LM3886 clipping behaviour

cs

Member
2005-06-04 7:37 pm
.
I am considering using some LM3886's in some active speakers, and wondered what the clipping characteristics are like. Do they clip cleanly and recover quickly ?

The datasheet shows some horrendous waveforms when the 'Spike' protection is operating, and I wouldn't like to think they behave that way when briefly driven into clipping ?

TIA,
Chris.
 

MJL21193

Disabled Account
2007-03-10 1:20 am
Hi,
I did an active 3-way amp here. It's in use now (I mean right now, I'm listening to Boston) and I can't tell you anything about how it sounds when it clips because I don't think I've come close.
And yes, I like it loud.

Maybe you should be considering ways to avoid clipping in the first place.
 

cs

Member
2005-06-04 7:37 pm
.
The drivers will be Jordan JX92S, 89dB/W/m, and I will use two of them per speaker, each driven by its own PA, running off +/-35V.

So, there shouldn't be any clipping most of the time, but there is still a chance that it might occur on large peaks, albeit very briefly.
 

cs

Member
2005-06-04 7:37 pm
.
leadbelly said:
It's silly to think you need that much power to drive a fullrange speaker.

Surely, it's even more important to have sufficient headroom with a full-range speaker, since the amp has to deliver the whole audio band.

The reason for my post is that with modest amp power (50W) and average efficiency speakers, clipping is bound to occur on transients, so it should do so in a benign manner. But if they clip badly (as has been stated above), then maybe I should avoid them.
 
cs said:

The reason for my post is that with modest amp power (50W) and average efficiency speakers, clipping is bound to occur on transients, so it should do so in a benign manner. But if they clip badly (as has been stated above), then maybe I should avoid them.
Not really. Say you like to listen at 86dB RMS, and have 96dB/W fullrange speakers. Normal listening voltage will be about 0.1W, or 0.9V into 8 ohms. Well recorded CDs will have full scale transients, and the signal RMS at -20dB. This will put your full scale transients at 10W, or 9V into 8 ohms, still well below clipping.

Do your own sums with the sensitivity of your drivers, what volume you listen at, and see whether your chipamp is ever likely to clip. Feel free to give yourself more than 20dB headroom if you wish.
 

cs

Member
2005-06-04 7:37 pm
.
cabbagerat said:
Not really. Say you like to listen at 86dB RMS, and have 96dB/W fullrange speakers. Normal listening voltage will be about 0.1W, or 0.9V into 8 ohms. Well recorded CDs will have full scale transients, and the signal RMS at -20dB. This will put your full scale transients at 10W, or 9V into 8 ohms, still well below clipping.

Do your own sums with the sensitivity of your drivers, what volume you listen at, and see whether your chipamp is ever likely to clip. Feel free to give yourself more than 20dB headroom if you wish.

Your figures sound about right for my Jordan drivers, as they are 89dB/W/m each. Using two of them (in phase) in the same cabinet gives another 6dB SPL, so 95dB/W/m overall.

However, my typical listening distance would be ~3m, so the power required would have to go up another 10dB. Full sale transients then become 100W, so I could expect occasional clipping using the LM3886.

Would using two LM1875 in parallel (via output current sharing resistors) be any better ? The data suggests that the device just uses simple output current limiting.
 
Why not "BIAMP" and have one lm3886 per speaker. Is there going to be a tweeter. Possibly use a lm1875 for the tweeter.

Then you will have MORE than enough headroom. two lm1875 in parallel wont be better than one lm3886.

Give each speaker a 220VA - 300VA toroidal and 2 x lm3886 + lm1875 for the tweeter.

Just my input :p
 
cs said:

However, my typical listening distance would be ~3m, so the power required would have to go up another 10dB. Full sale transients then become 100W, so I could expect occasional clipping using the LM3886.
Yes, you would probably see occasional clipping of transients with highly dynamic material. 100W into those speakers does sound like a lot, though. 116dB is very loud. Also, my normal listening volume comes out at about 70dB, instead of 80, but yours may vary.

My experiments suggest that transient clipping behaviour of the LM3886 isn't too bad, but I would need better equipment (a digital scope, instead of my analogue one) to be sure.

If the chips are going to clip at your chosen volume with your chosen material, you don't have much other option than to choose a different amplifier. Either one of the National chips that take discrete drivers, or another amp which clips above 100W.

tangmonster said:
Why not "BIAMP" and have one lm3886 per speaker. Is there going to be a tweeter. Possibly use a lm1875 for the tweeter.

Then you will have MORE than enough headroom. two lm1875 in parallel wont be better than one lm3886.
Two LM3386 in parallel will double the available current, allowing you to run them off higher rails. This will give about 3dB more headroom, which might or might not be enough.
 

MJL21193

Disabled Account
2007-03-10 1:20 am
cs said:
I am considering using some LM3886's in some active speakers,


Hi Chris,
I would do as Tangmonster suggested, and use one LM3886 per driver. This will give 50 watts/8 ohms at +/- 35V rails BUT the impedance for the Jordans are ~6 ohms. With the +/-35V rails, this will give extra power for transients (this is in actual fact how mine work - ~6 ohm driver).
 
Spike protection is not clipping, it is a thermal protection and your amp has to either be pretty hot or driving too low impedance speakers (<4 ohms). As long as you design the thermals right you should never see Spike protection. Clipping, as in when the voltage needed is more than possible for the amp to put out, is similar to most transistor based amps. There are a few quirks because of the all NPN output stage. Best option is to do as suggested here, plenty of headroom to avoid it or set a preamp to do it nicely and the LM3886 never clips.

-SL