Hello everyone.
I would like to solicit everyone's feedback on an upcoming build of mine. I am planning to build a 4-channel amp consisting of composite chip-amp modules, which happen to be SE input but I will be using balanced line receiver boards to allow balanced inputs. SE inputs to the amp will not be available.
Attached is a PDF with the proposed wiring layout. Thoughts?
I was reading through a bunch of stuff on Bonsai's site (namely this: https://hifisonix.com/wp-content/uploads/2019/02/Ground-Loops.pdf) and tried to apply the concepts where I could, but (for probably good reason) nearly all the material was focused on unbalanced setups, so I'm not sure exactly how these guidelines should translate to a fully balanced amplifier. The source will also be fully balanced FWIW.
What I would especially like to hear is what I can slack off on (i.e. not worry much about) without losing any real protection against noise & hum.
I would like to solicit everyone's feedback on an upcoming build of mine. I am planning to build a 4-channel amp consisting of composite chip-amp modules, which happen to be SE input but I will be using balanced line receiver boards to allow balanced inputs. SE inputs to the amp will not be available.
Attached is a PDF with the proposed wiring layout. Thoughts?
I was reading through a bunch of stuff on Bonsai's site (namely this: https://hifisonix.com/wp-content/uploads/2019/02/Ground-Loops.pdf) and tried to apply the concepts where I could, but (for probably good reason) nearly all the material was focused on unbalanced setups, so I'm not sure exactly how these guidelines should translate to a fully balanced amplifier. The source will also be fully balanced FWIW.
What I would especially like to hear is what I can slack off on (i.e. not worry much about) without losing any real protection against noise & hum.
I just realized that the PDF doesn’t render properly on at least some smartphones. Anyway.
The signal wires aren’t all twisted together. Only the two wires from each XLR jack are twisted together (so 4 twisted pairs). The 4 twisted pairs are loosely bundled together down the common runs.
Re: grounding; all the info seems to indicate XLR pin 1 must connect to chassis using as short a wire as possible.
The signal wires aren’t all twisted together. Only the two wires from each XLR jack are twisted together (so 4 twisted pairs). The 4 twisted pairs are loosely bundled together down the common runs.
Re: grounding; all the info seems to indicate XLR pin 1 must connect to chassis using as short a wire as possible.
Also, I'm thinking now that I would like to bring in just 4 wires from the external power supply instead of the 25 as shown; +36 VDC, 0 VDC (power supply common), -36 VDC, and safety earth, with terminal blocks inside the amplifier enclosure for distribution. Any concerns with that? From these terminal blocks the +36/0/-36 VDC would to go the 4 amps as well as a small power supply board for +15/0/-15 VDC that would then go to the amp aux (opamp) power inputs. Any concerns?
I do not claim to be an expert in this stuff, but I hope this helps: -
1. The overall layout within the housing looks good. You have taken the speaker returns back to each amplifier module, which is the correct way to do it.
2. The cross-channel ground loop problem is largely resolved using balanced inputs, but your layout is also good and minimizes it anyway because the total loop area between the inputs and the power rails is minimised, and the speaker returns back to each module. By running the balanced input cables close to the chassis, you are also minimizing common mode noise such that it exists within the housing.
3. I like the way you have taken the power (+- and 0V) back to the PSU for each module. Although this involves a lot of wiring since you have elected to have the PSU completely separate, it does minimise the opportunity for ground loops, cross-channel ground loops and common impedance coupling
4. For the balanced inputs, you only need nicely twisted cable pairs inside the housing - you will not need the shield. Using a screened 2-core cable is convenient, but if you do, do not connect the shield at the input connector side. You MAY get some benefit at HF by tying the other end of the shield to signal ground at the amplifier module.
5. Connect the GND of your XLR input socket directly to the chassis at the input connector - that's pin 1 on the XLR. This technique makes the amplifier housing, the XLR cable shield and the source component a single enclosure to RF (Google Neil Muncy and the 'Pin 1 Problem')
6. You do not discuss why you have +15V coming in separately - is that to power up the balanced input receivers? If this is the case, it is always much better to derive the low voltage supply off the main supply into each module because this will dramatically ease wiring, but also minimises loop areas and the opportunity for common impedance coupling. In your case, I'd tightly bunch the low-voltage and high-voltage supplies together to minimise loop areas.
7. Why do you need a low voltage 0V and a high voltage 0V? It might be better to just have a single 0V going to the modules and then tap off the signal 0V from that. Make sure if you do this, the signal 0V obeys the rules for common impedance coupling - see pages 45 to 48 in the 'Ground Loops' presentation. In your case, having a low voltage 0V and a high voltage 0V creates a large loop area - albeit minimised by twisting the cables - that runs from the PSU to the module and then back to the PSU again. Any flux that cuts the loop will generate a loop current and will generate a noise voltage between the power part of the circuit and the small signal area. So, if you stick with the 2 ground wiring to each module, make sure you do not have any common impedance coupling on the module PCB's.
