Hi hurtz, yes ideally you should have separate traces to the common ground point, or as you say separate wiring back to the PSU ground point. A lot of people say to run the speaker returns direct to the psu rather than to the amp circuit board.
I wouldn't call the pin 7 connection low level
It is the main power ground for the chip. The input ground is the low level ground (as is the feedback ground)
In my p2p LM3886 I shared the input ground and feedback ground, but had separate wiring for output compensation, power ground and speaker return (load return in the datasheet). Ideally you should have a separate trace on the PCB for each of these, all joining at a common point which should then run back to your PS 0V point.
Tony.
I wouldn't call the pin 7 connection low level
It is the main power ground for the chip. The input ground is the low level ground (as is the feedback ground) In my p2p LM3886 I shared the input ground and feedback ground, but had separate wiring for output compensation, power ground and speaker return (load return in the datasheet). Ideally you should have a separate trace on the PCB for each of these, all joining at a common point which should then run back to your PS 0V point.
Tony.
Sorry, but I don't understand what you mean! The signal Ground has to go somewhere, and if not the star ground, where else? And also the star Ground is on the AMP and not on the PSU
When I built my first amp, I don't have the star ground on the PCB. I use flying wires to connect the grounds to a "common point". That common ground is connected back to the PSU 0 volt. With this arrangement I can easily reconfigure the grounding topology. When I able to make the amp quiet, I transferred the whole thing to one PCB.
Your common ground is not locate at the connector Gnd point.
Your common ground is not locate at the connector Gnd point.
Let's give the signal hot and signal return their correct names and forget "ground".
The signal is part of a circuit. The flow half and the return half.
Those are the two wires in your interconnect.
The RCA has two connections to receive the two wires from the interconnect, the signal hot and the signal return.
The RCA passes it's signal on to the amplifier +IN and -IN terminals going via the two input filters that you have put in there.
The Negative FeedBack (NFB) must also feed into -IN and thus demands that this lower leg of the NFB must connect to Signal Return.
The input circuit is now complete. The current can flow around the circuit.
BUT,
there needs to be a reference for the amplifier to determine what to do with the output voltage. A voltage reference, that carries virtually no current is required between the Signal Circuit and the Speaker/Output Circuit. This connection is only a voltage reference it is not really a part of any circuit. A wire/trace, even better with a bit of resistance, is used to reference the Signal Return to the Speaker Return. Note the similarity here to an opamp. It has +IN & -IN and Load Return is off board but voltage referenced to the input signal. The opamp has no "ground". It is designed to to operate and amplify with a ground pin. The reference between output and input returns is done external to the opamp. The same as described above.
The signal is part of a circuit. The flow half and the return half.
Those are the two wires in your interconnect.
The RCA has two connections to receive the two wires from the interconnect, the signal hot and the signal return.
The RCA passes it's signal on to the amplifier +IN and -IN terminals going via the two input filters that you have put in there.
The Negative FeedBack (NFB) must also feed into -IN and thus demands that this lower leg of the NFB must connect to Signal Return.
The input circuit is now complete. The current can flow around the circuit.
BUT,
there needs to be a reference for the amplifier to determine what to do with the output voltage. A voltage reference, that carries virtually no current is required between the Signal Circuit and the Speaker/Output Circuit. This connection is only a voltage reference it is not really a part of any circuit. A wire/trace, even better with a bit of resistance, is used to reference the Signal Return to the Speaker Return. Note the similarity here to an opamp. It has +IN & -IN and Load Return is off board but voltage referenced to the input signal. The opamp has no "ground". It is designed to to operate and amplify with a ground pin. The reference between output and input returns is done external to the opamp. The same as described above.
Hmm the negative feedback line you're talking about is the one running through Rf?
And the signal return i.e. IN- should be tied in with(colse to) the speaker return thus being the reference line, or do you mean the left pin of Rin being tied close to the speaker return as reference?
Sorry if I'm bugging you >.< maybe its easier for me to understand if you just paint in the connections you're talking about, I still have some probems determining wich net is exactly which.
Cheers
And the signal return i.e. IN- should be tied in with(colse to) the speaker return thus being the reference line, or do you mean the left pin of Rin being tied close to the speaker return as reference?
Sorry if I'm bugging you >.< maybe its easier for me to understand if you just paint in the connections you're talking about, I still have some probems determining wich net is exactly which.
Cheers
To my eye your starGND does not look that bad...
I think (but I might be very wrong) that you could actually make this work.
But: If I was you, I would loose the two gnd connections between the Vn and Vp.
Instead I would put only one GND connection in the middle of your 'star', right above the Rz left connection point.
To that new gnd-everything (make it BIG) you connect your PSU and the crazy large reservoir capacitors and what ever you use to filter your power supply...
All with separate wires!
Your speaker return would be OK as it is. Your input return would be OK as it is.
Or you go with the Piggy's idea...
Or you go with Andrew...
But as I said, I might be very wrong here now... The grounding is a difficult thing to do perfectly.
I think (but I might be very wrong) that you could actually make this work.
But: If I was you, I would loose the two gnd connections between the Vn and Vp.
Instead I would put only one GND connection in the middle of your 'star', right above the Rz left connection point.
To that new gnd-everything (make it BIG) you connect your PSU and the crazy large reservoir capacitors and what ever you use to filter your power supply...
All with separate wires!
Your speaker return would be OK as it is. Your input return would be OK as it is.
Or you go with the Piggy's idea...
