They will work, and they are a lot easier to use since you don't have to mess around with insulating spacers or worry about isolating the heatsink from ground. They do however have a much higher thermal resistance, which will affect how much power they can deliver without overheating. Even if the heatsink is only warm, the die (chip) can be scortchingly hot, over 200 degrees C (or >390 F). The total thermal resistance of the die to air is the sum of the die-to-case, case-to-heatsink, and heatsink-to-air thermal resistances. I tried in vain to find the thermal resistance of the isolated package in the datasheets for several of the different chipamps that are available in this type of package. Finally I found some info in the National AN-1192 app note:
As you can see, the power dissipation capability of the isolated package is lower. The extra 1 deg C/W is significant when dealing with very high powers output (which causes very high instantaneous power dissipation in the device). The die may heat up enough to cause SPiKe to activate, which is a Very Bad Thing (TM). For comparison, using the non-isolated case with mica washer (and thermal compound) adds 0.4 deg C/W extra resistance, and using a silicone washer (dry) adds 0.5 deg C/W.
Thermal considerations are often overlooked and/or misunderstood by DIY'ers, because most of us are not engineers (although I am, BTW). Bottom line: The TF case is the easiest to use, and as long as you do not push it to it's limits, it should work just fine. Just make sure to use a big heatsink and a good quality thermal compound.
intereted in buying a pcb
HI thomas Im interested in buying a pcb. how much danish kr. im from denmark (silkeborg) I have some lm 3886tf But i need more what is the price for each?
If you are interested i know a place in dk where I buy 300w power supply with about 2X25 sec pris 300 kr. I had to modify but removing some turns.
hilsen jeppe dk.
HI thomas Im interested in buying a pcb. how much danish kr. im from denmark (silkeborg) I have some lm 3886tf But i need more what is the price for each?
If you are interested i know a place in dk where I buy 300w power supply with about 2X25 sec pris 300 kr. I had to modify but removing some turns.
hilsen jeppe dk.
The "U" clamp section
FWIW, Aluminium square tube sections are also commonly available that are much stiffer than a "U" section of the same size, with only a slight increase in weight and cost. Also may allow a smaller/cheaper section to work with the same stiffness (depending on the available size range). May be useful for this project?
FWIW, Aluminium square tube sections are also commonly available that are much stiffer than a "U" section of the same size, with only a slight increase in weight and cost. Also may allow a smaller/cheaper section to work with the same stiffness (depending on the available size range). May be useful for this project?
Gain mismatch
I'll be following this thread, waiting for the publication of the article.
I'll definitely be using the servo amps, as they allow the amps to be DC coupled, eliminating those nasty ecaps in the ground leg of the feedback divider (effectively in series with the audio, and causing all those known nasties).
I know that typically, adjacent 1% resistors from a reel tend to be quite close. But I'm still a little concerned when worst case 1% feedback resistors could cause up to 4% gain difference, which could easily cause one opamp to have to take the whole load. Maybe a 0.1% (or better) selection process is in order?
I'll be following this thread, waiting for the publication of the article.
I'll definitely be using the servo amps, as they allow the amps to be DC coupled, eliminating those nasty ecaps in the ground leg of the feedback divider (effectively in series with the audio, and causing all those known nasties).
I know that typically, adjacent 1% resistors from a reel tend to be quite close. But I'm still a little concerned when worst case 1% feedback resistors could cause up to 4% gain difference, which could easily cause one opamp to have to take the whole load. Maybe a 0.1% (or better) selection process is in order?
I came up with a better way to mount the LM3886T's to get the heat away...
I used 6mm thick flat bar which i mounted the LM3886's on directly.
Doing this allowed the heat to get away from the ICs much more quickly... i found that in normal designs, the insulator causes the heat problems.
I then insulated the alu bar from the heatsink using a sheet of thin plastic.
It works very well... ive been running bridged LM3886T's @ 42V at high-power on this for >2yrs now with no problems.
I used 6mm thick flat bar which i mounted the LM3886's on directly.
Doing this allowed the heat to get away from the ICs much more quickly... i found that in normal designs, the insulator causes the heat problems.
I then insulated the alu bar from the heatsink using a sheet of thin plastic.
It works very well... ive been running bridged LM3886T's @ 42V at high-power on this for >2yrs now with no problems.
MWP said:
Good concept - even 3mm of Al should increase the effective area of heat flux through the insulating film, and so reduce the thermal resistance substantially. I think I'll do something similar with a big amp I'm doing (1kW+) - Cover each of the two (big) heatsinks with 3mm Al sheets + epoxy + thin mylar film + epoxy (pressed hard and flat while curing; NPNs/+supply on one heatsink, PNPs/-supply on the other).