8. I don't know what the power supply end looks like, but you are going to have to manage the wiring terminations that end carefully to avoid common impedance coupling. See pages 56-60 in the Ground Loops presentation
Hope this helps - good luck with your project.
🙂
1. The overall layout within the housing looks good. You have taken the speaker returns back to each amplifier module, which is the correct way to do it.
2. The cross-channel ground loop problem is largely resolved using balanced inputs, but your layout is also good and minimizes it anyway because the total loop area between the inputs and the power rails is minimised, and the speaker returns back to each module. By running the balanced input cables close to the chassis, you are also minimizing common mode noise such that it exists within the housing.
3. I like the way you have taken the power (+- and 0V) back to the PSU for each module. Although this involves a lot of wiring since you have elected to have the PSU completely separate, it does minimise the opportunity for ground loops, cross-channel ground loops and common impedance coupling
4. For the balanced inputs, you only need nicely twisted cable pairs inside the housing - you will not need the shield. Using a screened 2-core cable is convenient, but if you do, do not connect the shield at the input connector side. You MAY get some benefit at HF by tying the other end of the shield to signal ground at the amplifier module.
5. Connect the GND of your XLR input socket directly to the chassis at the input connector - that's pin 1 on the XLR. This technique makes the amplifier housing, the XLR cable shield and the source component a single enclosure to RF (Google Neil Muncy and the 'Pin 1 Problem')
6. You do not discuss why you have +15V coming in separately - is that to power up the balanced input receivers? If this is the case, it is always much better to derive the low voltage supply off the main supply into each module because this will dramatically ease wiring, but also minimises loop areas and the opportunity for common impedance coupling. In your case, I'd tightly bunch the low-voltage and high-voltage supplies together to minimise loop areas.
7. Why do you need a low voltage 0V and a high voltage 0V? It might be better to just have a single 0V going to the modules and then tap off the signal 0V from that. Make sure if you do this, the signal 0V obeys the rules for common impedance coupling - see pages 45 to 48 in the 'Ground Loops' presentation. In your case, having a low voltage 0V and a high voltage 0V creates a large loop area - albeit minimised by twisting the cables - that runs from the PSU to the module and then back to the PSU again. Any flux that cuts the loop will generate a loop current and will generate a noise voltage between the power part of the circuit and the small signal area. So, if you stick with the 2 ground wiring to each module, make sure you do not have any common impedance coupling on the module PCB's.
8. I don't know what the power supply end looks like, but you are going to have to manage the wiring terminations that end carefully to avoid common impedance coupling. See pages 56-60 in the Ground Loops presentation
Hope this helps - good luck with your project.
🙂
Just to be sure, did you review the updated layout where only 4 wires are brought into the amplifier enclosure, vs the original 25?
Yes, if I do use a shield for the input signal twisted pairs, it will definitely not be connected at XLR jack pin 1, and instead would be connected to a reasonably safe 0 V at/near the balanced line receiver boards. But based on your comment I definitely will not do this - I don't enjoy pointless work. 🙂
The +15/0/-15 is produced to power 8 "devices": the 4 balanced line receiver modules and the 4 amplifiers (aux supply for the onboard op-amps - utilized for the "composite chip-amp" operation). If I'm understanding you correctly, you're saying that it's best to "make" the lower voltage at each amplifier module in order to simplify the wiring and minimize loop areas and such. Unfortunately, this is not an option at this point... mainly, I didn't order the amp kits with onboard supplies as I thought it would be cheaper. 🙂 And I'm still not sure that the amps with onboard supplies would have enough juice to also power the associated balanced line receiver.
So anyway, the low-voltage supplies run to pretty much the same locations as the higher-voltage supplies, so I will route the low-voltage supply wiring to run alongside the higher-voltage supply wiring as much as possible.
I assumed that if the low-voltage power supply provided a 0V at its output, it would just make sense to maintain that output 0V along with its +/- 15 VDC. It doesn't tickle my fancy to separately connect all the low voltage 0V end points back to the high voltage 0 V (star) and have these random green wires strewn about. However, for clarity, if I don't use a low-voltage 0V at all (i.e. don't connect the 0V at low voltage power supply output) and only run twisted pairs of +/- 15 VDC, and the devices that run on low voltage (amps and BLRs) have their low-voltage 0 V reference connected only to high-voltage 0V, this would in fact be an improvement? I guess this isn't really an issue for the amps, since I can readily jumper low-voltage 0 V to the high-voltage 0 V (it's the same board after all). I guess the nice thing about this is that it follows the star ground "greatest-to-smallest" concept of putting the larger return currents closer to the main T point, i.e. the high-voltage 0 V wire lands on the high-voltage 0 V module connection point and then jumpered over to the low-voltage connection point on the same module... ?
