After detailed diyaudio.com investigation for dozens of LM3886 topics, found one piculiar design here that would definitely build 🙂
Initially it uses 2 LM3886 per channel, however I suppose I will simplify it to a single chip due to the lack of space and 90db sensitivity of my speakers.
It uses regulated PS based on LT1038. As my speakers have 4Ohms, I will need some decent toroid of about 200-250VA 2 x 18-21 separate secondaries for both channels according to the posts related to Lm3886.
I have attached the schemes below and kindly ask to clarify few points:
1. Will 0.7uh inductor give any benefit along with several 1w resistors (R15)?
If yes, could I substitute it with 1uf or perhaps build one by wounding around the body of R15 as it's mentioned here.
2. the block marked green on the scheme is some kind of complication if compare to other LM3886 power amps. Should I leave it or abridge to 2.2-3uf mkp cap?
3. Is it worth building regulated PS? Or maybe the one from chipamp.com would suit more?
4. Is it worth to order some cheap factory made PCB and try to upgrade it according to the scheme or simply make the layout similar to shine7.com?? Perhaps, to build my own?
5. C12 & C14 really should be that big and close to the chip? All other modifications on the web stay with 100uf...
thanks in advance!
cheers
Initially it uses 2 LM3886 per channel, however I suppose I will simplify it to a single chip due to the lack of space and 90db sensitivity of my speakers.
It uses regulated PS based on LT1038. As my speakers have 4Ohms, I will need some decent toroid of about 200-250VA 2 x 18-21 separate secondaries for both channels according to the posts related to Lm3886.
I have attached the schemes below and kindly ask to clarify few points:
1. Will 0.7uh inductor give any benefit along with several 1w resistors (R15)?
If yes, could I substitute it with 1uf or perhaps build one by wounding around the body of R15 as it's mentioned here.
2. the block marked green on the scheme is some kind of complication if compare to other LM3886 power amps. Should I leave it or abridge to 2.2-3uf mkp cap?
3. Is it worth building regulated PS? Or maybe the one from chipamp.com would suit more?
4. Is it worth to order some cheap factory made PCB and try to upgrade it according to the scheme or simply make the layout similar to shine7.com?? Perhaps, to build my own?
5. C12 & C14 really should be that big and close to the chip? All other modifications on the web stay with 100uf...
thanks in advance!
cheers
Attachments
Well, you should probably use a paralleled chip amp if driving 4 ohm speakers. The Shine7 looks well made.
Either Regulated or CRC power supplies work well for audio amplifiers.
The chipamp.com power supply is neither.
Regulated needs a margin so your transformer would need a higher voltage than you would use otherwise.
It is difficult to compare the power output of Spike system chips, such as LM3886, because the clipping performance is useless (a screech), but in any case this means that if given the choice between 26vdc rails versus 37vdc rails, you'd really want the higher voltage option because of less clipping.
Parallel spike system chips make less of that spike noise because no two chips do that at the exact same instant in time and therefore parallel turns a lot of noise into a little bit of heat.
A fully optimized support circuit for LM3886 exists and it is named "myref" and these sound good because of current-pump/nested/composite topologies override the sonic signature of the power amp, thereby reducing noise.
Either Regulated or CRC power supplies work well for audio amplifiers.
The chipamp.com power supply is neither.
Regulated needs a margin so your transformer would need a higher voltage than you would use otherwise.
It is difficult to compare the power output of Spike system chips, such as LM3886, because the clipping performance is useless (a screech), but in any case this means that if given the choice between 26vdc rails versus 37vdc rails, you'd really want the higher voltage option because of less clipping.
Parallel spike system chips make less of that spike noise because no two chips do that at the exact same instant in time and therefore parallel turns a lot of noise into a little bit of heat.
A fully optimized support circuit for LM3886 exists and it is named "myref" and these sound good because of current-pump/nested/composite topologies override the sonic signature of the power amp, thereby reducing noise.
