Hi
I have already a stereo power amp for basses and I miss 4 other mono amps to acheive active tri-amping my speakers. The mediums are 8Ω nominal and tweeters are 6Ω. Some years ago I bought 16 MUR860, 4 heatsinks fisher sk92-100 (rated 1°K/W) and 4 LM3886T I'll use with thermal grease and no isolators (I know what I do).
Now I get some time to work again on the project.
I'd like to have the maximum power possible available. Living in France where mains are 230Vac guaranteed +/-10% and reading the chip and transfo datasheets I ended discarding +/-30Vac toroids I first selected unless I build a regulated supply what I'd rather not want.
Also, I would prefer the build is not dedicated to this first application but instead would be standard multipurpose amps, so I'm ready to forget energy consumption optimization related to different loads.
I hope I didn't forget something
... oh yes, all the 4 amps in a single case, and I already built 4 line receivers with THAT1206.
So what transfo(s) can I use? I read 160VA is ok for a stereo amp. So a single 300VA would provide the needed power plus a better so called "regulation", won't waste my money and make a lighter case. Or a pair of 300VA? Or 4 80VA or 160VA (in this last scenario I'll miss 16 rectifiers)?
Thank you for advises.
I have already a stereo power amp for basses and I miss 4 other mono amps to acheive active tri-amping my speakers. The mediums are 8Ω nominal and tweeters are 6Ω. Some years ago I bought 16 MUR860, 4 heatsinks fisher sk92-100 (rated 1°K/W) and 4 LM3886T I'll use with thermal grease and no isolators (I know what I do).
Now I get some time to work again on the project.
I'd like to have the maximum power possible available. Living in France where mains are 230Vac guaranteed +/-10% and reading the chip and transfo datasheets I ended discarding +/-30Vac toroids I first selected unless I build a regulated supply what I'd rather not want.
Also, I would prefer the build is not dedicated to this first application but instead would be standard multipurpose amps, so I'm ready to forget energy consumption optimization related to different loads.
I hope I didn't forget something
... oh yes, all the 4 amps in a single case, and I already built 4 line receivers with THAT1206.So what transfo(s) can I use? I read 160VA is ok for a stereo amp. So a single 300VA would provide the needed power plus a better so called "regulation", won't waste my money and make a lighter case. Or a pair of 300VA? Or 4 80VA or 160VA (in this last scenario I'll miss 16 rectifiers)?
Thank you for advises.
Did you read this ? LM3886 chip power amplifier power supply design.
Especially relevant data at the bottom of the page.
Especially relevant data at the bottom of the page.
I began with this one: Building a Gainclone chip amp power supply.
Thanks for your link man. I hope I'll find my honey.
Thanks for your link man. I hope I'll find my honey.
France is on the EU harmonised supply.
Your range is 230Vac +10% -6% i.e. 216Vac to 253Vac.
Four 60W chipamps delivering maximum power at the same time would work well with a 240VA to 480VA transformer.
But, I'll bet that all four will never be asked to deliver maximum power at the same time.
I suspect you could reduce that VA range to 160VA to 300VA and it will still work well.
You will either have all four operating from a common PSU with a common Main Audio Ground or you need to have all four isolated from each other and requiring 8 secondary windings. This last option is obviously unworkable.
That leaves the common PSU.
You now have a further choice. dual secondaries and dual rectifiers and then series connect to create a Main Audio Ground.
or
Centre Tapped secondary with a single bridge rectifier and the Centre tap becomes the Power Zero Volts that heads out to connect to the Main Audio Ground.
This last has just one bridge rectifier for the four channels.
However you need another bridge rectifier for the Disconnecting Network that connects the Main Audio Ground to Chassis. I use the double shorted version ~ shorted to ~ and + shorted to -
Then connect one corner to Chassis and connect an adjacent corner to Main Audio Ground.
You can use a ¼W 10r resistor across two adjacent corners and use a 100nF X7R across another pair of adjacent corners.
If you want to experiment with ground lifting, then add a switch across two adjacent corners.
Closing the switch defeats the Disconnecting Network. Opening the switch lifts the ground and makes the DN active.
But, I'll bet that all four will never be asked to deliver maximum power at the same time.
I suspect you could reduce that VA range to 160VA to 300VA and it will still work well.
You will either have all four operating from a common PSU with a common Main Audio Ground or you need to have all four isolated from each other and requiring 8 secondary windings. This last option is obviously unworkable.
