Hi! 🙂
***
I'm pretty familiar with what "default" implementations of LM3886, LM3875, and LM4780 (run with both of its internal "3886" channels in parallel) sound like, meaning that I know what they sound like when they are built according to something similar to the original Datasheet example schematics, and also properly powered and cooled as the Datasheet prescribes.
In their "default" form, Chip Amps tend to have, to me, a sound that is precise but still "undeniably analog", almost like a really well-designed and engineered single-ended tube amp that isn't "trying to sound warm". In contrast, "textbook" Class D amplifier implementations sound to me equally as precise as a Chip Amp, but a bit "drier" (for lack of a better term) especially in the vocal range and just above that. Chip Amps are probably all-around my favorite topology sound-wise: the best of many worlds IMO.
However, I am now considering building a new dual-channel "reference amplifier" with LM3886 as the basic building block, and I came across the extremely impressive-seeming, and well-documented, Modulus 86 from Neurochrome (and I also see that "Prof.Neurochrome" has a helpful presence in various Audio forums). Apparently the Modulus-86 implementation reduces distortion and noise over a "default" LM3886 implementation by a significant amount, edging closer and closer to any errors being non-detectable.
So, all this leads to my Question: are the "tiny errors" that Modulus-86 removes, in order to achieve its impressive distortion and noise figures, actually the subtle part of the sound that gives "Chip Amps" what I consider to be their "personality"? Put another way: would building a pair of Modulus-86 essentially leave me with a low-wattage amplifier (which, as an aside, is fine for me, thanks to Horn Speakers) that is completely indistinguishable from a Class-D amplifier except that it gets a bit warmer to touch when it plays?
***
--my sincere thanks for any thoughts! rs
***
I'm pretty familiar with what "default" implementations of LM3886, LM3875, and LM4780 (run with both of its internal "3886" channels in parallel) sound like, meaning that I know what they sound like when they are built according to something similar to the original Datasheet example schematics, and also properly powered and cooled as the Datasheet prescribes.
In their "default" form, Chip Amps tend to have, to me, a sound that is precise but still "undeniably analog", almost like a really well-designed and engineered single-ended tube amp that isn't "trying to sound warm". In contrast, "textbook" Class D amplifier implementations sound to me equally as precise as a Chip Amp, but a bit "drier" (for lack of a better term) especially in the vocal range and just above that. Chip Amps are probably all-around my favorite topology sound-wise: the best of many worlds IMO.
However, I am now considering building a new dual-channel "reference amplifier" with LM3886 as the basic building block, and I came across the extremely impressive-seeming, and well-documented, Modulus 86 from Neurochrome (and I also see that "Prof.Neurochrome" has a helpful presence in various Audio forums). Apparently the Modulus-86 implementation reduces distortion and noise over a "default" LM3886 implementation by a significant amount, edging closer and closer to any errors being non-detectable.
So, all this leads to my Question: are the "tiny errors" that Modulus-86 removes, in order to achieve its impressive distortion and noise figures, actually the subtle part of the sound that gives "Chip Amps" what I consider to be their "personality"? Put another way: would building a pair of Modulus-86 essentially leave me with a low-wattage amplifier (which, as an aside, is fine for me, thanks to Horn Speakers) that is completely indistinguishable from a Class-D amplifier except that it gets a bit warmer to touch when it plays?
***
--my sincere thanks for any thoughts! rs
Last edited:
I just got a Modulus-86 10th Anniversary Edition amp built by the man himself, so I can share some thoughts. I would say it does have the analog character you are hoping for. While precise, it is not dry or sterile like a class D amp.
The error correction was described to me as “negative feedback on steroids” and that is consistent with my listening impressions. It has excellent command of the bass. The background is pitch black, yielding nice dynamics. The midrange is smooth and the top-end is on the mellow side.
I had it paired with some already-mellow-sounding open baffles, and it was almost too mellow in the treble (partly my fault for building super dark/mellow speakers). I think it would pair really well with brighter and more lively speakers, like your horns.
I don’t know how it compares to a “default” LM3386 but I would say it is more like a high-end Yamaha (smooth and euphonic) then, say, a First Watt (lively and raw).
I’m really happy with it and Tom is really friendly and helpful so I bet he’ll chime in.
The error correction was described to me as “negative feedback on steroids” and that is consistent with my listening impressions. It has excellent command of the bass. The background is pitch black, yielding nice dynamics. The midrange is smooth and the top-end is on the mellow side.
I had it paired with some already-mellow-sounding open baffles, and it was almost too mellow in the treble (partly my fault for building super dark/mellow speakers). I think it would pair really well with brighter and more lively speakers, like your horns.
