I'm sending single ended signal from analog active crossover to Mod-86. Interconnect <12", output buffer is MC33079 opamp.
I see 3 possible choices for the input signal:
1. Bypass the balanced input and go straight to LM49710
2. Add differential output to the crossover, send to Mod-86 a balanced signal
3. Apply single ended signal to balanced input, shorting In- to ground as recommended in Mod-86 Design Doc.
#3 is easiest, but which will deliver the best sound quality?
I am happy with the SQ of Mod86's balanced input, when using a balanced signal, but I'm concerned that shorting a leg of THAT1200 will cause it to not sound as good. Their white paper says that unlike most balanced line receivers, theirs does not suffer performance with one leg shorted.
Thanks
Rich
I see 3 possible choices for the input signal:
1. Bypass the balanced input and go straight to LM49710
2. Add differential output to the crossover, send to Mod-86 a balanced signal
3. Apply single ended signal to balanced input, shorting In- to ground as recommended in Mod-86 Design Doc.
#3 is easiest, but which will deliver the best sound quality?
I am happy with the SQ of Mod86's balanced input, when using a balanced signal, but I'm concerned that shorting a leg of THAT1200 will cause it to not sound as good. Their white paper says that unlike most balanced line receivers, theirs does not suffer performance with one leg shorted.
Thanks
Rich
#1 removes the advantage of the THAT1200 receiver: common-mode rejection. Even with a pseudo-differential interconnect, there's substantial CMRR with the THAT1200.
#2 is the optimal solution from a technical perspective. I'm working on a differential driver board that can handle this task. The design is done and I've started the layout. Sadly, the project has stalled there as I'm pretty overloaded right now. I plan to have the board available this fall.
I implemented #3 in the MiniDSP + 4xMOD86 build I just completed. Sounds good to my ears. Measures well also. You do have to watch for ground loops, though.
Managing the grounds on the MiniDSP is a challenge and you will end up creating ground loops. This is no different that using the MiniDSP with any other amplifier. If I was to do this again, I'd select the version of the MiniDSP that has differential outputs.
Tom
Last edited:
Tom, beautiful craftmanship - question, what internal cables/wires do you use? I see they have a clear sheath - just some off the shelf speaker wiring possibly?
And OMFG Soongc, I can't believe you're still dwelling in this thread and filling it with nonsense - been ignoring your posts for near a year now >< I'm surprised bill hasn't lost his mind in rage yet lol
And OMFG Soongc, I can't believe you're still dwelling in this thread and filling it with nonsense - been ignoring your posts for near a year now >< I'm surprised bill hasn't lost his mind in rage yet lol
I'd connect the MiniDSP RCA outputs to +/- on the Mod86, and strap the Mod86 and MiniDSP power grounds together assuming they're not already connected together via the chassis/power supply/whatever.
Signal symmetry isn't necessary for good performance in a balanced system as the receiver is subtracting the two signals and rejecting any common mode content anyway. Especially with a good receiver like a THAT.
Sent from my Nexus 4 using Tapatalk
Signal symmetry isn't necessary for good performance in a balanced system as the receiver is subtracting the two signals and rejecting any common mode content anyway. Especially with a good receiver like a THAT.
Sent from my Nexus 4 using Tapatalk
I used regular 16 AWG (1.3 mm^2) zip cord for the speaker connections internal to the amp.
The power supply wires are 16 AWG as well. I would have liked 14 or 12 AWG (2-3 mm^2) to get slightly lower supply impedance, but my wire vendor didn't offer those gauges, so I used what he had and moved on. I doubt the difference matters greatly, it was more "just because" that I wanted the heavier gauge wire.
One could argue that a low supply impedance would result in lower output impedance and better damping ratio, though, the layout of the MOD86 board addresses that to a large extent. In the end, I think the 10' of AWG12 speaker cord will dominate in that game. I connect to the speakers with a 4xAWG12 SpeakON cable.
The inputs were routed from the MiniDSP to the MOD86 using twisted pair 20 AWG wire with a separate ground wire. No shield.
In the amp with XLR inputs and no internal XO, I used low-noise microphone cable from GLS Audio because I had it available. I usually pick up a 50' mic cable and cut the connectors off. It's often cheaper than buying the bulk wire.
Both amps measure mains hum well below -110 dB, so I must have done something right...
Tom
Last edited:
For quality wire in small quantity I quite like this guy on eBay. Have been buying from him for several years. He sells various jacket color assortment packs with 5-10 feet lengths of a given wire. Super handy in some situations.
Johns silver Teflon wire shop
http://stores.ebay.com/Johns-Silver-Teflon-Wire-Shop
PS. Loving my Mod for several months now. Thanks for great product Tom.
Johns silver Teflon wire shop
http://stores.ebay.com/Johns-Silver-Teflon-Wire-Shop
PS. Loving my Mod for several months now. Thanks for great product Tom.
Last edited:
I have put together 2 Modulus-86 and 2 Parallel-86, and I must say, they measure very well.
There is just one thing that might be improved imo. The return to ground from the load is now located on the PCB. When using more than 1 board on a power supply. there will be differences in ground potential on the different boards. It can be argued that, because the input to the boards is balanced, this will indeed mitigate the situation. However, the THAT 120X is referenced to ground, so there will not be complete immunity. And in case of quasi balanced connections, the situation will be worse, depending on the grounding point.
Therefore, better to avoid this situation by connecting the load return to the supply star ground and not through the board.
