Please no mention of, "Rap" in conjunction with this lovely amplifier.
Amen
CPU heat sinks
Somebody (Hi Tom!) sent me a PM asking about the Mac G5 heat sinks I am repurposing, and if a fan is needed -- am responding here just in case somebody else might be interested.
Many others have been using CPU heat sinks, there are posts about this all over the fora, including @xrk971 who showed some quite inspiring pictures, and many others.
It works brilliantly! Actually, much better than what I expected. So well indeed that I just ordered 2 more used G5 CPU modules with heatsinks on eBay -- for $4.98 a piece, can you believe that! (the shipping wasn't free though...). I'll build the Aleph J next, and this way I don't even have to mount/unmount my output "engine", which was the original plan, for less than $30!
Re/ the fan, it turns out, just having a low-RPM (inaudible) fan blowing into the general direction of the heat sink is good enough to get rid of the heat. This is obviously just for "demonstration" purposes. I built something slightly more sophisticated, with an Arduino/FWM-fans/Thermistor/relay-kind of contraption (am happy to share schematics and software, it's running on its own 12V power, no noise injected whatsoever).
To see what's going on, here's a graph/snapshot that shows how easy it is to keep the temperature at bay -- a SparkFun proMicro running a PID regulator (some Arduino-based software I adapted), read out by a Raspberry Pi and a little python monitoring script...obviously the Arduino can also just run on its own. The ups-and-downs were me playing around with positioning the fan(s) and thermistor etc.
I'm measuring right on the body of one of the MOSFETs (1 sink with the 2 MOSFETS per channel), and with the fan positioned just somewhere near the heat sink (see picture of my test listening setup, I moved the fan away to make the point, and currently only channel A exists), creating a slight breeze, the MOSFET body temperature (dark-blue curve) stays around 55C (and it's easy to get it below 50C by blowing right on to the sink). The orange curve shows the PID software sometimes ramping up the fans -- the green curve is just a blow-up of the temperature curve.
I have a few more fans when this all gets moved into the housing, and I don't expect any thermal problems.
Also, these fans at low-RPM are really quiet, impossible to hear from the listening position.
Somebody (Hi Tom!) sent me a PM asking about the Mac G5 heat sinks I am repurposing, and if a fan is needed -- am responding here just in case somebody else might be interested.
Many others have been using CPU heat sinks, there are posts about this all over the fora, including @xrk971 who showed some quite inspiring pictures, and many others.
It works brilliantly! Actually, much better than what I expected. So well indeed that I just ordered 2 more used G5 CPU modules with heatsinks on eBay -- for $4.98 a piece, can you believe that! (the shipping wasn't free though...). I'll build the Aleph J next, and this way I don't even have to mount/unmount my output "engine", which was the original plan, for less than $30!
Re/ the fan, it turns out, just having a low-RPM (inaudible) fan blowing into the general direction of the heat sink is good enough to get rid of the heat. This is obviously just for "demonstration" purposes. I built something slightly more sophisticated, with an Arduino/FWM-fans/Thermistor/relay-kind of contraption (am happy to share schematics and software, it's running on its own 12V power, no noise injected whatsoever).
To see what's going on, here's a graph/snapshot that shows how easy it is to keep the temperature at bay -- a SparkFun proMicro running a PID regulator (some Arduino-based software I adapted), read out by a Raspberry Pi and a little python monitoring script...obviously the Arduino can also just run on its own. The ups-and-downs were me playing around with positioning the fan(s) and thermistor etc.
I'm measuring right on the body of one of the MOSFETs (1 sink with the 2 MOSFETS per channel), and with the fan positioned just somewhere near the heat sink (see picture of my test listening setup, I moved the fan away to make the point, and currently only channel A exists), creating a slight breeze, the MOSFET body temperature (dark-blue curve) stays around 55C (and it's easy to get it below 50C by blowing right on to the sink). The orange curve shows the PID software sometimes ramping up the fans -- the green curve is just a blow-up of the temperature curve.
