Now that you have a heatsink go and look at the tables and graphs on P10 of the datasheet.
Rth j-c=1C/W
Rth c-s ~0.5 to 1C/W
Rth s-a=0.92C/W if the temperature difference between ambient and heatsink is ~ 70Cdegrees and if the heatsink face is evenly heated over it's whole area and if the air can flow past the fins as in the test apparatus.
The graph for 8r0 shows maximum dissipation of 44W when fed from +-40Vdc supply rails. This is the voltage when delivering the output to your load and will rise to >=+-42Vdc when in the quiescent condition.
Tambient=40degC
Ts=44*0.92+40=80.5degC
Tc=44*0.5+80.5=102.5degC
Tj=44*1+102.5=146.5degC. Getting very close to maximum operating temperature.
If 0.8 is substituted for 0.5C/W for Rth c-s then maximum Tj~160degC. The amp goes into thermal protection shutdown.
But, you will NEVER operate your chipamp in these conditions.
The heatsink will be ventilated or external to the case to ensure Ta<<40degC.
The average output power ~1/100 of maximum power leaving the remaining 59W as overhead for short term transients. Average dissipation is approximately quiescent dissipation+60% of average output power~ 5.9W+0.6W=6.5W
Ta~25degC
Ts~0.92*6.5*correction factor of ~1.3+25~32.8degC
Tc~0.8*6.5+34.1~38degC
Tj~1*6.5+40.2~44.5degC.
Now look at the table. Find 80V (+-40Vdc) and the intersection with the 8r0 load. Go across to the Ta=25degC. The recommended Rthc-s is 1.9C/W. But, this takes the junction temperature to the maximum that National specify and it also assumes the heatsink Rth c-s=0.2C/W.
I always recommend that you adopt a heatsink with double the dissipation capacity that National show in this table. The chipamp will operate better if you keep it cool. i.e use ~1C/W for a single channel when driving 8ohm from +-40Vdc and double this to 0.5C/W when two chipamps share the same heatsink.
Rth j-c=1C/W
Rth c-s ~0.5 to 1C/W
Rth s-a=0.92C/W if the temperature difference between ambient and heatsink is ~ 70Cdegrees and if the heatsink face is evenly heated over it's whole area and if the air can flow past the fins as in the test apparatus.
The graph for 8r0 shows maximum dissipation of 44W when fed from +-40Vdc supply rails. This is the voltage when delivering the output to your load and will rise to >=+-42Vdc when in the quiescent condition.
Tambient=40degC
Ts=44*0.92+40=80.5degC
Tc=44*0.5+80.5=102.5degC
Tj=44*1+102.5=146.5degC. Getting very close to maximum operating temperature.
If 0.8 is substituted for 0.5C/W for Rth c-s then maximum Tj~160degC. The amp goes into thermal protection shutdown.
But, you will NEVER operate your chipamp in these conditions.
The heatsink will be ventilated or external to the case to ensure Ta<<40degC.
The average output power ~1/100 of maximum power leaving the remaining 59W as overhead for short term transients. Average dissipation is approximately quiescent dissipation+60% of average output power~ 5.9W+0.6W=6.5W
Ta~25degC
Ts~0.92*6.5*correction factor of ~1.3+25~32.8degC
Tc~0.8*6.5+34.1~38degC
Tj~1*6.5+40.2~44.5degC.
Now look at the table. Find 80V (+-40Vdc) and the intersection with the 8r0 load. Go across to the Ta=25degC. The recommended Rthc-s is 1.9C/W. But, this takes the junction temperature to the maximum that National specify and it also assumes the heatsink Rth c-s=0.2C/W.
I always recommend that you adopt a heatsink with double the dissipation capacity that National show in this table. The chipamp will operate better if you keep it cool. i.e use ~1C/W for a single channel when driving 8ohm from +-40Vdc and double this to 0.5C/W when two chipamps share the same heatsink.