The IC should be fixed directly to a brass or copper bar with some addition of silicon grease, of course. Some people try to polish the back of IC with fine sand paper for better "sticking". Thus heat transfer is very effective, because there is any heat insulator between radiator insertion of IC and bar itself. Between the bar and radiator should be laid an insulator sheet with addition of some silicon grease. The bar may be mounted separately for each IC or one for all (depends on construction&electric connection between structure and radiator insertion). Thermal resistance is decreased because heat transfer is occurred through greater insulated surface than in previous case. The much effective cooling needs a fan.
MWP said:
I am thinking of doing exactly the same thing to mount three lm3886'sI've got a 1" x 1/4" copper bar). I understand how to attach the chips to the bar but can't figure out a good way to attach the bar to the heatsink, without electrical contact? I've considered nylon screws but I'm worried I can't tighten them enough to get good contact. I'm also worried that the elevated temperature of the heat sink might allow the nylon to stretch a bit making the problem even worse.
Thanks
I wouldn't reccomend nylon screws. Rather, use the plastic insulator bushes you can get to mount TO-220 packages. Unfortunately, these are only for M3 screws, but if you use a good few of them you should be able to get substantial pressure. Hex/allen drive screws are always nice
Hi all
Funny to see this tread come alive after 6 years of idle.
For the last couple of years I have been thinking (and working) about making a 4 x LM3886 PCB (non-bridged) without DC-servo and quite basic. The LM3886 would be standing up and the PCB-area would be alomst ½ of "Das Modul". This mean that it could be sold for 10-15 US$, if we get past the 100pcs.
This could be a cheap "buildingblock", where you would need two to make a bridged amp, but to that price it should be OK.
Let me know if there is interrest for this.
(this time I seems to be forced to write some kind of articel to go with it !)
Have fun
Thomas
Funny to see this tread come alive after 6 years of idle.
For the last couple of years I have been thinking (and working) about making a 4 x LM3886 PCB (non-bridged) without DC-servo and quite basic. The LM3886 would be standing up and the PCB-area would be alomst ½ of "Das Modul". This mean that it could be sold for 10-15 US$, if we get past the 100pcs.
This could be a cheap "buildingblock", where you would need two to make a bridged amp, but to that price it should be OK.
Let me know if there is interrest for this.
(this time I seems to be forced to write some kind of articel to go with it !)
Have fun
Thomas
Dear Thomas,
Wow I am glad you are still here as well! I discovered this topic way to late, so I hope I can still catch up. Better late then never they say...
Please allow me to ask some questions.1: Is there any drawing or picture of the PCB? I am curious how you construct this one
2: Did you use a whole unbroken ground-plane for the whole PCB?
I don't know why people are so against the DC servo's I think they really belong to the design and guarantee a trouble free operation. I've seen other designs with pot-meters on one of the input arms to adjust DC offset, but this isnt (in my opinion) a pretty solution, and most of those designs showed ringing on the scope.
With kind regards,
Bas
Thomas, the market here at diyaudio can't be saturated with good LM3886 pcb's so the interest is present, I'm sure of it.
BTW: I started this thread in 2003!
Last edited:
Hi all
With some luck we can make this tread into 10-year anv
Sebastian, to answer your questions.
1. In short no. I have just been "killing" Das-Modul and other PCB's testing out my topology. The last test turned out quite well, so I am rather sure of the concept.
The PCB will be around 14 x 4 cm (=5,5 x 1,57 inches), double sided. The 4 x LM3886 will be along the (long side).
Each of the LM3886 will receiver the same signal and then joint in parallel at the output, so we are taking about 4 x LM3886 in parallel.
(I am thinking about the option of "2x2", so one PCB could be used as a stereo-amp, each channel being 2 x LM3886 in parallel).
Ther will be no adjustments and I have made an effort to use standard components.
2. Yes
You are right about the DC-servo, I just wanted to try something more simpel with almost the same performance. If it can make more people try out the LM3886 in different configurations, my mission is completed
Have fun
Thomas
With some luck we can make this tread into 10-year anv
Sebastian, to answer your questions.
1. In short no. I have just been "killing" Das-Modul and other PCB's testing out my topology. The last test turned out quite well, so I am rather sure of the concept.
The PCB will be around 14 x 4 cm (=5,5 x 1,57 inches), double sided. The 4 x LM3886 will be along the (long side).
Each of the LM3886 will receiver the same signal and then joint in parallel at the output, so we are taking about 4 x LM3886 in parallel.
(I am thinking about the option of "2x2", so one PCB could be used as a stereo-amp, each channel being 2 x LM3886 in parallel).
Ther will be no adjustments and I have made an effort to use standard components.
2. Yes
You are right about the DC-servo, I just wanted to try something more simpel with almost the same performance. If it can make more people try out the LM3886 in different configurations, my mission is completed
Have fun
Thomas
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.