National semiconductor chipamp sound second attempt

[ionomolo]

Some time ago i started a thread called "gainclone sound differences" because i noticed serious differences in the sound of my LM3886 based chipamps and i thought it was very interesting to ask if other people have had the same experiences and how they did come arround and get good results.

My current experience:

LM4562: It's better not to place it in a circuit, i really hate its sound at low gains (unity or two) and this has been confirmed by many people here. At a gain of ten or so it's probably one of the best choices if you don't want to get things complicated and go with a discrete amplifier, but don't place it if you don't need it. NEVER use it as input buffer. It simply is not a buffer. There are designs out there that have unity gain as they are designed to have unity gain and not because of 140 dB's of NFB.

LM3875: I have not tried it.

LM3886: Sounds it's best at gains arround 28/30 V/V. Harsh at lower gains and bright at 47 and higher. It depends on the quality of parts but it does at least 10 times less than it's often said to do, instead, it depends on layout 10 times more than it's often said. It gives a compensation for a good job and can sound pretty well for most situations. It has been said that it depends a lot on the loudspeakers but since i don't have many to try i can't tell anything about that. It sounds definitely better with high power supply capacitances and this has been tried with the same board and leaving everything unchanged, connecting and disconnecting the power supply cap many times while playing and asking to other people who aren't in audio.

LM4702C : Not tried.

LM4702B + BJT : Not tried.

LM4702B + Lateral MOSFET (2sk1530/2sj201): Nice, more detailed and polite than the LM3886 and can go much higher volumes, but i find it to be rather euphonic and a bit woody, slow and with poor separation between instruments. It's an easy to listen amplifier, in fact a very good one but it doesn't make you feel like if you were in front of something really different. The mosfet version sounds mosfet. It sounds MUCH finer when pushed than when playing quiet (*)

LME49810 + Vertical MOS: I built it to try the driver before the lateral mosfets arrived and i found it had a strange sweet sound at lower levels, that sound was rather appealing but i didn't explore it more.

LME49810 + Lateral MOS: Faster amplifier ever heard. It's by far the best i've done with national ic's at all volumes except maybe the vertical one that went to junk. It sounds like if there were different circuits doing the highs, the mids and the bass, and like if they were powered by different power supplies connected to different power plants as they don't interfere at all. Sounds different up to the point i suggest to build one to everybody who says that gainclones have a huge dependency on capacitors to show what's a huge difference. It's not a perfect amplifier, may sound harsh and too revealing sometimes, but it's stunning. (*) It might benefit from a tube buffer but i havn't tried.

(*) Both with the same power supply and transistors.

Note: All of the amplifiers compared have been built by myself using the same parts grade from the same dealer, so there is no influence by the satisfaction of "having built it myself" or having used special parts. All amplifiers have been tested with the same loudspeakers at the same room. I have built many of them and i only write about things that were in common in all of them.

Excessive part dependency can be removed by avoiding the part or changing the design.

All national amplifiers do weird things (The most famous being humming and motorboating) if they are not properly bypassed or used in a bad layout. standard practices (short feedback loop, star ground, unbypassed decoupling near the amplifier and keeping a long distance between power and signal traces and between input and output traces) makes the problem dissapear and can be done with standard p2p and not requiring complicated things like simulations or a doctorate in electronics.

I hope not to be harsh with the following, but i would ask to avoid audiophoolery and this thread is very prone to it.

Statements like:

Invented by me right now
"The ic LM666 has 60% crossover, a PSRR of 2 dB's, a SNR of 5 dB's and a slew rate of one volt per hour, but if you bypass the power supply choke filter with some audio note silver caps to minimize it's parasitic inductance you will notice its magic inner detail and the air arround its deeper soundstage."

should better be avoided.

BTW i may leave this thread for a while as i have told all my experience so far. I would ask to keep some "scientific" point of view (Only talk about things that are repeatable, not about comparing a krell connected to a mobile phone speaker with a 6W chipamp tied to a sonus faber homage).

[AndrewT]

nice to see a fairly comprehensive comparison from one listener/builder.
Thanks.


BTW,
the 201/1530 are verticals, but of a low Vgs type.
Please confirm the 810+Lat Fet types.

[KSTR]

I have some differing experience with the LM3886 (and LM1875 for that matter) and low gains, in a inverting amp with gain x3 (but noise gain compensated to 30x for HF, to maintain nominal stability). Without knowing your exact schematic (standard non-inv gainclone?) it is hard to tell, and most probably a matter of personal taste.

Complete agreement on the layout vs parts quality issue, though.

- Klaus

[ionomolo]

Originally posted by AndrewT
the 201/1530 are verticals, but of a low Vgs type.
Please confirm the 810+Lat Fet types.

