You can download the APx500 software from AP and run it in demo mode without the AP hardware attached. As far as I know, the software has the full functionality in demo mode and you should be able to generate the multi-tone signal and store it in a WAV file.
I'll see if I can dig out the test signal at some point, though.
~Tom
Shame, you had some good contributions. But you know that you need to be able to take sound feedback. How else are you going to excel and making the perfect DUT?
Anyhow , the termal path from the upside is so much longer , ie insulation that the bar could be made of any alloy as it wont do much difference on the total.
It looks cool though.
I am not sure even if the actual chip is entirely cased .
It is for sure bonded to the tabplate, which in the TF version is entirely convered. But I wonder if there is space above it so that the bonding wires etc are located in free space. All drawings I have seen look like this. It would make sense to because the bonding wires needs to be suspended freely to absorb any thermal expansion difference between substrate, tab plate and casing.
@rikkitikkitavi
I just come back to replay because I think you are cool and smart guy (opposite to AndrewT).
Does other really think that photo I shown is just for sharing some stupid idea?
No, that approach really works and it is confirmed to gain enhancement in taking heat from IC to heat sink. It is just a one way to do that.
Does other think that I didn't read hundreds of documents and books about many different stuff in electronic including thermal design?
I mentioned in several posts that Audio is just one field which I personally likes, I design much much complicated staff (for over 20 years) in field of power electronic and industrial automation and for example when designing power devices in range up to 100kW I need to know the smallest secrets about thermal design, how to project, how to implement, how to measure, how to confirm results, material behaviors, cooling fluid behaviors, thermal dynamics and so on …
I just come back to replay because I think you are cool and smart guy (opposite to AndrewT).
Does other really think that photo I shown is just for sharing some stupid idea?
No, that approach really works and it is confirmed to gain enhancement in taking heat from IC to heat sink. It is just a one way to do that.
Does other think that I didn't read hundreds of documents and books about many different stuff in electronic including thermal design?
I mentioned in several posts that Audio is just one field which I personally likes, I design much much complicated staff (for over 20 years) in field of power electronic and industrial automation and for example when designing power devices in range up to 100kW I need to know the smallest secrets about thermal design, how to project, how to implement, how to measure, how to confirm results, material behaviors, cooling fluid behaviors, thermal dynamics and so on …
Thanks, I try to be open minded and acting upon facts, continously rework according to data if shown a better way. I think most here at Diyaudio will bend to fact and data regardless and I have learned proffessionally the hard way that backing up statements with facts wins the arguments most of the time, dont we all (except from talking to the better half...)
How big difference did it make?
Obiously the heat path can be split in two then if the chip has a good thermal path "upwards" two. My assumption then falls flat.
I never really considered but squeezing a flat chip or semiconductor between two heatsinks probably wont double the power loss capacity.
Anyhow it is well proven that to get a good thermal connection through the tab you need a bar pressing the chip to the heatsink, so milling a precise fit copper bar is not thaat much extra work. And as copper is soft, you need to make it thick, thus improving heat conduction... Perhaps even cheaper than Putin a bigger heatsink in? (where to optimize...)
Qurious about the gain , in numbers.
I have a couple of FETs in TO-247 that I really need to increase the cooling of. Or increasing the numbers to split the heat but it wont really make so much difference because there is a limit on heat sink size. But there is space for one on the other side...
Ps - those pics in the other thread about a free running LLC oscillator you posted... saw my question?
How big difference did it make?
Obiously the heat path can be split in two then if the chip has a good thermal path "upwards" two. My assumption then falls flat.
I never really considered but squeezing a flat chip or semiconductor between two heatsinks probably wont double the power loss capacity.
Anyhow it is well proven that to get a good thermal connection through the tab you need a bar pressing the chip to the heatsink, so milling a precise fit copper bar is not thaat much extra work. And as copper is soft, you need to make it thick, thus improving heat conduction... Perhaps even cheaper than Putin a bigger heatsink in? (where to optimize...)
Qurious about the gain , in numbers.
I have a couple of FETs in TO-247 that I really need to increase the cooling of. Or increasing the numbers to split the heat but it wont really make so much difference because there is a limit on heat sink size. But there is space for one on the other side...
Ps - those pics in the other thread about a free running LLC oscillator you posted... saw my question?
Last edited:
So if confirmed you have some data to share with us? Lots of words from you, very little information.
Instead giving you exact measuring results which we don't have for this particular example with LM3886 (we didn't want to lose time on that, that is just one unique AMP we build for one friend and some fun for us constructors Macola and me) I will try to explain how thermal conductivity can be measured with simple approach which can give quite good comparative results, even more realistic than doing heavy math (that is how we know that shown method can helps in some other applications!).
