LM3886 design

Hi Daniel

I'm going to check your design and I'll come back to you as soon as possible.

Best regards
José Luis Figueroa
Audio Applications Engineer | A-Team

Hello sir and thank you for your reply.

About the calculations I have to tell you that I am not an engineer. I found calculations for second order filters on a website (don't remember name). The cut frequency, according to this website should be: fc = 1 / (2pi SQRT((R2+R5) * R4 * C3 * C4)). Notations are in accordance with my schematic. Please correct me if I am wrong. I mention that I have already built some op amp circuit using this calculation and , yes, the -3dB reaches about 55KHz. I don't know how to calculate that (magic...).

About the band of the two amplifiers: it would be "politically correct" to have a feedback network that restricts the band of interest but flexibility might be helpful. You never know.

The amps that I try to replace now have the same wideband capability and they are ok. The trick is the active cross over (which has already been tested and works excellently) and I would like to restrict bands only there. After all I might make some other tests with these amplifiers so I might need their wideband.

Thank you again for the reply.

Thank you, sir.
I wait for your reply.

Thank you, sir.

Just another question if you don't mind: If, for some reason, one of the rails goes down, is there any danger of burning the chip? Should I make some design to prevent this?

Waiting for your reply.

Hi Daniel,

I understand your point completely. It's not a "bad thing" to have a wideband amplifier available. However, if you do decide to use these amps as such, as couple of points (yes, this gets into "audio opinions", I'll try and steer clear of anything that could rile anyone)...

1.) If the amps are for audio, for humans that is, there is nothing above 20(22...?) that is going to be informational. If this is the case, why amplify anything above what is capable of being heard?

2.) If you do decide you really want to go out to 100KHz., then at least use a good scope and function generator and run the amp out to it's "max" looking for oscillations and such. You'd be surprised at how many amps oscillate at higher freq.'s but never noticed because the speakers can't handle the freq. (you won't hear it as you would with oscillations in the audio spectrum). This then shows up as "fuzziness" in the sound, as well as higher, sometimes much higher power dissipation. These are the cases where someone says "My circuit draws too much current at idle", etc. There are software packages, some of them free, that will give you different audio tests, (chirp, swept, etc.) as well. In the case of using one of these software/computer hardware packages make sure you do a baseline first. Computer audio is seldom good enough for real tests of hardware.

3.) Regarding the question of dropping a rail; the device has under-voltage protection as well as thermal protection. If a rail isn't there the under-voltage should kick in, if not most likely the device would overheat and thermal kick in.

4.) Make sure you have good, well filtered power supplies. This often gets overlooked.

5.) Make sure you have a good PCB layout. The best place to ruin an otherwise good circuit is a bad PCB. If you need help with that, PM me.

Hi Daniel

It shouldn't have any problem because this device has Under-Voltage Protection, It doesn't allow internal biasing when |VEE| + |VCC| ≤ 12V, thus eliminating turn-on and turn-off transients.

Best regards
José Luis Figueroa
Audio Applications Engineer | A-Team

Hello, Mike!

1) Amps are for audio, right, and humans cannot hear above 20KHz, that's true. But...amps shoud reproduce the real sounds as acuratelly as possible, right? This is why they need a wider analog bandwidth, so they can reproduce steep sounds (or steps, I think they are called). As far as I know, a square wave signal needs a band 15 times wider than the fundamental frequency in order to remain square. At least this is what I remember from phisics. Well, I will use these amplifiers for movies too, and movies have a large dynamic range and usually steep sounds (I think it's about dV/dt here). So, why not use the huge slew rate of some 19V/uS that the amplifier offers? Of course, I am aware of the oscillation phenomenons that appear along with a wider band. That's why I asked for an opinion here. But, essentially, this is what I want to achieve.
And, as you say, we cannot hear those sounds above 20KHz but we can certainly hear the step in a sound. Imagine a trombone, a door slaming, an electric heavy metal guitar. They all have higher harmonics in their sound and the amplifiers should reproduce these.