That leaves the common PSU.
You now have a further choice. dual secondaries and dual rectifiers and then series connect to create a Main Audio Ground.
or
Centre Tapped secondary with a single bridge rectifier and the Centre tap becomes the Power Zero Volts that heads out to connect to the Main Audio Ground.
This last has just one bridge rectifier for the four channels.
However you need another bridge rectifier for the Disconnecting Network that connects the Main Audio Ground to Chassis. I use the double shorted version ~ shorted to ~ and + shorted to -
Then connect one corner to Chassis and connect an adjacent corner to Main Audio Ground.
You can use a ¼W 10r resistor across two adjacent corners and use a 100nF X7R across another pair of adjacent corners.
If you want to experiment with ground lifting, then add a switch across two adjacent corners.
Closing the switch defeats the Disconnecting Network. Opening the switch lifts the ground and makes the DN active.
Last edited:
Hi 00940 & AndrewT
Nick Whetstone (actually Rod Elliott) advised 5W for the 10Ω in the earth/ground loop breaker.
Also I spoke too quick when I said our mains are +/-10% waranteed. I found it there https://particuliers.edf.com/fichiers/fckeditor/Particuliers/Offres/CGV_CRE/CGV_PART_MCE.pdf but an other one says the standard is
http://fr.leonardo-energy.org/wp-content/uploads/2008/01/5_4_2_en50160.pdf so the % depends on whether we are customer or supplier in this matter. These parasite guys in european administration make me vomit with their tons of useless standards (but that's another story). Whatever all this makes transformer selection even hardest. I'll try to stick to advised if I find.
BTW, I didn't say 480VA, but instead 4*80 or 160.
Having yet the 10cm/4in tall heatsinks I imagine I'll put all the stuff in a 3U case I'll split in 2 floors. It is better to screw the 3886 to the low side of the heatsinks instead of centered or on the top, isn't it? If yes the audio part will live in the lower floor and the supply in the upper one. As toroids with lower VA are usually flatter I feel I'll use 2 or 4. Regards to your link to neurochrome I discovered the Antek brand which is really a great company with huge choice opened to us DIYers, light years in front of radiospares and their electrostatic/magnetic unshielded/chinacheapo. Antek even gives us an additional measurement for output voltage under mid load! <sad>Why don't have no more such a factory/shop in Europe, toroids weight will ruin our bucks in shipping cost</sad>?
neurochrome is really another great place. I even discovered the guy Tom Christiansen cooked a 3886 and even uses the THAT12xx A chance I didn't yet solder mines on my own pcbs http://www.diyaudio.com/forums/analog-line-level/225676-fully-balanced-line-1206-plus-1646-a.html .
Are its schematic/pcb drawing freely available somewhere?
Nick Whetstone (actually Rod Elliott) advised 5W for the 10Ω in the earth/ground loop breaker.
I'd rather stay away this topologic way to explain how to connect this loop breaker because I'd be afraid a newbie gets hands on a 35A bridge with AC leads on the same side (like 4pins DIP6 bridges are): then he would short earth to ground for more money than the needed simple wireAndrewT said:
Also I spoke too quick when I said our mains are +/-10% waranteed. I found it there https://particuliers.edf.com/fichiers/fckeditor/Particuliers/Offres/CGV_CRE/CGV_PART_MCE.pdf but an other one says the standard is
http://fr.leonardo-energy.org/wp-content/uploads/2008/01/5_4_2_en50160.pdf so the % depends on whether we are customer or supplier in this matter. These parasite guys in european administration make me vomit with their tons of useless standards (but that's another story). Whatever all this makes transformer selection even hardest. I'll try to stick to advised if I find.
BTW, I didn't say 480VA, but instead 4*80 or 160.
Having yet the 10cm/4in tall heatsinks I imagine I'll put all the stuff in a 3U case I'll split in 2 floors. It is better to screw the 3886 to the low side of the heatsinks instead of centered or on the top, isn't it? If yes the audio part will live in the lower floor and the supply in the upper one. As toroids with lower VA are usually flatter I feel I'll use 2 or 4. Regards to your link to neurochrome I discovered the Antek brand which is really a great company with huge choice opened to us DIYers, light years in front of radiospares and their electrostatic/magnetic unshielded/chinacheapo. Antek even gives us an additional measurement for output voltage under mid load! <sad>Why don't have no more such a factory/shop in Europe, toroids weight will ruin our bucks in shipping cost</sad>?
neurochrome is really another great place. I even discovered the guy Tom Christiansen cooked a 3886 and even uses the THAT12xx
A chance I didn't yet solder mines on my own pcbs http://www.diyaudio.com/forums/analog-line-level/225676-fully-balanced-line-1206-plus-1646-a.html .Are its schematic/pcb drawing freely available somewhere?
