Hi Owen,
I felt like I was on the right track till I started poking the thermocouple around. Glad that I did though! The heatsink temps looked reasonable and I could have easily just moved on after checking it was all okay electrically.
U-channel sounds like the go, putting it so that the bottom of the U is down would mean no need to drill/tap the heatsink again (which is appealing). (I assume that is the correct way to do it anyway, becasue applying pressure to the small surface area of the upside down U would risk damaging the devices I'm guessing).
Thanks!
Chris
I felt like I was on the right track till I started poking the thermocouple around. Glad that I did though! The heatsink temps looked reasonable and I could have easily just moved on after checking it was all okay electrically.
U-channel sounds like the go, putting it so that the bottom of the U is down would mean no need to drill/tap the heatsink again (which is appealing). (I assume that is the correct way to do it anyway, becasue applying pressure to the small surface area of the upside down U would risk damaging the devices I'm guessing).
Thanks!
Chris
I tried some alu U-channel without the centre bolt, works pretty well. Temps of the u-channel bar are 5-10degC warmer than the heatsink is near the power devices. I decided that since I had it both amps apart to fit the U-channel, I had better add a hole for a centre bolt but I don't have a suitable length M3 bolt for the middle at hand. The one in the pic attached is only 10mm long which means it only gets a few turns into the heatsink and isn't enough to get much force.
I will get a 16mm bolt which will be the same length as the two M4 and then measure temps again with the centre bolt later this week.
I will get a 16mm bolt which will be the same length as the two M4 and then measure temps again with the centre bolt later this week.
Attachments
since the pressure applied is spread in clamping rather just being using a screw and worst scenarios i feel sometimes the contact near the pins is not good with screws so i use a washer rings with the screws to get the contact much superior.
http://www.boatcraft.com/images/ProdPics/s-s-fender-washer.jpg
sil pads? why dont you prefer the mica pads with heatsink compound? sil i believe is not a great thermal coupler..
Last edited:
The pink stuff !That's Keratherm in my picture above ...
Ho,
can you see why the two bolt clamping failed?
The analysis is not that difficult to think through and when the penny drops you will never repeat that type of error in any future mechanical assembly, not just electronics.
Last edited:
Yeah, the penny dropped quite early on, before I powered it up I had noticed the flex in the bar. I backed off the pressure to reduce the flex but I wasn't sure if I'd still get enough pressure in the middle. In hindsight I was getting almost no pressure in the middle even with the bar not visibly flexed. All of that is why I was checking temps while also checking the board was electrically okay.
I think things are pretty good with two bolts and the U-channel now since the delta between main heatsink and the bar is much closer now. I'll take a more controlled measurement to see if I can quantify the difference 3bolt vs 2bolt arrangement on the u-channel clamp. I expect thst since the u-channel is only 1.5mm thick T5 alu there may be some improvement left there. The end result should be better than the standard screw attachment method and thats the main goal! If I'm going to try for high (up to 40W per device) bias then I want to be sure I've got everything right before it leaves the workbench. Especially when the surface area of heat source is small wrt surface of heatsink (heatsink Rth will be calculated with even input of heat across the surface)
Cheers,
Chris
Last edited:
hochopper,
If you only replaced that U-channel extrusion with a square tube with holes drilled for the screw to pass through the first wall you will see an increase in pressure on those devices. The U-channel with such thin walls still allows the channel to spread open and relieve the pressure across the extrusion. Even with the same thin wall type of construction you will see a difference with square tubing . If you have a micrometer you can measure across the open section of the U-channel before and after applying the pressure and see how far the opposing walls open up, that will show what is happening.
If you only replaced that U-channel extrusion with a square tube with holes drilled for the screw to pass through the first wall you will see an increase in pressure on those devices. The U-channel with such thin walls still allows the channel to spread open and relieve the pressure across the extrusion. Even with the same thin wall type of construction you will see a difference with square tubing . If you have a micrometer you can measure across the open section of the U-channel before and after applying the pressure and see how far the opposing walls open up, that will show what is happening.
