping to try and strike a balance between fullrange and multi-way. Haven't really started in earnest yet. Want to try and get the electronics nailed down first.
Hybrids utilize monolithic integrated circuits, discrete semiconductor chips and passive components as well as thick film technology built on a ceramic (usually aluminum or beryllium oxide) substrate which effectively acts as the circuit board. The circuit "traces" are made by screening "inks" made up of various metal alloys, often palladium silver and then the substrate is put in a kiln and fired, fusing alloy onto the ceramic surface.
Here's the innards of an Apex PA-16 power opamp:
An externally hosted image should be here but it was not working when we last tested it.
The two squares at the right are the discrete output transistors in raw die form. There's a monolithic differential opamp (also in die form) for the input and you can also see the various discrete chip resistors and capacitors that make up the rest of the circuit.
Only drawback is that they're considerably more expensive than the monolithic types. The PA-02 and PA-16 go for about $60 a pop in onesies twosies.
Hehehehe. That could be... interesting. 🙂
What? You mean you're not going to wait to announce it to the world on Larry King Live? 🙂
Hehehe. Not sure. From what I've heard, they basically use a gob of nested feedback loops.
se
Re: Re: Where can I get details to build one?
One remaining newbie question, which popped up when I read Scott Nixon's Webpage about his gainclone PCBs. What amp chip do you suggest one should use for best results? The 1875 or the 3875? And what's the difference(s), in brief?
Thanks.
Tarun
Andrewbee said:
Thanks Andrewbee and Dave.planet10 said:
One remaining newbie question, which popped up when I read Scott Nixon's Webpage about his gainclone PCBs. What amp chip do you suggest one should use for best results? The 1875 or the 3875? And what's the difference(s), in brief?
Thanks.
Tarun
Biggest difference between 1875 and 3875 is power. The 3875 also has more sophisticated protection circuitry.
You really don't need a huge amount of complexity to make a solid-state amplifier with perfectly adequate measured performance. And beyond "adequate measured performance" lies the murky realms of subjective evaluation where no two people can agree exactly.
To scoff at the LM3875 you have to either (a) demonstrate the performance numbers are deficient in some measurable way, or (b) appeal to subjectivism.
As far as I can tell, DrG is doing the latter. He objects to "zero scope for creativity" - in what sense does a circuit element with fixed voltage gain and low output impedance require creativity?
Now, maybe some of the appeal of the GainCard is the subjectivist pixie-dust garnered with those 9mm signal paths. But the argument seems to be that "subjectivist pixie-dust is an expensive substance, so it can't possibly be obtained in a $5 chip".
I'm up for a good subjective-vs-objective, double-blind-testing flame war any time night or day, but trying to dismiss one school of thought as "unscientific" with an argument which is equally unscientific seems to me, er, just puzzling.
Cheers
IH
To scoff at the LM3875 you have to either (a) demonstrate the performance numbers are deficient in some measurable way, or (b) appeal to subjectivism.
As far as I can tell, DrG is doing the latter. He objects to "zero scope for creativity" - in what sense does a circuit element with fixed voltage gain and low output impedance require creativity?
Now, maybe some of the appeal of the GainCard is the subjectivist pixie-dust garnered with those 9mm signal paths. But the argument seems to be that "subjectivist pixie-dust is an expensive substance, so it can't possibly be obtained in a $5 chip".
I'm up for a good subjective-vs-objective, double-blind-testing flame war any time night or day, but trying to dismiss one school of thought as "unscientific" with an argument which is equally unscientific seems to me, er, just puzzling.
Cheers
IH
Regarding the statement cediting that Kimura:
"2. Rocognize how good chip amps can be AND recognize one of the very best chips. It's easy to know this now that he discoved it. He had the confidence to believe in what he heard"
I'm not certain about the time line, but I think Nohr and Marchand got there first. They just didn't package (and PRICE) it right to get the over the top reviews from the high-end yogi's.
"2. Rocognize how good chip amps can be AND recognize one of the very best chips. It's easy to know this now that he discoved it. He had the confidence to believe in what he heard"
I'm not certain about the time line, but I think Nohr and Marchand got there first. They just didn't package (and PRICE) it right to get the over the top reviews from the high-end yogi's.
Hmmm, has anybody dismissed yet my previous "scientific" argument, that I read rising distortion with falling outout power for the LM chip amps datasheets... while say a Pass topology Class A will have falling distortion with falling output power...
By inference and dangerously extrapolated: using a chip for usual listening levels <1W, especially with high eff. speakers, practically guarantees to operate it in its highest distortion area. Combined with the benign distortion spectrum of the 38xx chips, this would lead to a sonic signature close to SET amps 😉 😉 😉 ... maybe this makes them so appealing
Corollary: operating an OPA548 into headphones sounded incredibly sweet to me... maybe all these harmonics made it so mushy... because same pre, same headphones into OPA2134/BUF634 sound extremely detailed and precise - but NOT sweet.