Or you go with Andrew...
But as I said, I might be very wrong here now... The grounding is a difficult thing to do perfectly.
Been following with interest as I've got a spare 3886 in my parts drawer. Here's my 2p's worth!.
Regarding the board in post 42 - looks like you've got provision for both a zobel network (Rz/Cz) and, I think, a Thiele network (R and L). Normally L is wound around R so removing L from your board means you've got room for Ci. I think I'm right in saying you don't use a zobel and a thiele, it's one or the other (or at least I've never seen both used together).
DaveR
Regarding the board in post 42 - looks like you've got provision for both a zobel network (Rz/Cz) and, I think, a Thiele network (R and L). Normally L is wound around R so removing L from your board means you've got room for Ci. I think I'm right in saying you don't use a zobel and a thiele, it's one or the other (or at least I've never seen both used together).
DaveR
Since a friend of mine wants to buy one assembled board from me, I'm eager to deliver a very good sounding build with this chip...
@palstanturhin: good idea, I relocated the GND's to a more convenient position.
I've found that the LM4780 datasheet is actually a much better source concerning the optional components (figure 5). Although I started with the minimal design approach of the point-to-point-lm3886 it seems when transferring this to a PCB the optional components should all be used, and as far as my knowledge goes, they're overall just a series of in-/out-/feedback filters that the p2p-design omits.
So thinkning about this I guess I might as well include Ci, and Rf2/Cf before going through 3 more design changes just to end up there anyways.
@DaveR: Hmm as I understand it, the zobel is "to stabilize the output stage by creating a pole that reduces high frequency instabilities." and the Thiele for "Provides high impedance at high frequencies so that R may decouple a highly capacitive load and reduce the Q of the series resonant circuit. Also provides a low impedance at low frequencies to short out R and pass audio signals to the load." to directly quote the datasheet.
Or in short terms as I understand it: Zobel: remove high frequency instabilities, and thiele to act as an output RF-filter.
AMP: v4.3
cheers!
@palstanturhin: good idea, I relocated the GND's to a more convenient position.
I've found that the LM4780 datasheet is actually a much better source concerning the optional components (figure 5). Although I started with the minimal design approach of the point-to-point-lm3886 it seems when transferring this to a PCB the optional components should all be used, and as far as my knowledge goes, they're overall just a series of in-/out-/feedback filters that the p2p-design omits.
So thinkning about this I guess I might as well include Ci, and Rf2/Cf before going through 3 more design changes just to end up there anyways.
@DaveR: Hmm as I understand it, the zobel is "to stabilize the output stage by creating a pole that reduces high frequency instabilities." and the Thiele for "Provides high impedance at high frequencies so that R may decouple a highly capacitive load and reduce the Q of the series resonant circuit. Also provides a low impedance at low frequencies to short out R and pass audio signals to the load." to directly quote the datasheet.
Or in short terms as I understand it: Zobel: remove high frequency instabilities, and thiele to act as an output RF-filter.
AMP: v4.3
An externally hosted image should be here but it was not working when we last tested it.
cheers!
Last edited:
no! Pin 7 does not carry much current and a reasonable track with, that can be home etched easily would be fine. Same goes for the input pins.
Touching unused pins with tracks is bad style, if not worse. Cutting of unused pins to have more room on the pcb is a really desperate measure.
I think you have painted yourself into a corner.
regards
http://img155.imageshack.us/img155/5302/ampv43.png
so pin 2,6,11 have vanished from your pcb by magic?
so pin 2,6,11 have vanished from your pcb by magic?
I find this design somewhat strange, for example using wider track widths at the input pins than at the amps output etc.
Last edited:
Yes I have removed the Pads for those pins, but when drilling everyone can decide whether he wants to drill holes for them or clip them. As for pin 11, the point-to-point method of wiring is considered to be OK, but clipping a pin thats not in use to decrease track length is not?
Also most of the tracks were made as big is possible, if only to save on etching solution. The Output-track has a with of 1.27mm at the moment.
And please cut me some slack, this is the first PCB I've ever designed...
Also most of the tracks were made as big is possible, if only to save on etching solution. The Output-track has a with of 1.27mm at the moment.
And please cut me some slack, this is the first PCB I've ever designed...
The small value decoupling caps wont do much as they require a short leads and a small loop, with the main ground going from the chip back to the star then the ground from the star to the caps the track inductance will be dominant, reducing the effective frequencys they work at. The small caps should be next to the pins and the tracks (power and ground) as short as possible.
I've noticed your progress and you certainly deserve lots of praise for it!
nevertheless..
+1
agree with most other comments ie like leaving off IC lead pins.
youre right about light IC ground currents, BUT decoupling need low inductance trace widths to chip amps internal s.
TIP > one or two well placed wire jumpers can make or break a single sided layout. used on signals or outputs can free up valuable routing area.
well the OP started with a design based partly on marketing speak ie minimal parts gain clone. He was then directed to NS data sheets where an engineer still needs to be able to read between the lines to seperate the wheat from the chaff so to speak, but it's all in there. All chip amps have to deal with the same physics or as you say non-ideal qualities. I'd rather include more parts for a robust amp design than some other thing with all it's limitations.
And I was close to thinking I'm just to stupid for this
Well, I can't talk about "well" placed, but the capacitor-loop is definitely shorter now.
Thanks
I'm still working on the nevertheless...Still, clipped pin 11 stays, if only for individualistic sake.
AMP: v4.4
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.