Those calculations are for the unisolated T package.
The TF package is probably not a good choice to use with ±40 V rails. You can derive that from AN-1192, where National examined the LM3886 thermally. They came to the conclusion that the TF package cannot dissipate more than 30 W without the help of a fan. The LM3875 has the same case, so it is safe to assume that the same limitation applies for it.
±40 V for the T-package is okay. For the TF-package 10 % less would be a safer choice.
The TF package is probably not a good choice to use with ±40 V rails. You can derive that from AN-1192, where National examined the LM3886 thermally. They came to the conclusion that the TF package cannot dissipate more than 30 W without the help of a fan. The LM3875 has the same case, so it is safe to assume that the same limitation applies for it.
±40 V for the T-package is okay. For the TF-package 10 % less would be a safer choice.
TF Package:
Perhaps 35vdc rails for power?
Artic Silver for thermal paste on the plastic TF package to speed dissipation?
EDIT: This was assuming 8 ohm speakers, although generalized amplifier designs would probably have wider tolerances, such as using 32vdc rails or less.
EDIT2: Having a peek on the topic of "what does it do?" could be good.
Considering voltage, considering load, and considering heatsink. . . you need to shoot for a seemly and safe compromise between two extremes and those two extremes are fire hazard (too much voltage) versus clipping (not enough voltage). After this determination, purchase the appropriate heatsink.
Perhaps 35vdc rails for power?
Artic Silver for thermal paste on the plastic TF package to speed dissipation?
EDIT: This was assuming 8 ohm speakers, although generalized amplifier designs would probably have wider tolerances, such as using 32vdc rails or less.
EDIT2: Having a peek on the topic of "what does it do?" could be good.
Considering voltage, considering load, and considering heatsink. . . you need to shoot for a seemly and safe compromise between two extremes and those two extremes are fire hazard (too much voltage) versus clipping (not enough voltage). After this determination, purchase the appropriate heatsink.
westers151 said:
OK, this calculation is in error.
The 0.08 number is the TOTAL thermal resistance for 10 degree rise. Since you already have 1 degree/Watt junction-case the target junction temperature is unachievable with 125W dissipation. You will have 125 degrees rise in temperature before any additions due to the finite resistance of the heatsink (to ambient). Pick a more realistic junction temperature. Remember, no-one can touch the junction, what the numbers mean is that the junction can be @ 150 degrees while the case is @ 25. For case temperatures >25 degrees the device must be derated.
w
Ok, I'm totally confused about how to calculate this. All the examples posted have lost me completely because I don't understand what all the abbreviations stand for.
Can someone please show me a calculation for the 3875 chip assuming 24V rails, but please explain in simple terms how you've done the calculation. Remember, physics isn't second nature to me, so I need it explained in non jargon terms.
I really want to understand and learn, but I'm lost at the moment.
Thanks
Can someone please show me a calculation for the 3875 chip assuming 24V rails, but please explain in simple terms how you've done the calculation. Remember, physics isn't second nature to me, so I need it explained in non jargon terms.
I really want to understand and learn, but I'm lost at the moment.
Thanks
Hi,
Greek symbol theta = thermal resistance = Rth.
This resistance is composed of lumped elements, Junction to Case (j-c), Case to heatSink (c-s) and Sink to Ambient (s-a). There can be others dependent on the mechanical assembly being used.
Add the temperature drop across each resistance to find the total temperature drop (Greek symbol delta temperature, deltaT). Add deltaT to the ambient temperature to find the individual temperatures of the lumped elements.
C/W is the heat transfer capability of an interface, eg. case to sink.
But, be careful, most Rth s-a ratings are based on the whole backplate at isothermal temp (all at equal temperature) and that the s-a deltaT of a fixed value is buried in the specification, usually 70Cdegrees or 75Cdegrees. This Rth s-a MUST be adjusted to take account of actual use conditions and the bigger manufacturers give data and graphs to allow this de-rating to be done, giving a fairly accurate mathematical model.