I have a few more fans when this all gets moved into the housing, and I don't expect any thermal problems.
Also, these fans at low-RPM are really quiet, impossible to hear from the listening position.
Attachments
G5 heatsink
This is very interesting.
Do you think a single heat sink per MOSFET would enable you to run without any fan support?
I saw some of the posts others have made on the use of CPU heatsinks and fans, but to be honest, I am not wild about using fans, regardless of how slow or fast, or how compact the packaging.
These things are kind of interesting, in that you can conjure an M2X amp that has a very industrial vibe to it.
Appreciated the pix, especially the one with the M2X board laying atop the two heat sinks. Gives a pretty good idea of size and packaging parameters.
Where did you get the black, finned plates that screw down over the MOSFETs?
This is very interesting.
Do you think a single heat sink per MOSFET would enable you to run without any fan support?
I saw some of the posts others have made on the use of CPU heatsinks and fans, but to be honest, I am not wild about using fans, regardless of how slow or fast, or how compact the packaging.
These things are kind of interesting, in that you can conjure an M2X amp that has a very industrial vibe to it.
Appreciated the pix, especially the one with the M2X board laying atop the two heat sinks. Gives a pretty good idea of size and packaging parameters.
Where did you get the black, finned plates that screw down over the MOSFETs?
Last edited:
Almost all of the expense when building an M2x, is in the chassis, the heatsinks, the power supply, and the gold plated RCA conenctors + gold plated speaker terminals. The electronic components are relatively few and relatively cheap.
Thus it's not an especially costly "failure" if you perform an experiment which cooks your M2x amp channel boards with excess heat. It's particularly cheap if the only destroyed parts are the output MOSFETs: considerably less than USD 20 total.
The probability that you'll injure the input stage daughter cards through inadequate heatsinking, is negligibly small in my opinion. That's important if you've got $50 worth of genuine Toshiba JFETs in a pair of Ishikawas, or $200 worth of labor in Norwoods that you paid somebody else to solder cuz you're intimidated by surface mount parts.
Thus it's not an especially costly "failure" if you perform an experiment which cooks your M2x amp channel boards with excess heat. It's particularly cheap if the only destroyed parts are the output MOSFETs: considerably less than USD 20 total.
The probability that you'll injure the input stage daughter cards through inadequate heatsinking, is negligibly small in my opinion. That's important if you've got $50 worth of genuine Toshiba JFETs in a pair of Ishikawas, or $200 worth of labor in Norwoods that you paid somebody else to solder cuz you're intimidated by surface mount parts.
This is very interesting, LATB. Please share the schematics and software. I like using MCU's and if you've sorted out the PID's, even better.
.
Looks like I'm being volunteered to do some experiments I have indeed a few replacement parts lying around, at ~$2.50 a piece that's no big deal.
I doubt that one can get away using these heatsinks without having something to move air around, though. Maybe there is enough convection if I arrange them vertically, but I can't imagine this to work in an enclosure. I'll play a bit over the weekend.
The black plates that hold the MOSFETs down are small heatsinks, Mouser 567-527-45AB (which happen to be out of stock, but Mouser 567-527-24AB should also work). I didn't have the tools to make nice holes through the fins (end mill?). The mechanics of these coolers are quite nice, the screws go into some spring loaded nuts that exert just the right amount of pressure on the MOSFETs to hold them in place.
I have the feeling that Kapton or Mica and paste would work much better than the Sil-pads I'm currently using (I really don't want to use paste though). The temp gradient to the copper insert in the sink is a bit high, which is ok as long as one blows some air on top of the MOSFETs.
One nice advantage (maybe more important for the F6 which has no real bias regulation circuitry): this system goes into thermal equilibrium within minutes, and then stays where it is. No waiting for everything to get nice and warm before the music starts to flow -- it's all right there a few moments after you switch on.
