Remove LM3886 harsh sound

[soundNERD]

I've been working quite hard on my quad LM3886 bridged amplifier. I've got one board done, which are two chips running in standard non-inverted mode. They will be, once that board is done, controlled by a DRV134 for the bridging. With the two chips done I fired it up in standard non-brigded mode, and it sounded great - I don't think I've ever seen any manufacturer call something as powerful as this 68W, because it sounds more like 300 Sony-Watts. Since I've got the case and boards designed for bridging, I'm still going to do it that way, and I'm sure I'll be blown away-literally-by the power.

Anyway, it sounds great, but I find it quite harsh with higher frequencies. Bass and mid-range is great, and so is treble, but if the song contains any very high-pitched voices, I notice it's very harsh and not natural sounding.

The only thing I can think of is either my PSU or the choice of parts on the amp itself. I used all metal film resistors in the amplifier, plus one 22uf non-polarized tantalum Kemet brand capacitor from feedback to ground. All of the resistors are NTE brand 1% tolerance.

In the PSU I have a single 2200uf capacitor per rail, as I've been told countless times that the lower the capacitance, the better the treble/mid, and the higher the capacitance, the better the bass. The goal of this amp is the best overall quality I could do, so I made a comprimise between them by using 2200s. Maybe I need more? Maybe I need less? And also how would adding some 0.1uf caps on the PSU board and/or the amplifier boards help as far as my harsh sound issue goes?

The setup I'm running now is:
Phillips portable CD player (Also tried my Creative Zen Xtra with some 320k full stereo MP3's, same issue)
B&W DM580 floorstanding speakers
~34-0-34 VDC at 400-500VA (not 100% sure)
10K input impedence (if it matters)

Can anybody provide any suggestions as to what I can do to calm down the amplifier? It would sound amazing if not for that slight issue.

BTW I'm using the schematic from the first page of National Semiconductor's PDF for the 3886.

Thanks,
Mike

[aletheian]

Do you have a filter network on the output? If not, give it a try, and then tweak it for your speakers. I've had good results without one, but it is just an option to try.

I don't like tantalums though... I doubt it makes a huge difference, but you might want to ditch that cap for a film.

[sam9]

I don't think I've ever seen any manufacturer call something as powerful as this 68W, because it sounds more like 300 Sony-Watts

Before proceeding to the real question, I feel compellede that based on my Sony experience, that ain't saying much.

Now to be more helpful. Do you have an oscilloscope? Can you borrow. There is a fair chance that in bridged mode your amplifier may be clipping. You could verify this pretty quickly with a 'scope. The common observation seems to be that LM3886's sound pretty nasty when clipping.

LM3886 sounds un-reasonably (for cost and size) good when used with 8 ohm speakers where all you need is 50W. In fact, if your needs are limited to a good 50W ampl for 8-ohm speakers, I think it is hard to justify anything more complex. (I notice that Siegfried Linkwitz is now suggesting 3 LM3886's for his dipoles -- that should say something right there!) When pushed beyond this, I think you may be asking a bit more from the poor thing than is reasonable.

[MBK]

Hi,

I second Aletheian's suggestion. Don't omit the filters suggested ("optional" on the data sheet. In some configurations, the amp may go into bursts of oscillation without them.

One question of course is the impedance plot of your speakers and their sensitivity. If this is a difficult and inefficient load, you possibly do have a clipping issue right there.

[soundNERD]

Currently I only have the board that will become the left channel done, and since I don't have the DRV134 board done as well, I've got the amps running in standard non-bridged mode.

The speakers are 8-ohm and I belive are 87-89dB sensativity. Nothing that should be too hard for the chips to handle, especially since I'm not even close to pushing it to it's limits.

These aren't the speakers I will be driving with it in the end. I will be driving two in-wall 8-ohm speakers. I figured I'd bridge it in case those in-walls are abandoned and larger floor-standers are put in. Also the 3886 is, in a way, designed for 4-ohm, I figured I'd run it bridged to give each amp the 4-ohm impedence it wanted.

I do have a scope, and at one time had it hooked up when I was driving some old, cheap 4-ohm speakers. Sure, it would clip when the volume was cranked, but I couldn't detect anything before that. Then again the tweeters were trash compared to the B&W speakers. I couldn't detect anything unusual on it, though, when it was hooked up to those speakers. I can plug it in with the new speakers connected if needed.

I've always found the B&W's to be very bright, so I was expecting bright sound, so that may be part of it, but it just seems excessively harsh.

It could also be possible that the amp needs a few hours to break in? I've heard of drastic changes to sound quality once the chips break in.

Thanks again for the great, prompt help,
Mike

[I_Forgot]

[i]
In the PSU I have a single 2200uf capacitor per rail, as I've been told countless times that the lower the capacitance, the better the treble/mid, and the higher the capacitance, the better the bass. The goal of this amp is the best overall quality I could do, so I made a comprimise between them by using 2200s. Maybe I need more?

Can anybody provide any suggestions as to what I can do to calm down the amplifier? It would sound amazing if not for that slight issue.
Mike [/B]

Ignore the nonsense you've heard and try putting a decent amount of capacitance in the power supply. That Gain-clown BS about using 500uF or 1000 uF in the power supply is just silly. There's no way the chip can provide decent low frequency output when the power supply rails are sagging and the ripple is high. Bypass the big caps with smaller caps, as is standard practice when using big caps, and your troubles will disappear.

I_F

[soundNERD]

Thanks.

What would you suggest as being "decent"?

[sam9]

What would you suggest as being "decent"?

One common rule of thumb is 10kuF per 100W per rail. I.e., a 100W amp would have a pair of 10kuF caps. Double that for two channels. I doubt more than that would serve any purpose and personally, I've used less (6.8kuF rather than 10kuF) without hearing any difference.

Re my earlier comment, since you have eliminated clipping as a culprit, I suspect the other comments offered may com closer. The anti-zobel, anti-output inductor school may have a point but I think there iare so many pemutaions of speaker and cable reactiance that are possible, it is best to assume they are required until proven to your satifaction that they are not required in circumstance you have.

[aletheian]

Anything around or over 10KuF should be fine. Switching from 1000uF to 6800uF is a big difference in sound, and from 6800 to 10000, it is less so. I have a stash of 15,000uF caps that I use in these things, but I would be happy with 10000uF if I didn't have those.

[lgreen]

Try some different speakers and see if it still sounds that way; I don't think its break-in or the capacitance, mine had only 2,000 uF per channel when I made the 1st version and it was fine. Might want to bypass the electrolytics with some small value films. Might want to look at the scope, look at a square wave and see if there is excessive ringing. Try a zobel (will take about 10 minutes) on the output.

If your ultimate implemention is bridged, you might get less noise, because if the problem you are having is common mode, it will be eliminated by the balanced design. So hope for the best.

Hey, I remember when you first posted on this, congratulations on getting this far!