Tj = Junction temperature.
Tc = Case temperature
Ts = heatSink contact face temperature.
Ta = Ambient temperature.
Note the use of degreesC or degC as indicating a temperature, but Cdegrees indicating a difference in temperature, (deltaT). Do not confuse the two, they are different, even though some will argue that Cdegrees does not exist. I know that a sink to ambient deltaT of 10Cdegrees means a temperature difference of 10Cdegrees (sometimes abbreviated to degrees). To call up deltaT as 10degC implies that both sides of the interface are at 10degC, a complete nonsense if heat is flowing across the interface.
Greek symbol theta = thermal resistance = Rth.
This resistance is composed of lumped elements, Junction to Case (j-c), Case to heatSink (c-s) and Sink to Ambient (s-a). There can be others dependent on the mechanical assembly being used.
Add the temperature drop across each resistance to find the total temperature drop (Greek symbol delta temperature, deltaT). Add deltaT to the ambient temperature to find the individual temperatures of the lumped elements.
C/W is the heat transfer capability of an interface, eg. case to sink.
But, be careful, most Rth s-a ratings are based on the whole backplate at isothermal temp (all at equal temperature) and that the s-a deltaT of a fixed value is buried in the specification, usually 70Cdegrees or 75Cdegrees. This Rth s-a MUST be adjusted to take account of actual use conditions and the bigger manufacturers give data and graphs to allow this de-rating to be done, giving a fairly accurate mathematical model.
Tj = Junction temperature.
Tc = Case temperature
Ts = heatSink contact face temperature.
Ta = Ambient temperature.
Note the use of degreesC or degC as indicating a temperature, but Cdegrees indicating a difference in temperature, (deltaT). Do not confuse the two, they are different, even though some will argue that Cdegrees does not exist. I know that a sink to ambient deltaT of 10Cdegrees means a temperature difference of 10Cdegrees (sometimes abbreviated to degrees). To call up deltaT as 10degC implies that both sides of the interface are at 10degC, a complete nonsense if heat is flowing across the interface.
westers151 said:
Use the Overture Design Guide Daniel has pointed you to.
Here are the results for 8 Ohm speakers with 24 V rails:
1% THD output: 27,07 W
Power Dissipation per IC: 15,31 W
Recommended Heat Sink Size for 25 °C ambient temperature
TA package: 7,16 K/W
TF package: 6,16 K/W
This leads to delta T 15,31 * 7,16 = 109,6 K and a peak heatsink temperature of 25 °C + 109,6 K = 134,6 °C. More than the datasheet allows, because that is the peak figure, while your amplifier will not always run at peak power.
Nevertheless you don't really want the heatsink to become 134,6 °C hot ever. Let us assume heatsinks with 70 °C max at 25 °C ambient. 70 - 25 = 45. That is your goal for delta T. Divide that by the power dissipation. 45 / 15,31 = 2,94. So your heatsink should have a thermal resistance of 2,94 K/W or less with the TA package. For the TF package you have to subtract 1 K/W increased junction-to-case resistance and need a heatsink with 1,94 K/W or less.
Your original calculation was not that far out in principle, you just fell into a beginner's trap. Because the unrealistic values you chose produced the results they did, you mistakenly subtracted the smaller number from the larger, whereas what you should have done is:-
0.08 - 1 = -0.92
You needed a number bigger than 1 for the total so that you'd have something left when you subtracted the 1 for the junction-case.
Having a negative number as a result should clue you in to the fact that you had a problem - an unachievable target junction temperature. If you read the datasheet in detail you will see that the chip is commonly run up close to the max. junction temperature.
Although this is physics, it's quite straightforward, you're just thinking in detail about something we take for granted, that heat flows from hot to cold.
In this case cold is the atmosphere, and hot is the chip die, which is bonded to the package. Heat is being continuously generated in the chip by the flow of current. If we think initially about a constant current, it keeps things simple.
