| Circlotron |
I was looking at the internal cct of an LM3875 today and got to thinking about the fact that the positive supply rail has somewhat better supply rejection than the negative rail. A look at the cct shows that there is a whole pile of things referenced to the neg rail. Some go through current sources at least while others go through just a resistor. This being the case, it is no wonder that the neg rail has more of an influence on the internal goings-on.
Then I thought - how about an amplifier that has the input signal referenced to the neg rail instead of zero volts? Of course it would mean the output would have to be capacitively coupled to the load, but it would also mean that the internal circuitry would have the chance to operate in a more "pure" manner. The neg rail could jump around all it likes, but provided it is connected to the input signal earth it will not function as an unintended "input" to the chip. What do others think? |
|
|
| Circlotron |
| quote: | Originally posted by Christer
[B]Or maybe it would help to just do the decoupling "right" |
Ah, yes. Better get me a pair of 1000uF caps then. :dodgy:
Thanks for the link, Christer. :) |
|
|
| Circlotron |
Well, then. I'm going to go ahead with this project even if no-one else is excited about it. :bawling: First question concerns the tab of a non-isolated LM3875 - the datasheet says that if the tab is connected directly to a heatsink (i.e. no insulation) then the heatsink should be connected to the negative rail. What I want to know is this - is the tab actually a direct connection to the negative rail pin of the package? It would take 5 seconds to test with a multimeter but all 3 of mine are the isolated type package. Could someone fill me in here?
If the tab *is* in fact a direct path to the negative supply connection on the die, then it would have to be the shortest, most direct, lowest inductance path possible. If the tab were bolted and /or soldered to a copper groundplane + heatsink, with a surface mount ceramic cap bypassing the positive rail down to the groundplane right where the lead comes out of the plastic case then we have the possibility of some seriously good earthing. :smash:
Everyone talks about short feedback paths for getting the best out of a gainclone. I even read somewhere about cutting away part of the plastic case to make the connection shorter still! Seeing the supply rail rejection on the negative rail is only fairly good, then perhaps there is just a little to be gained ( :rolleyes: ) by using a similarly meticulous approach to earthing and bypassing.
I've decided to go for the bridge approach because I couldn't stomach the idea of an output coupling capacitor. I didn't want to be 'straining out gnats but gulping down camels'. |
|
|
| Christer |
| quote: | Originally posted by Circlotron
I've decided to go for the bridge approach because I couldn't stomach the idea of an output coupling capacitor. I didn't want to be 'straining out gnats but gulping down camels'. |
That sounds like a good idea. You ought to have some DC
protection circuitry with a relay in series with the speaker
though. It could be a bit hot for the speaker if one of the
amps fail or just decides to do a thermal shutdown or
something. Always a good idea to include anyway, I guess,
but it seems more necessary here. |
|
|
| Circlotron |
| quote: | Originally posted by Circlotron
First question concerns the tab of a non-isolated LM3875 - the datasheet says that if the tab is connected directly to a heatsink (i.e. no insulation) then the heatsink should be connected to the negative rail. What I want to know is this - is the tab actually a direct connection to the negative rail pin of the package? It would take 5 seconds to test with a multimeter but all 3 of mine are the isolated type package. Could someone fill me in here? |
Could they, please? |
|
|
| pinkmouse |
| quote: | Originally posted by Circlotron
If the tab *is* in fact a direct path to the negative supply connection on the die, then it would have to be the shortest, most direct, lowest inductance path possible. . |
I seem to remenber that the datasheet notes that you shouldn't use the negative tab as a supply rail, might be worth checking... |
|
|
| Circlotron |
| Yeah, I did look at the data sheet, but nowhere did it tell me what I really wanted to know, and that is - is the negative supply rail pin a direct connection to the tab??? |
|
|
| pinkmouse |
| quote: | Originally posted by Circlotron
Yeah, I did look at the data sheet, but nowhere did it tell me what I really wanted to know, and that is - is the negative supply rail pin a direct connection to the tab??? |
Might be worth dropping a line to National's tech support then, I might be confusing it with the OPA548/9s ;) |
|
|
| Circlotron |
| Aw, come on you blokes! Can't one of you just do a little measurement with their multimeter? Huh? Huh? Huh? |
|
|
| pinkmouse |
The problem is that you may get conductivity, but the connection may be too fine to pass much current through!:smash: :bawling:
But I will look for you tonight! ;) |
|
|
| sreten |
| quote: | Originally posted by Circlotron
Well, then. I'm going to go ahead with this project even if no-one else is excited about it. :bawling: First question concerns the tab of a non-isolated LM3875 - the datasheet says that if the tab is connected directly to a heatsink (i.e. no insulation) then the heatsink should be connected to the negative rail. What I want to know is this - is the tab actually a direct connection to the negative rail pin of the package? It would take 5 seconds to test with a multimeter but all 3 of mine are the isolated type package. Could someone fill me in here? |
Its probably just connected to the substrate, which needs to
be at least as negative as the most negative part of the circuit.