In general we need to know temperature rise, ∆T/t.
Next we need to have some way to "pump" controlled / constant amount of energy into DUT (IC, transistor, diode, whatever) which will be transferred into heat, usually with constant voltage or current.
We also need two "magic" components, water and ice, to have same starting point/temperature for different tests.
So, we mount our DUT on heat-sink with method we want to test, for example, different thermal grace, with/without mica insulator, with/without copper in between DUT and heat sink, different clamping force, different packaging of DUT etc.
Next we put whole heat sink into icy water, NOT filling over DUT. Water need to be mixed/shuffled all the time while we conducting test for better heat exchange.
Finally we start test and observe temperature rise of water over time (water is crucial for proper exchange of heat into thermal probe).
And that's it! Faster temperature rise = better thermal conductivity and vice versa.
Also, longer we conducting test we can have more precise results.
I think that I give you (all users who will read this theme) more useful advice than arguing about some stuff.
Anyone can try this at home and compare its own cooling/mounting methods.
In general we need to know temperature rise, ∆T/t.
Next we need to have some way to "pump" controlled / constant amount of energy into DUT (IC, transistor, diode, whatever) which will be transferred into heat, usually with constant voltage or current.
We also need two "magic" components, water and ice, to have same starting point/temperature for different tests.
So, we mount our DUT on heat-sink with method we want to test, for example, different thermal grace, with/without mica insulator, with/without copper in between DUT and heat sink, different clamping force, different packaging of DUT etc.
Next we put whole heat sink into icy water, NOT filling over DUT. Water need to be mixed/shuffled all the time while we conducting test for better heat exchange.
Finally we start test and observe temperature rise of water over time (water is crucial for proper exchange of heat into thermal probe).
And that's it! Faster temperature rise = better thermal conductivity and vice versa.
Also, longer we conducting test we can have more precise results.
I think that I give you (all users who will read this theme) more useful advice than arguing about some stuff.
Anyone can try this at home and compare its own cooling/mounting methods.
And again... Lots of words. No data. How much of an improvement did you get in the thermal resistance from the IC die to the heat sink?
Distributing the clamping pressure evenly across the IC is a good idea. All you need for that is a bar across the IC package that you clamp to the heat sink. You don't have to use fancy materials or fancy shaped clamps for this.
The block you showed is more likely to make the overall system performance worse as it forces the decoupling capacitors further away from the IC and requires more space in the layout.
If you want advice on how to mount power semiconductors, I suggest looking at ON Semiconductor's App. Note: http://www.onsemi.com/pub_link/Collateral/AN1040-D.PDF
I linked to it earlier as well. They explain the physics needed to understand how to get a good thermal interface. They also back up their words with data.
The reason I put the words "(with data)" in the thread title was to encourage sharing of data. Data. Not fluffy marketing speak or hypotheses.
~Tom
Distributing the clamping pressure evenly across the IC is a good idea. All you need for that is a bar across the IC package that you clamp to the heat sink. You don't have to use fancy materials or fancy shaped clamps for this.
The block you showed is more likely to make the overall system performance worse as it forces the decoupling capacitors further away from the IC and requires more space in the layout.
If you want advice on how to mount power semiconductors, I suggest looking at ON Semiconductor's App. Note: http://www.onsemi.com/pub_link/Collateral/AN1040-D.PDF
I linked to it earlier as well. They explain the physics needed to understand how to get a good thermal interface. They also back up their words with data.
The reason I put the words "(with data)" in the thread title was to encourage sharing of data. Data. Not fluffy marketing speak or hypotheses.
~Tom
1) In the upper right corner, select Sequence Mode.
2) About half way down on the left side, select Add Measurement and select Multitone Analyzer from the list that shows up.
3) The generator in the Multitone Analyzer defaults to "AP 32-tone". Click on the drop-down in the generator settings where it says "AP 32-tone" and select Create New...
4) In the window that shows, select Start Definition From: AP 32-tone. Click OK to the default settings and select Create WAV File in the final dialog.
Sorry... Not the most obvious process.
Attached are 1-second tones of the AP 32-tone test signal created in this fashion. One is in 16-bit format, the other 24-bit. I zip'ed the two files as DIYA doesn't accept WAV files directly. The sampling rate is 192 kHz.
For the thermal tests, I looped the sound bites to form a continuous test signal. The AP software does the looping automatically.
~Tom
2) About half way down on the left side, select Add Measurement and select Multitone Analyzer from the list that shows up.
3) The generator in the Multitone Analyzer defaults to "AP 32-tone". Click on the drop-down in the generator settings where it says "AP 32-tone" and select Create New...
