Is the LM386 Noisy ?

My subjective opinion of the LM386 driving a 2.5" 8ohm loudspeaker is "really noisy" although that was in a quiet workbench area, in somewhere full of whiring IT equipment I might not have noticed a thing.

A graph on the datasheet shows significant THD above 0.2w (1.2v into 8ohm) and I recall a test sinewave starting to sound nasty well below maximum volume.

At high output the THD reduces the s/n ratio, I'd say 80dB using your measurement. If you want volume in reserve at normal listening levels for a device that might be used somewhere noisy then you might be down to 60dB.

Anything i am likely to to build is small volume high margin or a hobby project and I'l happily pay a couple of euro's more for something better even if I could get away with mediocre performance.

Bob

** Groper IDIOT Alert

** Is 100% complete s**te.

A noise level of 144uV ( = 86 dB below 1 watt) is not audible via a normal loudspeaker.

...... Phil

And this ignores any noise from the filter, which of course reduces the actual noise of the LM386.

Nice job, Phil.

Cheers

PeteS

** The audio band filter ??

Totally irrelevant since it come after 60dB of pre-amp gain & operates at nearly line level.

** What drivel.

Care to rethink you words ??

** Shame you tried to spoil it.

........ Phil

Phil,

Thanks, that more or less confirms what I see. I remeasured the output noise yesterday and it corresponds more or less with that (the figures I gave before were from (incorrect) memory). I measure approx. 1 mV peak-peak, and since for gaussian noise RMS ~ peak-peak/6 that implies about 166 uV RMS.

In any case that is clearly audible as a hiss in the headphones.

So we can close that discussion...

greetings, Tom

Very useful measurement. 87dB is damn good for a lot of applications, and would've certainly qualified as "hi-fi" 50 years ago. (30 years ago, probably too!)

Many hobbyist-type uses of LM386's allow a lot of power supply feedthrough and have poor layout (large loops) that allow all sorts of

60Hz and RF hash to enter and get mixed/demodulated into what is nearly just noise. This is part of its bad reputation. Lack of any rational gain budget in a lot of hobbyist or cookbook type circuits don't help either.

Even hobbyists who know how to and work hard to lay out a good PC board to support a 600MHz $30 digi-RF IC will often neglect doing a good layout for a part that costs pennies :-).

Tim.

"Tom"

** YOU seem to be after a 1 watt ( low duty cycle) amp with about 120 dB s/n ratio.

Dead simple.

Use a low noise op-amp like the NE5534 with about 5 times gain = circa 3 uV output noise.

Add a pair of current boost transistors to the output - ie emitter followers, one NPN the other PNP.

The op-amp can deliver 15 mA, the followers get you up to 500mA peak.

....... Phil

"Tim Shoppa"

** Consider that hi-fi amps vary from about 20 watts to 200 watts per channel and that their s/n ratios are invariably expressed relative to FULL power.

so 87 dB rel 1 watt

= 97 dB rel 10 watts

= 107 dB rel 100 watts

Or, looked at another way - hi-fi speakers have an electo-acoustic conversion efficiency of about 92 dB for a 1 watt noise input at 1 metre. Driving such a speaker with an LM386 creates a noise of 5 dB SPL at 1 etre - which is WAY below the threshold of audibility even in a quiet room.

The OPs wacky problem is BECAUSE what he is doing with that earpiece is SOOOO damn dangerous.

It (allegedly) delivers over 150 dB SPL for 1 watt of input.

Be like supplying a hi-fi speaker with 650,000 watts !!!!

...... Phil

Care to post a url where this topology is discussed. I've always had a bit of a problem with quiescent current in this circuit, especially in single supply operation (which is what the op postulated as a requirement).

Jim

I think this is incorrect. The noise at the output is EIN x gain + output noise. With only 20dB gain it is quite possible the output noise dominated. I suggest you try again with a much higher gain. Note that 144uV is about -80dBu which, depending on the value of feedback resistor, is probably close to its normal output noise.

Ian

"RST Engineering (jw)"

** The OP has no need for low distortion at high SPLs - may well be able to use the followers in zero bias mode.

** The EIN of a circuit is DEFINED as being the measured output noise level divided by gain.

The "E" stands for equivalent !!!

** Gobbledegook.

....... Phil

That wasn't the question.

Jim

"RST Engineering (jw)"

** Can you stop top posting ?

You will get no replies from me until you comply with normal etiquette.

........ Phil

"Phil Allison"

** Maths correction:

Natsemi do not supply a noise spec for their LM386 audio amp.

This has become a sore point with a poster called " Tom " .

Having a spare LM386N ( 8 pin DIL) in my bits box meant is was easy for me to test one.

I used the most basic circuit with a gain of * 26 dB * and fitted a 47uF bypass cap on pin7.

Input pins 2 & 3 were grounded to pin 4, DC supply came from a 9 volt radio battery.

The output signal on pin 5 was fed first to a 60dB gain mic pre-amp and then a 22Hz to 20 kHz audio band filter with 12 dB/ oct roll-off slopes.

The noise signal was observed on a scope and found to be hum free, then measured with a True RMS voltmeter with over 100kHz bandwidth. The reading was 144mV rms.

So, the LM386's output noise level was 144 uV rms.

Hence the EIN = * 7.2 uV rms.*

Input noise density = * 50 nV/ rt Hz * ( !!! )

Little wonder it is NOT speced.

Relative to a 3 volt rms output level, the s/n ratio is around 93 dB.

Fine for the intended applications.

....... Phil

Fair enough. It does not invalidate my point.. It does therefore depend on the gain of the amplifier and has two components viz: the ACTUAL noise at the input x the gain plus the ACTUAL output noise of the op amp. Try making EIN measurements at different gains and you will get difference answers.

Ian

On a sunny day (Mon, 22 Jan 2007 09:21:38 +1100) it happened "Phil Allison" wrote in :

That could be wrong. If this is the thing I think it is, like they use to test your hearing, a distorded wave at X amplituide may rattle the headphones mechanically so although the person under test will not hear the original tone, he / she will hear 'something'. Having a pure undistorded wave is needed, I expect, that is why the 2 x EL34 (OK or the 50kW tubes). LOL

"Ian Bell"

** You simply had no identifiable point

- just a half baked, personal suspicion.

** Shame that assertion is not backed up with any support. ** Sure - with systems consisting of several gain stages.

Not one tiny bit relevant to a single chip device.

BTW.

Note that I have corrected my previous calculation.

The most unfairly maligned Natsemi LM386 is 6 dB quieter than earlier posted.

...... Phil

"Jan Panteltje"

** The earth may very well be flat ............

** No.

The OP's INSANE Nazi inspired device is intended to punch brain numbing SPLs into the auditory nerve cells of the VICTIM - to make them walk straight or - SHIT THEMSELVES in pain.

Take your pick.

........ Phil