According to the datasheet calculations, you're looking at about 45-50 watts. This is complicated by the requirement of getting the heat off such a tiny area as the back of a LM4780.
You will need a pretty beefy heatsink. The 4780 itself is 0.8, add 0.5 for the insulator/compound, that's already 1.3. Even a 0.7 C/W takes the total to 2 C/w, which means a temp rise of 100 degrees for the 50 watt heat, which will trip the thermal protection with a 25 degree ambient. Ideally 0.2-0.3 is the range you're looking at, and even then the chips will get hot at worst case dissipation.
On the other hand, the worst case results will not always apply, and you can get away with a slightly smaller sink. The amp should be fine for music without having the thermal protection kick in, though it might not pass a full power square wave test without tripping SPiKE.
Edit: I run a set of 4780 chips in a bridge/parallel setup. The parallel chips help keep the dissipation in the individual chips down, and I run on ~ 2C/W heatsinks for each chip. They run warm but not uncomfortably so.
You will need a pretty beefy heatsink. The 4780 itself is 0.8, add 0.5 for the insulator/compound, that's already 1.3. Even a 0.7 C/W takes the total to 2 C/w, which means a temp rise of 100 degrees for the 50 watt heat, which will trip the thermal protection with a 25 degree ambient. Ideally 0.2-0.3 is the range you're looking at, and even then the chips will get hot at worst case dissipation.
On the other hand, the worst case results will not always apply, and you can get away with a slightly smaller sink. The amp should be fine for music without having the thermal protection kick in, though it might not pass a full power square wave test without tripping SPiKE.
Edit: I run a set of 4780 chips in a bridge/parallel setup. The parallel chips help keep the dissipation in the individual chips down, and I run on ~ 2C/W heatsinks for each chip. They run warm but not uncomfortably so.
That is understandable, because it is impossible to heatsink that configuration.
Two channels with ±30 V bridged into an 8 Ohm load result in a heat dissipation of ~94 W. That limits the junction temperature to ~74 °C. The LM4780 has 0,8 K/W. 94 W * 0,8 K/W = 75,2 K, so your heatsink + isolation washers + thermal grease must remain below 74 °C - 75,2 K = -1,2 °C and still dissipate 94 W.
You stand a chance to heatsink an LM4780 with 4 Ohm on both channels, if you reduce the supply voltage to ±22 V. Or you replace the LM4780 with two LM3886T (not TF!), but even then it will be an effort to keep them cool with 4 Ohm load per IC. Or add another LM4780 and convert it into a BPA-200 configuration. Still the same effort to heatsink.
Two channels with ±30 V bridged into an 8 Ohm load result in a heat dissipation of ~94 W. That limits the junction temperature to ~74 °C. The LM4780 has 0,8 K/W. 94 W * 0,8 K/W = 75,2 K, so your heatsink + isolation washers + thermal grease must remain below 74 °C - 75,2 K = -1,2 °C and still dissipate 94 W.
You stand a chance to heatsink an LM4780 with 4 Ohm on both channels, if you reduce the supply voltage to ±22 V. Or you replace the LM4780 with two LM3886T (not TF!), but even then it will be an effort to keep them cool with 4 Ohm load per IC. Or add another LM4780 and convert it into a BPA-200 configuration. Still the same effort to heatsink.
Well it will work if he keeps each channel on a 0.3 C/W heatsink (this means about 45 watts per heatsink, not altogether unreasonable given that a lot of the Class A monsters are dissipating 100 watts at idle.
Obviously two channels cannot share a reasonably sized heatsink, it is practically impossible without forced cooling of some kind. But then again there are possibilities to cater to that. A modern Quad-core CPU pushes over a hundred watts of heat out of an area much smaller than the 4780. Inspite of the tightly coupled heatspreader and much lower allowable junction temperatures (under 100 degrees in most cases), the chips manage fine. They only need some reasonable airflow to stay under 60 degrees under peak loads.