Last edited:
thanks! so I will stay then with 2 LMs per channel as on the diagram above.
I found that for 4 ohm speakers LM needs about 18-21VAC what equals to ~28VDC on the output of PS. 25-28VAC (up to 35VDC) would be suitable for 8 ohm speakers.
thanks! I've glaced at this endless thread and it looks a bit more complicated, hence need more space and pcb would be difficult to get. However found one quite similar pcb on ebay, perhaps it could be used as well?
To recapitulate, building LM3886 power amp according to Shine7.com scheme is not the best solution in comparison with "myref" design?
Parallel: Each chip spots half the load.
Parallel amp running 4 ohm speakers uses same operating voltage as solo amp running 8 ohm speakers. In both cases, the chips spot an 8 ohm load at their outputs. Personally, I'd give the amp the normal pair of rails within the range of 32vdc to 37vdc. Under-volting = clipping.
Solo amp running 4 ohm speaker load is the case where you decrease the operating voltage dramatically. . . but that is not directly applicable to a parallel amplifier where the load per chip is half.
Is a Parallel LM3886 or a MyRef better for you? I do not know that answer.
However, both are better than standard/common single LM3886 designs.
P.S.
Electronic items on eBay are usually more suitable for selling but less suitable for using.
The real PCB for MyRef is not difficult to get, and if you want it, inquire at the MyRef thread, since they do have boards.
Parallel amp running 4 ohm speakers uses same operating voltage as solo amp running 8 ohm speakers. In both cases, the chips spot an 8 ohm load at their outputs. Personally, I'd give the amp the normal pair of rails within the range of 32vdc to 37vdc. Under-volting = clipping.
Solo amp running 4 ohm speaker load is the case where you decrease the operating voltage dramatically. . . but that is not directly applicable to a parallel amplifier where the load per chip is half.
Is a Parallel LM3886 or a MyRef better for you? I do not know that answer.
However, both are better than standard/common single LM3886 designs.
P.S.
Electronic items on eBay are usually more suitable for selling but less suitable for using.
The real PCB for MyRef is not difficult to get, and if you want it, inquire at the MyRef thread, since they do have boards.
Last edited:
does it mean that parallel design bring less heat to LMs or it's not connected at all? 🙂
Parallel amplifiers will produce less heat compared to an normal amp provided that both are running on the same rail voltage driving say a 4 ohm load.
Most speaker systems available today are nominal 6 ohms, which may go down to 3 ohms , so a parallel amplifier will definitely have some advantage over a normal amp.
If compared driving the same load. . . Parallel LM3886 has half the heat into EACH chip than solo LM3886.
But the total amount of heat into the heatsink, had no significant change.
So, it is still necessary to double the heatsink size when driving 4 ohm speakers instead of 8 ohm speakers.
Likewise, since you are driving 2 chip, you do still need to double your transformer amperage rating when driving 4 ohm speakers. These apply because a 4 ohm speaker load is double that of an 8 ohm speaker load. Parallel is not magically more efficient--parallel is merely able to do double the work load.
The chip is made with the expectation of 8 ohm speakers, and so if you double the work load, then you either cut the operating voltage dramatically or double the number of chips (because of double the work load).
The concept of parallel:
1 Egyptian house cat moving a 1/2 pound weight at 27 mph versus 500 Egyptian house cats tandem teamed to move a 250 pound weight at 27mph. The 500 parallel (tandem) example moved 500 times the load but emitted 500 times more body heat, ate 500 times more cat chow and drank 500 times more water. OOPS!!! Reciprocity error--cats don't have brakes and 0 brakes times 500 is still 0 brakes. That won't do, except for a jfet amp.
I'll try again.
Solo = 1 person carrying a 50 pound weight but Parallel = 2 people carrying a 100 pound weight. Twice the weight was moved by Parallel, but twice body heat was emitted and twice the burgers and beer was consumed.
Observe that parallel kept the individual's heat and workload within tolerances.
