what is the surface area?
i.e. are fins oriented along the 4 inch length or along the 12 inch direction?
i.e. are fins oriented along the 4 inch length or along the 12 inch direction?
The fins are oriented lengthwise and there are 16 or them. The fins are evenly spaced and about 1/8" thick and 3/4" long. The base of the heatsink is about 1/4" thick. For a total size of 12" long x 4" wide 1" deep.
The last time I measured after running at high volume for six hours the temp on the heatsink was 48 deg C. Thats cooler than my CPU runs usually. 🙂 I use the amp in a home theater where it drives the FL/FR/SL/SR/SRL/SRR channels. My center channel is self powered with an additional LM4780 and an Avel 200VA toroid. My sub has its own 500watt plateamp. All of the speakers are DIY "Cryolites" as seen here Cryolite
The sub is a 15" Dayton Quattro in a 3.6 ft3 sealed enclosure. All of this is driven by the preouts from my Denon 3805. Sounds really good. 🙂
The sub is a 15" Dayton Quattro in a 3.6 ft3 sealed enclosure. All of this is driven by the preouts from my Denon 3805. Sounds really good. 🙂
GCSS operates in a "bridged" format. Thus a 8 ohm impedance is really like a 4 ohm to each package. Keep this in mind. Certainly you can operate the package at lower voltages but I believe you need to keep this above 20V for optimal biasing and apparently the packages sound better at higher voltages, but I could be proven wrong. This points to using two separate packages to keep power desnsity low and minimize the SPIKE protection kicking in during transients.
What I would be cautious about is the power density and the ability to transfer that heat out of the package in a very short time before the SPIKE protection kicks in. National clearly brings this up. I believe if one were to use a 4780, a copper heatsink surface might actually be technically beneficial.
You can't really judge the actual inside temp of the package by touching the heatsink. A short transient can heat up the package and trigger SPIKE without the heatsink even getting warm. You need to carefully look at the junction thermal resistances. There was a discussion on Plastic Power Trans vs Metal TO3 packgaes that might prove interesting.
What I would be cautious about is the power density and the ability to transfer that heat out of the package in a very short time before the SPIKE protection kicks in. National clearly brings this up. I believe if one were to use a 4780, a copper heatsink surface might actually be technically beneficial.
You can't really judge the actual inside temp of the package by touching the heatsink. A short transient can heat up the package and trigger SPIKE without the heatsink even getting warm. You need to carefully look at the junction thermal resistances. There was a discussion on Plastic Power Trans vs Metal TO3 packgaes that might prove interesting.
But you would not be bridging the channels in one chip. You would be parraleling them so one channel of SUSYGC would use two LM4780. For a given load the two LM4780s should dispate heat better than 2 LM3886s since the surface area of the metal contact surface of the LM4780 is larger. You might get better results with 4 LM3886s but I doubt by much.
My brother Mark and I have run the LM3886s and LM4780s very hard, and with an oscilliscope the only time I have seen SPIKE protection kick in is when we artificially heated a LM4780 chip with a heat gun 😀. Of course I have never run them bridged. The SPIKE temperature is very high indeed.
My brother Mark and I have run the LM3886s and LM4780s very hard, and with an oscilliscope the only time I have seen SPIKE protection kick in is when we artificially heated a LM4780 chip with a heat gun 😀. Of course I have never run them bridged. The SPIKE temperature is very high indeed.
OK, some lively discussion happening, all to the good!
For those not entirely familiar with the +/- of the different approaches, here is some data to consider:
LM3875 Specs
56W continuous power into 8 ohms
100W instantaneous peak output power
6 A output current limit
120dB open loop gain.
Amplifier circuit would require 2 chips per channel, and be capable of 100W into 8ohm, 72W into 4ohm, and 36W into 2ohm
LM3886 Specs
68W continuous power into 4 ohms (38W into 8ohm)
135W instantaneous peak output power
11.5 A output current limit
115dB open loop gain.
Amplifier circuit would require 2 chips per channel, and be capable of 75W into 8ohm, 135W into 4ohm, and 130W into 2ohm
LM1875 Specs
30W continuous power into 4 ohms (38W into 8ohm)
Instantaneous peak output power unspecified.
4 A output current limit
90dB open loop gain.
Amplifier circuit would require 4 chips per channel, and be capable of 150W into 8ohm, 120W into 4ohm, and 64W into 2ohm
LM4780 Specs
60W continuous power into 8 ohms (per channel)
instantaneous peak output power unspecified
11.5 A output current limit (per channel)
115dB open loop gain.
