Hi All,
I'm new here and have been experimenting with a Class D design that is based on the IRAUDAMP7s reference design. It has been working well for about 6 months as a stereo amp (both amp channels on one PCB as in the reference design). The power supply is a linear design with a 25A bridge, 750VA transformer and 44,000uF per rail. I'm running +/-50V rails. Signal GND is connected to safety ground through a 25A bridge, 100nF cap and a 100ohm resistor. Safety GND is connected directly to the chassis, as required, close to the IEC inlet with a heavy solid core wire. This amp is totally free of hum and buzz with a very low noise floor. Based on the quality of this amp I decided to build a couple more boards and make a 6-channel version with all boards sharing the same linear supply. I have applied "textbook" star grounding with the following directly connected to it: Transformer centre tap, power supply GND, each amp board power GND, each speaker GND, and the chassis GND disconnecting network. Input GND and shield are connected directly to each amp input and the RCA inputs are isolated from the chassis. No matter what I do I get a slight hum and some buzz with an increase in overall noise floor when I connect the outputs of my AVR. I have tried taking the input GND/shields and removing them from the amp boards and connecting them to the star ground point. Noise increases when I do this. I have removed the disconnecting network and tied everything to safety GND. Noise increases. I tied all of the input GNDs directly to their respective speaker GNDs at the speaker binding posts and noise reduced slightly. Finally I have tied the transformer centre tap to the power supply GND directly instead of to the star GND. Noise reduces very slightly. Is there any special technique I should be looking at that specifically applies to Class D amps? This must be a grounding issue as everything is silent if I only have inputs connected to one amp board.
Many thanks for your help!
I'm new here and have been experimenting with a Class D design that is based on the IRAUDAMP7s reference design. It has been working well for about 6 months as a stereo amp (both amp channels on one PCB as in the reference design). The power supply is a linear design with a 25A bridge, 750VA transformer and 44,000uF per rail. I'm running +/-50V rails. Signal GND is connected to safety ground through a 25A bridge, 100nF cap and a 100ohm resistor. Safety GND is connected directly to the chassis, as required, close to the IEC inlet with a heavy solid core wire. This amp is totally free of hum and buzz with a very low noise floor. Based on the quality of this amp I decided to build a couple more boards and make a 6-channel version with all boards sharing the same linear supply. I have applied "textbook" star grounding with the following directly connected to it: Transformer centre tap, power supply GND, each amp board power GND, each speaker GND, and the chassis GND disconnecting network. Input GND and shield are connected directly to each amp input and the RCA inputs are isolated from the chassis. No matter what I do I get a slight hum and some buzz with an increase in overall noise floor when I connect the outputs of my AVR. I have tried taking the input GND/shields and removing them from the amp boards and connecting them to the star ground point. Noise increases when I do this. I have removed the disconnecting network and tied everything to safety GND. Noise increases. I tied all of the input GNDs directly to their respective speaker GNDs at the speaker binding posts and noise reduced slightly. Finally I have tied the transformer centre tap to the power supply GND directly instead of to the star GND. Noise reduces very slightly. Is there any special technique I should be looking at that specifically applies to Class D amps? This must be a grounding issue as everything is silent if I only have inputs connected to one amp board.
Many thanks for your help!
Noise
Ground planes are essential for mulitchannel modules.
Each module MUST have its own ground plane with plus and minus rails running directly to their respective buss capacitors. The ground for each module must also return to a "T" ground. This is done quite simply. The main ground busbar between the +Vcc caps and -Vcc caps is (and should be) a single high current busbar arrangement. The transformer's centre tap must be connected at the centre point of the busbar via a "T: section. This means that the transformer wires themselves actually "tee" into the centre point of the bus bar.
Now ALL your modules ground returns are run separately to a true "T" stub which physically protrudes from the centre point of the busbar. Let's assume it is 25mm long. Your module's ground returns are soldered onto this T only. The high ground currentls flowing from the plus capacitors to the minus capacitors CANNOT escape the busbar and "jump" onto the "T" section. So no noise can enter the grounds.
The speaker grounds MUST only be taken from each class D module itself and NOT from the "T" point.
Keep input leads away from speaker and rail leads. Twist the rail wires for each module as well
Steve Mantz
Zed Audio
Ground planes are essential for mulitchannel modules.
Each module MUST have its own ground plane with plus and minus rails running directly to their respective buss capacitors. The ground for each module must also return to a "T" ground. This is done quite simply. The main ground busbar between the +Vcc caps and -Vcc caps is (and should be) a single high current busbar arrangement. The transformer's centre tap must be connected at the centre point of the busbar via a "T: section. This means that the transformer wires themselves actually "tee" into the centre point of the bus bar.
Now ALL your modules ground returns are run separately to a true "T" stub which physically protrudes from the centre point of the busbar. Let's assume it is 25mm long. Your module's ground returns are soldered onto this T only. The high ground currentls flowing from the plus capacitors to the minus capacitors CANNOT escape the busbar and "jump" onto the "T" section. So no noise can enter the grounds.
