My web page at http://home.pacific.net.au/~gnb/audio/lm4780.html#gndpins has some words to say about the GND pin on the LM4780 (the dual LM3886). I recommend that it be connected to the power supply star earth, and not the input signal earth.
An externally hosted image should be here but it was not working when we last tested it.
I use that PCB layout for 4 different project. They work very robust and noiseless. I recommend you to connect pin7 to signal gnd. Because It doesnt carry anykind of current. Its for op-amp reference point. I guess!
Dxvideo said:
On such a tight layout where the ground traces are so short, it wouldn't really matter where the GND pin is located on the trace.
The GND pin has to carry some type of current, otherwise it wouldn't be necessary! Its of the order of 1mA, which won't add any measurable distortion on traces which are a few MicroOhms long.
Hi Dxv,
I like the location of CS1 & 2, with the ground pins very close together. A short and effective HF loop with minimal inductance.
Have you located both these caps on the reverse side?
Have you tried connecting a rail to rail cap across pins 4 to 5?
Have you snipped off the unused pins?
I like the location of CS1 & 2, with the ground pins very close together. A short and effective HF loop with minimal inductance.
Have you located both these caps on the reverse side?
Have you tried connecting a rail to rail cap across pins 4 to 5?
Have you snipped off the unused pins?
From the datasheet and internal schematic of the LM3886 it can be seen that the GND pin has several functions.
1 - its voltage vs. the V- pin is monitored, keeping the chip in mute mode when it is less than 9V.
2 - it defines the output potential in mute/startup mode. Note the 10k on the left side of the inner diff amplifier. This diff provides a closed feedback loop in mute mode with the right hand 10k and the clamping diodes. These diodes go to ground, thus in normal operation there is a current path from the output to GND, via the 10k, about 0.1mA per Volt of output voltage.
3 - it's the reference potential for the mute logic. When current is drawn from the mute pin (towards V-) in the ON mode, the startup diff's current source gets turned off by the grounded base transistor on the left. The GND pin provides the base current for it.
4 - there must be something missing here, because from the single-supply schematic I realize that the GND pin shall only sink current in normal operation. OTOH, that schem might be wrong/flawed.
Especially point #2 indicates that the current into/out of the GND pin should not be routed to signal GND, in theory, as this nonlinear current might contaminate the signal GND. In fact it's so small (well below 10mA) that this is not a problem in a proper design/layout. Likewise, if this GND pin carries some noise, it won't affect operation -- but avoid excessive HF noise, of course.
So, my 2 cts:
If you have the option, don't bond it to signal GND. But don't sacrfice a good layout just for that tidbit.
Regards, Klaus
1 - its voltage vs. the V- pin is monitored, keeping the chip in mute mode when it is less than 9V.
2 - it defines the output potential in mute/startup mode. Note the 10k on the left side of the inner diff amplifier. This diff provides a closed feedback loop in mute mode with the right hand 10k and the clamping diodes. These diodes go to ground, thus in normal operation there is a current path from the output to GND, via the 10k, about 0.1mA per Volt of output voltage.
3 - it's the reference potential for the mute logic. When current is drawn from the mute pin (towards V-) in the ON mode, the startup diff's current source gets turned off by the grounded base transistor on the left. The GND pin provides the base current for it.
4 - there must be something missing here, because from the single-supply schematic I realize that the GND pin shall only sink current in normal operation. OTOH, that schem might be wrong/flawed.
Especially point #2 indicates that the current into/out of the GND pin should not be routed to signal GND, in theory, as this nonlinear current might contaminate the signal GND. In fact it's so small (well below 10mA) that this is not a problem in a proper design/layout. Likewise, if this GND pin carries some noise, it won't affect operation -- but avoid excessive HF noise, of course.
So, my 2 cts:
If you have the option, don't bond it to signal GND. But don't sacrfice a good layout just for that tidbit.
Regards, Klaus
Mick_F said:Of course, the ground pin of the LM should be connected to signal ground, not to power ground.
Look at the reference board of the LM4780! Very useful
Mick
I agree with KSTR's comments. The schematic does not show everything. NS do not want to give away proprietary design information !
Just because the reference / evaluation board does something it a certain way, it doesn't mean that they have researched it to be the absolute best way ! It may have just suited them at the time.
It would be interesting to add a 10 Ohm resistor between the clean signal earth and the dirty power supply earth, and then measure the output THD etc when connecting the GND pin to either the clean or dirty earth. It would prove the point one way or the other and provide a magnitude of the effect. It may vindicate or eradicate my assumptions !
Dear Andrew,
Sorry for late answering. But I am working on my new Class A project as you know.
Anyway,
Yes as you see CS1&2 is on bottom layer. I saw a lot of benefits to keep there close to the power pins, especially on HF interferences.
For rail to rail caps; I've tried that with many different values of caps; 47uF + 100nF, 100uF + 470nF however I couldnt hear any difference on sound. Some diyers recomment that way very strongly but I cannot. For example in my LM4702 project I applied the original factory design for MOSFETs. And in this configuration there was a 100nF poly between rails. I tried the circuit first with that cap and I feel some bass weakness, after then tried without it and the same bass weakness was still exist. But when I put extra 1000µFs on each rail to gnd then bass response was improved A BIT but not much.
And for the unused pins. Generally I trim that pins, so they never be soldered to the PCB anyway.
...
Dear Glenn,
I havent noticed that pin 7 is sinking 1mA current. Thats interesting. Is this the main difference between LM3875 and 3886? Because as you know, there is no GND pin on LM3875.
Thx for the reminding.
Sorry for late answering. But I am working on my new Class A project as you know.
Anyway,
Yes as you see CS1&2 is on bottom layer. I saw a lot of benefits to keep there close to the power pins, especially on HF interferences.
For rail to rail caps; I've tried that with many different values of caps; 47uF + 100nF, 100uF + 470nF however I couldnt hear any difference on sound. Some diyers recomment that way very strongly but I cannot. For example in my LM4702 project I applied the original factory design for MOSFETs. And in this configuration there was a 100nF poly between rails. I tried the circuit first with that cap and I feel some bass weakness, after then tried without it and the same bass weakness was still exist. But when I put extra 1000µFs on each rail to gnd then bass response was improved A BIT but not much.
And for the unused pins. Generally I trim that pins, so they never be soldered to the PCB anyway.
...
Dear Glenn,
I havent noticed that pin 7 is sinking 1mA current. Thats interesting. Is this the main difference between LM3875 and 3886? Because as you know, there is no GND pin on LM3875.
Thx for the reminding.
Maybe more effective might be to couple some small test signal (100Hz bw-limited square) directly into the GND pin and see how much feeds through to the output, with no signal at the input. Most easily seen with a high noise gain setup (like 100 or so).glennb said:
Makes me think about the option to generate a seperate GND potential for that pin that is stable with reference to the neg supply. When there are internal current sources etc. attached to the GND pin that also might explain the weak -PSRR curves of the chip, in case this isn't mainly due to the miller cap -- creating the amps dominant pole -- which semms to reference to the V- supply. And maybe the Miller cap is actually hooked to the GND pin. A lot of vague guesswork...
Regards, Klaus
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.