That matches my math (linked to in Post #2) for sine wave operation. 63 VA per channel (8 Ω) and 121 VA per channel (4 Ω).
However, few people listen to pure sine waves at full blast, so unless you intend your amp to operate that way, you can reduce the transformer requirements quite a bit.
For a music signal with 14 dB crest factor, you'll need 27 VA per channel (8 Ω) and 51 VA per channel (4 Ω).
Above numbers assume ±28 V rails.
One could make some assumptions about the frequency distributions of various kinds of music, crossover frequencies of the speakers involved, and calculate the amount of power delivered per channel. You may very well find that all channels deliver considerable power more or less continuously.
Considering such a network would violate the electrical code in most countries, I'd shy away from making such a recommendation.
There's nothing mystical about the grounding of a chip amp. Some like to think it's some sort of black art, but the art is really only black until someone turns on the light. It's perfectly possible to design an LM3886-based amplifier which doesn't suffer from hum issues.
Tom
For good toroids in Europe, check Transformatory toroidalne - Producent transformatorów Toroidy.pl
You can also get custom transformers from audiophonics.fr . I don't know what their big ones are worth but I've no issues with those, up to 120VA.
You can also get custom transformers from audiophonics.fr . I don't know what their big ones are worth but I've no issues with those, up to 120VA.
I couldn't answer the transformer question without more information.
What is the wattage rating of the bass amplifiers?
What is the crossover frequency between the bass and midrange amplifiers?
What is the load and efficiency of the bass drivers?
What is the efficiency of the 8 ohm midrange drivers?
Hi Daniel
Sorry I didn't reply sooner, I was (and I'm still) writing a reply post to 00940 & tomchr about the transfo voltage sizing...
The bass amp is the home made early 80's Crescendo (1kVA supply), 2*140Wrms @ 8Ω
The cabs are home made too, in the same old days.
L 8Ω 80W 90dB@1W@1m) > LP 1to2kHz
M 8Ω 80W 92dB@1W@1m) same good flat from 1k to 4kHz (0 to 30° off axis)
H 6Ω 100W 90dB@1W@1m) > HP 2to4kHz
I want to start this stuff again, but I removed the passive cross-over (weird, I think because of inductors ferrites I randomly choose in these years).
I'll start with a behringer dcx2496 I bought some years ago when I got some parts for the 3886.
Whatever I'd prefer the 3886 build limitations/tradeoffs would not take into account an eventual undersizing of the bass amp, so that the build could take place in any other larger association, or even [a lower?] as standalone with basses handled.
These are info you missed I hope. IIRC there was some info in a THAT datasheet or application note which talked about the real music power spread along the audio frequency range.
Sorry I didn't reply sooner, I was (and I'm still) writing a reply post to 00940 & tomchr about the transfo voltage sizing...
The bass amp is the home made early 80's Crescendo (1kVA supply), 2*140Wrms @ 8Ω
The cabs are home made too, in the same old days.
L 8Ω 80W 90dB@1W@1m) > LP 1to2kHz
M 8Ω 80W 92dB@1W@1m) same good flat from 1k to 4kHz (0 to 30° off axis)
H 6Ω 100W 90dB@1W@1m) > HP 2to4kHz
I want to start this stuff again, but I removed the passive cross-over (weird, I think because of inductors ferrites I randomly choose in these years).
I'll start with a behringer dcx2496 I bought some years ago when I got some parts for the 3886.
Whatever I'd prefer the 3886 build limitations/tradeoffs would not take into account an eventual undersizing of the bass amp, so that the build could take place in any other larger association, or even [a lower?] as standalone with basses handled.
These are info you missed I hope. IIRC there was some info in a THAT datasheet or application note which talked about the real music power spread along the audio frequency range.
Last edited:
I remember that design. Elektor Electronics from the early/mid 80ies.