The speaker grounds MUST only be taken from each class D module itself and NOT from the "T" point.
Keep input leads away from speaker and rail leads. Twist the rail wires for each module as well
Steve Mantz
Zed Audio
Thank you for the detailed response Steve!
The ground planes on the amp PCBs are identical to the IR reference design with a large, central ground plane running through the middle of the board with a straight path from input signal ground to power supply ground. The ground returns for the speaker outputs have their own planes connected to the central ground plane through the heat spreader bar. I have a small power supply PCB with a large central ground plane surrounded by the plus and minus rail traces, which are on either side of the ground. The main smoothing caps are connected to this ground buss, along with the transformer center tap and the ground returns form each amp PCB. I'm not sure I understand the "T" concept you are speaking about. Is this something you could sketch? Interesting that the speaker ground must come directly from the amp boards. This is a point that is contradicted by many of the companies selling DIY class D amps! I'll try this out and report back.
Thanks!
The ground planes on the amp PCBs are identical to the IR reference design with a large, central ground plane running through the middle of the board with a straight path from input signal ground to power supply ground. The ground returns for the speaker outputs have their own planes connected to the central ground plane through the heat spreader bar. I have a small power supply PCB with a large central ground plane surrounded by the plus and minus rail traces, which are on either side of the ground. The main smoothing caps are connected to this ground buss, along with the transformer center tap and the ground returns form each amp PCB. I'm not sure I understand the "T" concept you are speaking about. Is this something you could sketch? Interesting that the speaker ground must come directly from the amp boards. This is a point that is contradicted by many of the companies selling DIY class D amps! I'll try this out and report back.
Thanks!
Always follow the current - it doesn't always flow the way you expect.
Here's a picture of a very simplified D-Amp. I don't know if this picture will be clear or not, but here goes - you want the contain the high-current so it stays within the output driver only. Look at the picture, and consider a positive going transition - S1 would be on, S2 off. You want the current to flow to the speaker, and back to C1 in the exact path sketched (Bob Pease styel 🙂)
R1, R2 and R3 represent the finite impedance you have in traces and/or planes back to the supply. *If*, you chose to have the speaker return path going to the chassis-ground point, you can get a voltage drop across R2 - or another way to think of it is that there is some wiggle at R2.
Now consider what happens when multiple channels are running at the same time. They all have a common return path, and the shared current creates cross-talk. I believe this is what Steve is saying.
On the input side - you also want to pay attention to the current flow. Your AVR, RCA cables - presumable single-ended coax, ground + signal? Or are you using differential inputs? From your source, the signal must go to its load (the input buffer on your amp?) and return to the AVR. The return path is must be through that very same cable.
Other stuff - make sure your AVR and D-Amp are plugged into the very same outlet.
hope this helps
gene
Here's a picture of a very simplified D-Amp. I don't know if this picture will be clear or not, but here goes - you want the contain the high-current so it stays within the output driver only. Look at the picture, and consider a positive going transition - S1 would be on, S2 off. You want the current to flow to the speaker, and back to C1 in the exact path sketched (Bob Pease styel 🙂)
R1, R2 and R3 represent the finite impedance you have in traces and/or planes back to the supply. *If*, you chose to have the speaker return path going to the chassis-ground point, you can get a voltage drop across R2 - or another way to think of it is that there is some wiggle at R2.
Now consider what happens when multiple channels are running at the same time. They all have a common return path, and the shared current creates cross-talk. I believe this is what Steve is saying.
On the input side - you also want to pay attention to the current flow. Your AVR, RCA cables - presumable single-ended coax, ground + signal? Or are you using differential inputs? From your source, the signal must go to its load (the input buffer on your amp?) and return to the AVR. The return path is must be through that very same cable.
Other stuff - make sure your AVR and D-Amp are plugged into the very same outlet.
hope this helps
gene
Thank you Gene! Your explanation makes perfect sense. I changed the speaker ground return connections to the amplifier PCBs and the background noise level dropped slightly. This was mainly a reduction in "buzz". I would say that the level of 120Hz hum is still far higher than I'd like. I quickly tried changing the power supply filtering to a CRC configuration with 0.22R resistors. Happily this resulted in a significant reduction in the residual hum. Being picky, I'd still like to make the noise floor even lower. Knowing that the PSRR of this amp design is quite good, I'm surprised that the change to the power supply made such a big improvement, even though there is no sign of any of this noise with only one stereo amp board running. My AVR is plugged into the same circuit as the amp and I'm using standard RCA cables. I have doubled checked that all connections are good. Do you (or anyone else) have any idea where I should next concentrate my efforts?
PSRR will not necessarily reject noise on ground, it definitely sounds to me as if you have a ground loop via your AVR. have you looked at the system grounding article here on the forum? can you draw up a schematic to describe your setup? or some photos?
Last edited:
you say the supply is linear - so the transformer is torroid or other? Maybe the problem lies there and it needs to be shielded? Torroid provides some self-shielding but iron core probably not.
- Status
- Not open for further replies.