As others have suggested it probably is connected to -ve
internally, but not very robustly, as it doesn't need to be.
Your right about the negative rail and PSRR, its inherent in the
topology of the amplifier, due to the VAS sitting on the -ve rail.
Feedback references the VAS to the input earth so -ve PSRR
tracks the open loop gain of the amplifier, as shown by D.Self.
:) sreten. |
|
|
| sreten |
| quote: | Originally posted by Circlotron
The neg rail could jump around all it likes, but provided it is connected to the input signal earth it will not function as an unintended "input" to the chip. What do others think? |
Well there will not be a "negative" rail to jump about as it
will be the earth rail. You have to generate a 1/2 +ve
reference point for the input but this is easy to decouple
from the pertubations of the positive rail.
:) sreten. |
|
|
| MWP |
| quote: | Originally posted by Circlotron
Aw, come on you blokes! Can't one of you just do a little measurement with their multimeter? Huh? Huh? Huh? |
From memory they are.
I cant test it though (i dont have any spare).
I blew the cr@p out of 8 LM3886's by shorting the tabs with ground accidently.
Bits of ceramic went flying everywhere :xeye: |
|
|
| HDTVman |
You know of course that what you are talking about here is the single supply version shown in the National app. note right? The idea to bridge the circuit to eliminate the output cap is a good one. I think that would be worth while.
I have done the single supply version of the 3875 shown in the app note to replace the amp stage in an old Dynaco SCA80 which used a +72 VDC power supply, and it worked great. I used the output cap coupling with good results.
Later BZ |
|
|
| Circlotron |
| Did anyone make a measurement for me? |
|
|
| peranders |
| quote: | Originally posted by Circlotron
Yeah, I did look at the data sheet, but nowhere did it tell me what I really wanted to know, and that is - is the negative supply rail pin a direct connection to the tab??? |
Probably yes, because the tab is usually in one peice with the normal supply pin.
Hey, wait...
My LCR bridge PRISM Elelctronics says 6.7 mohms between tab and V- (pin 4) => My statement seems to be true. :nod:
I tested an LM3875T.
I see no reason why you couldn't have V+ as signal ground and V- as supply. Of course you must have AC coupled out and input. |
|
|
| jackinnj |
| quote: | Originally posted by Circlotron
Did anyone make a measurement for me? |
The tab on the LM3875 is attached to pin 4.
Jack |
|
|
| Circlotron |
| quote: | Originally posted by peranders
My LCR bridge PRISM Elelctronics says 6.7 mohms between tab and V- (pin 4) |
| quote: | Originally posted by jackinnj
The tab on the LM3875 is attached to pin 4.
Jack |
:confused: :confused: :confused: |
|
|
| peranders |
| quote: | Originally posted by Circlotron
:confused: :confused: :confused: |
| quote: | Originally posted by Circlotron
Yeah, I did look at the data sheet, but nowhere did it tell me what I really wanted to know, and that is - is the negative supply rail pin a direct connection to the tab??? |
Yes!
Did you wonder about something more? |
|
|
| jackinnj |
| quote: | Originally posted by Circlotron
:confused: :confused: :confused: |
Pin 4 is the neg rail --- we were just having fun with you
:) |
|
|
|