Switched off, the chip starts at the same temperature as the air. When the current starts to flow the chip heats up. It will get hotter until the amount of heat entering the air matches the amount of heat entering the die. The equilibrium temperature can be determined from a notional figure called the thermal resistance to ambient whose dimensions are degrees/Watt. Obviously if the chip is enclosed in shuttle heatshield it will get hotter than if it is bolted to a large chunk of aluminium. The numbers are just a reflection of how 'easily' the heat flows away from the chip and into the surrounding air.
The datasheet gives a figure for the thermal resistance to ambient of the device without a heatsink - 43 degrees/Watt. This indicates that, ignoring any other considerations, you could get away without a heatsink @ 21 degrees ambient if you only run 3 watts.
The datasheet also states that the thermal resistance from junction-case is 1 degree/Watt. This is to permit you to calculate the temperature rise of the junction when a heatsink is employed by adding to this any thermal resistances between the case and ambient, ie the case-heatsink, and heatsink-ambient.
This cosy picture is somewhat complicated by 2 principal factors. one being the heat distributuon within the heatsink itself and it's convection cooling, because it may only work as advertised with quite a high differential between it and ambient.
The other is that the averaged power of a music signal is different from it's peak power. You can't just take the headline dissipation number off the datasheet, you have to poke a bit deeper into it.
Issues such as supply voltage variation further complicate the situation but are less significant. Fortunately, in this case, the manufacturer supplies precalculated values which reduce your design task to little more than reading from a table. You go to page 10 of the datasheet, find the junction between your load impedance and supply voltage in the nomogram, and read horizontally across to find the recommended HS for a number of different ambients. 48V @ 8 ohms ...looking across to 25 ambient gives a recommended 7.1 degrees/Watt (minimum). Job done.
w
You did good sticking with it this far, now REALLY read the datasheet. All this is in there.
0.08 - 1 = -0.92
You needed a number bigger than 1 for the total so that you'd have something left when you subtracted the 1 for the junction-case.
Having a negative number as a result should clue you in to the fact that you had a problem - an unachievable target junction temperature. If you read the datasheet in detail you will see that the chip is commonly run up close to the max. junction temperature.
Although this is physics, it's quite straightforward, you're just thinking in detail about something we take for granted, that heat flows from hot to cold.
In this case cold is the atmosphere, and hot is the chip die, which is bonded to the package. Heat is being continuously generated in the chip by the flow of current. If we think initially about a constant current, it keeps things simple.
Switched off, the chip starts at the same temperature as the air. When the current starts to flow the chip heats up. It will get hotter until the amount of heat entering the air matches the amount of heat entering the die. The equilibrium temperature can be determined from a notional figure called the thermal resistance to ambient whose dimensions are degrees/Watt. Obviously if the chip is enclosed in shuttle heatshield it will get hotter than if it is bolted to a large chunk of aluminium. The numbers are just a reflection of how 'easily' the heat flows away from the chip and into the surrounding air.
The datasheet gives a figure for the thermal resistance to ambient of the device without a heatsink - 43 degrees/Watt. This indicates that, ignoring any other considerations, you could get away without a heatsink @ 21 degrees ambient if you only run 3 watts.
The datasheet also states that the thermal resistance from junction-case is 1 degree/Watt. This is to permit you to calculate the temperature rise of the junction when a heatsink is employed by adding to this any thermal resistances between the case and ambient, ie the case-heatsink, and heatsink-ambient.
This cosy picture is somewhat complicated by 2 principal factors. one being the heat distributuon within the heatsink itself and it's convection cooling, because it may only work as advertised with quite a high differential between it and ambient.
The other is that the averaged power of a music signal is different from it's peak power. You can't just take the headline dissipation number off the datasheet, you have to poke a bit deeper into it.
