I have been mulling over the thermal question to. I'm very tempted to build these, but one of my pairs of speakers is a solid 6Ohms, the other 4 and neither are very efficient (84dB/W area). Room is small, I don't headbang so should be fine, but would be nice to give the chips the best chance.
An internal heatsink with forced cooling does give the heat removal, but finding a quiet enough fan so as to not add to the room noise floor becomes a challenge. Water block cooling as used in bling (sorry high performance) PCs might work but again there is a mechanical component to keep quiet.
I might be over thinking it and just a copper spreader plate with a large surface area of insulator might be good enough for the job...
An internal heatsink with forced cooling does give the heat removal, but finding a quiet enough fan so as to not add to the room noise floor becomes a challenge. Water block cooling as used in bling (sorry high performance) PCs might work but again there is a mechanical component to keep quiet.
I might be over thinking it and just a copper spreader plate with a large surface area of insulator might be good enough for the job...
You seem to be assuming both the source and THAT 1200 have at least an order of magnitude less offset than the control amp, which is low by a couple orders of magnitude. Have a look at typical output buffer behaviours and the 120x datasheet. Might also be a good idea to parameter sweep the parts on your demoboard to their worst case specs.So the use of the OPA could avoid the DC Servo.
A small heatsink will keep the lm3886 happy without fans. Something like 5"*3"*1.5" will work for cooling. Make sure the base of the heatsink is not too thin. 3/8" base is a big improvement from 1/4". Heatsinking this chip is not a hard thing to do.
Bill goto the chip amp photo gallery and see pics of real working builds. My power into efficient speakers wont matter to you. If you view that thread you'll know you dont need class A sized sinks. My recommendation is a good one and I'm sure you can run full blast on that size sink but I'm not asking you to take my word without checking. See the thread and it will be obvious that heatsinking is not usually a limiting factor in this build.
Folks typically overestimate swing requirements and underestimate C/W requirements. The errors in both cases are often substantial but tend to cancel each other. This is fortunate for music reproduction but unfortunate for accurately specifying heatsinks. The fix is pretty easy; measure the required swing and do the math to determine the needed C/W of the heatsink. (You can also check your work with the dissipation curves in the 3886 datasheet.)
The alternative to doing the design up front is guessing and potentially iterating, which tends to be more work in the long run. Heat sinks of the physical size udailey's suggesting around 1.7 C/W; OK if one's looking for a few volts peak on ±28 V, high by a factor of at least two if the goal's maximum swing.
The alternative to doing the design up front is guessing and potentially iterating, which tends to be more work in the long run. Heat sinks of the physical size udailey's suggesting around 1.7 C/W; OK if one's looking for a few volts peak on ±28 V, high by a factor of at least two if the goal's maximum swing.
@Udailey and twest820
Either I am misreading you or you are misunderstanding. The problem is not the heatsink size, but getting heat OUT of the 3886 into it. As Tom said, the isolated package runs into trouble at 30=35W in 4Ohms. The non-isolated form is better, but you either have to run the heasink at 24V or find a suitable insulator that actually gains you cw the plastic package. Once that is solved sizing the heatsink is for the static case.
Either I am misreading you or you are misunderstanding. The problem is not the heatsink size, but getting heat OUT of the 3886 into it. As Tom said, the isolated package runs into trouble at 30=35W in 4Ohms. The non-isolated form is better, but you either have to run the heasink at 24V or find a suitable insulator that actually gains you cw the plastic package. Once that is solved sizing the heatsink is for the static case.
Working through the maths should help with understanding the design space and the relationships between package C/W, pad or grease C/W, and heatsink C/W. The continuum of options is much larger than the relatively, hmm, shall we say somewhat one sided solution points you mention. If you'd prefer to consider just those options that's certainly your choice. But I think what udailey and I are both saying is we'd suggest more nuanced approaches.
I've not seen the Mod86 manual so I'll let Tom comment as to what it provides in the way plug and play 4 ohm options.
Package DAPs are Vee, BTW, not Vcc. You'll fry the 3886, crowbar the supply, and blow a fuse with a non-isolated mount to a sink at Vcc.
I've not seen the Mod86 manual so I'll let Tom comment as to what it provides in the way plug and play 4 ohm options.
Package DAPs are Vee, BTW, not Vcc. You'll fry the 3886, crowbar the supply, and blow a fuse with a non-isolated mount to a sink at Vcc.
Working through the maths should help with understanding the design space and the relationships between package C/W, pad or grease C/W, and heatsink C/W. The continuum of options is much larger than the relatively, hmm, shall we say somewhat one sided solution points you mention. If you'd prefer to consider just those options that's certainly your choice. But I think what udailey and I are both saying is we'd suggest more nuanced approaches.
The continuum is indeed quite wide. It would appear that no one has explored it in any great depth either. No problem as I can do this and don't want my personal research into whether this board is applicable in my application to add to the noise on this thread. Luckily the TI datasheet has a lot on the thermal aspects, mainly because this chip can be twitchy.
