LM3886 input voltage

[lowell]

Hey all,
I'm having a difficult time deciphering what the max input voltage is for this chip amp. I have built one that has a PSU from a 21-0-21 PT and full wave bridge rectifier. Gives me +/-31VDC power rails. There doesn't seem to be anything in the data sheet stating the max input. Finally, does +/-31VDC rails give me ((31*.7)^2)/8 = 58watts into 8ohms? Not sure my RMS and wattage calculation is correct.

thanks!

[nashvillebill]

I'm certainly no expert, but the chipamps web site has a section in their assembly manual on selecting the transformer and rail voltages for the LM3886 http://www.chipamp.com/docs/User%20G...3886%20Kit.pdf

[tedmich]

It should be fine, and this voltage will sag a bit when running. Up to +/-42V is well tolerated driving 8 ohms, >47V is where they POP. At higher voltages you can't drive 4 ohm loads, but you really don't want the ever drive <8 ohm loads with the LM3886, IME.
here's a pretty good treatment
LM3886 chip power amplifier power supply design.
You can probably make 100w but these sound much better when not pushed so hard, enjoy!

[lowell]

Thanks guys. My question, though maybe not clearly stated, is what is my maximum INPUT voltage into this thing before clipping?

Secondly would a 4ohm load give me twice the wattage? ((31*.7)^2)/4 = 116watts into 4ohms. I am aware you wouldn't want to do this as it would probably fry the chip, but is the math correct theoretically speaking?

[g1]

Theoretically, your math is correct. Wattage will double into half the impedance. Practically, there will always be more sag in the supply rails when delivering more current, so you won't get the full doubling of power.

I believe the input voltage will vary depending on the supply rails used. On page 3 of the pdf, it states that the differential input voltage max. is 60V when using +/-40V rails. I think it would clip long before that .

[tedmich]

A typical supply voltage (and gain) you are safe putting 250-350 mV at the LM3886's input before clipping starts.

Making 68w at 4 ohms means the output is about 4.1 A at 16.5 V
typical gain setting= 33db = 44.7 gain ratio (V/V)
16.5V/44.7= 370 mV

I find this page helpful and the math seems sound:
http://www.sengpielaudio.com/calcula...lification.htm

[J M Fahey]

The Math is fine, but it needs some adjustments because the amplifier and power supply are both real World and have losses, so actual measured power will be lower.

1) yes, 21+21 VAC will give you no load rails of 21*1.41=29.6V .
To which we should substract 2 x 0.7V , the drop across diodes , so you should have +/- 28V rails (by the way, a datasheet recommended LM3886 supply voltage for 4 ohm loads).

Yet you measure +/- 31V

Simple, either you have slightly higher wall voltage than expected or, quite common transformer winder trick (I do that myself ) : it's standard practice to wind 5% more than strictly calculated.

It's not to "cheat" but to compensate for two >real World< problems many ignore:
a) there's not *perfect* coupling between windings
b) a little of the magnetic flux is wasted magnetizing the core itself

both can be calculated , like anything else, but in practice amount to some 2 or 3% loss, thus the common 5% overwind.
Similar to the old British formula to brew tea: "one teaspoon per cup ... plus one for the pot".

2) as soon as you start pulling current from that supply, available voltage will drop

a) because of resistive losses.
Mainly in the winding wire but also in plain external wiring, connections (even if soldered), the *fuse* , etc.

b) because of ripple in the caps, so average measured voltage will drop.

3) this one is important : active elements (tubes and transistors) have a minimum unavoidable voltage drop across them to function.
In power transistors it's about 2V but since they usually have drivers, total loss per rail is around 4V

So a corrected real World value to be used in calculations is:

30V rail *0.9 (supply drop 10%) -3V (again, supposing LM3886 designers did some magic and diffused very high quality transistors inside that chip) = 24V peak = 17V RMS = 70W RMS@4 ohms .

How much do they claim? ...... 68W? ...... guess my Math is quite accurate

In fact they mention 68W but *guarantee* "60 W min." ... just in case some Lawyer wants to complicate their life.

So yes, your Math is fine, simply in practice you must derate some values a little.

There's other limits (literally) which are mentioned in the datasheet .

Suppose you heatsink your chipamp very well, use a fan, water cooling, whatever, and lower load impedance to get higher power ... the chipamp is short protected and has an internal current limiter.

Datasheet states "typical" 11.5A ... followed by minimum guaranteed 7A Peak ... and that for 10 milliseconds .

To boot, LM3886 has an extra protection layer, called "SPIKE" , a catchy name for an internal thermal cutoff ... which starts limiting peak current depending on chip temperature.

So in a nutshell: serious datasheets such as those from NS, TI, ON, ST, Fairchild, etc. *can* be trusted and should be followed.

And read the fine print and "the other pages" besides the main ones, they explain *how* did they measure or what do they actually guarantee.

That said, having 60 or 70W is practically the same, so no need to obsess with that.

EDIT: almost forgot, the input signal

It's a big Op Amp so gain calculation is accurate, typical NFB is (20k/1k)+1=21X.
Input sensitivity to *just* reach clipping is:
17V RMS/21=800mV RMS

If it's a guitar amp and has mixed feedback , the calculation is more complex and depends on speaker impedance, but is usually chosen to again need around 1V RMS, simply for convenience.

[R.G.]

Thanks guys. My question, though maybe not clearly stated, is what is my maximum INPUT voltage into this thing before clipping?

It's a 100W opamp. The voltage before clipping depends on (1) the size of the power supply voltage and (2) the gain you have set up with the feedback and input resistors.

The output clipping level will be a few volts below the power supply, so if you have +/-30V power rails, it will push an output to about +/-26-28V peak, less at heavier loads.

The input voltage before clipping is then this peak voltage - call it 26V for the made-up example - divided by the the gain. So if it has a gain of 24, which is a common number for such amps, your peak input voltage before the output clips is 26V on the output divided by a gain of 24, or 1.083V peak. If the gain was changed to 30, then the peak input voltage before clipping changes to 26V/30 = 0.87V.

If you used a +/-40V supply and a gain of 30, the peak input before clipping is then ~36/30 = 1.2V for the same gain; the bigger power supply raises how much output voltage you can get.

Secondly would a 4ohm load give me twice the wattage? ((31*.7)^2)/4 = 116watts into 4ohms. I am aware you wouldn't want to do this as it would probably fry the chip, but is the math correct theoretically speaking?

Yeah, kind of. The real-world changes to that are reflected in the idea that at heavier loads, the output won't go quite as high and also that the power supply voltages will sag under loading.

The right answer to both questions is ... it depends.

[lowell]

Thanks for the enlightening posts guys. I completely follow your explanations regarding the output clipping point to be just below the supply rails and then working backwards via the gain calculation to find the max input. Makes perfect sense.

[lowell]

So i figured out I made a mistake when building this LM3886 PA many years ago. I had 1k for Rin!!!! Haha whoops. This is the Zinput. Swapped for 100k, adjusted feedback loop for Rf 100k and Rfb2 3.3k (fb R divider to ground). Gets me at approximately 50watts output with line level 750mv ish. This is great because now I can use anything with that output level to drive it!

[lowell]

And thanks JM, probably never woulda known had you not said it shouldn't require a 13v input.

[J M Fahey]

Cool

Now you have a very useful bench amp, good for testing anything with a bench speaker, specially preamp or mixer output when you do not trust their amps ... or they are dead , also for signal tracing.

In between, good for some background music

[lowell]

Exactly! Perfecto. These things are so easy and so solid I'm gonna make a few just for use around the shop and the house etc... Maybe even for my tour van