Obviously two channels cannot share a reasonably sized heatsink, it is practically impossible without forced cooling of some kind. But then again there are possibilities to cater to that. A modern Quad-core CPU pushes over a hundred watts of heat out of an area much smaller than the 4780. Inspite of the tightly coupled heatspreader and much lower allowable junction temperatures (under 100 degrees in most cases), the chips manage fine. They only need some reasonable airflow to stay under 60 degrees under peak loads.
bridged gives double the power into double the impedance.
Using a pair of 30Vdc powered amps is the equivalent to using 4ohm loads.
30Vdc and 4ohm loads gives the maximum plus a bit of ~60W into 4r0.
The bridged pair will give 120W into 8r0.
The dual chip must be designed as two amps, each driving 60W into 4ohms.
That is what is difficult. A misunderstanding of what bridging is doing to your overloaded chipamps.
Using a pair of 30Vdc powered amps is the equivalent to using 4ohm loads.
30Vdc and 4ohm loads gives the maximum plus a bit of ~60W into 4r0.
The bridged pair will give 120W into 8r0.
The dual chip must be designed as two amps, each driving 60W into 4ohms.
That is what is difficult. A misunderstanding of what bridging is doing to your overloaded chipamps.
I would use a nice large heatsink and slap a couple of quiet fans on it. I really don't see any issues with anything over 0.3 C/W, or a simple 0.5C/W heatsink and a couple of fans. It is possible to get these chips to work fine with those rails and that load as long as heat is taken away fast enough.
Max power output is *not* = Max dissipation. For most Class AB amps, the worst case dissipation is around 35% of power output, IIRC but I don't have a reference for this. Even the graphs in the chip's datasheets corroborate this, though I don't have an explanation for it myself.
You might want to check with Russ/Brian on this specific question, they will probably point you to something they use. With all due respect, if the 4780 is rated up to 4 ohm loads at 30V rails, it should be able to deal with 8 ohm bridged, as each half of the chip will be looking 'into' a 4 ohm load. It's right at the spec limit for the chip - to be safer 25 volts is what you should be running at but I don't think there'll be an issue as long as the heatsinking is good enough.
@pb - where did the 74 degree limit come from? The LM4780 datasheet rates the bridge amplifier into an 8 ohm load up to 30 volts. I understand the dissipation is very high - around 90 watts as you state and not 45 like I calculated (my error came from using 30V instead of 60V as I should have taken). If you assume 90 watts an 0.8 Tj-c, the case temperature is already at 72 degrees (I assume this is what you meant). Add another 0.5 Tc-a and we get 1.3 C/W Tj-a, with 90 watts of heat this is >100 degrees C. Very very high, but within 150 C limit of the chip. A good CPU air cooler is able to deliver thermal resistance (with a fan) of under 0.2 C/W, which would be able to keep the chip under the limits. A large enough passive sink would also do.
I agree that it is heavily overloaded, but also worst case dissipation is not frequent in its occurrence anyway, and usually not at full power. Plus the chip has protection from temperature (which actually works, I know this the hard way) rise, so though it will run hot it will survive.
Max power output is *not* = Max dissipation. For most Class AB amps, the worst case dissipation is around 35% of power output, IIRC but I don't have a reference for this. Even the graphs in the chip's datasheets corroborate this, though I don't have an explanation for it myself.
You might want to check with Russ/Brian on this specific question, they will probably point you to something they use. With all due respect, if the 4780 is rated up to 4 ohm loads at 30V rails, it should be able to deal with 8 ohm bridged, as each half of the chip will be looking 'into' a 4 ohm load. It's right at the spec limit for the chip - to be safer 25 volts is what you should be running at but I don't think there'll be an issue as long as the heatsinking is good enough.
@pb - where did the 74 degree limit come from? The LM4780 datasheet rates the bridge amplifier into an 8 ohm load up to 30 volts. I understand the dissipation is very high - around 90 watts as you state and not 45 like I calculated (my error came from using 30V instead of 60V as I should have taken). If you assume 90 watts an 0.8 Tj-c, the case temperature is already at 72 degrees (I assume this is what you meant). Add another 0.5 Tc-a and we get 1.3 C/W Tj-a, with 90 watts of heat this is >100 degrees C. Very very high, but within 150 C limit of the chip. A good CPU air cooler is able to deliver thermal resistance (with a fan) of under 0.2 C/W, which would be able to keep the chip under the limits. A large enough passive sink would also do.