There is just one thing that might be improved imo. The return to ground from the load is now located on the PCB. When using more than 1 board on a power supply. there will be differences in ground potential on the different boards. It can be argued that, because the input to the boards is balanced, this will indeed mitigate the situation. However, the THAT 120X is referenced to ground, so there will not be complete immunity. And in case of quasi balanced connections, the situation will be worse, depending on the grounding point.
Therefore, better to avoid this situation by connecting the load return to the supply star ground and not through the board.
You've got it backwards; local return is optimal to sense and correct for the error voltage you're concerned about. I suggest spending time with section 4 of DIY Audio's grounding article for clarity on use of signal and power grounds. THAT's design collateral will likely also be helpful in understanding how difference amplifiers work.
Twest, you got it wrong yourself. The conclusion of the article you refer to is
-quo-
Rule 1: Each of the following must be connected to the system star ground by one and only one route.
The earthing of the load should be connected to system star ground, second bullet. Rule 4 applies as well. Rule 7 doesn't. This is not a Bruno Putzey's 'G-Word' - 'follow the ground situation'.
Yeez, try to think before you shoot from the hip.
-quo-
Rule 1: Each of the following must be connected to the system star ground by one and only one route.
- All signal references
- All power commons
- Shields of non-galvanically isolated single-ended inputs and outputs
- Safety ground and chassis. The safety ground and chassis should be thought of as a single entity.
- The connection may be direct, or indirect through a star-of-stars or buss. This is expanded upon below.
- The safety ground and chassis may be connected to the system star ground through a Safety Loop Breaker Circuit.
- The “one and only one” part of this rule precludes ground loops. There is no excuse for a ground loop within a single component.
The earthing of the load should be connected to system star ground, second bullet. Rule 4 applies as well. Rule 7 doesn't. This is not a Bruno Putzey's 'G-Word' - 'follow the ground situation'.
Yeez, try to think before you shoot from the hip.
Last edited:
The only way to get potential difference between the star ground and the amp ground is by steering load return to the amp. That is why you should not do it.
If the load is not connected to the PCB but to star ground, the only currents to ground on the PCB are housekeeping for LM3886 and reference for THAT (see fig. 8 of PDF). Not enough to cause any potential differences with star ground.
If the load is not connected to the PCB but to star ground, the only currents to ground on the PCB are housekeeping for LM3886 and reference for THAT (see fig. 8 of PDF). Not enough to cause any potential differences with star ground.
Yes, that's good advice. As someone in their third decade of designing for and measuring grounding related errors I recommend taking it---one does not arrive at Modulus performance levels without careful analysis and validation. Though, if it's more important to argue over incorrect understanding of circuit behavior, I'm content to leave you to it.
Let me show a picture to make clear what I mean:
Figure A is the grounding setup that is best imo. Fig. B is what Twest advocates.
The reason this is sub-optimal is that return current from the load will flow through the PC ground return, as well as through the signal ground return. This is called a ground loop.
Since the signal is balanced, one might opt to float the ground on either side of the connection between star ground and signal ground, but this is no perfect solution either. Thus, the ground loop is cut, but still a potential will develop between PCB ground and star ground.
Now, this is the important point, also in the case of a balanced connection, the + and - signals are still referred to ground by necessity. Look at Fig. 8 of the THAT PDF. Fluctuations in ground potential are identical to common mode swings. These will be well suppressed by a device like the THAT, but why have them in the first place?
Therefore, scheme A is the safe way to do it, scheme B will only work by virtue of the common mode reduction capability of the THAT input chip, which is excellent but not perfect.
Figure A is the grounding setup that is best imo. Fig. B is what Twest advocates.
The reason this is sub-optimal is that return current from the load will flow through the PC ground return, as well as through the signal ground return. This is called a ground loop.
Since the signal is balanced, one might opt to float the ground on either side of the connection between star ground and signal ground, but this is no perfect solution either. Thus, the ground loop is cut, but still a potential will develop between PCB ground and star ground.
Now, this is the important point, also in the case of a balanced connection, the + and - signals are still referred to ground by necessity. Look at Fig. 8 of the THAT PDF. Fluctuations in ground potential are identical to common mode swings. These will be well suppressed by a device like the THAT, but why have them in the first place?
Therefore, scheme A is the safe way to do it, scheme B will only work by virtue of the common mode reduction capability of the THAT input chip, which is excellent but not perfect.
This is also incorrect.
In precision implementations everything needs special consideration.
What about the current going to and from the capacitors on the amp board?
In fig. A there are current loops from: Local Caps > LM3886 > LOAD > PSU > Local Caps.
And current loops for charging: PSU Caps > Local Caps.
Both of these currents flow though the Ground connecting the PSU and the AMP boards.
Last edited:
You are right, it is even better to hang those caps between the power lines so that ground is just there for references and housekeeping of the LM3886.
In class AB, except the bias current, which is small in this power chip, all currents are asymmetrical. In other words, when one rail draws power, the other one is hard and can be used for the caps to push against.
I use this concept with success in a power amp with 14 LM3886 producing 6 channels. Just bare chip amps, which have much worse PSRR than the Modulus-86, so problems show up much earlier.
In class AB, except the bias current, which is small in this power chip, all currents are asymmetrical. In other words, when one rail draws power, the other one is hard and can be used for the caps to push against.
I use this concept with success in a power amp with 14 LM3886 producing 6 channels. Just bare chip amps, which have much worse PSRR than the Modulus-86, so problems show up much earlier.