Well, have fun. Though, really, it's a plug and chug exercise once requirements are defined---probably the most explorative part is settling the interaction between heat sink and chassis shape. The 3886 isn't twitchy. It's just that folks tend to want one chipamp to have the same thermal handling as the four discrete devices found in most 100W discrete amps. This despite most home audio operating at milliwatt RMS levels with less than 10dB crest. The predictable result is much DIYer energy expended upon headline numbers of little practical significance.
When physics collides with expectations desire loses. 😉
When physics collides with expectations desire loses. 😉
Bill, I have done my own design with these chips, and have run them at continuous headbanging levels, without problems. Normal sensitivity speakers and impedance, but run at deafening levels for extended periods. Intelligent heatsinking is the answer, and here is where size does matter - a piddly bit of metal will be useless for keeping problems at bay.
Mounting it correctly is just part of the equation, as people have pointed out - take care in how it's done, and you should be right ...
Mounting it correctly is just part of the equation, as people have pointed out - take care in how it's done, and you should be right ...
you say plug and chug, yet seem happier to write lots of words than throw down a few numbers. Most curious. And I will have fun. That would be the only reason for going this route. Isn't that the point of this hobby.
Intelligent heatsinking is the answer, and here is where size does matter - a piddly bit of metal will be useless for keeping problems at bay.
Size cannot help you if Theta CS is 0.5C/W. You have to keep this down. There are a wide range of ways to do this, some easier than others. I'm looking at a blank sheet packaging, so why not.
Until requirements are known specific numbers cannot be thrown. So instead there are links to a complete design methodology and a pointer to a second one. Should it be preferable not to turn the crank on either of these Tom offers consulting services. And, certainly, it'd be unseemly of me to infringe upon his business in his thread. So I'm afraid you'll have to accept I'm not going to do the project on your behalf. 'tis is DIY Audio, not HaveSomebodyElseDoItForYou Audio, no?you say plug and chug, yet seem happier to write lots of words than throw down a few numbers
Further, look back through the last page or two. A few numbers are, as you put it, thrown down. The solution to the theta_cs concern is mentioned too; note support for it's called out in the Modulus-86 feature list.
Bill, you're misunderstanding how this heat transfer thing works - it's all about what the left hand side of "0.5C/W" means: if the metal object that the chip is mounted to remains at 25C no matter what temperature the chip tries to rise to, from internal heat generation, then the higher the temperature of the chip package goes the more the heat is transferred from the chip to the heatsink, even with relatively poor mounting. If the heatsink, however, heats up very quickly from the heat being fed into it by the chip working, then it becomes useless as a means of drawing the heat away - a huge lump of metal sitting there, laughing at the amount of heat that the chip is throwing at it, is extremely effective at doing the job of keeping the temperature of the chip package under control ...
Last edited:
@tom: Really sorry I didn't mean for all this noise.
@twest820. I know what I want to do. I only started off asking if anyone had some novel ideas for non grounded heatsinks. I can work all this out and will, so no idea why you feel the need to try and score points with your posts. You don't want to help and have nothing useful to add, so why are you bothering?
@twest820. I know what I want to do. I only started off asking if anyone had some novel ideas for non grounded heatsinks. I can work all this out and will, so no idea why you feel the need to try and score points with your posts. You don't want to help and have nothing useful to add, so why are you bothering?
Sometimes it takes a couple exchanges to understand where someone's coming from and what they're looking for. One might call it an operational hazard of posting on the internet. Bill, you've two other folks trying to help don't seem to be getting much traction either. So I think all I can do is wish you good luck and hope you can eventually step out of the mode of seeing this as confrontation where someone must score points to win. Go forth, build cool stuff, be happy. It really is that simple.
Bill, ALL the replies you have received have been accurate & helpful.@twest820. I know what I want to do. I only started off asking if anyone had some novel ideas for non grounded heatsinks. I can work all this out and will, so no idea why you feel the need to try and score points with your posts. You don't want to help and have nothing useful to add, so why are you bothering?
Whether improving case to sink is useful depends on loadsa factors which are addressed by the various methods mentioned by these helpful people. No one is trying to score points.
There is no need to diss people who are trying to help. Tom himself has a good discussion about isolated vs non-isolated case for 3886 on exactly this subject.
The methods are quite simple so a 'plug & chug' description is appropriate. But they DO need an understanding of the factors involved.
Douglas Self has a good treatment in his books.
__________
However, if you do reduce case/sink to less than 0.5 C/W, we would all be pleased to know how you have done this ... whether its necessary or not 🙂
__________
My $0.02 is that the TOP of the plastic case is often very hot so it's worthwhile getting a hunk of Cu/Al/heatsink in contact.
Last edited:
And sonics wise too, have you tried out other opamps like ad797, ada4627b, opa827, etc? 🙂
I performed an extensive survey of the opamps available when I designed the circuit. I am not able to find an opamp that beats the LME49710 when used in a composite LM3886-based amp. I optimized for the lowest THD and the highest loop gain possible.
Random op-amp "rolling" is not recommended for composite amplifiers as the controlling opamp impacts the stability of the circuit directly.
~Tom
Member
Joined 2006
- Home
- Vendor's Bazaar
- Modulus-86: Composite amplifier achieving <0.0004 % THD+N.