lm3886 implementation advice needed

[Fast Eddie D]

I've never owned a high-end amplifier, just some consumer A/v receivers, and I've played around with some Class D amps. Last year, I built an amp based on the XY LM1875 board, and I was really impressed with the sound quality - it sounds better than everything else I own. In layman's terms, how are chip amps deficient to the best audiophile amps? Or is it as simple as saying that a well implemented chip amp can be on par with the best? In other words, how much better can it get?

You probably did a good job on the bypassing, parts selection, power supply, etc. I built prototypes of the LM1975 and 3886, fairly crude on perfboards no less. I put 220 uF bypass capacitors to each chip (rail to rail, no ground) and soldered physically large MKP capacitors directly to the power supply pins of each chip. The rest was basic design with quality parts. Power supply was external (wires as short as possible) and overkill. I was totally surprised by both chips and knew they had huge potential for improvement. I have focused my efforts on line level circuitry and never did anything with it.

 

[tomchr]

You probably did a good job on the bypassing, parts selection, power supply, etc. I built prototypes of the LM1975 and 3886, fairly crude on perfboards no less. I put 220 uF bypass capacitors to each chip (rail to rail, no ground) and soldered physically large MKP capacitors directly to the power supply pins of each chip.

220 uF is less than half of what the LM3886 data sheet recommends. According to the application section in the LM3886 data sheet, you need 470 uF || 10 uF tantalum or electrolytic || 100 nF ceramic as a minimum. I've seen the LM3886 oscillate due to poor decoupling.

I've build point-to-point circuits with the LM3886 as well. While it likely is possible to get the data sheet performance of the LM3886 that way, odds are you won't. For more detail (and measurements) see here: LM3886 PCB vs Point-to-Point (with data)

Tom

 

[tomchr]

So the lm3886 needs error correction for best results.

Many circuits can be improved by adding error correction of some sort.

Are there any chips that have high performance and don't need it.

Depends on what you consider "high performance". If you want -120 dB THD, you'll need error correction. If you're content with -90 or -100 dB THD midband and -60 dB THD in the higher end, the LM3886 will do just fine.

You can always follow one of the suggestions from earlier in this thread: Get an eBay board and stuff it with genuine parts. If that meets your needs, great. If not, consider replacing the eBay circuit with one that has error correction.

The largest cost of your project will be the power supply, heat sinks, and chassis. As long as you stay with the LM3886, you can reuse all of that. For example, you can build the XY eBay board. If you don't like it, take it out and put my Modulus-86 in instead. Everything else is reusable (assuming you leave enough room in the chassis to fit the Modulus-86 as it's bigger than the XY board.

Tom

 

[Fast Eddie D]

220 uF is less than half of what the LM3886 data sheet recommends. According to the application section in the LM3886 data sheet, you need 470 uF || 10 uF tantalum or electrolytic || 100 nF ceramic as a minimum. I've seen the LM3886 oscillate due to poor decoupling.

I've build point-to-point circuits with the LM3886 as well. While it likely is possible to get the data sheet performance of the LM3886 that way, odds are you won't. For more detail (and measurements) see here: LM3886 PCB vs Point-to-Point (with data)

Tom

It was a prototype. If I built a permanent circuit I'd beef it up.

Placement of the caps on the board affects stability. Closer = better, just like op amps. Many commercial applications fall far short of this.

I've looked at your work, Tom. If I was going to do it for real I'd pretty much emulate what you do. I haven't seen better optimization of the 3886 than what you offer.

 

[tomchr]

It was a prototype. If I built a permanent circuit I'd beef it up.

Nothing wrong with that. 🙂

Placement of the caps on the board affects stability. Closer = better, just like op amps. Many commercial applications fall far short of this.

Yep. Supply inductance matters. But so does that 10 uF (min) electrolytic cap in the bypass circuit. Remove it or lower it and you'll find the LM3886 oscillates once the output swings above about 30 V.

The LM3886 has a lot or warts, but if you treat it right (i.e. follow the directions in the data sheet and design a reasonable PCB layout) it can really sing. A lot of the stability challenges can also be avoided by using a large heat sink and keeping the supply voltage at ±24 V or below.

I've looked at your work, Tom. If I was going to do it for real I'd pretty much emulate what you do. I haven't seen better optimization of the 3886 than what you offer.