Issues such as supply voltage variation further complicate the situation but are less significant. Fortunately, in this case, the manufacturer supplies precalculated values which reduce your design task to little more than reading from a table. You go to page 10 of the datasheet, find the junction between your load impedance and supply voltage in the nomogram, and read horizontally across to find the recommended HS for a number of different ambients. 48V @ 8 ohms ...looking across to 25 ambient gives a recommended 7.1 degrees/Watt (minimum). Job done.
w
You did good sticking with it this far, now REALLY read the datasheet. All this is in there.
Most enjoyable reading!
I wish to highlight two areas for consideration:
Junction:
You can use clamp bar (across face of chip), a heatsink with a thick shiny base, and Artic Silver for the TF (preinsulated) chip. That will speed up the thermal transfer to decrease the junction temperature. It works in practice.
Voltage:
If lm3875 is derated (undervolted) far enough to cool it considerably, this makes a clipping-only amplifier; therefore, small voltage differences are unlikely to be helpful. Do please stay within spec.
However, since pure DC is unlikely to be present during actual operation, I'd like to mention that cleaner dc runs cooler than dirty dc. . .
Consider that the limited power supply of gainclone style will require a larger heatsink than a similar amplifier with an elaborate power supply. A cost comparison either way is quite similar, albeit a very slight percentage in favor of the gainclone, despite the larger heatsink needed.
Don't use a minimum theoretical size heatsink. Use AndrewT's "double-size it" suggestion for heatsink selection.
That might show that, while heatsink is important indeed, there are many other important factors that are worthy of consideration. . . and that is becoming overdue. Its time to purchase your heatsink.
I wish to highlight two areas for consideration:
Junction:
You can use clamp bar (across face of chip), a heatsink with a thick shiny base, and Artic Silver for the TF (preinsulated) chip. That will speed up the thermal transfer to decrease the junction temperature. It works in practice.
Voltage:
If lm3875 is derated (undervolted) far enough to cool it considerably, this makes a clipping-only amplifier; therefore, small voltage differences are unlikely to be helpful. Do please stay within spec.
However, since pure DC is unlikely to be present during actual operation, I'd like to mention that cleaner dc runs cooler than dirty dc. . .
Consider that the limited power supply of gainclone style will require a larger heatsink than a similar amplifier with an elaborate power supply. A cost comparison either way is quite similar, albeit a very slight percentage in favor of the gainclone, despite the larger heatsink needed.
Don't use a minimum theoretical size heatsink. Use AndrewT's "double-size it" suggestion for heatsink selection.
That might show that, while heatsink is important indeed, there are many other important factors that are worthy of consideration. . . and that is becoming overdue. Its time to purchase your heatsink.
wakibaki said:
..on the following conditions
- - TA package is used, TF package is not used.
- the heatsink is mounted outside of the amplifier case and the amplifier is not placed in a rack or similar, where air circulation is limited.
- the heatsink is impossible to touch during operation to avoid burning injuries, which of course contradicts the former condition.
- the amplifier will not be used at full throttle during hot summer days.
- the amplifier will never be used with loads below 8 Ohms.
danielwritesbac said:
It will not actually speed thermal transfer up. Arctic Silver can have a slightly lower thermal resistance than other thermal greases, if properly applied. The difference is however not significant.
Be also aware that Arctic Silver and similar products are electrical conductors and must therefore not be used with non-insulated packages, only with TF type.
It is important that you use (any) thermal grease. Without it the thermal transfer will not work well and all the above calculations will be rendered useless.
Voltage, theory versus practice:
The 48vdc rails don't work in practice. The amplifier will break.
The maximum in practice is 41vdc rails. Even so, its unwise to design real amplifiers with such poor tolerances. Therefore, it is unnecessary to quote figures that could result in poor design.
Use average 32vdc rails.
Thermal, theory versus practice:
Instead of electroshock, use an insulator with the "T" chip. For heatsink size, there is no practical difference when either chip version is used insulated. Therefore, it is unnecessary to quote heatsink figures for an uninsulated chip.
