Kapton W/m•K is not good.
Ceramic is very good.
Kapton HN PA, aka Polyimide tape or film
0.025mm
0.16W/m•K Thermal conductivity
Kapton MT
0.46W/m•K Thermal conductivity
Aavid 4180G Aluminium Oxide Ceramic thermal pad
15.06W/m•K @ 75°C Thermal conductivity.
However, are those W/m•K ratings for one meter of the stuff or specific to the thickness of each different pad? That Kapton is particularly thin at 0.025mm??
If indeed expressed as in your examples, changing that unit (thermal conductance) to include the thickness of the material is easy:
Divide the TC by the thickness in [m]. You will get a new (standard) unit [W/m2K] which tels you the thermal transfer (i actually use this unit a lot (construction technics in house building, heating energy calculations)). The bigger the number, the better (but the smaller the better in house construction).
F.ex: The 1mm thick AluOxide pad 15/0.001=15kW/m2K
0.025mm Kapton, 6.4kW/m2K
However, I think the thermal unit of the pad is usually expressed as a thermal resistance [K/W] (or W/K (which of course is 1/(K/W)). This is usable and comparable as is. You probably want as small (thermal) resistance as possible...
By that reckoning:
Keratherm Red
Thickness: 0.25mm.
86/82:
6.5W/m•K Thermal conductivity.
0.09°C/W Thermal resistance Rth at clamping force of 120 N/cm².
6.5 / 0.00025 = 26kW/m•K.
That's far and away the best that I've found so far.
Look like I can get a 100mm x100mm sheet from Rapid (Conrad). Part number: 54-4246 about £11.00 + post which will make about 40 pads for these LME chips.
Works out a bit expensive if I only need a few though. First I'll try it with something else to see how warm the LME is on the 1.9°C/W sink.
Otherwise, are there any Keratherm Red pads still available from earlier is this thread or elsewhere in small quantities?
Keratherm Red
Thickness: 0.25mm.
86/82:
6.5W/m•K Thermal conductivity.
0.09°C/W Thermal resistance Rth at clamping force of 120 N/cm².
6.5 / 0.00025 = 26kW/m•K.
That's far and away the best that I've found so far.
Look like I can get a 100mm x100mm sheet from Rapid (Conrad). Part number: 54-4246 about £11.00 + post which will make about 40 pads for these LME chips.
Works out a bit expensive if I only need a few though. First I'll try it with something else to see how warm the LME is on the 1.9°C/W sink.
Otherwise, are there any Keratherm Red pads still available from earlier is this thread or elsewhere in small quantities?
Even if it was free post, at £12.30 for the ten, cut into 20 pads of 11.5mm x 19mm pads, with some over that'll probably be wasted, it's about £0.64 per pad. £1.07 per pad with normal delivery.
Vs. £0.43 per pad from Conrad, including VAT and delivery.
Or, if I had a lot more to make, the 190mm x 190mm sheet is about twice the price for nearly four times the area, so it's nearer £0.20 per pad, inc VAT & del.
Vs. £0.43 per pad from Conrad, including VAT and delivery.
Or, if I had a lot more to make, the 190mm x 190mm sheet is about twice the price for nearly four times the area, so it's nearer £0.20 per pad, inc VAT & del.
After a few years with the parts for these amps I've finally started to stop stuffing around aimlessly and finish a few of these projects where I've got all the parts already.
Here are a few pics of where I'm up to so far (see attached)
2 x LME49830 amps with DPS600 SMPS, 2 x LPUHP v1, aussie amps mains controller in the middle
When I got everything back out after 12months on the shelf I decided to change direction with layout and that's left me with a bit of a messy back panel unfortunately.
Chris
Here are a few pics of where I'm up to so far (see attached)
2 x LME49830 amps with DPS600 SMPS, 2 x LPUHP v1, aussie amps mains controller in the middle
When I got everything back out after 12months on the shelf I decided to change direction with layout and that's left me with a bit of a messy back panel unfortunately.
Chris
Attachments
Hey Owen,
How much noise would a DC/DC buck converter create? I'm more worried about radiated noise than conducted, I'll run it off the LME supply so the PSRR should help there. A linear reg dropping from 65V to 5V is too inefficient so I'm looking at other options. Thinking about supply options for that stupid bias switching board I had in mind a long while back.
Chris
How much noise would a DC/DC buck converter create? I'm more worried about radiated noise than conducted, I'll run it off the LME supply so the PSRR should help there. A linear reg dropping from 65V to 5V is too inefficient so I'm looking at other options. Thinking about supply options for that stupid bias switching board I had in mind a long while back.
Chris
Owen tell me when this gets too far OT okay?
I tried to simulate tonight one of the buck converters that I picked out. Just to see how it looks. I'm very green at anything SPICE, never really used it for more than about 20mins previously. I found a model on LT website for the converter I'd picked and adjusted it a little to suit my load and output voltage requirements and input voltage.
Came up with the attached. Is probably a bit oversimplified to make any real conclusion right?
I need to finish drawing up the circuit for control but it's pretty basic. 555 timer delay to allow startup of amp and it's power supply under very low bias current - after say 2 or 3 seconds - then switch relay to add/remove parallel resistor for the preferred bias current level with an Omron G6K relay (I've got a bag of 100 of them so that's how I made that choice). I can also use the second contact in the relay to start the amp muted but this might not be necessary with this amp.
I'm not sure about adding LM335, comparator and a 2nd relay for a temperature adjusted bias level - to select a lower bias as ambient temp increases. Adds a bit of complexity for not a huge gain.
Thoughts?
I tried to simulate tonight one of the buck converters that I picked out. Just to see how it looks. I'm very green at anything SPICE, never really used it for more than about 20mins previously. I found a model on LT website for the converter I'd picked and adjusted it a little to suit my load and output voltage requirements and input voltage.