I don’t know how it compares to a “default” LM3386 but I would say it is more like a high-end Yamaha (smooth and euphonic) then, say, a First Watt (lively and raw).
I’m really happy with it and Tom is really friendly and helpful so I bet he’ll chime in.
Haha, now that you mention it: "High End Yamaha Receiver in Piano Black Finish" is exactly how I'd describe the character of a "default" LM3886 implementation. Thanks so much for the assurances here. I thought I was headed in the right direction but wanted to make sure. --thanks again! rs
I'd say that's a good description. The distortion numbers for the LM3886 itself are already decent in the mid bass region (few hundred Hz). But the distortion rises towards the ends of the audio band, including a significant rise towards the high end. I believe that's what's driving the "chipamp sound" that many find a bit objectionable, "good enough", or sorta "mid-fi, not bad, better than Bose".
The Modulus error correction removes this distortion. You'll find that the THD+N vs frequency of the Modulus-86 is dominated by the noise (+N) of the test setup up to about 2 kHz. At higher frequencies the amp does start to contribute some THD, but it's still orders of magnitude below audible.
The resulting precision in the sound is hard to describe. The bass is tight, firm, and well-controlled but more lively or warm than what I get from the best Class D amps (Purifi 1ET400A for example). The mids and highs are crystal clear and fatigue-free. There's no Class D glare.
I've listened to many amps, including some marquee names. I keep returning to the low distortion amps for that sound of effortless amplification. The Modulus amps are just amps. They amplify. They don't add to the sound. In my view that's what an amplifier should be. But then again, I can easily be accused of bias.
Depends on what you consider to be "low-wattage". If the 40 W of the Modulus-86 isn't enough, I do have the Modulus-286 (65 W) and Modulus-686 (240 W). All powers into 8 Ω at <0.01 %. THD, unlike many of the Class D amps that quote power numbers at 10 % THD.
Tom
Thank you for taking the time to reply!
***
A few years back, I was reading some random general musings about modern Op Amps on a forum or something. The thrust of the text was that modern high-precision Op Amps have become “technically perfect” for all practical purposes, and that it was something like a modern-day miracle that this level of near-perfection can now be had for just a few dollars. But, the crucial point was that they’re only so cheap because there was literal tens of millions in R&D put into each one, first.
I carry that same notion of “R&D
investment” over to the work of Neurochrome: you’ve gotten infinitesimally close to a “technically perfect” LM3886 implementation, evidently over years (and even several versions). From that perspective, it would make little sense for me to attempt to guess and overengineer my way into “accidentally doing better”.
I’m building this upcoming pair of monoblocks in a very deliberate and meditative fashion, but chances are that Modulus-86 will be the platform I select. (And I do believe that the “analog character” stays intact despite the error correction).
***
—thanks again! I respect your work! rs
***
A few years back, I was reading some random general musings about modern Op Amps on a forum or something. The thrust of the text was that modern high-precision Op Amps have become “technically perfect” for all practical purposes, and that it was something like a modern-day miracle that this level of near-perfection can now be had for just a few dollars. But, the crucial point was that they’re only so cheap because there was literal tens of millions in R&D put into each one, first.
I carry that same notion of “R&D
investment” over to the work of Neurochrome: you’ve gotten infinitesimally close to a “technically perfect” LM3886 implementation, evidently over years (and even several versions). From that perspective, it would make little sense for me to attempt to guess and overengineer my way into “accidentally doing better”.
I’m building this upcoming pair of monoblocks in a very deliberate and meditative fashion, but chances are that Modulus-86 will be the platform I select. (And I do believe that the “analog character” stays intact despite the error correction).
***
—thanks again! I respect your work! rs
Yeah. Precision circuit design is not a stochastic process. Rather, it involves developing an understanding of physics, engineering, and how electronic devices work so you can combine them into a circuit that performs well. It then takes experience and more engineering to turn that into a product.
Tom
Tom
I run a set of 4 of Tom's older 86p chip amps arranged as two bridged monoblocks, I guess they're 4p by his current nomenclature. They're sweeter than the hypex/ purifi in the mids and treble, but not class A amp sweet, and a touch more rounded but with marginally less slam in the bass. I use them with ns1000m with raal ribbons replacing the yamaha tweeters, so I guess I'm hearing what the mid and treble really offers.
I've back to backed them with the AHB2 benchmark and various ncore/ purifi. I haven't swapped yet....
I've back to backed them with the AHB2 benchmark and various ncore/ purifi. I haven't swapped yet....
Wow. That's a strong endorsement. Thank you!