Use the figures for "TF" chip.
I believe that it would be best to use figures that work well in practice and also allow for more generous tolerances.
The 48vdc rails don't work in practice. The amplifier will break.
The maximum in practice is 41vdc rails. Even so, its unwise to design real amplifiers with such poor tolerances. Therefore, it is unnecessary to quote figures that could result in poor design.
Use average 32vdc rails.
Thermal, theory versus practice:
Instead of electroshock, use an insulator with the "T" chip. For heatsink size, there is no practical difference when either chip version is used insulated. Therefore, it is unnecessary to quote heatsink figures for an uninsulated chip.
Use the figures for "TF" chip.
I believe that it would be best to use figures that work well in practice and also allow for more generous tolerances.
Daniel, you got it wrong again.danielwritesbac said:
The 48V is the sum of the dual polarity voltages. i.e.+24Vdc.
The TF chip is not as good as a T chip + insulator.
I agree with the generally forgotten conditions that National hide in the literature and highlighted by Pacific. That 7.1C/W does not work well. <=3.5C/W will keep the chip much cooler and as a result the Spike will trigger less often improving sound quality.
If it's two chips on the same sink, then <1.8C/W is required for a pair of T packages for good performance.
AndrewT said:
Thank you for the clarification on the voltage! Those figures were just too high.
In my own use, the "TF" chip + artic silver (or other metalized compound) has no practical difference -versus- a "T" chip with Kapton/Mica (or other insulator) with artic ceramique or GC Waldham Type 44.
By practical, I mean a physically measurable difference in the dimensions of the heatsink.
My apologies for assuming that a proper plastic-to-metal filler compound would be used for the "TF" chip while a non-conductive thermal compound would be used for the "T" chip. I should not have assumed usage of optimal compounds when making that statement previously. This is a most ironic error to make during complaining about too-tight tolerances, and again, I apologize.
It does at least demonstrate that more-generous tolerances (larger heatsink) are probably beneficial.
Guys
Thanks for the continued advice. I've not responded before as I've been mulling the advice over and reading the spec sheets - at last things are becoming clearer
Whilst the maths is still confusing me slightly, (although I think it's more the terminology that's confusing - I much prefer to get someone to do explanations via sketching out the situation, plus the maths. Then I can visually understand what the terminology means) I'm starting to understand the data sheets and how to read them.
For the LM3875 I have deduced that for 24V rails (48V in total) then a heat sink of 3c/w or less will be good for one chip. Note, I'm using the "double it" rule of thumb here.
If I then assume that putting the heat sink inside a case will make it slightly less then optimal, then using something that gives 1.5 c/W or less should be more than enough to keep things cool. It will also allow for higher ambient temperatures than 21c (does happen in the UK...sometimes ).
Thanks
Ian
Thanks for the continued advice. I've not responded before as I've been mulling the advice over and reading the spec sheets - at last things are becoming clearer
Whilst the maths is still confusing me slightly, (although I think it's more the terminology that's confusing - I much prefer to get someone to do explanations via sketching out the situation, plus the maths. Then I can visually understand what the terminology means) I'm starting to understand the data sheets and how to read them.
For the LM3875 I have deduced that for 24V rails (48V in total) then a heat sink of 3c/w or less will be good for one chip. Note, I'm using the "double it" rule of thumb here.
If I then assume that putting the heat sink inside a case will make it slightly less then optimal, then using something that gives 1.5 c/W or less should be more than enough to keep things cool. It will also allow for higher ambient temperatures than 21c (does happen in the UK...sometimes
).Thanks
Ian
Ok, another dumb question 
If my rail voltage is, for example, 35V and each chip is on a separate heat sink, why would I use a voltage of 70V (2*35V) when reading the National graph for determing heat sink values?