Third time lucky ...
I did a few informal tests and after 20 mins or so the U-channel with 3 bolts and the rectangular cross section tube both have pretty similar heatsink temps. But this looks much cooler 
With ~70W bias and ambient temp of 26degC I get a heatsink temp of around 54degC with both approaches. Probably average heatsink temp is a few degrees less than that.
To be honest I think I could get a bit more torque on the bolts but without a torque wrench I'm hesitant because I don't want to overtighten and strip the threads.
Thanks,
Chris
I did a few informal tests and after 20 mins or so the U-channel with 3 bolts and the rectangular cross section tube both have pretty similar heatsink temps. But this looks much cooler
With ~70W bias and ambient temp of 26degC I get a heatsink temp of around 54degC with both approaches. Probably average heatsink temp is a few degrees less than that.
To be honest I think I could get a bit more torque on the bolts but without a torque wrench I'm hesitant because I don't want to overtighten and strip the threads.
Thanks,
Chris
Attachments
Last edited:
Looks good!
That's a very healthy amount of bias if you're dissipating 70W! Are you running at 500mA or are they just very high rails?
It makes perfect sense that the U channel would perform the same as the box section. The added strength comes from the sidewalls, not from the top and bottom. You can easily verify this by grabbing a 1m length of 19mm x 1.5mm bar stock. Try bending it in the flat direction (probably bends under its own weight) and then try bending it lengthwise, which you will find is nearly impossible. Multiply this effect by 2 for a U channel, and that's why it works. Adding the top for a box section probably only increases strength a few percent at best. If you've already got it though, might as well keep it!
The only other thing I could think of would be swapping materials to something stronger like carbon steel, but you're definitely into the realm of diminishing returns at this point
Very curious to see how this all turns out! I love the attention to detail as it provides useful information that we can all draw on to improve our builds!
Cheers,
Owen
That's a very healthy amount of bias if you're dissipating 70W! Are you running at 500mA or are they just very high rails?
It makes perfect sense that the U channel would perform the same as the box section. The added strength comes from the sidewalls, not from the top and bottom. You can easily verify this by grabbing a 1m length of 19mm x 1.5mm bar stock. Try bending it in the flat direction (probably bends under its own weight) and then try bending it lengthwise, which you will find is nearly impossible. Multiply this effect by 2 for a U channel, and that's why it works. Adding the top for a box section probably only increases strength a few percent at best. If you've already got it though, might as well keep it!
The only other thing I could think of would be swapping materials to something stronger like carbon steel, but you're definitely into the realm of diminishing returns at this point
Very curious to see how this all turns out! I love the attention to detail as it provides useful information that we can all draw on to improve our builds!
Cheers,
Owen
I'm using the DPS-600 SMPS with 65V LME and 55V MOSFET rails. I did the test with a bench supply that maxes out at 30V so I was more interested in the power dissipation than the bias current at the time. I have just hooked the first on up to its DPS600 and set its bias to 350mA nothing let out any magic smoke, so I'll hook it up to a dummy load and DSO tomorrow after I cut the other channels leads to length and attach connectors for the SMPS. But I just remembered that I wanted to take them apart and solder down the LME heatsink now that everything is okay, I guess I'll do that in the morning too. We're out of town for the weekend so realistically I won't be listening to the amp until next week, but its getting close now!
In summer I am guessing that even with a couple of 140mm fans in the back of my equipment shelf I'm going to see air temps in there of ~35-40degC. The Denon receiver that I use runs pretty hot too. So in summer I'm working on a bias for less than 20deg rise in avg heatsink temp, winter I could go with 10 or 15 degC more heat. I know 350 to 400mA bias is pretty much point of diminishing returns for performance ...
The high rails on the SMPS are why I want to stuff around with the bias switching circuit that I mentioned, in our summer 60-70W just isn't a great idea for keeping things at sane temperatures.