The LM3886 by itself is not so hot really. Put it into a composite loop and it really shines. Then of course I'm biased.
I seem to recall the frequency distribution being about 1/f for classical music and 1/(f^2) for compressed heavy rock. I could have the two distributions reversed, but at least that gives you an idea of what you'll be dealing with.
Tom
Thanks all!
Very good specs from Toroidy! The guy Tomasz Lachowski from them even sent me more easy to use data (full load=nominal voltage, no-load=+5%/coil, half-load=+2.5%/coil) than the AnTek ones (which are mismatched even inside a narrow family e.g AS12xx serie as they only give no-load/rough 66%-load/rough 120%-load that never match the nominal, although nominal may be about at ~50% load which would be good to know when we know Pd/Ps min is also ~50% at 6dB crest factor if we sometime ever want to ~ear~ deathmetal ~music ?~ at full power ...and even the Tom Christiansen guy added 50% headroom for the transfo rating ...
Very good news for me, Toroidy offers 1V step output voltages from 10 to 50V, 10VA steps from 20 to 60VA, 20/30/50/100 steps above: nice choice
I'm still somewhat puzzled on selecting the transfo because I can't see the bridge voltage drop in the spreadsheet maths and how the mains condition somewhere called "highline" are taken into account.
Whatever, after reading the Tom's 3886 thermal design taming I think I've got the good beginning: I already have the heatsinks, so this is will be my starting point. This afternoon I checked the 1°K/W Rth as specified by the manufacturer with the "taming" way #1 : @ ~50°K differential temp: Pd~48W, Tamb=(273+17)°K, Tsink=(273+67)°K...=> 1.05°K/W.
The resistors, 2 old TV 10*9*75mm wirewound models 9,1Ω//13Ω=5,35Ω gave 2,98A@16,03V from an old drawer 65VA toroid, clamped to the heatsink with thermal grease between pieces of wood, climbed up to 180°C... maybe not such good contact surfaces (although it is 15cm², 15 times higher than Aavid reference), or just overloaded although they didn't fuse for the afternoon (only a clamp plastic tooth melt in few seconds on first trial
). Some guess here: a screw tightened through the hole of the TA11B case small area will give a hugely higher pressure, then a far better heat transfer to the heatsink, hence a higher Tsa , so a lower Rsa, all fine. On the other hand, if I had a good Rcs, this means there is a problem with this method (maybe just in the way I understand it, or also because my test resistors have a Rca value so huge or low that it makes the method inaccurate).
Then, should I dare call this maths, I decide to give me headroom by "setting" Tamb to 40°C if I ever run the box under the summer sun.... so if I understand correctly, single chip per heatsink:
Tj max=150°C>40+(1.05+0,2+1)Pd gives me Pd max, isn't it? 0,2 being Rcs (thermal compound LM3886TA11B without isolating washer) and 1 beeing 1°C/W Tjc from 3886 datasheet. Rounding 1.05 to 1.1 for my possible temperature measures errors, Pd<110/2.3~47.8W (oh, nice close match with my test), no matter how hot is the heatsink, as long as any error in reasoning doesn't drive in frying the chip. If I'm not happy with that heat, I'll have to change the heatsinks
.
2nd step: although at a first glance I'll use 6 and 8Ω speakers, I want things simple (single equation for my 4 trivialized amps project) and allow me to use 4Ω if I want some day, and I want I'll never want less than 4Ω.
Looking at the spreadsheet, it seems I have to do the walk from right to left until I find the Vac from Vcc=1.1*Vac*√2-2V (1.1=highline;2V=bridge V drop) <=> Vac=(Vcc+2V)/1.1/√2.
... and also clean (read unclipped) hell's metal, i.e. C=6dB ...@ knob-11 'cause any setting below 11 turns hell in lullaby.
I tried to rewrite the spreadsheet formulas so that typing 47 (for 47,8W) in K row (Pd cell) would return Vcc in A row, but I fail in loop error: I'm a poor player at maths game.
Also, contributing to the difficulties I meet, Vdrop is just a constant in the spreadsheet, when it varies from 2.2V to 6V in DS fig.14. So I wrote a formula to stick to this figure (nested IF statements with linear functions of 5Vcc range each), because I prefer setting the sanity headroom in parameters that make sense to my common brain (e.g. Tamb) instead than leaving it in other uncontrolled ones like Vod:
line 6 below because I play with C=6dB as a constant.
