I have been doing some calculations about bridged LM3875 chips.
It seems that this chip is almost always thermaly limited, rather than limited by it's current output.
If it has an efficency of 60%, where if it is outputting 60W of power, it is disipating 40W. It's max continuous output current is 4A. So, lets say that the maximum output of each chip is 60W due to thermal limitations. Now lets put together 4 chips, all in inverting standard 10K input and 220K feedback. Just use an invering buffer to drive the bridged chips.
Drive a 4 ohm load, they all see 4 ohms, can supply a total of 240W into the 4 ohms, again due to thermal limitations. Same thing with 8 ohms, they can deliver up to 240 watts continuous due to thermal limitations of only being able to disipate 60Watts each, and 240 watts total.
If we use a +-25V PSU, we can swing up to 50V due to the bridged amplifier config minus some losses because it cant quite swing all that (i cant find the max swing percentage on the datasheet) and some loss due to power supply regulation. P=V^2/R, so voltage wise, we have about 625Watts into 4 ohms, and 312.5 into 8 ohms.
If you look at current, we can supply a maximum of 16Amps of continuous current. So Current wise, we have a 1024watts into 4 ohms, and 2048 watts into 8 ohms.
Therefore, I have a couple questions about bridging and just the 3875 in general.
1) How much power can this thing really disipate?
2) Why on earth dosn't national ever mention that the 3875 can drive 4 ohm loads? it seems that it would be perfectly possible, just as long as you keep the supply around +/-25V.
3) Are my calculations correct?
4) How much power, given a decent HS with about .85 *C/Watt can the LM3875 dissipate?
Thank you for your time,
Paul Hilgeman
It seems that this chip is almost always thermaly limited, rather than limited by it's current output.
If it has an efficency of 60%, where if it is outputting 60W of power, it is disipating 40W. It's max continuous output current is 4A. So, lets say that the maximum output of each chip is 60W due to thermal limitations. Now lets put together 4 chips, all in inverting standard 10K input and 220K feedback. Just use an invering buffer to drive the bridged chips.
Drive a 4 ohm load, they all see 4 ohms, can supply a total of 240W into the 4 ohms, again due to thermal limitations. Same thing with 8 ohms, they can deliver up to 240 watts continuous due to thermal limitations of only being able to disipate 60Watts each, and 240 watts total.
If we use a +-25V PSU, we can swing up to 50V due to the bridged amplifier config minus some losses because it cant quite swing all that (i cant find the max swing percentage on the datasheet) and some loss due to power supply regulation. P=V^2/R, so voltage wise, we have about 625Watts into 4 ohms, and 312.5 into 8 ohms.
If you look at current, we can supply a maximum of 16Amps of continuous current. So Current wise, we have a 1024watts into 4 ohms, and 2048 watts into 8 ohms.
Therefore, I have a couple questions about bridging and just the 3875 in general.
1) How much power can this thing really disipate?
2) Why on earth dosn't national ever mention that the 3875 can drive 4 ohm loads? it seems that it would be perfectly possible, just as long as you keep the supply around +/-25V.
3) Are my calculations correct?
4) How much power, given a decent HS with about .85 *C/Watt can the LM3875 dissipate?
Thank you for your time,
Paul Hilgeman
Paul,
There is a difference between bridge operations and parallel operations wrt load.
If you use a bridge with 4Ohms load, each side actually sees 2 Ohms. This is because if you riase the output voltage say 1 V, the other side of the load goes 1 V down, so the current increases twice what you would expect with 1V rise, so this acts as half the resistance.
In parallel operation, say 2 chips, again say one GC raises the output by 1V, the other also rises by 1 V, which means that each provides half the current for a 1v rise, which means it acts as twice the resistance, 8 Ohms in this case.
Jan Didden
There is a difference between bridge operations and parallel operations wrt load.
If you use a bridge with 4Ohms load, each side actually sees 2 Ohms. This is because if you riase the output voltage say 1 V, the other side of the load goes 1 V down, so the current increases twice what you would expect with 1V rise, so this acts as half the resistance.
In parallel operation, say 2 chips, again say one GC raises the output by 1V, the other also rises by 1 V, which means that each provides half the current for a 1v rise, which means it acts as twice the resistance, 8 Ohms in this case.
Jan Didden
Some of the calculations are a bit optimistic as far as I can see. The power in 4/8 should probably be RMS so that is P=V*V/2R giving 312.5 or 156.25 W respectively instead of the 625 or 312.5 you had. Furthermore the current into an 8 Ohm load will not be 16 A (you mean 4 chips per half giving a total of 8 chips per channel, right?) as the maximum swing 50 V will be voltage limiting the current to 50/8 = 6+ A peak and some 4.4 A RMS giving the same result as the previous.
And as mentioned the datasheet clearly indicates the use of a 4 Ohm load.
And as mentioned the datasheet clearly indicates the use of a 4 Ohm load.
OK, I see this now on the datasheet.
I forgot about the RMS thing, that clears it up a bit.
I fully understand how bridging and paralleling affects the load impedence, and voltage across the load. I guess I dont understand how it affects the current supplied to the load. Correct me if I am wrong: Bridging doubles the current delivered into the same load, paralleling it stays the same. Though, through paralleling, twice the current is available, which is why I could drive a load impedence half of the original value.
Sorry if it was confiusing how I presented it. I was looking at voltage and current separately, I do have an EE degree. I just forgot about the RMS.
So in conclusion, it really is true that in order to do any bridged gainclone into 4 ohms, you must parallel at least two chips. You could if you wanted to bridge two chips to do an 8 ohm load, but it would get WAY too hot and probably not be able to supply the current to drive a 4 ohm load unless played quietly.