It doesn't make sense to me - if I were putting two chips on one heat sink then I could understand the 70V value, but not for one chip per heat sink.
If my rail voltage is, for example, 35V and each chip is on a separate heat sink, why would I use a voltage of 70V (2*35V) when reading the National graph for determing heat sink values?
It doesn't make sense to me - if I were putting two chips on one heat sink then I could understand the 70V value, but not for one chip per heat sink.
westers151 said:For the LM3875 I have deduced that for 24V rails (48V in total) then a heat sink of 3c/w or less will be good for one chip. Note, I'm using the "double it" rule of thumb here.
If I then assume that putting the heat sink inside a case will make it slightly less then optimal, then using something that gives 1.5 c/W or less should be more than enough to keep things cool. It will also allow for higher ambient temperatures than 21c (does happen in the UK...sometimes
Pdmax = Vcctot²/((2*PI)²*Rl) = 48V²/(2*PI²*4Ohm) = 29,2 W
So with a 1,5 K/W heatsink you will get a heatsink DeltaT of 29,2 W * 1,5 K/W = 43,8 K. The heatsink will be 43,8 degrees hotter than the surrounding air.
The IC's Delta T depends on which package you use.
LM3875T (unisolated) DeltaT = 29,2 W * (1,5 K/W + 0,2 K/W + 1 K/W) = 78,84 K
LM3875TF (isolated) DeltaT = 29,2 W * (1,5 K/W + 0,2 K/W + 2 K/W) = 108,04 K
The maximum permitted IC temperature is 150 °C. That means the LM3875T would be working fine up to an ambient temperature of 150 °C - 78,84 K = 71,26 °C, and the LM3875TF up to 150 °C - 108,04 K = 41,96 °C.
If you mount the heatsink outside of the case, then you need to make sure that nobody burns his fingers on it. The heatsink temperature should be kept below 60 °C then. 60 °C - 43,8 K = 16,2 °C. Either use a heatsink with a lower thermal resistance or move the heatsink into the case. Take into account that the ambient temperature inside of the case will be higher than the ambient temperature in the rest of the UK. You need to make corrections for that. I. e. either use a bigger heatsink again or use the LM3875T.
This is all assuming the amplifier is working on worst case conditions. A real music load will lead to less average dissipation. And if you use 8 Ohm speakers the situation becomes more relaxed as well.
One more remark. All those calculations were based on the assumption that the heatsink has 1,5 K/W at Pd 29,2 W. More often than not the thermal resistance is given at the highest reasonable temperature and the actual thermal resistance at your load is higher than 1.5 K/W. To determine that there are datasheets for heatsinks as well.
as an example I have a 3886 running on +-21.5Vdc at the moment driving an 8ohm speaker.
Ta=18degC (underfloor heating)
Tsink is too hot to be comfortable.
Tc burns within about 3seconds. I'm guessing it's about 55degC to 60degC.
But the National design table says a 11C/W sink is OK when ambient is 25degC and the T version is directly connected to the sink.
I have used 8.6C/W and running it 7Cdegrees cooler and yet it still burns. I wonder how far the SPIKE trigger points have been temperature de-rated to account for the unnecessarily high case and junction temperatures?
BTW, it sounds strained and the average output is 2Vac to 2.4Vac ie. 500mW to 700mW into a 88dB/Wm 8ohm speaker.
Ta=18degC (underfloor heating)
Tsink is too hot to be comfortable.
Tc burns within about 3seconds. I'm guessing it's about 55degC to 60degC.
But the National design table says a 11C/W sink is OK when ambient is 25degC and the T version is directly connected to the sink.
I have used 8.6C/W and running it 7Cdegrees cooler and yet it still burns. I wonder how far the SPIKE trigger points have been temperature de-rated to account for the unnecessarily high case and junction temperatures?
BTW, it sounds strained and the average output is 2Vac to 2.4Vac ie. 500mW to 700mW into a 88dB/Wm 8ohm speaker.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.