3886x3 vs 49830

This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I will be building a set of tri amped speakers and was just wondering about the tradeoffs of the two designs. What are the performance differences between the two approaches.


The 3886 seems like it would be an easier design but I've also read that it can be problematic when driving some speakers.

I could probably add some circuitry to smooth the speaker impedance so I would think the stability could be overcome.

Is it just preference?
smoothing the speaker impedance usually requires a parallel Zobel. This method of smoothing reduces the impedance seen by the amplifier. It is wasteful of power and adds extra stress to the chipamp.

If a speaker manufacturer wants a speaker that can be driven by a wide variety of amplifiers, then they will often resort to using Zobel correction. It also helps smooth out the frequency response when the speakers are coupled through a passive crossover.
All this goes out the box when you design for active coupling.
Thanks for the response. I guess what I was trying to say is that if there is some instability caused by the speaker load then I can figure something out to solve it.

The real question I'd like to know is what the difference is between the two deisgns. The only thing I can figure is that it may be easier for the 49830 design to achieve higher power, particularly for high impedance loads. The bridged parallel design shown for the 3886 seems almost more complicated than the multi FET design for the 49830.

After so many hits, I'm surprised no one has an opinion!

Thanks again.
OK, after doing a number of calculations I think I have a handle on the differences. The big problem with the 3886 is the lower max voltage which makes it tough to get to high power levels. With my 3 ohm speakers, I could barely get enough power without going bridged since they are relatively inefficient.

The real problem, for me anyway, is that we get a lot of power surges out here so if I'm at the very edge of the voltage spec and get a 2x hit, I'll be replacing a lot of chips! :bawling:

If I go bridged, ala BPA200, that makes the design complex enought that it seems easier to go with the 49830 and multiple FET pairs. That should give me plenty of headroom for the power hits and still have enough current to drive low impedance loads.

I'll do a little more work before I make a final decision which way to go.
Basic differences:

LM3886 - handful of external components and done, FULL protection. Limited in power typically by voltage but if 3 ohm then power dissipation. PCB is pretty easy, lots out there to use.

LME49830 - more components, need to understand setting bias for correct thermal compensation (if using FETs that need such), NO protection. Much higher voltage headroom (+/-100V) and no limits on current drive just adds cost to add more FET pairs in the output stage. More complicated PCB design, stability may be an issue depending on PCB, less out there to just buy to download and use.

If I was tri-amping I would probably go with LM3886. You can set your power supply to keep from blowing up chips. Remember, 50W into a tweeter is actually a lot of power. Might need to use parallel or bridge for the woofer but nothing else.


One of the things I was thinking of was to try the BPA-300 approach done on a thread here where he just did two PA150's and drove them differentially.

Since I am doing a differential input anyway, I could just swap input pins on one channel and drive them in parallel to basically do what he did.

The nice thing about this approach is that I could try the PA150 as is and if I ran into problems I could just add the second board.

The only bugaboo is the surge problem. I've already lost thousands of dollars in equipment due to the power company and they make it really difficult to get compensation.

I guess I could make a really robust power supply with precision hard clamping and fuses as well as some whomping TVS's

I know 50W continuous into a tweeter is loud but shouldn't one plan for handling peaks? Perhaps it's just the engineer in me overdesigning. :xeye:

I guess I'll have to whip up a schematic and see what the pro's here think.


The '30 is a mono, FET optimized version. The app note for the 4702 does have some excellent performance specs.

Do you know how much you can draw through a FET without affecting performance? I think the FET's they show are good for about 16 amps or so. If they are good to any where near that, it would mean some awesome power output with even one pair.

I do have a couple of subs that will handle 1300W with dual 1.8 ohm VC's. It would be nice to build something to handle that! :D

The biggest plus is that the chip and FET's are good to more than double the supply voltage so they would be much more robust for power hits.

I am not that familar with higher power applications. But I am pretty sure that for your big subwoofers, you will want to consider the power dissipation of the FETs. The IRF540 FET data sheet calls out 150 watts of heat dissipation but recommend it for "50 watt applications". So for continuous high dBs you will want multiple FETs. If you want to play the 1812 Overture I would guess that one pair might do well for the occasional canon blasts.

I have looked at BJTs more and one 15 amp transistor lost too much gain beyond 5 amps. I wouldn't expect this problem with FETS but I am not sure.

I am assuming that you will be feeding one driver with one amp: this means you will get the full sensitivity of the speaker. With a 90 spl/1 meter speaker, a kilowatt will be 150 db, a heck of a lot of noise. I would suggest building your first amp with one pair of FETs. If it is loud enough on the woofer then you are set to build the rest. If not, try it on the mid-range. If still not loud enough, then it will work on the tweeter. Also when a negitive feedback amp (such as these) powers one speaker driver the inductive impedence shouldn't be a problem and no correction components are necessary.
I have done plenty of testing on the bench with LM3886 and just invert one of the input signals to create BTL (LM4780 is great for this). Your PA boards used this way should work fine.

With FETs SOA basically comes down to power dissipation. If you want loads of power then you will want the highest power dissipated rated FETs or lots in parallel. For example, the Renesas FETs used in the LM4702 FET app note are only rated for 150W while the Toshiba ones are rated for 200W.

I'd post in the power supply form about your problem and see what solutions you can get. Going LME49830 with FETs will give you lots of headroom but keep in mind the no protection problem, one blown speaker driver can be worth a bit of $$$. Designing, building, testing and proving a LME49830 based board with some protection will take considerable time and cost. Start with the PA150 board (assuming you have these done) and see if you think you need more power as suggested by other post. One solution to the voltage problem is to run all the LM3886 amps in BTL and back the supply voltage down to say 25V. Then you have lots of headroom and the BTL design gets you the power you want from a lower supply, hopefully.

That is what I was thinking with regard to power, I'll need a big heat sink for the subwoofer amp. It looks like I'll need 19A@65V to generate the 650W per winding which says that I could get away with two FET pairs assuming appropriate heat sinking. With four FET pairs I could run both windings off of one amp.

The sub is not that efficient, running more like 87dB and according to WinISD Pro Alpha 1KW is only 117dB which would only be 106 at the listening position. So, if I were listening at an 86dB level that would give me a 20dB headroom for cannon fire, etc.

Since I am doing a PCB, I think I'll design for four pairs so that I can add pairs as I need them. My guess is your're correct that I can get away with a single pair for the tweeter and mid and perhaps the woofer with two to four pairs for the sub.

I'll not be putting protection on the amp itself, my concern is the supply going high enough to kill the 3886. My previous life was spent doing switching power supplies for ultra reliable applications, (voice/data sytems in high reliability/availability environments). Worst case I could do some low noise switchers.

I think I'll also do a PA150 design with a differential input so I can play with a BPA topology via swapping differential leads. If I have an input impedance of 20K differrential and parallel two, that would still give a 10K differential which should be adequate.

One nice thing about being retired is that I can take the time to play a little. :D
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.