Using the LibreOffice Calc Solver, target K6 Pd as static 47 value, by changing cells Vcc A6, limiting conditions 10.1<A6<39.9999 and B6=$'Vod vs Vcc @ 4Ω'.$A$1 (Vdrop) fails in "No solution - not a linear model"
Here where I am at the moment.
Have all a good day guys (no girls here?)
Very good specs from Toroidy! The guy Tomasz Lachowski from them even sent me more easy to use data (full load=nominal voltage, no-load=+5%/coil, half-load=+2.5%/coil) than the AnTek ones (which are mismatched even inside a narrow family e.g AS12xx serie as they only give no-load/rough 66%-load/rough 120%-load that never match the nominal, although nominal may be about at ~50% load which would be good to know when we know Pd/Ps min is also ~50% at 6dB crest factor if we sometime ever want to ~ear~ deathmetal ~music ?~ at full power ...and even the Tom Christiansen guy added 50% headroom for the transfo rating
...Very good news for me, Toroidy offers 1V step output voltages from 10 to 50V, 10VA steps from 20 to 60VA, 20/30/50/100 steps above: nice choice
hey tomchr, I see, more the maple leaf, you are the designer! Great!tomchr said:
I'm still somewhat puzzled on selecting the transfo because I can't see the bridge voltage drop in the spreadsheet maths and how the mains condition somewhere called "highline" are taken into account.
Whatever, after reading the Tom's 3886 thermal design taming I think I've got the good beginning: I already have the heatsinks, so this is will be my starting point. This afternoon I checked the 1°K/W Rth as specified by the manufacturer with the "taming" way #1 : @ ~50°K differential temp: Pd~48W, Tamb=(273+17)°K, Tsink=(273+67)°K...=> 1.05°K/W.
The resistors, 2 old TV 10*9*75mm wirewound models 9,1Ω//13Ω=5,35Ω gave 2,98A@16,03V from an old drawer 65VA toroid, clamped to the heatsink with thermal grease between pieces of wood, climbed up to 180°C... maybe not such good contact surfaces (although it is 15cm², 15 times higher than Aavid reference), or just overloaded although they didn't fuse for the afternoon (only a clamp plastic tooth melt in few seconds on first trial
). Some guess here: a screw tightened through the hole of the TA11B case small area will give a hugely higher pressure, then a far better heat transfer to the heatsink, hence a higher Tsa , so a lower Rsa, all fine. On the other hand, if I had a good Rcs, this means there is a problem with this method (maybe just in the way I understand it, or also because my test resistors have a Rca value so huge or low that it makes the method inaccurate).Then, should I dare call this maths, I decide to give me headroom by "setting" Tamb to 40°C if I ever run the box under the summer sun.... so if I understand correctly, single chip per heatsink:
Tj max=150°C>40+(1.05+0,2+1)Pd gives me Pd max, isn't it? 0,2 being Rcs (thermal compound LM3886TA11B without isolating washer) and 1 beeing 1°C/W Tjc from 3886 datasheet. Rounding 1.05 to 1.1 for my possible temperature measures errors, Pd<110/2.3~47.8W (oh, nice close match with my test), no matter how hot is the heatsink, as long as any error in reasoning doesn't drive in frying the chip. If I'm not happy with that heat, I'll have to change the heatsinks
.2nd step: although at a first glance I'll use 6 and 8Ω speakers, I want things simple (single equation for my 4 trivialized amps project) and allow me to use 4Ω if I want some day, and I want I'll never want less than 4Ω.
Looking at the spreadsheet, it seems I have to do the walk from right to left until I find the Vac from Vcc=1.1*Vac*√2-2V (1.1=highline;2V=bridge V drop) <=> Vac=(Vcc+2V)/1.1/√2.
... and also clean (read unclipped
) hell's metal, i.e. C=6dB ...@ knob-11 'cause any setting below 11 turns hell in lullaby.I tried to rewrite the spreadsheet formulas so that typing 47 (for 47,8W) in K row (Pd cell) would return Vcc in A row, but I fail in loop error: I'm a poor player at maths game.
Also, contributing to the difficulties I meet, Vdrop is just a constant in the spreadsheet, when it varies from 2.2V to 6V in DS fig.14. So I wrote a formula to stick to this figure (nested IF statements with linear functions of 5Vcc range each), because I prefer setting the sanity headroom in parameters that make sense to my common brain (e.g. Tamb) instead than leaving it in other uncontrolled ones like Vod:
line 6 below because I play with C=6dB as a constant.