I agree that it is heavily overloaded, but also worst case dissipation is not frequent in its occurrence anyway, and usually not at full power. Plus the chip has protection from temperature (which actually works, I know this the hard way) rise, so though it will run hot it will survive.
Why not just lower the voltage a little? According to the Overture Design Guide, +/- 30V into a bridged LM4780 with 8 ohm speakers puts out about 167W of power and needs a heatsink rated at 0.53C/W. If you lowered your supply to +/- 25V, you would have about 115W and would need a heatsink of 1.10C/W. Much more manageable.
Of course the power outputs above depend on the values used to set the gain.
Of course the power outputs above depend on the values used to set the gain.
I must publish a retraction.
At 40 degree ambient and with a thermal coupling resistance of 0.4 C/W, we require a heatsink of 0.01 C/W to keep the chip under 150 degrees C at worst case dissipation for 30V/8/bridged 4780. This is obviously not possible.
Here's a quick excel sheet I threw together for calculations. Varying the rails and load impedance will give required results (please ignore everything except the 4780 results - the others have to have their values fed in from their datasheets, I just used values from the top of my head) for Rth of the heatsink.
I have assumed 0.4 C/w for the interface resistance, this is in one of the hidden columns and can be tweaked. I have also hidden the automated power dissipation column. The ambient temperature is a changeable setting, 40 degrees is common for most enclosed electronics.
At 40 degree ambient and with a thermal coupling resistance of 0.4 C/W, we require a heatsink of 0.01 C/W to keep the chip under 150 degrees C at worst case dissipation for 30V/8/bridged 4780. This is obviously not possible.
Here's a quick excel sheet I threw together for calculations. Varying the rails and load impedance will give required results (please ignore everything except the 4780 results - the others have to have their values fed in from their datasheets, I just used values from the top of my head) for Rth of the heatsink.
I have assumed 0.4 C/w for the interface resistance, this is in one of the hidden columns and can be tweaked. I have also hidden the automated power dissipation column. The ambient temperature is a changeable setting, 40 degrees is common for most enclosed electronics.
rhysh said:
To convert from AC to DC, multiply by 1.4 (normally multiply by 1.41 then subtract 0.7 for diode drop, but 1.4 is close enough). That would be the loaded voltage, and unloaded it will be a little higher due to transformer regulation.
In your case, you will get 30V or 31V under load, a bit above that with no load depending on transformer regulation.
The datasheet specifies the maximum heat dissipation with 125 W @ 25 °C. The maximum operating temperature is 150 °C, where the IC can dissipate 0 W. Analog to the derating curve of a transistor you can draw a line between those two points and interpolate the permissible temperature for each power dissipation.sangram said:
If you follow the 125 W rating you get even lower than 74 °C. The 74 °C are based on the Overture Design Guide that calculates the possible heat dissipation with 155 W @ 25 °C. This seems logical, because the LM4780 is bigger than the LM3886, which is rated at 125 W and with a 125 W rating the LM4780 would be next to useless. So the 125 W is probably a typo in the datasheet or result of Copy-&-Paste.
You may be able to use the LM4780 with a big fan-supported heatsink at civilized listening levels, but not...
rhysh said:
Hi,
Okay so it looks like my rails will be around 29v - 31v under load.
The most suitable heatsink i have found for my enclosure, is rated at 0.85°C/W , which is 200mm wide, 40mm tall and 75mm deep. This is the biggest sink i can fit, but the problem is it must sit inside the enclosure not the outside.
If i was to use a fan to blow air out the back i should be ok right?
Im pretty certain this is a lot bigger then what most use for the 4780?
Thanks for all the help so far everyone!
Okay so it looks like my rails will be around 29v - 31v under load.
The most suitable heatsink i have found for my enclosure, is rated at 0.85°C/W , which is 200mm wide, 40mm tall and 75mm deep. This is the biggest sink i can fit, but the problem is it must sit inside the enclosure not the outside.