P.S.
Other function of parallel op-amps, even really big size op-amps like LM3886, is that when run parallel, current noise (unpleasant sonic) is reduced so that you get lavish sounding current headroom if using parallel amplifiers. This effect is beneficial regardless of speaker load.
Instead of parallel, it is possible to simply double the size of the output devices. For example a BD911 with an 8 ohm speaker versus MJL21194 with a 4 ohm speaker.
However, the output devices for LM3886 are sealed inside the chip where we can't change them, so to double the output devices size, we double the number of chips to create a parallel amplifier capable of appropriate performance with a 4 ohm speaker.
However, the output devices for LM3886 are sealed inside the chip where we can't change them, so to double the output devices size, we double the number of chips to create a parallel amplifier capable of appropriate performance with a 4 ohm speaker.
Last edited:
Paralleling has however other downsides besides increased heatsink size/power requirements, paralleling is tricky as current sharing is always an issue. Current sharing resistors are needed at the output , feedback resistors etc need to be matched carefully to balance the load equally. Careful pcb design is also required for a parallel amp to perform to its true potential.
I will expect to see 'burgers and beer consumption' listed in the specs for all future chip amps.....
Here's some examples:
Solo with 8 ohm speaker
Bridged with 16 ohm speaker
Parallel with 4 ohm speaker
In all three examples, each individual chip sees the same load as an 8 ohm speaker.
. . .
Here's two more examples:
Solo with 8 ohm speaker
Parallel with 8 ohm speaker
The heatsink, transformer requirements, and power output are identical.
Solo with 8 ohm speaker
Bridged with 16 ohm speaker
Parallel with 4 ohm speaker
In all three examples, each individual chip sees the same load as an 8 ohm speaker.
. . .
Here's two more examples:
Solo with 8 ohm speaker
Parallel with 8 ohm speaker
The heatsink, transformer requirements, and power output are identical.
Last edited:
There is no significant increase.
Regardless if parallel or solo, it is actually the work load that determines heatsink and transformer requirements.
daniel,
thank you very much for detailed explanation!
Frankly speaking I didn't even think that this parallel design would require to double the transformer. In this case I'll probably stay with myRev as the maint point of my requirements besides the quality is to enclose everything in this tiny box. And as it's only 150x250x55mm I'm pretty sure there will be simply no space for huge 400VA toroid + 4 LM3886s, 4 x 10k uf caps. However it's still possible to put 200-230VA toroid, some 2 x 10k uf caps and myRev pcb. Unfortunately this pcb doesn't have regulated PS as described in shine7 version and that is superior as mentioned here, though from the other perspective, no need for additional heatsinks for the regs.
Not sure if that is good or bad point, weird that myRev designers didn't stay with regulated PS... anyway, hope that the case itself would serve as an adequate heatsink for 2 LMs...
thank you very much for detailed explanation!
Frankly speaking I didn't even think that this parallel design would require to double the transformer. In this case I'll probably stay with myRev as the maint point of my requirements besides the quality is to enclose everything in this tiny box. And as it's only 150x250x55mm I'm pretty sure there will be simply no space for huge 400VA toroid + 4 LM3886s, 4 x 10k uf caps. However it's still possible to put 200-230VA toroid, some 2 x 10k uf caps and myRev pcb. Unfortunately this pcb doesn't have regulated PS as described in shine7 version and that is superior as mentioned here, though from the other perspective, no need for additional heatsinks for the regs.
Not sure if that is good or bad point, weird that myRev designers didn't stay with regulated PS... anyway, hope that the case itself would serve as an adequate heatsink for 2 LMs...
Last edited:
If space is a constraint you should try paralleling LM4780 instead of LM3886, because LM4780 is actually two 3886 in one package. Heat dissipation will be an issue but it isn't that bad. I have used LM4780 in parallel configuration to drive 4 ohm speakers and it works quite well without getting too hot.
It doesn't. Transformer amperage is because of speaker load.