Amplifier circuit would require 1 chip per channel, and be capable of 120W into 8ohm, est. 135W into 4ohm, and 130W into 2ohm
So far, not including my vote, the opinion poll is at:
LM3875 1 vote
LM3886 1 vote
LM1875 0 votes
LM4780 5 votes
For those not entirely familiar with the +/- of the different approaches, here is some data to consider:
LM3875 Specs
56W continuous power into 8 ohms
100W instantaneous peak output power
6 A output current limit
120dB open loop gain.
Amplifier circuit would require 2 chips per channel, and be capable of 100W into 8ohm, 72W into 4ohm, and 36W into 2ohm
LM3886 Specs
68W continuous power into 4 ohms (38W into 8ohm)
135W instantaneous peak output power
11.5 A output current limit
115dB open loop gain.
Amplifier circuit would require 2 chips per channel, and be capable of 75W into 8ohm, 135W into 4ohm, and 130W into 2ohm
LM1875 Specs
30W continuous power into 4 ohms (38W into 8ohm)
Instantaneous peak output power unspecified.
4 A output current limit
90dB open loop gain.
Amplifier circuit would require 4 chips per channel, and be capable of 150W into 8ohm, 120W into 4ohm, and 64W into 2ohm
LM4780 Specs
60W continuous power into 8 ohms (per channel)
instantaneous peak output power unspecified
11.5 A output current limit (per channel)
115dB open loop gain.
Amplifier circuit would require 1 chip per channel, and be capable of 120W into 8ohm, est. 135W into 4ohm, and 130W into 2ohm
So far, not including my vote, the opinion poll is at:
LM3875 1 vote
LM3886 1 vote
LM1875 0 votes
LM4780 5 votes
metalman:
I favor the use of the LM4780 but why not use two per channel and get around 200WPC into 4 ohms (I am not sure if my math is right).
I favor the use of the LM4780 but why not use two per channel and get around 200WPC into 4 ohms (I am not sure if my math is right).
Here here. If you sample them from National, you can get 4 for free, so let's use 4 4780's. I have four sitting on my workbench. That's reason enough for me 😉
http://www.tech-diy.com/LM4780_bridgedamp.htm
Real experience of Bridging with a 4780 with dual rails of 26 volts. I think it's clear to me that power dissipation becomes a crucial issue unless the rail voltages are run very low. As for a fan? sorry, I have enough issues with those on PCs..
Real experience of Bridging with a 4780 with dual rails of 26 volts. I think it's clear to me that power dissipation becomes a crucial issue unless the rail voltages are run very low. As for a fan? sorry, I have enough issues with those on PCs..
You counting my vote for the 4780? 😉
I mean, it's not like I'm not already working on it or anything. Already have half the parts in hand.
The only unknown I have at this point is whether to do the PCB design I already have done as a one-off, to send it out for samples for a bunch, or to give up on the whole thing and use someone elses.
The one thing I did do is decide that the chip should not extend over the edge of the board to allow for adequate power traces (perhaps overly adequate). This means either a heat-sink will need to be slotted to fit the PCB (if the HS makes up part of the case or IS the case) OR the HS has to sit above the PCB.
C
I mean, it's not like I'm not already working on it or anything. Already have half the parts in hand.
The only unknown I have at this point is whether to do the PCB design I already have done as a one-off, to send it out for samples for a bunch, or to give up on the whole thing and use someone elses.
The one thing I did do is decide that the chip should not extend over the edge of the board to allow for adequate power traces (perhaps overly adequate). This means either a heat-sink will need to be slotted to fit the PCB (if the HS makes up part of the case or IS the case) OR the HS has to sit above the PCB.
C
cjd, I think your design is a great starting point. I wonder if Brian has any comments/suggestions. I do like the chips being on the edge, so I would vote that that be changed.
Peter, very glad to see you join in. The idea of the 2SK389 is interesting, and probably worth some investigation. However, the bipolar input differential was selected so that the SuSy feedback portion operates as current feedback, which in this topology magnifies the SuSy benefit.
First your math is close, two 4780's per channel would give you closer to 250W into 4 ohms.
I'm a little hesitant to start paralleling chips for a few reasons.