Came up with the attached. Is probably a bit oversimplified to make any real conclusion right?
I need to finish drawing up the circuit for control but it's pretty basic. 555 timer delay to allow startup of amp and it's power supply under very low bias current - after say 2 or 3 seconds - then switch relay to add/remove parallel resistor for the preferred bias current level with an Omron G6K relay (I've got a bag of 100 of them so that's how I made that choice). I can also use the second contact in the relay to start the amp muted but this might not be necessary with this amp.
I'm not sure about adding LM335, comparator and a 2nd relay for a temperature adjusted bias level - to select a lower bias as ambient temp increases. Adds a bit of complexity for not a huge gain.
Thoughts?
Attachments
Last edited:
I did a LME49830 with the same bias and the output stage but im getting very problematic bias issue.. can you please guide me what has actually went wrong in this?
https://dl.dropboxusercontent.com/u/7421969/lme49830 Lmos.png
The bias increases as the psu voltage increases as low as 2V of psu voltage I see .5Amp bias and even with the bias pot tried at its min and max the bias current through the fuse is so high and i see displacement on the cone showing the DC offset. Its just very frustrating can anybody guide me if anything might have gone wrong here..
https://dl.dropboxusercontent.com/u/7421969/lme49830 Lmos.png
The bias increases as the psu voltage increases as low as 2V of psu voltage I see .5Amp bias and even with the bias pot tried at its min and max the bias current through the fuse is so high and i see displacement on the cone showing the DC offset. Its just very frustrating can anybody guide me if anything might have gone wrong here..
I had the second channel from my amp build on the bench today giving it its first taste of electricity. All went well, electrically.
Since I'm going to be trying to go for quite high bias (especially if I put together the bias switching idea I'm working on) I thought I'd poke around and check some temperatures.
I have done some calcs on heatsink temperatures and whatnot when checking what bias I will be able to handle. I've checked my calcs against what was done in this article - http://linearaudionet.solide-ict.nl/sites/linearaudio.net/files/Didden LA V3 PK lr.pdf - by EUVL explaining the modelling/calcs done in the design for the F5X enclosure design.
Everything seems above board on paper, but I wanted to check with real world data.
My concern now is how I physically attach the output devices to the heatsink. In that article I got the idea of using a clamp rather than a bolt through the hole on the device due to the package requiring more pressure further down the chip. On re-reading it today it seems to be more along the lines of being in addition to the usual method. See attached for how I've got it at the present.
I'm currently seeing ~60degC temps for the clamping bar and ~45degC temps for the heatsink when sitting at about 60W idle dissipation. I'm assuming that inside a case this will be worse. I presume this means that I'm not getting ideal coupling of the device to the heatsink through the keratherm red.
So, should I:
1. drill two more holes in the bar and two more into the heatsink and tap those for 4 bolts through the bar into heatsink; or
2. discard the bar and drill two more holes, mount the output devices normally and get on with life.
Thoughts?
Edit: I previously had just a split washer between the bolt and bar, I think things are better after adding the flat washer as well but the numbers above are with the arrangement pictured.
Since I'm going to be trying to go for quite high bias (especially if I put together the bias switching idea I'm working on) I thought I'd poke around and check some temperatures.
I have done some calcs on heatsink temperatures and whatnot when checking what bias I will be able to handle. I've checked my calcs against what was done in this article - http://linearaudionet.solide-ict.nl/sites/linearaudio.net/files/Didden LA V3 PK lr.pdf - by EUVL explaining the modelling/calcs done in the design for the F5X enclosure design.
Everything seems above board on paper, but I wanted to check with real world data.
My concern now is how I physically attach the output devices to the heatsink. In that article I got the idea of using a clamp rather than a bolt through the hole on the device due to the package requiring more pressure further down the chip. On re-reading it today it seems to be more along the lines of being in addition to the usual method. See attached for how I've got it at the present.
I'm currently seeing ~60degC temps for the clamping bar and ~45degC temps for the heatsink when sitting at about 60W idle dissipation. I'm assuming that inside a case this will be worse. I presume this means that I'm not getting ideal coupling of the device to the heatsink through the keratherm red.
So, should I:
1. drill two more holes in the bar and two more into the heatsink and tap those for 4 bolts through the bar into heatsink; or
2. discard the bar and drill two more holes, mount the output devices normally and get on with life.
Thoughts?
Edit: I previously had just a split washer between the bolt and bar, I think things are better after adding the flat washer as well but the numbers above are with the arrangement pictured.
Attachments
Hi Chris,
I'd consider the following three options:
1. Spread the fets apart just enough to get another screw in between them
2. lose the bar stock and use U channel instead. It's much stronger, and won't flex as much in the middle.
3. Drill two more holes in the bar stock and run bolts through the bar stock and the fet holes for more clamping force in the middle of the bracket.
I certainly agree with the mounting methodology overall. It's much better to have the force applied directly over the die as opposed to the screw hole above the die. Paradigm did this in all their amplifiers with steel U-channel, and the thermal coupling with good sil pads was exceptional.
Regards,
Owen
I'd consider the following three options:
1. Spread the fets apart just enough to get another screw in between them
2. lose the bar stock and use U channel instead. It's much stronger, and won't flex as much in the middle.
3. Drill two more holes in the bar stock and run bolts through the bar stock and the fet holes for more clamping force in the middle of the bracket.
I certainly agree with the mounting methodology overall. It's much better to have the force applied directly over the die as opposed to the screw hole above the die. Paradigm did this in all their amplifiers with steel U-channel, and the thermal coupling with good sil pads was exceptional.
Regards,
Owen