If compared using the same speaker, Parallel LM3886 versus MyRef have this difference in needed transformer amperage: Zero difference. None.
Resources relates to workload. The amplifier has merely to withstand this.
P.S.
The problem of transformer amperage and heatsink size expenses is caused by using a 4 ohm speaker instead of an 8 ohm speaker. In this case, changing your amplifier without changing your speaker. . . is not effective in altering transformer amperage and heatsink size figures. . . at all.
P.P.S.
For reference,
The amplifier that automatically forces double the transformer amperage regardless of speaker, is a bridged amplifier because bridge forces double the power (current) output and "no free lunch" applies.
Last edited:
That tiny box? For a 100w per channel to 120w per channel amplifier? Heck no!!!
The only class AB amp that can do that is TDA7294, and even then I'm not sure if it will take it. Given that much heat pooling in the confines of such a small oven, it would be a contest to see if either the caps or the chips would die first.
Therefore, we'd also have to dramatically decrease operating voltage down to about a 15+15vac transformer, and this isn't powerful, but there would be a bit less heat in that inappropriately tiny amp enclosure.
Suggestion: Fetch a free or low cost amplifier enclosure from a second hand store or donation center or even a dumpster, since this will come with a big enough box and usually some heatsinks. Sometimes it even comes with a free transformer albeit rarely at exactly the voltage you'd want. After that, you need a can of black spray paint (not attempt shiny) and a can of clear spray paint (runs in clear do not show much).
The only class AB amp that can do that is TDA7294, and even then I'm not sure if it will take it. Given that much heat pooling in the confines of such a small oven, it would be a contest to see if either the caps or the chips would die first.
Therefore, we'd also have to dramatically decrease operating voltage down to about a 15+15vac transformer, and this isn't powerful, but there would be a bit less heat in that inappropriately tiny amp enclosure.
Suggestion: Fetch a free or low cost amplifier enclosure from a second hand store or donation center or even a dumpster, since this will come with a big enough box and usually some heatsinks. Sometimes it even comes with a free transformer albeit rarely at exactly the voltage you'd want. After that, you need a can of black spray paint (not attempt shiny) and a can of clear spray paint (runs in clear do not show much).
thanks, but I've already ordered 4x LM3886, will use 2 with myRev_c or all of them if stay with parallel.
did you have a chance to take some temperature measurements from the chips?
thanks, Daniel, so I've got the point:
1. using 4 ohm speakers will require more grave heatsink
2. transformer stays the same for 4 and 8 ohm speakers, the only thing that changes (according to other chip amp threads) is V and A in accordance to some 120-150VA transformer for a single channel using 3886.
to find the box is not an issue. The thing is that I need to make it rather narrow and small to match with my Preamp/Dac box. From the design perspective it is possible however to place all the stuff there:
-200VA toroid (130x130x50mm)
-2x 10k uf caps (30x50mm)
-myRev board with LMs attached to bottom of the case.
heat is the only issue. I wonder if I have an amp capable of up to 2x68w on 8 ohms on LM3886, what would be the real value I use, some trivial 10-20 watts? Consequently, I doubt case will be heating like hell, though not sure 🙂
there is another box, that is a bit higher and of te same type that I used with the DAC. Unfortunately it doesn't have any holes drilled.
For the worst case, I think that the box could be enhanced with these narrow heatsinks at the bottom and perhaps on the sides if I find appropriate black heatsinks.
Last edited:
I was testing the amp on a very small heat sink 100 X 100 MM with fins around 20 MM height, after playing for around 2 hours with ambient temperatures ~ 27C heat sink temperature was around 38C. I am not using a servo to correct the offset, I have an offset of around 12mV. The offset by itself will heat the amps a bit so if you're using a servo your amp might run cooler.
There's a small advantage in using 4780 for paralleling as both the amps are located in the same chip they are better matched than say paralleling two discrete 3886.
- Status
- Not open for further replies.