1) This topology has a tendency towards instability due to the huge open loop gain of the chips and the existance of feedback loops within a larger feedback loop. With my original prototype, it took me months of troubleshooting and iterations of changes to get the circuit stability nailed down, and the other builders of my design have all run into that problem, albeit to much lesser extents. The small variations between the behaviour of each chip would likely aggravate the stability issue, requireing another few rounds of development work to get things working properly.
2) I've already been playing with boosting the power output by adding extra chips, I haven't had a lot of success with paralleling chips and have had better success with bootstrapping. That being said, I still don't have a workable high power design yet. In short, it isn't quite as simple as paralleling chips in the basic GC designs, and I'm no electronics guru. I'm just an enthusiastic DIY'er rummaging around in the darkness with a goodly amount of persistance.
3) For my first kick at producing a GP product, I'd like to keep things simple and avoid potential problems, and I suspect the majority of the interested people will find a chipmap that produces 100+ watts into almost any speaker load more than enough.
4) SuSy works best with simple circuits, and in this GP I want to showcase in the best possible way the improvement SuSy makes to chipamps.
5) I do plan on continuing development of a higher powered version, and will make that available as a GP once I've got it sorted out.
Cheers, Terry
Or insert a copper heatspreader between the chip and the heatsink to bring it flush. How far in are you mounting it?
I might be able to swing a GB on some 1/4" copper pieces. I'd have to find the appropriate copper and coax my brother into cutting it up for me. Maybe even predrill some through holes.
I might be able to swing a GB on some 1/4" copper pieces. I'd have to find the appropriate copper and coax my brother into cutting it up for me. Maybe even predrill some through holes.
LM4780 has same delayed on/off feature. And is actually cheaper, and simpler. Only downside I see to the LM4780 is that those pins are so dang close together. 🙂 But on a PCB thats a non-issue.
Well, since we know that both will eventually be done, I guess it's more a question of which gets done first.
The PCB I did puts the chip about 1/10" in from the edge. The power traces are .2". If thicker copper is used on the PCB there is probably not any pressing reason why this couldn't be dropped to .1" but I prefer to over-engineer especially when I'm not quite sure of the limits (or rather, my math).
On the other hand, a copper spacer/spreader may not be a bad idea either given the thermal properties and how this all works. In fact, it may improve performance, at least WRT heat dissipation. Of course, I know some people might insist copper will muddy the sound. 😉 Given that the chips probably need to be clamped anyhow, ... The downside is that this complicates any GB options. Or simplifies I suppose, if it means there is only ONE viable means of mounting these.
The latest changes I've done improves the ground star a bit along with a through-hole for connection to chassis star (though it puts one cap "backwards" from the rest). Also, matching the trace lengths on the inputs and outputs (which may or may not make a big difference). And I'm looking to take two caps out of the picture and leave a bit more room for other options at the power input.
However this works out, I'm willing to accept change suggestions along with just letting others do their own thing.
The other thing to consider here is the power requirements - Traditional chip-amp logic would suggest that 500VA is required per channel with the LM4780. 😉
C
On the other hand, a copper spacer/spreader may not be a bad idea either given the thermal properties and how this all works. In fact, it may improve performance, at least WRT heat dissipation. Of course, I know some people might insist copper will muddy the sound. 😉 Given that the chips probably need to be clamped anyhow, ... The downside is that this complicates any GB options. Or simplifies I suppose, if it means there is only ONE viable means of mounting these.
The latest changes I've done improves the ground star a bit along with a through-hole for connection to chassis star (though it puts one cap "backwards" from the rest). Also, matching the trace lengths on the inputs and outputs (which may or may not make a big difference). And I'm looking to take two caps out of the picture and leave a bit more room for other options at the power input.
However this works out, I'm willing to accept change suggestions along with just letting others do their own thing.
The other thing to consider here is the power requirements - Traditional chip-amp logic would suggest that 500VA is required per channel with the LM4780. 😉
C
I'v been following this thread with great intrest for quite some time, and would like to commend MetalMan and Mr. Pass on work well done 😉
Now that the poll on power ic is running, i vote for the lm3886, I feel its not a smart move ruling out those need more than the lm3875 can provide, me being one of them. 🙄
interesting amp, looks like a nice transition towards SS. on the complexity scale.
when i have more some more free time on my hands 😀
regards
Marius
Now that the poll on power ic is running, i vote for the lm3886, I feel its not a smart move ruling out those need more than the lm3875 can provide, me being one of them. 🙄
interesting amp, looks like a nice transition towards SS. on the complexity scale.
when i have more some more free time on my hands 😀
regards
Marius