But it gets uglier with the 0 V for the BLRs since I'd then have to daisy chain LV 0V wires (or "split" into two branches off each high-voltage 0 V connection point at each module; 1 to LV 0V point on same module and 1 to BLR LV 0V point). OTOH, the BLRs would only ever get one 0 V wire anyway since they take only a single bipolar supply (vs the amps which take two bipolar supplies).
I'm not sure exactly which apply to me the most, but I will follow the "T" style of star ground (producing my "high-voltage 0 V") with sequential connection of returns from largest currents to smallest currents, where largest currents are closest to the middle of the two central decoupling caps... that, and generally following the layout diagrams which I think I have done.
Thanks very much again!
For what it's worth, I don't think there are going to be any transformers in the amplifier enclosure. They should all be in the external enclosure, which will be underneath the amplifier enclosure but shielded by at least a 1.5 mm (mildly perforated) steel sheet together with a ~2 mm aluminum sheet, both of which will be grounded. The power supply providing the high-voltage is also going to be a pair of SMPS so that should also produce less EMI compared to a big ol' toroid?
What will be inside the main enclosure:
What will be inside the main enclosure:
- 1x capacitance multiplier (pair)
- 3x distribution terminal blocks, one each for +36 VDC, "HV 0 V", and -36 VDC
- 1x low-voltage power supply (makes +15/0/-15 VDC from +36/0/-36 VDC)
- (2 or 3)x distribution terminal blocks, one each for +15 VDC, maybe one for "LV 0 V", and -15 VDC
- 4x balanced line receiver modules (takes +/- 15 VDC)
- 4x amplifier modules (takes +/- 36 VDC and +/- 15 VDC)
- The requisite external input and output connectors
Last edited:
Hi there,
I looked at the second PDF that you posted up and commented on that.
Re the two 0V (high power 0V and clean 0V): Its better to run a thick 0V to each module and then extend that and call it the clean 0V and take that to the small signal stuff - see (2) below. Make sure you tap it off the correct place to avoid the common impedance coupling I referred to earlier. If you run a separate 0V for the low signal level stuff you have 2 problems: a loop in the ground connection and common impedance coupling - this is shown in (1) below.
I'm assuming when you say 1 capacitance multiplier pair per amp module you mean + and - per module. I would take the capacitance multiplier 0V to the main power 0V at each module.
I suggest the next step is to redraw your proposed wiring and put that up and everyone can take another look - a few eyes on the problem is always better than one or two and its easy to overlook some of this stuff 🙂
I looked at the second PDF that you posted up and commented on that.
Re the two 0V (high power 0V and clean 0V): Its better to run a thick 0V to each module and then extend that and call it the clean 0V and take that to the small signal stuff - see (2) below. Make sure you tap it off the correct place to avoid the common impedance coupling I referred to earlier. If you run a separate 0V for the low signal level stuff you have 2 problems: a loop in the ground connection and common impedance coupling - this is shown in (1) below.
I'm assuming when you say 1 capacitance multiplier pair per amp module you mean + and - per module. I would take the capacitance multiplier 0V to the main power 0V at each module.
I suggest the next step is to redraw your proposed wiring and put that up and everyone can take another look - a few eyes on the problem is always better than one or two and its easy to overlook some of this stuff 🙂
Last edited:
Are you using 1 ripple eater aka capacitance multiplier per amp module or are you sharing 1 between 2 amp modules?
One for the whole thing. Although technically speaking my cap multiplier modules, as I understand, will be 1 per rail (so two modules but they are the mirrored to generate one bipolar supply).
Coming off that [pair of] cap multipliers, the power will directly feed 4 amp modules as main power, and also feed the low-voltage power supply.
However, I will likely have a pair of large capacitors on the output of the cap multiplier. Not sure yet. I suspect they'll be 2x 35 mm dia 4,700 uF or 10,000 uF each. The star ground would originate from the connection between those two central decoupling/bulk caps.
Coming off that [pair of] cap multipliers, the power will directly feed 4 amp modules as main power, and also feed the low-voltage power supply.
However, I will likely have a pair of large capacitors on the output of the cap multiplier. Not sure yet. I suspect they'll be 2x 35 mm dia 4,700 uF or 10,000 uF each. The star ground would originate from the connection between those two central decoupling/bulk caps.
Last edited: