Help me taming the upper mids of LM3886 chipamp

[Edits]

Hello all
It has been quite some time since I finished building my chipamp (LM3886). When I started it, I thought that if the result is good then I'll keep it and move ahead building nice chassis for it. If the sound is bad then I'll just throw it away.

The reality is not that simple. The result is not that good that I start building as beautiful chassis as I can, yet it is not bad either. In fact I feel it is so close to a beautiful amplifier that I began spending more time try to tweak it. And the only resources is this helpfull forum of course.

First impression: after testing it with a scrap driver, I put my Philips compo bookshelf speaker and I was sooo surprised. I put the speakers on the floor yet the music floats at my eye level sitting 1.5 m in front of them. With some programs this soundstage was so wide I could not help thinking there must be some kind of sound processor inside the chip (I know it was not true of course, or is it?). Vocals sounded big, foward, so close to me. Piano and strings sounded so firm and fast. Bass was nothing special, I thought it was normal with such small speaker. With my B&W (an inefficient 84db sensitivity) everything was even better except the soundstage was not as dramatic.

Just one condition that prevented me to hurry up building beautiful chassis for it.
It was not hum (that I dreaded most). Actually there was no hum, hiss, or thump. It was The upper mids!
It was too strong or dominant, or should I say aggressive? Was it "the angry woman" symptom? Vocals with mostly low to middle mids like Don Williams or Anne Murray sounded so nice. But many of Pavarotti's that I love so much sounded tiring almost unbearable. It happened with all CD so it must not be the recording.

So I began lurking frequently and silently to this forum, finding everything I can understand. Then the tweaking began.

My chipamp was based on standard schematic from National and configured as dual mono. The PSU each is a simple CT 25-0-25 150VA EI transformer, packaged rectifier bridge and 4700 uF caps for each rail.I tried 100, 220, 1000, 2200 uF for Cs(at the amp pcb); 2.2, 4.7 uF BP for Cin. for Ci I tried 47, 100 uF, 100+100 uF back to back (BP config;dont ask me why, just a wild guess). Remove the zobel (I use 10R and 100nF), add RL at the out put. Add another 4700 uF per rail at the PSU (I wanted to try CarlosFM recipe but my transformer is not suitable). Lowering the voltage (using 18-0-18), change Rf/R3 from 22k/1K to 20K/680R (maintain the gain at around
30), lower the gain using 20k/1k for Rf/R3. Add 22k/50pF as Rf2/Cf (used in the single supply version on the datasheet).
After weeks of torturing the pcb, I build another chipamp, same schematics, and the values used are:

Cin: 2.2 uF BP, Ci: 100 uF, Cs: 2200uF||100nF, Zobel: 10R+100nF, no RL at the output, Rf/R3: 20k/680, Rf2/Cf: 22k/50pF, PSU caps: 2x4700 uF, Voltage: 18-0-18.

The result:
Adding Rf2/Cf seems to soften a bit the harshness or hardness of the upper mid, but also weaken the crispiness of the upper (sorry for my unusual description but that really is the way I perceive).

Using 2200 uF Cs give "louder" bass but not the bass that I like most. I like the bass that is felt not heard(the bass on "Hotel California" sounded funny) But this has the effect of "compensating" the wild upper mid, so it is ok for the time being even at the cost of lack of presence.
I could not detect significant sonic changes with other tweaks; bad thing was that I did not make record of the result of certain combination.

BUT the beast is still there!!!! Is it the speaker? Even my soft sounding AR94 sounded hard at upper mid. Is it me, my ear, or my listening habit? Is there anything else that I can try before I resort to building preamp with tone-control? Is this phenomena happen most of the time? From what I read in this forum, most complain is the lack of bass or else too bright. I don't mind a little bright at the top, though.

One thing for sure, this saga will continue. I feel so close, I can not stop now. So please help me find the way. How to tame the upper mids.

[analog_sa]

A very cool post.

A few observatons:

- not all caps sound the same
- not all resistors sound the same
- not all rectifiers sound the same

You don't seem to mention specific brands but only values.

A combination of Panasonic HA, BG-N, Riken and some nice discrete fast recovery diodes is unlikely to have harsh mids. You may also find that mechanical construction/heatsink material is also able to influence the tonal balance.

[fizzard]

Have you considered an equalizer, or adding a filter?

[AndrewT]

hi,
does the harsh mid lessen at lower SPL levels?

I ask in case the 84dB B&W are asking for too much current and causing the amp to clip or protect/limit.

[danielwritesbac]

Oh, I feel your pain.

First off, let's get rid of the whole Ci and dc offset mess. . .
See my circuit on this post: http://www.diyaudio.com/forums/showt...33#post1498833

The 3.3R groundlift won't fit, but the rest will.
In the diagram of that post, you can vary the 470R to sink the mids.

You can also add a very small value ceramic cap or polyester cap as a "bypass cap" to your input filter cap, CIN. See the discussions on bypass cap practice and extra "air" effects. That can make the upper mids less apparent by boosting frequencies that are higher than the upper mids. About 4.7nF (code 472) is a good spot to start, and then try progressively larger values until the "highs" appear.

For CIN (input filter cap) there's also a botique capacitor option, Enla Cerafine, and its not expensive at about $1.
Capacitors vary by voltage rating as well as vary by capacitance.

Next up. . .

Choose between a pair of 220uF, or 330uF, or 470uF for the power caps at the amplifier board.
This will "make a difference" in the audio; however, since capacitors vary greatly, then it is up to you to figure it out.

Summary of voicing:
The 470R (you may vary this value) between the input filter cap and the potentiometer.
The 330uF (you may vary this value) for the power caps at the amplifier board.
The "bypass cap" practice to raise the highs for the input filter cap.

How easy is that?


P.S.
James Baxandall made a nifty tone circuit that is in the majority of all audio equipment. Its there for a reason. That reason has a name--Fletcher Munson medical research, that explains why your ear cannot be expected to agree with your scope, and why hifi <> burglar alarm. EDIT: So, while the amplifier may have been working perfectly, there was nothing in the system to accomodate your ear.

[Edits]

Thanks for the responses, I feel like having new energy

Analog_sa:
Yes, I read a lot about different sonic qualities from different brand of components. Right now I should say everything is standard. I use all metal film resistors (tried standard/generic, vishay (brown), TKD). Only carbon 2w restors for Rz and/or RL. Cin and Cs are Elna, Nichicon(?), and some Rubicon. The 100n block capacitor has no brand on it so I dont know where they come from. I tried 3 type: one is yellow some 10x4 mm thickness, the other also yellow but quite small some 6x3 mm, another one is green and not a perfect block in shape. I did not yet try some exotic component because they are so expensive. I heard that a Riken (gold) can cost a decent lunch a piece? But I will try that if that is surely the reason.

Fizzard:
Like I said, I am ready even if I have to build a pre-amp, or may be some buffer? But I am lost. Do you or anyone can share experience on it? I do that as my final resort when everything else failed. I just can not stop.

AndrewT:
the sonic color (harshness) is the same even at moderately low level, but of course more bearable. Is it possible that clipping happens at certain frequency only? Or severe distortion occurs at that part of the spectrum? I dont think my B&W is the problem because this happened also on my "easy" compo speaker as well as my AR (86 (or 89?)dB)

Daniel:
I must try your recommendation. It sounded full of hope
I do remember about "air" effect you mentioned in some of your thread, now I'll have the chance to see what it really is.

How easy is that?

Lol, I do hope so. Remember that you have gone through a lot more trial and tribulation (and survived), while I am still just begin to put my feet in the water. So I do need this kind of encouragement.

I'll do something this weekend, and report back here as I am sure there so many other just like me.

[Edits]

Daniel, I've look into the link you mention. Before I do anything silly, please check if I get it correctly.

Based on typical application from ref. sheet, I will jump the pot from slider(?) to ground with 1M, then change the Rb from 1k to 470R followed with 4.7uF in series. Shunt the end of this cap to ground with 15K then to the input pin 10 (LM3886). As for feedback, I will use 27k/820 as Rf/R3 and no Ci. Zobel will stay 10R and 0.1uF. Cs will be varied starting from 220uF. Is it correct?

I look forward into it

by the way, how do I can identify wich cap is polyester, polypropelen, polystearine etc ? I know well ceramic disk though

[eketehe]

Hi Edits,
your ears must be tired... that was happen to me.

pls try your drivers with a different amp for simple,
if their OK,
you can try daniel's, tht was works with mine ( Cin 3.3uF by passed by 100nF, all WIMA MKP10/160V ), if your existed Rf is metalfilm, other option is 22K carbon.
thats what i did for my amp.

you can try an equalizer, but if your problem highlighted 'vocals', equalizer will effect to other instrument at the same freq, such saxophone or strings which maybe we like em bright.
once in other test, for my 3way, i change the plastic tube midrange with the ugly and old and must be cheap toho 6' paper midrange, then solved. also reducing the annoying sharp ssssssssssssss yes not yesssss, without losing the detail. the soft dome tweeter must be its perfect pair.
thats what i did for my drivers,

I recognize your schematic, maybe same as mine ( we're both indonesian ).
I found that our PCB is a bit different, i don't know what are designer's idea.
Heres i attach 2 schematic of our local designer, which i have em all beside the original schematic, but still need time and experiment to make a personal review. except the no. 2, regarding my personal taste that was outperformed by my little 1875 hahaha.

Yours must be pic no. 1 which there is no decoupling caps ( stated 2 x 1000uF on longside of PCB )

[soongsc]

Basically the original design in the data sheet is pretty good. But it needs to be understood that these do not sound good when you are driving them to the power limit. Another thing to be aware of is to make sure your speaker impedance curve is pretty flat. Normally if you are having problems in the mid could be because the impedance in that range peaks up, that is the place to fix.

[Edits]

Hi Eketehe, I usually drive my speakers with Yamaha, one is an old HT with DSP, the other is a receiver. They sounds nice and sweet albeit too soft may be. I'm sure it is not my speaker. Glad to hear you get success with Dan's recipe; sure I'll try that.
I can not recognize the diagram that you attach, but I think I got the massage.
How do you evaluate the 1875 vs 3886?

By the way, is the electronic parts market is still there in Cikapundung?