By inference and dangerously extrapolated: using a chip for usual listening levels <1W, especially with high eff. speakers, practically guarantees to operate it in its highest distortion area. Combined with the benign distortion spectrum of the 38xx chips, this would lead to a sonic signature close to SET amps 😉 😉 😉 ... maybe this makes them so appealing
Corollary: operating an OPA548 into headphones sounded incredibly sweet to me... maybe all these harmonics made it so mushy... because same pre, same headphones into OPA2134/BUF634 sound extremely detailed and precise - but NOT sweet.
Irrelevant, I think. First, remember that this is not a THD graph, it's a THD + N spec. So one would expect, for a constant noise voltage, this number to rise with decreasing power. Second, consider magnitudes; I don't have Pass's comparable specs, but I'll bet the numbers overall are higher. Third, both sets of numbers are undoubtedly well below the threshold of detection by ear. If someone can, in a valid listening test, detect their presence in a signal path, that's a data point. But AFAIK, no one has done so, and I suspect no one will.
Regarding the "discovery" of the NS chips, this is pretty funny stuff. It's not like someone figured out that, say, a disk head driver or an RF modulator worked really well as a power amp. These chips are DESIGNED to be high-quality audio power amps. And someone "discovered" that they indeed work well as audio power amps! How creative! How brave! Thank goodness, we all have been given permission to use them as intended by their designers!
Hi MBK,
If you want sweeties go for a single ended class-A. Your reasoning is correct concerning low level distortion but it is not that much in agreement with what you actually hear out of the LM chips and no they do not sound like SET amps at all IMO. They are NOT the best amps of the world but they do the job quite “sweet” also with 90 dB efficient speakers at low listening levels.
Cheers
If you want sweeties go for a single ended class-A. Your reasoning is correct concerning low level distortion but it is not that much in agreement with what you actually hear out of the LM chips and no they do not sound like SET amps at all IMO. They are NOT the best amps of the world but they do the job quite “sweet” also with 90 dB efficient speakers at low listening levels.
Cheers
SY said:
Yes, but irrespective of noise one would also expect to see that number increase with decreasing power due to their class AB output biasing.
By the way, I can't think of any reason why one couldn't use the same class A biasing trick with these devices that is sometimes used on small signal opamps. You could effectively end up with a little single-ended class A amplifier.
Has anyone tried this yet?
se
Pjotr said:
Which, unless I'm overlooking something, could be accomplished with a chip amp. See above.
se
OK, my lack of depth in solid state amplification is now showing; wouldn't you expect the opposite, since at lower levels the amp is effectively running class A? I'm assuming that the designers were smart enough to choose a bias point that avoids crossover distortion.
SY said:
Sure, but the "lower levels" in this case will be in the milliwatt range.
I don't see how you can avoid crossover distortion short of pure class A. Unless you're going to operate in pure class A, it would seem that the best approach would be to keep it biased as close as possible to class B.
I believe Douglas Self showed that as you start going into class AB, crossover distortion actually increases because the discontinuity is occurring at higher levels.
Oh, hey, just thought of a great name for a chip amp biased into single-ended class A. Call it the ChipSET. 😀
Of course the T will have to stand for Transistor rather than Triode, though technically, I suppose a transistor is also a triode.
se
SY said:
yes, this is classic:
"Hmmm, I ronder if National has any undiscovered chips that I can ruse in a high quality audio amp?"
(sound of browser opening , and navigation to National.com. Further sounds of search term "audio amp" being keyed into search box.)
"Well! here's one that probabry hasn't been tried by anyrone, the LM1875. I think I'll build a high qrarity amp with it, make thousands of dorrars, and be a hero in the eccentric, stuck up Hi Fi community, because if I charge at least 2k, people will assume it is a technical marvel that uses some new, undiscovered chip to produce all that sweet sweet mrusic"
(sounds of press release being typed announcing the revolutionary GainCard...)
To Be Continued......
SY,
I have only seen one THD vs. power plot on Pass amps, found it on the Pass DIY site. They measured THD+N so we can compare to the data sheets.
According to this,
THD vs. f: At 1 W the Zen V4 achieves 0.05% THD+N quite equally from ca. 20 Hz to 3000 kHz, thus covering the most audible range of 100-3k.
THD vs. power: For 1 kHz, Pass shows ca. 0.03% at 0.1 W and ca. 0.3% at 10W.
According to LM3886 data sheet,
THD vs. f: NS doesn't give data for 1 W, but at 30W the chip achieves ca. 0.006% at 20 Hz, falling to ca. 0.0015% at 200 Hz, rising to ca. 0.005% at 3 kHz. From THD vs. Power, extrapolated, aggregate THD at 1 W should measure ca. 0.008%, hard to compare since THD changes a lot over f but still, very low indeed.
THD vs. power: For 1 kHz we get ca. 0.01% at 0.1W and ca. 0.015% at 10W.