If necessary, you could parallel 4 chips and then bridge them, being able to drive a 1 ohm load with a +/- 25V swing. That would be crazy.
-Paul Hilgeman
I forgot about the RMS thing, that clears it up a bit.
I fully understand how bridging and paralleling affects the load impedence, and voltage across the load. I guess I dont understand how it affects the current supplied to the load. Correct me if I am wrong: Bridging doubles the current delivered into the same load, paralleling it stays the same. Though, through paralleling, twice the current is available, which is why I could drive a load impedence half of the original value.
Sorry if it was confiusing how I presented it. I was looking at voltage and current separately, I do have an EE degree. I just forgot about the RMS.
So in conclusion, it really is true that in order to do any bridged gainclone into 4 ohms, you must parallel at least two chips. You could if you wanted to bridge two chips to do an 8 ohm load, but it would get WAY too hot and probably not be able to supply the current to drive a 4 ohm load unless played quietly.
If necessary, you could parallel 4 chips and then bridge them, being able to drive a 1 ohm load with a +/- 25V swing. That would be crazy.
-Paul Hilgeman
PaulHilgeman said:
no. you can not change the current into a given load. that would defy Ohm's Law.
adding devices, whether bridged or paralleled (or both), simply changes how much of what each chip is asked to provide to the load.
bridging doubles the voltage available to the load; it splits the voltage delivered between the two devices. it means into a given load for a given power level, when you go from single-ended to bridged, each device is providing the exact same current as before, but they are delivering half the voltage. so, the impedance each device "sees" is half. if an amp circuit was current-limited to begin with (as these chips are), it means there isn't much to be gained bridging; if they were voltage limited (e.g., run low supply rails), then you can get a very large power boost. the old Krell balanced amps are a good example of the latter, they can deliver a lot of current so when you bridge them, they deliver a ton of power.
actually i've been told that bridging a pair of chips is still useful, albeit with kind speaker loads (never going much below 8 ohms). some people have pulled it off in the Bridgeclone thread, and i think the device temperature is reasonable is the supply voltage is kept on the low side (e.g. +/- 22V).
i'm going to try bridging w/o paralleling. i have high impedance speakers (9 ohm minimum), and i really don't want to get into the whole parallel scenario... too many issues that detract from the "beauty" of the gainclone concept.
Bridging would provide many benefits, in theory. No actual ground connection on the speaker, and the DC offset would be greatly reduced (in offset would i nessence be the difference in offsets of the two chips).
But then again you get an added IC, but since the LM already is an IC packed with components adding another IC (balanced line driver) would only add some more to that already large number, so should cause two much harm.
As for I, bridging ain't my cup of tea, more parts, more heatsinking, an extra PSU (that is where I see the most posibility for loss in sound quality, a poor PSU power the line driver will ruin everything) and more space in the chasis (which I do not have, have to bend the pins on my bridge rectifiers to fit everything in).
But then again you get an added IC, but since the LM already is an IC packed with components adding another IC (balanced line driver) would only add some more to that already large number, so should cause two much harm.
As for I, bridging ain't my cup of tea, more parts, more heatsinking, an extra PSU (that is where I see the most posibility for loss in sound quality, a poor PSU power the line driver will ruin everything) and more space in the chasis (which I do not have, have to bend the pins on my bridge rectifiers to fit everything in).
//Lures Paul away to the dark side //
//////////////////
/////////////////
I know you want to make the uber
power amplifier... but howabout
a classic design like the 1kw Holton amp.
1.6kw @ 2 ohms short term, I want
to use more output transistors.
http://www.aussieamplifiers.com/1kwamp.htm
Build doc.
http://www.aussieamplifiers.com/downloads/AV800.pdf
Bridge mode = 2x more power.
I'm lagging - haven't built my prototype
yet...... sigh...
////////////////// /////////////////
//////////////////
/////////////////I know you want to make the uber
power amplifier... but howabout
a classic design like the 1kw Holton amp.
1.6kw @ 2 ohms short term, I want
to use more output transistors.
http://www.aussieamplifiers.com/1kwamp.htm
Build doc.
http://www.aussieamplifiers.com/downloads/AV800.pdf
Bridge mode = 2x more power.
I'm lagging - haven't built my prototype
yet...... sigh...
//////////////////
/////////////////
PaulHilgeman said:
i'm not sure there's much wiggle room for wording when it comes to Ohm's Law. "Bridging doubles the current delivered into the same load" is most definitely incorrect. sorry, not meaning to be a troll, just trying to make sure the matter is perfectly clear. i think your interpretation of bridging was a little off, but what you said about parallel devices is correct.
right, so you meant "bridging doubles the voltage available into the same load" correct?
(sorry, i know i'm nitpicking, this is purely for the benefit of others reading the thread... i know this matter of bridging gets confusing for some people and i've seen a lot of misinterpretation on the subject.)
(sorry, i know i'm nitpicking, this is purely for the benefit of others reading the thread... i know this matter of bridging gets confusing for some people and i've seen a lot of misinterpretation on the subject.)
thylantyr said:
if you mean my own webpages, unfortunately not, sorry. i was going to set up a whole diy section on my site, but i got distracted (and lazy) and never got to it.
i haven't really done anything in the last 6 months or so, just came back to this site last week to check up on how things were going in Peter's gainclone thread... whew i still haven't caught up there. my last project was the Borbely/White JFET buffer, which has been doing some funky things as of late so i stopped using it... it was supposed to be the first phase of development of my "ultimate multichannel preamp," but then i got a MSB unit that sounds so good it took a lot of steam out of my project. - Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.