Code:
New sheet [Vod vs Vcc @ 4Ω]
B1=Sheet1.A6 (Vcc)
A1=IF(B1<15;((2.55-2.2)*B1+15*2.2-10*2.55)/5;IF(B1<20;((3-2.55)*B1+20*2.55-15*3)/5;IF(B1<25;((3.45-3)*B1+25*3-20*3.45)/5;IF(B1<30;((4-3.45)*B1+30*3.45-25*4)/5;IF(B1<35;((4.8-4)*B1+35*4-30*4.8)/5;(6-4.8)*B1+40*4.8-35*6)/5))))Here where I am at the moment.
Have all a good day guys (no girls here?)
Last edited:
The 3886 can be run at upto 150°C and not be damaged.
One can use the worst case continuous dissipations with the heatsink and if the Tc never exceeds 150°C then you know you cannot damage the chipamp.
When driving music into speakers, the average dissipations are well below the worst case dissipations. As a result the chipamp and the heatsink operate at very much lower temperatures.
But be wary.
Look at the curve in the datasheet: Figure 35 Power Dissipation vs Output Power
for 4ohms with ±30Vdc, the highest dissipation is ~47W. But can the PSU hold 30V while the chipamp is delivering 42W to the load and dissipating 47W to the sink. I suspect not. So let's drop the PSU voltage down to ±25V while worst case dissipation is happening. We see 32W to the sink and 30W to the load. A total of 62W from the PSU, while it holds 25V at the supply pins.
Now reduce the output power to 10W, the dissipation goes down as well and the result is that the supply rails go back up closer to the quiescent value of ~±28Vdc.
I'll guess that ±27.5Vdc from the PSU while delivering 10W output and that dissipation is now around 30W. The chip and heatsink are still going to be very hot if the output is ONLY 10W continuous into 4r0.
Reduce the output power to 2W and assume the supply is now @ ±28Vdc. The graph indicates that the dissipation is very approximately 10W (your calculation should be able to confirm what this dissipation really is)
The total draw from the PSU is 12W even though the average output power is only 2W into 4r0. It is slightly worse if the load is reactive.
It is for this reason (the chip still dissipating significant heat) that I came to the conclusion that the National recommended heatsink (Figure 34) be doubled.
This doubling of the dissipation helps keep the chipamp internal temperature much lower, particularly when average output power is quite low. This has the effect of not allowing the "Spike" protection to trigger early due to increased internal temperature.
So looking at Figure 34 we see that @ ±28Vdc with a 4r0 load that National recommend a 1.9C/W when Ta=25°C
I recommend 0.95C/W for that operation. 1C/W would be close enough.
Note how close that doubled dissipation figure is to your 1.05C/W
Using the doubled National value gives a cool running chipamp that should not trip "Spike" too often due to the cooler running temperatures.
I saw an example of this while testing two chipamps on a common heatsink inside a two channel active crossover amp. One ran at 2W into 8r0 and the other ran at 1W into 8r0. This test was simulating average output when operating VERY LOUD.
After a couple of minutes I touched one of the chipamps. It nearly burnt my finger.
I rechecked figure 34, it shows 3C/W for ±35Vdc using an 8r0 load. Two chipamps on a common sink with my doubled allowance would require a 0.75C/W sink. I was using a 0.9C/W sink. The chips do indeed need a lot of cooling for low average output powers.
One can use the worst case continuous dissipations with the heatsink and if the Tc never exceeds 150°C then you know you cannot damage the chipamp.
When driving music into speakers, the average dissipations are well below the worst case dissipations. As a result the chipamp and the heatsink operate at very much lower temperatures.
But be wary.
Look at the curve in the datasheet: Figure 35 Power Dissipation vs Output Power
for 4ohms with ±30Vdc, the highest dissipation is ~47W. But can the PSU hold 30V while the chipamp is delivering 42W to the load and dissipating 47W to the sink. I suspect not. So let's drop the PSU voltage down to ±25V while worst case dissipation is happening. We see 32W to the sink and 30W to the load. A total of 62W from the PSU, while it holds 25V at the supply pins.
Now reduce the output power to 10W, the dissipation goes down as well and the result is that the supply rails go back up closer to the quiescent value of ~±28Vdc.
