Most power cords I encounter are polar in an end over end sense because they have different shaped plugs on each end and polar in a handedness sense due to keyed plug or socket geometry. I assume polarity here means the reversal of how hot and neutral connect to an audio component's primary permitted by nonpolarized types.
It's hard to quantify the amount of bounce incurred by charging pulses into the reservoir caps. Transformers are rarely specified for leakage inductance unless for series LLC supplies, stray capacitances are by definition poorly controlled, variability in cable and mains impedances large, and it can be quite difficult to collect measurements without perturbing the system. I generally get microvolts best case and in design for large pro audio installs there's something of a de facto convention of 1V worst case. So, you know, only a 120dB range. Maybe 160dB with bad music festival grounding, mains wiring errors, or for two free floating class II devices before a cable is connected between them.
I'm guessing probably you were measuring below single digit millivolts in home environments before the error cancellation which occurs when components bounce in the same direction at the same time?
It's hard to quantify the amount of bounce incurred by charging pulses into the reservoir caps. Transformers are rarely specified for leakage inductance unless for series LLC supplies, stray capacitances are by definition poorly controlled, variability in cable and mains impedances large, and it can be quite difficult to collect measurements without perturbing the system. I generally get microvolts best case and in design for large pro audio installs there's something of a de facto convention of 1V worst case. So, you know, only a 120dB range. Maybe 160dB with bad music festival grounding, mains wiring errors, or for two free floating class II devices before a cable is connected between them.
I'm guessing probably you were measuring below single digit millivolts in home environments before the error cancellation which occurs when components bounce in the same direction at the same time?
Actually the voltage differences were a few volts, and I think we were able to get them to about one volt. It has been at least over 30 years ago. In those days I did not have millivolts measurement capability.
The issue is that you are a true believer in sighted listening and substituting the latest guru claptrap on a thread about an amplifier that stands on its objectively superb measurements. Like a fish out of water, a fart in a space suit or a pork sausage at a jewish wedding. You have no use here other than free BTT so more people see what Tom is selling.
My mental state is very self secure thank you very much. You are the one I worry about.
Yah, sounds like measurement issues. 1V with 40dB CMRR from balanced is 10mV error, so approaching 0dB SNR at typical power amp input levels for home use. That'd be pretty horrid to have to listen to.
You mean the ground between two components was two volts with them unconnected? Yes, this varies with a lot of things, transformer primary-to-secondary capacitance and such, and so the "source" is such a high impedance that the voltage will vary greatly depending on the resistance of the voltmeter you use. I don't see this as a useful measurement.
The better measurement is the current between the grounds when they are connected, and also the receiving device's sensitivity to ground current in its input connector. I recall that Bill Whitlock proposed some standard to measure how sensitive a device's ground connections are to 60Hz current. No doubt, due to its balanced input device, the topic of this thread has VERY low sensitivity to such currents.
Current, voltage, impedance. Ohm's law and that...
There are a couple of issues with audio systems in general:
1) The equipment is usually grounded to a common rail (power strip).
2) In single-ended systems (RCA connectors), the individual components are connected via multiple ground paths.
These set up many ground loops. There's really no way around that, aside from using optical interconnects.
The ground loops themselves are really not all that problematic. OK, some current flows. Big deal! The trouble begins when this current sets up an error voltage that is in series with the signal source. Then the amplifier output will be: Av * (Vsignal + Verror), where Av is the voltage gain of the amp and Vsignal and Verror are the signal voltage and error voltage, respectively. That is the crux of the problem with single-ended circuits. It would not at all surprise me if most consumer audio gear, even so-called high end gear, operates just a hair better than the audible threshold for the error term.
One way to reduce the error term is to minimize the ground impedance, i.e. throw more copper at the problem. A quality power cord could be one way of accomplishing this. Now, by "quality power cord" I mean a power cord with half-decent connectors, conductors of a reasonable cross-sectional area, that is no longer than it needs to be. I see no point in buying fu-fu brand power cords. Just buy a good pair of connectors, some 12-14 AWG (2.5-4 sq mm) cable, and make a power cord. This may be enough to push the error term from the audible threshold to just below the audible threshold. Note that this is just speculation and not backed by data. Though, I think the theory holds water. In my opinion, beefing up the power cord is still a band-aid.