By "86p" I'm thinking you mean the Parallel-86. The modern equivalent is the Modulus-286. Two of them in a bridge would sorta make a "Modulus-486". I say "sorta" because if I was to make a Modulus-486 I would follow the same architecture as the Modulus-686 where I use a fully differential amp to drive the bridged sets of parallel LM3886. That would be even better than two Modulus-286 bridged.
Tom
Thanks again folks for the thoughts and votes of confidence!
I'm not familiar with the sound of Class A power amplifiers, beyond the old Motorola "Squawker" in my dad's Work Van in the 1980's. But, I am familiar with the "sound" of such things as the all-discrete input and output stages in a UREI 1176 RevA, or the old Bozak DJ mixers. When those low-level Class A stages are driven hard, my experience is that they do a certain subtle "bloom" or "splat" thing that I can imagine would be extremely appealing to some -- especially for the appreciation of certain genres of music (anything "ballad", for starters).
Based on my mental model of the "Class A" sound, I think I slightly prefer the "cool but not frozen" sonic profile of the LM chips, especially for most music made after, say, 2010. It may be, for all practical purposes, as accurate a version of "the truth" as is attainable.
--with all the usual caveats that we're all extremely fortunate people, who are blessed enough to be comparing Ferraris and Porsches here to begin with! --rs
I'm not familiar with the sound of Class A power amplifiers, beyond the old Motorola "Squawker" in my dad's Work Van in the 1980's. But, I am familiar with the "sound" of such things as the all-discrete input and output stages in a UREI 1176 RevA, or the old Bozak DJ mixers. When those low-level Class A stages are driven hard, my experience is that they do a certain subtle "bloom" or "splat" thing that I can imagine would be extremely appealing to some -- especially for the appreciation of certain genres of music (anything "ballad", for starters).
Based on my mental model of the "Class A" sound, I think I slightly prefer the "cool but not frozen" sonic profile of the LM chips, especially for most music made after, say, 2010. It may be, for all practical purposes, as accurate a version of "the truth" as is attainable.
--with all the usual caveats that we're all extremely fortunate people, who are blessed enough to be comparing Ferraris and Porsches here to begin with! --rs
@tomchr One Question for you (if you are at liberty to say): Does the "10th Anniversary Edition" demonstrate the architectural path forward (using the OPA627B chip for correction), or is that just an "interlude" with the standard continuing to be the LM4562 for correction, in any future numbered versions? --thx! rs
The 10th Anniversary Edition is a limited edition of the Modulus-86. It's a further development of Modulus-86 Rev. 3.0 that uses the OPA627 for the controlling opamp.
The OPA627 is no longer made in DIY-friendly package options (DIP and TO-99 metal can). I was able to get my hands on some of them and I include them in the Kit. Once I've sold out there will be no more 10th Anniversary Edition. I currently have eight kits left with the TO-99 metal can and 40+ of the DIP option.
The 10th Anniversary Edition provides better performance than Rev. 3.0 due to the OPA627 and a further optimized PCB layout. In particular, the transient response is cleaner. Here's one builder's subjective experience of the 10AE: https://www.diyaudio.com/community/...er-achieving-0-0004-thd-n.262273/post-7789400 and another: https://ncforo.com/index.php?threads/modulus-86-10th-anniversary-edition.70/post-728
What does the future hold for the Modulus-86? That's a great question. I honestly don't know. To squeeze more performance out of it I really have to use surface mounted parts. These parts allow me to reduce the loop areas in some parts of the circuit. I also get the sense - based on measurements - that SMD resistors are now better than PTH resistors, so I would likely use SMD resistors in some critical places. Then the question becomes what to do about the rest of the parts. Do I go all-SMD and basically bring back the Modulus-186? Or do I deliver partially assembled boards for folks to then solder the PTH parts? I've written about mixing and matching PTH and SMD parts here: https://www.diyaudio.com/community/threads/modulus-86-build-thread.267802/page-286#post-6839949. Those were my thoughts three years ago (almost to the date). I can certainly take another look and ask my assembly house what they'd charge for a partially assembled board so I can get an idea of the pricing, but I suspect it'll be prohibitively expensive. And there is no way that I will entertain DIY assembly of all-SMD circuits. My thoughts on that topic haven't changed since I wrote this: https://www.diyaudio.com/community/threads/modulus-86-build-thread.267802/page-229#post-5738981
To summarize: Odds are that the 10th Anniversary Edition and Rev. 3.0 are the end of the line for the Modulus-86 ... or at least a long, long way from the next station. I just don't see any way to improve upon them in a way that allows for DIY assembly. I will keep them going until I run out of OPA627 (for the 10AE) and as long as people keep buying them (Rev. 3).