I was completely sure they were laterals, but now i've looking arround and you are absolutely right . I won't buy any 2sk1058/2sj162 because the last time i got fakes and i'm not willing to take the risk again. The faults of 2sk1530/2sj201 are high capacitance and positive tempco, but the lme49810 has enough juice to overcome the capacitance problem (is that what's making my LM4702 slow and dull? [ I forgot to say it was a plain AN-1652 without driver transistors ]).

I forgot one very important point in the comparison between LME49810 and LM4702, i get hooked to the lm4702 more than to the LME. Maybe the detail gets a bit painful after a while, but i don't like the idea of not hearing all that's on the record.

I would really love to hear other people's experiences since mine may be dependant on other parameters like source or loudspeakers or my ears.

EDIT: I also tried it with darlingtons (TIP142/147), the first prototype did not work, the second lasted for some 10 seconds and the last played for 2 minutes. I don't remember how they sounded, not bad for sure, but my opinion about bjt's at the moment is limited to "these transistors smell really bad". I have used 8 2sK1530/2SJ201 and 6 IRFP240/IRFP9240 and i havn't burned a single one.

[ttan98]

"LME49810 + Lateral MOS: Faster amplifier ever heard. It's by far the best i've done with national ic's at all volumes except maybe the vertical one that went to junk. It sounds like if there were different circuits doing the highs, the mids and the bass, and like if they were powered by different power supplies connected to different power plants as they don't interfere at all. Sounds different up to the point i suggest to build one to everybody who says that gainclones have a huge dependency on capacitors to show what's a huge difference. It's not a perfect amplifier, may sound harsh and too revealing sometimes, but it's stunning. (*) It might benefit from a tube buffer but i havn't tried."

I am currently building an amp based on LME49810,

The real difference between this chip and using chipamp(LM3886*) is mainly(not all) the o/p transistors or mosfets are discrete o/ps, the amp(LME48910) is much better in delivering peak current more readily hence the perceived sound improves, better dynamics and stronger bass(Chipamp main complain is bass deliver performance) and better clarity(you reported this).

Another diff. amp build from LME49810, sound perception can be changed by altering the the slew rate capacitance, if you leave it at 10pf the slew rate amounts to about 50V/uS, the sound output perceived as more forward and faster, I would increase capacitance to about 22pf equals to about 20V/uS.

The toplogy of the LME49810 with o/p transistors is very similar to the Destroyer's HRII using discrete transistors, I will summise the 2 amps can sound fairly similar with some tweaking on my part.

* the die from chipamp is too small to able to deliver peak current effectively.

[ionomolo]

Originally posted by KSTR
I have some differing experience with the LM3886 (and LM1875 for that matter) and low gains, in a inverting amp with gain x3 (but noise gain compensated to 30x for HF, to maintain nominal stability). Without knowing your exact schematic (standard non-inv gainclone?) it is hard to tell, and most probably a matter of personal taste.

Yes, I was talking about standard noninverting amps in all designs. I will try the one with low audio gain and high noise gain if you tell it sounds good.

I belive the one with a gain of 30 sounded the best because it had less defects that i belive they wouldn't be liked by anyone, but this does not mean that designs with a more complicated feedback loop don't sound well at lower gains.

My comments weren't intended to be complete, they are only about what i've tried! I started a thread and not a website so people can report other experiences and maybe contradictory ones (Some parameters may depend on the actual system/room and not only on the ic).

Originally posted by ttan98
Another diff. amp build from LME49810, sound perception can be changed by altering the the slew rate capacitance, if you leave it at 10pf the slew rate amounts to about 50V/uS, the sound output perceived as more forward and faster, I would increase capacitance to about 22pf equals to about 20V/uS.

This is one thing i havn't thought, lower slew rate may sound better because it would be less forward, have you actually built it and compared?

[ttan98]

Originally posted by ionomolo

This is one thing i havn't thought, lower slew rate may sound better because it would be less forward, have you actually built it and compared?

Someone has try it on this amp(he changed 10pf to 20pf), I have tried on another amp. By changing the 1 cap which is easy to do you will DEFINITELY notice the DIFFERENCE. Do try.

[ionomolo]

Originally posted by ttan98
Someone has try it on this amp(he changed 10pf to 20pf), I have tried on another amp. By changing the 1 cap which is easy to do you will DEFINITELY notice the DIFFERENCE. Do try.

But this difference could be an improvement? I belive that lower slew rate would only add distortion, albeit being too forward is one of the worst problems of solid state.

[ttan98]

high slew rate is not the only criteria to determine good sound reproduction. Mike Bitnner"s amp has slew rate about 20V/uS. His amp sounds quite good.

It is easy to change back if you find the sound not to your liking.By changing the capacitance the sound changes choose the sound you like best NOTHING ESLE matters not even the lowest distortion you can get.

The choice is yours.

[ionomolo]

I will give a try to both options, add a driver to increase the slew rate of the LM4702 and increase capacitance of the LME49810 to decrease it and see if any gets better, but both sound very good already!