A more effective solution is to throw some engineering at the problem and use a differential circuit. Differential interconnects minimize the error voltage. The amp output, for a differential amp, will be: Av * (Vdiff + Vcm/CMRR), where Vdiff is the differential voltage, Vcm the common-mode voltage, and CMRR the common-mode rejection ratio. For a differential circuit, the ground loop error voltage appears as a common-mode voltage and the desired signal as differential mode. Hence, Vdiff = Vsignal and Vcm = Verror. CMRR is typically very high. In the Modulus-86 and Parallel-86, about 90 dB (= 31600x). Hence the amp output for the MOD/PAR86 becomes: Av * (Vsignal + Verror/31600). In other words, the error term resulting from ground loops is reduced by a factor of 31600 compared to an amp with a single-ended input. In other words, the differential input of the MOD/PAR86, effectively, removes the ground loop from the signal path. Even when using a pseudo-differential connection, there is a significant benefit from using the differential input. That's why I included it on both the MOD86 and the PAR86.
Subjectively, I've also perceived an improvement in sound quality when switching from single-ended to differential signalling. I had the opportunity to test drive a stereo consisting of a pair of Parasound A23 amps biamped to drive a pair of Dali Suite 2.8. The A23s were fed by a Parasound P3 preamp. Both the pre and the power amps supported both single-ended and differential signalling. I swapped from single ended to differential, honestly not expecting any difference at all. I was quite surprised to experience a significant improvement in the subtle details of recordings I thought I knew inside out. From that point on, I've been differential all the way...
Now, of course, this is my subjective experience in a sighted test. The usual cognitive biases apply.
Tom
There are a couple of issues with audio systems in general:
1) The equipment is usually grounded to a common rail (power strip).
2) In single-ended systems (RCA connectors), the individual components are connected via multiple ground paths.
These set up many ground loops. There's really no way around that, aside from using optical interconnects.
The ground loops themselves are really not all that problematic. OK, some current flows. Big deal! The trouble begins when this current sets up an error voltage that is in series with the signal source. Then the amplifier output will be: Av * (Vsignal + Verror), where Av is the voltage gain of the amp and Vsignal and Verror are the signal voltage and error voltage, respectively. That is the crux of the problem with single-ended circuits. It would not at all surprise me if most consumer audio gear, even so-called high end gear, operates just a hair better than the audible threshold for the error term.
One way to reduce the error term is to minimize the ground impedance, i.e. throw more copper at the problem. A quality power cord could be one way of accomplishing this. Now, by "quality power cord" I mean a power cord with half-decent connectors, conductors of a reasonable cross-sectional area, that is no longer than it needs to be. I see no point in buying fu-fu brand power cords. Just buy a good pair of connectors, some 12-14 AWG (2.5-4 sq mm) cable, and make a power cord. This may be enough to push the error term from the audible threshold to just below the audible threshold. Note that this is just speculation and not backed by data. Though, I think the theory holds water. In my opinion, beefing up the power cord is still a band-aid.
A more effective solution is to throw some engineering at the problem and use a differential circuit. Differential interconnects minimize the error voltage. The amp output, for a differential amp, will be: Av * (Vdiff + Vcm/CMRR), where Vdiff is the differential voltage, Vcm the common-mode voltage, and CMRR the common-mode rejection ratio. For a differential circuit, the ground loop error voltage appears as a common-mode voltage and the desired signal as differential mode. Hence, Vdiff = Vsignal and Vcm = Verror. CMRR is typically very high. In the Modulus-86 and Parallel-86, about 90 dB (= 31600x). Hence the amp output for the MOD/PAR86 becomes: Av * (Vsignal + Verror/31600). In other words, the error term resulting from ground loops is reduced by a factor of 31600 compared to an amp with a single-ended input. In other words, the differential input of the MOD/PAR86, effectively, removes the ground loop from the signal path. Even when using a pseudo-differential connection, there is a significant benefit from using the differential input. That's why I included it on both the MOD86 and the PAR86.
Subjectively, I've also perceived an improvement in sound quality when switching from single-ended to differential signalling. I had the opportunity to test drive a stereo consisting of a pair of Parasound A23 amps biamped to drive a pair of Dali Suite 2.8. The A23s were fed by a Parasound P3 preamp. Both the pre and the power amps supported both single-ended and differential signalling. I swapped from single ended to differential, honestly not expecting any difference at all. I was quite surprised to experience a significant improvement in the subtle details of recordings I thought I knew inside out. From that point on, I've been differential all the way...
Now, of course, this is my subjective experience in a sighted test. The usual cognitive biases apply.