2) I will mak tests on more than one oscilloscope after everything is mounted and functional, of course. And I will try signals out of this band to see how the amplifiers behave. Hopefuly the cross over will stop everything out of the band.

3) Thank you for this point. I ommited it, of course.

4) 10mF on every rail and 0.1uF near amplifier's pins. And, of course, wide regions, as hey are called in Altium.

5) I will do my best with the layout as I know how important it is.

Thank ou for your reply.

Thank you, Jose.

I ommited to read that carefully, of course. Thanks for poining it out!

Hi Daniel,

LOL - I was trying to avoid this part of the conversation; it gets into an area where opinions more than facts seem to cloud what people do.

I agree with you in principle, only to say that if the source for audio is a CD/DVD the additional "bandwidth" is already accounted for in the digitizing of the original source; no audio sampler that I am aware of samples at less than 44.1KHz, 16 bits. Let's just say for expediency that you do require additional bandwidth beyond what the human ear can hear; is not twice the source frequency enough (assuming a "signal" of 22KHz. using a sampling freq. of 44.1KHz.)? Consider the playback "source" (CD/DVD/Blueray); do they output signals over 50KHz.? In consumer electronics, I doubt it (in fact, unless it's a "good" piece of equipment, I'd say you're lucky to get more than 12-15KHz.). If we were to truly "sink into audiophile territory", we could start teasing apart all aspects of the argument; phase shift, slew rate (as mentioned), linearity, etc.,etc. I would venture to say that when one uses a wideband amp where frequencies above what the human brain can hear are present, one is setting themselves up for IMD issues more than anything else.

I'm going to avoid a dissertation on "audio reproduction" when someone has done it so already. Please see the two videos presented at: http://xiph.org/video/ . "Monty" and his team know a bit about sound reproduction being the guys behind the OGG VORBIS codec. I like these videos as they present the material in a format that anyone can "grok".

MT

Hi Mike,

Nice videos. Worths the time even if I already grasped those things, still they are explained very simple and efficient.

Today I re-read a handbook from TI where the author is talking exactly about phase problems if you limit the bandwidth of an amplifier. On the other hand, in the datasheet of LM3886, the model design described has a pole at 100KHz and, as far as I understand, a flat response in the respective band. Am I missing something?

However, if something goes wrong with my design, I have plenty of pads to replace parts and decrease the bandwidth. But I would rather have a larger bandwidth than I need so I can make comparative tests to the original design.

By the way, this design is to replace some Audiophile M-Audio BX5a "guts". Not that it would matter, but there is no secret about it.

One more question that came up: the GND pin on the chip, should it be connected to small signal analog ground (same ground that I use for input)?

Yes.

If using a PCB, use a copper pour. If doing point-to-point maintain a "star" pattern ground (bring all ground wires back to one point, usually the power supply entry point).

Mike T.

Tank you very much, Mike. I will follow your advice.

You're welcome.

For what it's worth, this is a very popular device with the "Gainclone" crowd. If you're not familiar with this, in a nutshell, years ago there was a Japanese guy that made amps using this device. He hand wired the minimum amount of components directly across the leads of the device, put it in a box and charged people enormous amounts of money for this "perfect" amp. People raved (still do in some circles) about these things - all sorts of mystical properties; phase of the moon, fairy dust kind of thing. Then the "clones" started. Don't get me wrong, this is a great little beast; I use them in my "bench" amp (though with the advent of the newer Class-D amps I'm re-thinking my bench amp setup).

If you search on "gainclone" you will find a dizzying number of amps built out there based on this "premise". One thing you'll notice is the size and number of the power supply capacitors. Some of these guys go to an absurd degree using sometimes as much as a Farad (or more). Too bad no one ever pointed out a little math to them...

Good luck and have fun!

I dug up a link: http://diyaudioprojects.com/Chip/LM3886_CA/LM3886_CA.htm 

I misspoke; the "original" was based on a an LM3875. I've only worked with the LM3886...

Mike T.

OK, you made me laugh a lot. I had no idea about this "clone" stuff.
Well, that part of phisics that is non-intuitive is called magic!