Tom
The OPA627 is no longer made in DIY-friendly package options (DIP and TO-99 metal can). I was able to get my hands on some of them and I include them in the Kit. Once I've sold out there will be no more 10th Anniversary Edition. I currently have eight kits left with the TO-99 metal can and 40+ of the DIP option.
The 10th Anniversary Edition provides better performance than Rev. 3.0 due to the OPA627 and a further optimized PCB layout. In particular, the transient response is cleaner. Here's one builder's subjective experience of the 10AE: https://www.diyaudio.com/community/...er-achieving-0-0004-thd-n.262273/post-7789400 and another: https://ncforo.com/index.php?threads/modulus-86-10th-anniversary-edition.70/post-728
What does the future hold for the Modulus-86? That's a great question. I honestly don't know. To squeeze more performance out of it I really have to use surface mounted parts. These parts allow me to reduce the loop areas in some parts of the circuit. I also get the sense - based on measurements - that SMD resistors are now better than PTH resistors, so I would likely use SMD resistors in some critical places. Then the question becomes what to do about the rest of the parts. Do I go all-SMD and basically bring back the Modulus-186? Or do I deliver partially assembled boards for folks to then solder the PTH parts? I've written about mixing and matching PTH and SMD parts here: https://www.diyaudio.com/community/threads/modulus-86-build-thread.267802/page-286#post-6839949. Those were my thoughts three years ago (almost to the date). I can certainly take another look and ask my assembly house what they'd charge for a partially assembled board so I can get an idea of the pricing, but I suspect it'll be prohibitively expensive. And there is no way that I will entertain DIY assembly of all-SMD circuits. My thoughts on that topic haven't changed since I wrote this: https://www.diyaudio.com/community/threads/modulus-86-build-thread.267802/page-229#post-5738981
To summarize: Odds are that the 10th Anniversary Edition and Rev. 3.0 are the end of the line for the Modulus-86 ... or at least a long, long way from the next station. I just don't see any way to improve upon them in a way that allows for DIY assembly. I will keep them going until I run out of OPA627 (for the 10AE) and as long as people keep buying them (Rev. 3).
Tom
That is a great description of the sound. I swapped the Modulus-86 in for a fairly well regarded LM3886 kit (that I thought completely lacked the chipamp sound) and found the Modulus-86 to have noticeably better defined bass and a more airy, clean top end. The other one is good enough that when I'm not comparing it to the Modulus I still enjoy it, but when I finish the anniversary build I will replace it with the non-anniversary Modulus.
Compared to 'regular' chip amp implementations that I've tried it is night and day different.
Last edited:
I swapped in Modulus 86 in place of GaN class-D, which is supposed to be smoother and rounder than typical class D. I would say Modulus 86 actually had a bit more of analogue texture and was more revealing of recording quality. I did end up putting in a Pass Korg Nutube in front of it to add that tube flavor, so if pleasant colorations are needed, one could always do something like that..
Tom if you keep the smd parts a sensible size most of us are happy soldering them. Keep the caps TH.
I was going to suggest the same. For me it would be a positive learning experience to do more SMT work, and a good (if not great) skill to develop 🙂
There is no way I will support DIY SMD assembly. I explain why here: https://www.diyaudio.com/community/threads/modulus-86-build-thread.267802/page-229#post-5738981
In case clicking the link is just too much work, here are my main points:
The best way for DIY assembly of SMD builds is to use a solder paste stencil to deposit the paste in the right spots (and in the right amounts). Then place the parts and bake the board in a toaster oven while keeping a keen eye on the temperature to make sure the oven follows the prescribed reflow profile. I use this method for prototypes. I still would not rely on it for anything that comes close to mass production.
Tom
In case clicking the link is just too much work, here are my main points:
The best way for DIY assembly of SMD builds is to use a solder paste stencil to deposit the paste in the right spots (and in the right amounts). Then place the parts and bake the board in a toaster oven while keeping a keen eye on the temperature to make sure the oven follows the prescribed reflow profile. I use this method for prototypes. I still would not rely on it for anything that comes close to mass production.
Tom
I, for one, entirely respect your experience with SMD production. It's enlightening to know this (I wasn't aware until now, since I hadn't read the entire Modulus86 build thread).
One thought: would you consider making the Anniversary boards available for people who get lucky and can source their own OPA627s?
One thought: would you consider making the Anniversary boards available for people who get lucky and can source their own OPA627s?
Last edited:
I might be able to do that. I found a few more OPA627AP. Enough for 20-some kits. So I need to order more boards. If I get any spares a few lucky few might be able to get boards. That's not something I'm planning to stock, though.
Tom