If i was to use a fan to blow air out the back i should be ok right?
Im pretty certain this is a lot bigger then what most use for the 4780?
Thanks for all the help so far everyone!
The point is that you can only use 5-10 of the 150 W a bridged application gives in theory. Above that power the IC will overheat, whatever heatsink you use.rhysh said:
You would probably be better off to convert the amplifier from bridged to single configuration. That halves the heat dissipation, and you would get away with your 0,85 K/W heatsink even without fan most of the time. You get around 40 W you can really use. Makes more sense than 150 W, which you cannot use.
.85 is way too high. With a fan this could drop a bit, but it depends on the heatsink fin configuration. With a 0.85 C/W heatsink you can dissipate 50-60 watts from a single 4780 to keep temperatures under 100C, this corresponds to one being used in stereo configuration as PB says earlier. Also depending on the fin configuration it might be better to channel airflow to the sink, rather than away from it.
@pb: It was my understanding that the temperature derating curve applies not to operating temperature but to ambient temperature. With higher ambient temperature the heatsink will have to become bigger as the temperature delta between heatsink and ambient decreases.
I see the logic of what you're saying, but the latest datasheet does not have the power derating curve at all. I'll check back with the ones downloaded earlier.
@pb: It was my understanding that the temperature derating curve applies not to operating temperature but to ambient temperature. With higher ambient temperature the heatsink will have to become bigger as the temperature delta between heatsink and ambient decreases.
I see the logic of what you're saying, but the latest datasheet does not have the power derating curve at all. I'll check back with the ones downloaded earlier.
Both is true. If the ambient temperature is higher the difference to the IC's max temperature is smaller and the heatsink must be bigger, if the dissipation remains the same.sangram said:
And with higher dissipation the IC's maximum allowable temperature sinks.
Don't bother checking. the only datasheet with a derating curve is the LM1875's. The others give the figure at 25 °C and imply that dissipation is 0 at 150 °C. You would have to draw the curve yourself with those numbers or use this simple formula Top = Tmax - Pd * (Tmax - Tnom) / Pdmax, wheresangram said:
Top = permissible operating temperature at junction
Tmax = 150 °C from datasheet
Tnom = 25 °C from datasheet
Pdmax = maximum dissipation at 25 °C from datasheet
E. g. Top = 150 ° - (94 W * (150 °C - 25 °C) / 155 W) = 74,2 °C
In real life you rarely use more than 1 W average, everything above that is dynamic headroom for short peaks. As long as you listen to the Simpatico at "normal" loudness levels, it will work fine.rhysh said:The thing is, i have bought a pair of mono boards which are intended to be used for this purpose. Are you telling me there is no way to run these above 40W ?
The thermal protection will start to bother as soon as the average output power goes up, e. g. if you try to fill a party with sound. Or also, if you try to reproduce a church organ with its long sustained and powerful bass notes at realistic levels. Long orchestra tutti can be an issue as well as pop music with high bass content. That leads to the funny situation that the amplifier with the higher nominal power will sound better at lower listening levels, due to its dynamic headroom, while the version with less power sounds better at higher listening levels, because it does not shut off due to overheating.
I use my Sympatico with this heatsink:
http://www.jaycar.com/productView.asp?ID=HH8546&CATID=45&form=CAT&SUBCATID=751
Two channels on one sink. I drive a pair of 4-ohm speakers (Cryolites). Listening at very high levels (wife and kids not home) for +1 hour, then heatsink is moderately warm. Spike protection never engages. The chips are mounted with Bergquist K10 Kapton pads. Transformer for each channel is a 400VA 22V+22V Antek.
http://www.jaycar.com/productView.asp?ID=HH8546&CATID=45&form=CAT&SUBCATID=751
Two channels on one sink. I drive a pair of 4-ohm speakers (Cryolites). Listening at very high levels (wife and kids not home) for +1 hour, then heatsink is moderately warm. Spike protection never engages. The chips are mounted with Bergquist K10 Kapton pads. Transformer for each channel is a 400VA 22V+22V Antek.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.