I have indeed a few replacement parts lying around, at ~$2.50 a piece that's no big deal.I doubt that one can get away using these heatsinks without having something to move air around, though. Maybe there is enough convection if I arrange them vertically, but I can't imagine this to work in an enclosure. I'll play a bit over the weekend.
The black plates that hold the MOSFETs down are small heatsinks, Mouser 567-527-45AB (which happen to be out of stock, but Mouser 567-527-24AB should also work). I didn't have the tools to make nice holes through the fins (end mill?). The mechanics of these coolers are quite nice, the screws go into some spring loaded nuts that exert just the right amount of pressure on the MOSFETs to hold them in place.
I have the feeling that Kapton or Mica and paste would work much better than the Sil-pads I'm currently using (I really don't want to use paste though). The temp gradient to the copper insert in the sink is a bit high, which is ok as long as one blows some air on top of the MOSFETs.
One nice advantage (maybe more important for the F6 which has no real bias regulation circuitry): this system goes into thermal equilibrium within minutes, and then stays where it is. No waiting for everything to get nice and warm before the music starts to flow -- it's all right there a few moments after you switch on.
For those who prefer no-Arduino, no-PWM, no-software,
but still want a temperature controlled, variable speed fan using thermistor sensor:
Fan inside audio chassis: variable speed, temperature controlled, analog. No PWM.
but still want a temperature controlled, variable speed fan using thermistor sensor:
Fan inside audio chassis: variable speed, temperature controlled, analog. No PWM.
G5 heatsink
LATB:
I certainly wasn't trying to volunteer you to experiment! I am merely thinking out loud about how these might behave if cooling one MOSFET per heatsink. From the look of things, I suspect it ought to do pretty well, just judging from the massive size of these, and the amassed cooling area with two of them, especially within the context of the standard approach, of a 4U chassis, or even using a 3U chassis. I would guess the cooling coefficient of both combined can't be too far off the pace.
I am thinking of using of these one per MOSFET so that I can go passive cooling. The issue is next how to package this so that it is visually acceptable and not a FrankenM2x. I personally don't need something with a specific aesthetic, but I do need to preserve domestic peace... I think that can be done, just need more thought. In my case, those heatsinks would be outside of any enclosure for the boards. And I think it may be possible to make this surprisingly visually engaging.
Thanks for the mouser part numbers and tips!
LATB:
I certainly wasn't trying to volunteer you to experiment! I am merely thinking out loud about how these might behave if cooling one MOSFET per heatsink. From the look of things, I suspect it ought to do pretty well, just judging from the massive size of these, and the amassed cooling area with two of them, especially within the context of the standard approach, of a 4U chassis, or even using a 3U chassis. I would guess the cooling coefficient of both combined can't be too far off the pace.
I am thinking of using of these one per MOSFET so that I can go passive cooling. The issue is next how to package this so that it is visually acceptable and not a FrankenM2x. I personally don't need something with a specific aesthetic, but I do need to preserve domestic peace... I think that can be done, just need more thought. In my case, those heatsinks would be outside of any enclosure for the boards. And I think it may be possible to make this surprisingly visually engaging.
Thanks for the mouser part numbers and tips!
My day job is in experimental physics, so I'm fine with trying and breaking things ;-)
I think the main issue is to get good heat transfer to the cooler, and I don't think Sil-pads would cut it if you don't use a fan.
With Iq of 1.4A you should be ready to remove 35W in idle, per MOSFET. I think under full load the max power per MOSFET should be 70W-80W?
The IRFP240 specs say
- Maximum Junction-to-Case (Drain) 0.83 degC/W
- Thermal resistance ratings Case-to-Sink, Flat, Greased Surface 0.24 degC/W
Without grease and Sil-pads I should have at least 0.5 C/W heat resistance, that's just a guess though. That would mean a temperature difference of less than 50 degC between the junction and the copper heat pipes, at 35W in idle -- still, that could go up to 100 degC at full load.