I'll guess that ±27.5Vdc from the PSU while delivering 10W output and that dissipation is now around 30W. The chip and heatsink are still going to be very hot if the output is ONLY 10W continuous into 4r0.
Reduce the output power to 2W and assume the supply is now @ ±28Vdc. The graph indicates that the dissipation is very approximately 10W (your calculation should be able to confirm what this dissipation really is)
The total draw from the PSU is 12W even though the average output power is only 2W into 4r0. It is slightly worse if the load is reactive.
It is for this reason (the chip still dissipating significant heat) that I came to the conclusion that the National recommended heatsink (Figure 34) be doubled.
This doubling of the dissipation helps keep the chipamp internal temperature much lower, particularly when average output power is quite low. This has the effect of not allowing the "Spike" protection to trigger early due to increased internal temperature.
So looking at Figure 34 we see that @ ±28Vdc with a 4r0 load that National recommend a 1.9C/W when Ta=25°C
I recommend 0.95C/W for that operation. 1C/W would be close enough.
Note how close that doubled dissipation figure is to your 1.05C/W
Using the doubled National value gives a cool running chipamp that should not trip "Spike" too often due to the cooler running temperatures.
I saw an example of this while testing two chipamps on a common heatsink inside a two channel active crossover amp. One ran at 2W into 8r0 and the other ran at 1W into 8r0. This test was simulating average output when operating VERY LOUD.
After a couple of minutes I touched one of the chipamps. It nearly burnt my finger.
I rechecked figure 34, it shows 3C/W for ±35Vdc using an 8r0 load. Two chipamps on a common sink with my doubled allowance would require a 0.75C/W sink. I was using a 0.9C/W sink. The chips do indeed need a lot of cooling for low average output powers.
Last edited:
Even a much smaller mids&treble amplifier can keep up easily if only needed at 1khz and higher. So there's no need to compromise fidelity on a push for power build--quite the opposite is true.
Tom's and AndrewT's VA figures will do fine in this case.
The transformer voltage you're looking for is 22+22vac or 22-0-22vac, for your LM3886.
Even then, it will be far more powerful than needed for that job.
Error in my formula: should be
I'm still wandering if, from the PSU point of view, Pload is included in Pd, or beside?
PSrms=Pidle+Pd
or
PSrms=Pidle+Pd+Pload
Code:
A1=IF(B1<15;((2.55-2.2)*B1+15*2.2-10*2.55)/5;IF(B1<20;((3-2.55)*B1+20*2.55-15*3)/5;IF(B1<25;((3.45-3)*B1+25*3-20*3.45)/5;IF(B1<30;((4-3.45)*B1+30*3.45-25*4)/5;IF(B1<35;((4.8-4)*B1+35*4-30*4.8)/5;((6-4.8)*B1+40*4.8-35*6)/5)))))Yes/No, doesn't the answer instead depends on the PSU sizing?AndrewT said:
I'm still wandering if, from the PSU point of view, Pload is included in Pd, or beside?
PSrms=Pidle+Pd
or
PSrms=Pidle+Pd+Pload
Which means 32.22Vcc, which in turn gives Vdrop=4.35 and ends in Pd=55.75W, higher than my limit.
It seems you forgot highline condition isn't it?
National have done the calculation and plotted the results on the 4r0 and 8r0 graphs.
You look up your supply voltage that exists while you are delivering the output power you require to investigate and read off the power dissipation.
I gave two examples. Read them off the graph.
The graph only shows the supplies in 5V increments. If you want something else then do the long hand calculations, or interpolate between the given curves.
If you want a better estimate of the dissipation at very low output levels, then once again long hand calculation will give the answers.
The outcome is that supply voltage and load current determine how hot the chipamp will run. Note, in all cases that maximum dissipation does not occur when maximum output is being delivered.
The 3886 is allowed to go as high as 84V with some signal present.
Allow 6% to 10% reduction to cater for mains supply voltage variations above the nominal supply voltage. 84V-7% ~ ±39Vdc
Then predict/model the transformer Vac to PSU Vdc to be no more than that safe value.
You'll find that 230:26+26Vac 7% regulation operating from 253Vac into a single bridge rectifier will be just under 84Vdc when a little bit of sag occurs as the quiescent current passes to the connected amplifier. The smoothing capacitors could see ~86V when the amplifier/s are not connected.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.