Thank you. Yeah... I'd say I know the LM3886 pretty well now. 🙂

Tom

 

[Fast Eddie D]

Yep. Supply inductance matters. But so does that 10 uF (min) electrolytic cap in the bypass circuit. Remove it or lower it and you'll find the LM3886 oscillates once the output swings above about 30 V.

Tom

Now I remember your tutorial where you showed the impedance vs frequency curve of the three caps in parallel. It's been as long time since I looked.

I used a 0.22 uF, 600 volt MKP capacitor soldered under the board right to the power supply pins. The cap is physically almost as large as the chip (larger than an 1875). I've used similar schemes when modifying consumer equipment. In one case (My old dumpster Pioneer receiver) this audibly cleaned the sound up. (It's a discrete circuit but same concept.)

It amazes me that consumer equipment drops the ball on details like this. It's so easy and cheap, and makes such a large difference. It's like $5-10 worth of parts makes the difference between mediocre and awesome. So many consumer circuits I've seen use the datasheet circuit verbatim, even though changing just a few parts can make a huge difference. To me the datasheet is just a start.

 

[tomchr]

It amazes me that consumer equipment drops the ball on details like this. It's so easy and cheap, and makes such a large difference. It's like $5-10 worth of parts makes the difference between mediocre and awesome.

$5-10 worth of parts (plus their assembly cost) will add ~$100 to the final price of the product, assuming brick-and-mortar retail margins. That's a steep increase for a $200 receiver.

To me the datasheet is just a start.

It's usually a good start. It's remarkable how few read the data sheet, though.

Tom

 

[Fast Eddie D]

$5-10 worth of parts (plus their assembly cost) will add ~$100 to the final price of the product, assuming brick-and-mortar retail margins. That's a steep increase for a $200 receiver.

Yeah, but even using better parts in strategic places can make a difference. For example using higher precision and higher stability resistors in the feedback loop of a power amp can make performance more consistent. Lowering low frequency poles can clean up muddiness. It adds up to a couple dollars per unit.

My Pioneer SX2600 was a good exercise in turd polishing. Besides adding bypass caps to the board, I replaced the antique power supply caps (the big cans) with arrays of 2200 uF, 100 volt high ripple current caps. I made boards and mounted them vertically where the old caps were. BAM! Did the bass ever hit after that! I didn't change the transformer or diodes. BIG difference. I also bypassed the line level circuitry (it was terrible) and used my own. I bypassed the power amp and put RCA jacks on the back. My preamp went from tape monitor to the power amp. This was in my main system for 2 years while I built better stuff. I still use the tuner section. This unit came out of the dumpster 5-6 years ago.

 

[tomchr]

Yeah, but even using better parts in strategic places can make a difference. For example using higher precision and higher stability resistors in the feedback loop of a power amp can make performance more consistent. Lowering low frequency poles can clean up muddiness. It adds up to a couple dollars per unit.

Yeah, but that means lower profits and cranky shareholders. Sucks, but that's business.

I have come across an amp designed following your philosophy, though. The Parasound A23 uses a low temperature coefficient resistor in the feedback path (where it matters). I think that's the only 'special' component. That said, the amplifier is designed very well. It's also considerably more expensive than your average A/V receiver. So good designs do exist. You just have to look at the >$1k range. 🙂

Tom

 

[Fast Eddie D]

Yeah, but that means lower profits and cranky shareholders. Sucks, but that's business.

I have come across an amp designed following your philosophy, though. The Parasound A23 uses a low temperature coefficient resistor in the feedback path (where it matters). I think that's the only 'special' component. That said, the amplifier is designed very well. It's also considerably more expensive than your average A/V receiver. So good designs do exist. You just have to look at the >$1k range. 🙂

Tom

I use pretty much all "special" components. I use Dale "CMF Industrial Series" wirewound resistors pretty much everywhere in the audio chain. These are precision, high stability, low temperature coefficient, high resonance resistors. I use the Nichicon "Muse" bipolar caps for coupling. I'm careful not to stack my low frequency poles.

It works. Maybe the resistors are overkill but I buy 10 at a time so not so bad. My preamp is nice and clear and beats everything that's been in my shop. Now I have to build a power amp using the same principles.

It's a moot point though if the decoupling is lousy and the op amps are the cheapest of the cheap.