However, I'm mostly happy just using the First Watt mostly staying around idle, and as long as I measure something below maybe 80deg on top of the MOSFET cases, I should be fine.
In my case, with both MOSFETs on one sink, using Sil-pads, my measurement show that at idle (dissipating ~70W), with fans running at full, the sink body feels cold to the touch (20 deg room temperature) -- so I'm removing all the heat -- while the top of the MOSFETs are still at around 60 deg. Then delta-T is ~40deg, and at 70W that's a thermal resistance between MOSFET body and the heat fins of ~0.6C/W.
Allowing some of the airflow to go to the MOSFET bodies immediately improves the situation and the fans can go to the lowest setting for a body temp of <55deg.
In summary, I can blow as much air through the heat sink as I want, the MOSFET bodies will stay at around 60 deg. I'll measure what happens if I shut off the fans, as long as the top plate stays below 80 deg, and will show the result here.
In principle the specs for these sinks should be available somewhere, but googling it seems Apple has removed all the tech specs for the old G5, possibly out of embarrassment... They were clearly power hogs, people measured 170W idle and 400W under load. Back then that was a lot for a desktop machine.
I think the main issue is to get good heat transfer to the cooler, and I don't think Sil-pads would cut it if you don't use a fan.
With Iq of 1.4A you should be ready to remove 35W in idle, per MOSFET. I think under full load the max power per MOSFET should be 70W-80W?
The IRFP240 specs say
- Maximum Junction-to-Case (Drain) 0.83 degC/W
- Thermal resistance ratings Case-to-Sink, Flat, Greased Surface 0.24 degC/W
Without grease and Sil-pads I should have at least 0.5 C/W heat resistance, that's just a guess though. That would mean a temperature difference of less than 50 degC between the junction and the copper heat pipes, at 35W in idle -- still, that could go up to 100 degC at full load.
However, I'm mostly happy just using the First Watt
mostly staying around idle, and as long as I measure something below maybe 80deg on top of the MOSFET cases, I should be fine. In my case, with both MOSFETs on one sink, using Sil-pads, my measurement show that at idle (dissipating ~70W), with fans running at full, the sink body feels cold to the touch (20 deg room temperature) -- so I'm removing all the heat -- while the top of the MOSFETs are still at around 60 deg. Then delta-T is ~40deg, and at 70W that's a thermal resistance between MOSFET body and the heat fins of ~0.6C/W.
Allowing some of the airflow to go to the MOSFET bodies immediately improves the situation and the fans can go to the lowest setting for a body temp of <55deg.
In summary, I can blow as much air through the heat sink as I want, the MOSFET bodies will stay at around 60 deg. I'll measure what happens if I shut off the fans, as long as the top plate stays below 80 deg, and will show the result here.
In principle the specs for these sinks should be available somewhere, but googling it seems Apple has removed all the tech specs for the old G5, possibly out of embarrassment... They were clearly power hogs, people measured 170W idle and 400W under load. Back then that was a lot for a desktop machine.
Mustn't forget Papa's fan controller.
Attachments
G5 heatsink
G5 data are hard to come by.
There IS an outdated summary on Apple's site, but I didn't download it, as it wasn't terribly useful for my needs here: I was looking for the physical dimensions of these heatsinks (good luck finding that on the web), and the thermal resistance specs. The page DID have power dissipations for various CPUs used in the G5, though. Low to mid 100's of watts, AFAIR.
Your latest post offered pretty decent data from which some basic heat transfer estimates can be made. I will be very interested in the results of the measurements you outlined!
G5 data are hard to come by.
There IS an outdated summary on Apple's site, but I didn't download it, as it wasn't terribly useful for my needs here: I was looking for the physical dimensions of these heatsinks (good luck finding that on the web), and the thermal resistance specs. The page DID have power dissipations for various CPUs used in the G5, though. Low to mid 100's of watts, AFAIR.