4) In the window that shows, select Start Definition From: AP 32-tone. Click OK to the default settings and select Create WAV File in the final dialog.
Sorry... Not the most obvious process.
Attached are 1-second tones of the AP 32-tone test signal created in this fashion. One is in 16-bit format, the other 24-bit. I zip'ed the two files as DIYA doesn't accept WAV files directly. The sampling rate is 192 kHz.
For the thermal tests, I looped the sound bites to form a continuous test signal. The AP software does the looping automatically.
~Tom
@tomchr
I don't know what is wrong with you and what is a reason for such cynical writing?!
How anyone can have a DATA if they do not conduct an test with some method and one very trusty method I explained.
I only can guess that your writing is with intention to kick me off from the forum (and you did it) and probably strongly related to your Modulus-86 which for some reason have some idea VERY SIMILAR to BatoMM design which is done 2-3 years before yours!?
You can reply now whatever you want, I don't care ...
This is my final reply, I do not want to be in company with such people which only goals in life are profit.
It's a shame for other users of this forum who will not have a chance to read some useful advices, but they can come to my (noncommercial) forum and we can talks and exchange ideas without any pressure. Anyone interested can contact me on PM.
All the best and good luck!
I don't know what is wrong with you and what is a reason for such cynical writing?!
How anyone can have a DATA if they do not conduct an test with some method and one very trusty method I explained.
I only can guess that your writing is with intention to kick me off from the forum (and you did it) and probably strongly related to your Modulus-86 which for some reason have some idea VERY SIMILAR to BatoMM design which is done 2-3 years before yours!?
You can reply now whatever you want, I don't care ...
This is my final reply, I do not want to be in company with such people which only goals in life are profit.
It's a shame for other users of this forum who will not have a chance to read some useful advices, but they can come to my (noncommercial) forum and we can talks and exchange ideas without any pressure. Anyone interested can contact me on PM.
All the best and good luck!
Last edited:
Yeh, some people whom do not go by the book messes up the thought those who live by it. Going by the book is not a bad start. The ancient Chinese had ceased to innovate when they started to rely on only following the books. Sometimes some ideas may look crazy, but we just have to accept that some people tend to do things differently. Personally I like to understand the process more than just looking at data. Test methods tell you more if you really understand what is basically going on.
Was this a reply to my post? Anyway, I agree, the process or idea is where it starts - no doubt about that. And after the dust settles, it's the data that tell you whether the idea was worthwhile or just nonsense....
Often people throw up ideas and it is a shame that even they themselves are apparently not interested to know how good (or bad) it really is. What a waste!
Jan
Often people throw up ideas and it is a shame that even they themselves are apparently not interested to know how good (or bad) it really is. What a waste!
Jan
True, we finally might need to look at the data. For LM3886 application in the home, I have only found the need for to really work on heat sinks when you start giving it DC closer to the maximum. Not sure what the big fuss is about. Lots of LM3886 designs working out there will give a good indication what size works, I would spend valuable time elsewhere.
Last edited:
To even improve test results I would recommend to use distilled water, or even better, deionized water. With mentioned water and ice, very stable 0°C thermal point can be achieved.
Best Regards,
Aleš
The big fuss is about LM3886 and its thermal protection. When temperature of LM3886 rises, degradation in sound occurs, on the low end and dynamics. It is important to keep LM3886 temperature close to the ambient temperature, rather then having large ∆T between heatsink and ambient. How big this ∆T is I don't know, but when I was testing my PCB on a small heatsink, first few minutes the dynamics were very good, but after when the heatsink got warm/hot that sound appeared to be thin.
BR,
Aleš
Well if that is your concern, the very first step would be to chuck the isolated package and go back to the bare tab (are these still available?). You're not concerned per se with the heatsink temp, you're concerned with the chip die temp! And to get that as low as possible in a given situation,the bare tab does much better!
Jan
Wouldn't it be easier to just use the largest heat sink you feel you can accept? I still do not understand the big fuss? Unless you are going to really find out what temperature and power you will get unperceivable listening difference, and then the whole discussion goes into whether you want to do a double blind test protocol or not...
I have done that on my build and gone OTT on heatsinking, mainly because they heatsinks were cheap and I am activating my speakers so have space for them. The ribbons are 3.1Ohms DC so I am on the edge of the 3886 capability.
When looking at what one has to play with, your thetaJC is a given so you have thetaCS to gain you anything. My rough calculations showed you have under perfect conditions a 0.4C/W max gain so at 40W dissipation can drop junction temp 16C, but to get that is a lot of work and the net result is unlikely to be audibly louder. But a possibly interesting exercise when you want to push limits because you can. As I used to work on applications where thermal management was a huge issue I do get tempted to push the envelope for lols.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.