Tom
Last edited:
And those posters who use unbalanced stuff and are measuring these huge EVIL currents are probably listening to a LOT of crap. 😱
No wonder they hear differences between mains cables 😀
They should switch to Modulus-86 and banish these EVILs. 🙂
Tom, nice expansion on post 1476. I detect an echo too... 😉
Considering the difference between an unbalanced receiver and a basic balanced receiver is two resistors and maybe a cap it's not something I've ABXed much---cheaper and way more effective to implement balanced receive. I have gotten noticeable discrimination between AWG16 and AWG9 but it wasn't statistically significant (p < 0.05). Subjectively I wouldn't say one or the other sounded better or worse, just some wrongness in the background that was difficult to describe. The various ABXes I've done between balanced and unbalanced have offered easier discrimination and subjective preference for balanced.
It depends on the spectrum of the charging current pulses as shaped by the reservoir supply caps and whatnot. Along with how critical the listener is and where that places the threshold of detection. Generally speaking, there isn't a threshold. Also, in simple wired pairs pushing resistance down tends to increase inductance. So increasing gauge can trade improvement in psychoacoustically well masked low frequencies for degradation in the more perceptible top of the mains spectrum around 1kHz. Too much going on, IMO, to make much of call.
Considering the difference between an unbalanced receiver and a basic balanced receiver is two resistors and maybe a cap it's not something I've ABXed much---cheaper and way more effective to implement balanced receive. I have gotten noticeable discrimination between AWG16 and AWG9 but it wasn't statistically significant (p < 0.05). Subjectively I wouldn't say one or the other sounded better or worse, just some wrongness in the background that was difficult to describe. The various ABXes I've done between balanced and unbalanced have offered easier discrimination and subjective preference for balanced.
It does beg the question as to why people don't chuck a THAT1200 in everywhere! OK it's probably the limiting part in the Mod-86, but given how few pre-amps match that level of performance it seems a no-brainer!
Even better to use Sowter transformers .. truly hand-carved from solid Unobtainium by Ipswich virgins with $$$ to match. None of THAT cheapo stuff. 😀
One reason is a not insignificant level of knowledge is required to understand the value proposition for parts like the 120x series. A couple more are THAT is not especially well known as an audio component supplier and the 120x are all priced out of consumer and prosumer markets.
In regards to high end design it also depends on what one's optimizing for. The 1200 offers a compelling set of tradeoffs but it's just one of several points on what one might call the efficient frontier of design.
Sad if a $6 part is too expensive for domestic audio, but guess that adds $100 to the selling price.
Surely high end design is optimizing for marketing copy?
Surely high end design is optimizing for marketing copy?
Yeah, well... I did have the echoes of that post ringing in my head as I typed. 🙂
Depends... The highest charging pulses are found in the power amp. They should be pretty easy to keep out of the interconnect and mains grounds. They should stay in the power amp power transformer secondary circuit.
One could argue that some amount of the charging pulses from the source's power supply may trickle through as signal. Those are more likely to be lower in magnitude.
It costs money. Every penny (or fraction of a penny) counts for consumer electronics.
The Jensen JT11 transformers are quite good too. I use them in my DG300B... $75 transformer ... $5 THAT1200... If ±15 V is already available, the THAT1200 is a no-brainer.
ROFL!
Tom
The next step for Neurochrome is to develop a preamp board with THAT inputs and THAT outputs, with gain optimized for the MOD86 and Parallel 86. As an older guy, I would also like the option of a high frequency boost-only shelving to lift the treble by 2-4 db.
next step for mod86
I'd prefer to see a BPA-style Mod86 amp, with 4 LM3886 slots built onto the same board for more amperage capability. It would be cheaper than stacking all those boards, duplicate THAT components, etc...
Is there a limit to how many LM3886's one LME49710 could handle?
Boba
I'd prefer to see a BPA-style Mod86 amp, with 4 LM3886 slots built onto the same board for more amperage capability. It would be cheaper than stacking all those boards, duplicate THAT components, etc...
Is there a limit to how many LM3886's one LME49710 could handle?
Boba
I'll check with the virgins and see if they can cut you one of those unobtanium trafos where the primary isn't coupled to the secondary. Conservation of energy and all that. 😉
For real trafos, primary side pulses induce voltages across neutral back to wherever the neutrals between the components join. These are developed over the neutral impedances. There will be some amount of bounce due to coupling within the trafo too---primarily from leakage inductance (commonly uH) as coupling through interwinding capacitance (commonly a few pF) is typically negligible in comparison.