Your latest post offered pretty decent data from which some basic heat transfer estimates can be made. I will be very interested in the results of the measurements you outlined!
Mark, thanks for posting the link the fan speed controller circuit. I may have a need for that.
Others: Looks aside, the CPU style force-air heatsinks are way overkill for the M2x. Mine gets by just fine with a relatively small 2.5U, 400mm deep chassis. The output devices will run at +/– 22.5V with the recommended transformer of 18V secondaries and conventional bridge rectifiers. Using LT4320-based rectifiers will increase that to about +/– 23.4V, which is as high as some of the front-end circuits can stand for long term reliability. Given a bias current of about 1.3A, which is set by the opto-coupler circuit independently of the rail voltage, total power dissipation isn't very much.
The main issue with the M2x is maintaining sufficient physical separation between the power transformer(s) and the Edcor signal transformers. To that end, a 400mm deep chassis is recommended.
Others: Looks aside, the CPU style force-air heatsinks are way overkill for the M2x. Mine gets by just fine with a relatively small 2.5U, 400mm deep chassis. The output devices will run at +/– 22.5V with the recommended transformer of 18V secondaries and conventional bridge rectifiers. Using LT4320-based rectifiers will increase that to about +/– 23.4V, which is as high as some of the front-end circuits can stand for long term reliability. Given a bias current of about 1.3A, which is set by the opto-coupler circuit independently of the rail voltage, total power dissipation isn't very much.
The main issue with the M2x is maintaining sufficient physical separation between the power transformer(s) and the Edcor signal transformers. To that end, a 400mm deep chassis is recommended.
I agree -- Mark's or even Nelson's fan control would be completely sufficient, and going with the "standard" chassis is the obvious way forward.
This other stuff for me is just about the fun of re-using some old and pretty cool looking stuff laying around in the house in my endeavor to understand and experience Nelson's ideas and genius and fun designs, and as Mark pointed out, it doesn't even cost much to try out all these different designs!
The other intriguing thing for me is moving the whole output stage to this weird looking contraption, that modulates the current on its way from the power supply to the speakers, that's almost it. It's like adding this little "music listening engine" to my large and very hefty power supply (I'll show pictures some time), this engine just made of a couple of MOSFETS and 4 or 6 resistors -- genius in its simplicity -- almost a material manifestation of this idea that you're "listening to the power supply". What a fun and easy way of thinking about a power amp! And it sounds great, too!
I mean, this is genius, and I really appreciate Nelson and all of you sharing these idea, and I'm have a lot of fun exploring. Thank you!!!
This other stuff for me is just about the fun of re-using some old and pretty cool looking stuff laying around in the house in my endeavor to understand and experience Nelson's ideas and genius and fun designs, and as Mark pointed out, it doesn't even cost much to try out all these different designs!
The other intriguing thing for me is moving the whole output stage to this weird looking contraption, that modulates the current on its way from the power supply to the speakers, that's almost it. It's like adding this little "music listening engine" to my large and very hefty power supply (I'll show pictures some time), this engine just made of a couple of MOSFETS and 4 or 6 resistors -- genius in its simplicity -- almost a material manifestation of this idea that you're "listening to the power supply". What a fun and easy way of thinking about a power amp! And it sounds great, too!
I mean, this is genius, and I really appreciate Nelson and all of you sharing these idea, and I'm have a lot of fun exploring. Thank you!!!
I disassembled my old Apple PowerMac G5 Dual 1.8GHz (circa 2004). The two other ones I ordered from eBay (for my future Aleph J built), which seem to look exactly the same, are these (only one was left this morning, but there are other offers): Apple PowerMac G5 630-6606 630-6422 630-4957 1.8GHz Processor with Heatsink.
There are a ton of others that would work, too. I think @xrk971 used some old Dell workstation coolers, etc.