Note that the measured ZenV4 was designed for DIY and putting it up against optimized production chips I stretch the limits of fairness of course, but i like to discuss this with real numbers even when I just partly strained my whole argument 🙂
So, anyway - yes, the chip achieves lower THD than this particular non production Pass DIY design, but 1- frankly the discrete design achieves remarkable numbers where it matters most, contrary to popular opinion on discrete designs, especially SE, and 2- the distortion of ZenV4 develops more evenly than in the chip.
What difference IMD, input overload, overshoot recovery and clipping characteristics would make to the perceived sound we can only speculate.
Pjotr,
I quite like chip amps and have used them, but some perceived artifacts (on transients) of my implementatioon with OOA58 made me look further. I actually don't know how SET's sound: I inferred from listener descriptions that sounded quite like what I heard using an OPA548 as a headphone amp - it sounded very sweet, but not very precise. Through speakers the OPA548's didn;t sound sweet. But now I built an Aleph design and find them "smoother" and "more precise" (in low level detail), yet not "sweet" or mushy.
As most DIY'ers I have trouble with accurte, replicable measurements or DBX tests. I would like to perform them but either I can't or I find myself too lazy - once I like a sound / a design and don't find anything wrong with it, I keep it until I start finding something wrong again 😉 ... so all my statements should be read as opinions, not replicable scientific results. [as a scientist I would not equate even "replicable scientific results" as "facts", just as "temporary state of the art opinion" 😎 ]
I have only seen one THD vs. power plot on Pass amps, found it on the Pass DIY site. They measured THD+N so we can compare to the data sheets.
According to this,
THD vs. f: At 1 W the Zen V4 achieves 0.05% THD+N quite equally from ca. 20 Hz to 3000 kHz, thus covering the most audible range of 100-3k.
THD vs. power: For 1 kHz, Pass shows ca. 0.03% at 0.1 W and ca. 0.3% at 10W.
According to LM3886 data sheet,
THD vs. f: NS doesn't give data for 1 W, but at 30W the chip achieves ca. 0.006% at 20 Hz, falling to ca. 0.0015% at 200 Hz, rising to ca. 0.005% at 3 kHz. From THD vs. Power, extrapolated, aggregate THD at 1 W should measure ca. 0.008%, hard to compare since THD changes a lot over f but still, very low indeed.
THD vs. power: For 1 kHz we get ca. 0.01% at 0.1W and ca. 0.015% at 10W.
Note that the measured ZenV4 was designed for DIY and putting it up against optimized production chips I stretch the limits of fairness of course, but i like to discuss this with real numbers even when I just partly strained my whole argument 🙂
So, anyway - yes, the chip achieves lower THD than this particular non production Pass DIY design, but 1- frankly the discrete design achieves remarkable numbers where it matters most, contrary to popular opinion on discrete designs, especially SE, and 2- the distortion of ZenV4 develops more evenly than in the chip.
What difference IMD, input overload, overshoot recovery and clipping characteristics would make to the perceived sound we can only speculate.
Pjotr,
I quite like chip amps and have used them, but some perceived artifacts (on transients) of my implementatioon with OOA58 made me look further. I actually don't know how SET's sound: I inferred from listener descriptions that sounded quite like what I heard using an OPA548 as a headphone amp - it sounded very sweet, but not very precise. Through speakers the OPA548's didn;t sound sweet. But now I built an Aleph design and find them "smoother" and "more precise" (in low level detail), yet not "sweet" or mushy.
As most DIY'ers I have trouble with accurte, replicable measurements or DBX tests. I would like to perform them but either I can't or I find myself too lazy - once I like a sound / a design and don't find anything wrong with it, I keep it until I start finding something wrong again 😉 ... so all my statements should be read as opinions, not replicable scientific results. [as a scientist I would not equate even "replicable scientific results" as "facts", just as "temporary state of the art opinion" 😎 ]
I'll get around to building a couple of LM**** stages into a dead amp chassis (the STK Ic's are NLA) and see what they are like for myself.
The in person reports that I have heard are that the LM amplifiers are actually very good up until the overload/protection circuitry cuts in and then they fall to pieces badly.
I am also reliably informed that some of the TDA15** series amplifiers are far better under all conditions.
Eric.
The in person reports that I have heard are that the LM amplifiers are actually very good up until the overload/protection circuitry cuts in and then they fall to pieces badly.
I am also reliably informed that some of the TDA15** series amplifiers are far better under all conditions.
Eric.
Of course it's been tried and found to work for some but not for others. The 1875 with 70mA idling is probably the better candidate. In the experiments i've heard about, the negative rail bias is quite low - about 15mA. Whatever it achieves is certainly not class A. Realistically pure class A will only be possible on cans.
analog_sa said:
What's the idling current to do with it?
I'm talking about slapping a current source on the output between the output and the negative supply rail and shoving a few amps of current through the upper half of the output stage.
Off the top of my head, you should be able to achieve what amounts to a single-ended (the bottom half of the output stage will be effectively out of the picture) output stage which should get you about 20 watts into an 8 ohm load.
se
- Status
- Not open for further replies.