I thought something like 300/350mA when the ambient temperature inside the amp chassis is above 30degC and 600mA otherwise should keep things within limits.
Cheers,
Chris
It makes perfect sense that the U channel would perform the same as the box section. The added strength comes from the sidewalls, not from the top and bottom. You can easily verify this by grabbing a 1m length of 19mm x 1.5mm bar stock. Try bending it in the flat direction (probably bends under its own weight) and then try bending it lengthwise, which you will find is nearly impossible. Multiply this effect by 2 for a U channel, and that's why it works. Adding the top for a box section probably only increases strength a few percent at best. If you've already got it though, might as well keep it!
The only other thing I could think of would be swapping materials to something stronger like carbon steel, but you're definitely into the realm of diminishing returns at this point
Forgot to mention ...
Yeah I bought 1m length of 3mm bar, 1.5mm U-channel and 2mm rectangular tube. I do think if I really ratched up the torque the U-channel would start to bend a little on the short axis which would mean the pressure wouldn't be even. The top of the rectangular tube keeps this in check. I don't think I'm quite getting to that point yet. There was some mention in EUVL's F5X chassis design article of higher pressure's helping too. My goal for now was to get the pressure even and in the right place, I'm pretty happy with the result.
I had a bit of fun looking at the power consumption of the board changing as I was watching the temperatures today too. I found that after I adjusted the bias to 54W at room temp then ran for 20mins to increase by a bit under 17deg or so the bias had dropped to 51.5 watts. 1.25W per voltage rail. I then adjusted to 70W over the next 15minutes the heatsink rose a further 8degC and the consumption dropped by almost 1W. (0.5W per voltage rail). I'll be sure to run the amps for 20mins or so ... that should help me sleep well at night.
Cheers,
Chris
This is the beauty of lateral fets! No thermal bias compensation and no emitter resistors and they provide a perfect mild negative temperature coefficient all on their own! It's a pretty sweet deal.
Accomplishing the above with a temp compensated bias circuit takes a lot of trial and error, and even then it would not retain as good a thermal tracking profile as the lateral fets can do naturally. The lateral parts will only creep into positive temp co at very low bias.
Cheers,
Owen
^^ that was meant to read:
I'll be sure to run the amps for 20mins or so before doing the final bias adjustment ... that should help me sleep well at night.
A few finishing touches left for next week after I go out of town this week. But the internal wiring is now finished. Need to trouble shoot why one of the LME ICs isn't getting power (mostly likely culprit is my connector for the DPS600 pin header output).
Attachments
Hochopeper,
Nice work! I'm considering building the LM. Can I ask you why you put in two LPUHP headphone amps in your system? At least, based on what you said, these are what they are.
For the switch driver? Why did you use such a design instead of going with just a switch driver like AMB does (AMB.org).
Thanks.
Bertrand.
Nice work! I'm considering building the LM. Can I ask you why you put in two LPUHP headphone amps in your system? At least, based on what you said, these are what they are.
For the switch driver? Why did you use such a design instead of going with just a switch driver like AMB does (AMB.org).
Thanks.
Bertrand.
Thanks,
The LPUHP is a 15W speaker amp. I'm building it as a 4ch amp for my active speakers and I will hook this up to the speakers using 4 pole speakon connectors. The LPUHP will power the B&C DE250 and the LME49830 amps will power a pair of Peerless 8in nomex woofers each side. Have you seen the LPUHP thread - http://www.diyaudio.com/forums/soli...gh-perfromance-lpuhp-16w-power-amplifier.html - there are some discussions from people thinking about using them for headphones but for most cases Owen's BAL-BAL headphone amp is an equivalent choice (and my personal favorite).
The switch driver circuit does a few extra things, DC offset protection (disconnect mains power on speaker amp going DC) and has a 12V trigger input as well as allowing me to avoid the need to run mains wires to the front panel switch.
Chris