These couple into green wire in several ways. One is conducted. Since there is some leakage across the trafo a small current must flow into safety ground and return to neutral at the earth point to complete the circuit. This is negligible from the standpoint of the mains--even if it's driven by 1mV it's more than 120dB down from the mains voltage. But that's only 20dB down from a typical-ish 10mV input signal to a power amp. More practically, my estimates tend to come in around 10uV range, which is at that 60dB down probable edge of audibility in typical home audio use cases. Given the quirkiness of historical wiring and the complexity of what all is attached to the mains it's not difficult to construct cases where it might be 100uV or more. Those should be audible to critical listeners.
There is also field coupling. Stray capacitance between neutral in green wires is a few tens of pF per meter. Multiply by the total wiring in the local circuit and the impedance between neutral and green comes up around a few megaohms. Charging pulses bounce neutral by a few volts (that 3% nominal mains THD) which means a few microvolts will develop on green. Smaller than the direct conducted term, but not necessarily by an order of magnitude. It's also probably not the dominant capacitance. Where power line filters with X2Y caps referenced to safety are present there will be nF coupling rather than adding up pF the hard way. Cancellation does occur from the corresponding motion of hot, but only within the matching of the filter capacitors. So this error term could be below 1uV but, say, 100uV is not implausible either. In some cases I've read about it's been so bad pro studios using all balanced gear have had to take apart all their equipment and modify the inlet filters.
You've mentioned similar kinds of "leakage" paths in whatever's feeding the power amp---I believe we're both aware of incidents where they've proven audible even in decently implemented preamps. Many amplifiers have voltage reference management problems larger than those fixed between v1.0 and v2.0 of the Modulus, too. Plus current Modulus versions deliberately implement leakage paths. Notably the referencing of the Y caps on the RF inlet filter to safety ground. This creates another path from the source's secondary to green. It also creates one on the Mod since the 1200 servos the node the cap's connected to. I think this one is only a nV error, so probably negligible even before the 1200's CMRR, but other inlet filters are frequently not so carefully designed. (Note to folks who are wondering what I'm talking about: Tom's stated in this thread the Modulus uses the inlet filter of figure 11 in the THAT 1200 datasheet. Specifically, C3 creates the paths in question.)
All this adds up to a squirrely, funky, variable, and kind of unpredictable problem. Should be familiar to anyone who's dealt with hum issues.
Also it does not, in general, hold a power amp pulls more current or power than the preamp or source in typical use. It's usually loafing along at mW, meaning an amp channel tends spend most of its time operating within the usual 50-100mA bias. This is comparable to the quiescent dissipation of 10-20 op amps and many DACs, for example, consume around 25mA per channel. Often that adds up to more pull than the power amp when all channels in the source/preamp are considered. If the source is a computer sound card or audio interface the source may consume an order of magnitude more power. There is some gearing to work through for the use of different trafo ratios and SMPSes and potentially PFCs but you get the point; the two things on either end of the interconnect are going to inject different voltages on their neutrals with different timings, magnitudes, shapes. That means some funky coupled voltage on green wire reference to slosh across the interconnect.
You may be interested in the Parallel-86. Drives 4 Ω loads from a ±35 V rail. If you need to drive a 2 Ω load, I recommend lowering the supply voltage to ±28 V to stay within the guaranteed current limit of the LM4780 (commonly considered to be 2x LM3886).
Yes, but I bet you'll run out of money before the LME49710 runs out of current drive.
So far, nobody has requested a bridged-parallel board and only one potential customer has indicated that he'd bridge multiple boards. This leads me to believe that the market is rather small.
Most are happy with the MOD86. For those who want a bit more power or need to drive low impedance loads, there's the PAR86. For even more power, one can bridge two PAR86 modules...
Good point. It would be interesting to set up a sim, including the transformer and leakage terms.
Another reason to take the ground circuit out of the signal path by using a differential input.
Thanks for the analysis, by the way. You've clearly thought this through further than I have.
Tom
Well... I've been mumbling about a "preamp" of sorts for a while now.
I'm working on a board which will allow you to take a single-ended input and turn it into a differential output. Add a volume control and you have a preamp. The design is done and I've started on the layout. The project has been stalled there for about a month as I've had quite a few other irons in the fire lately. I hope to make some progress on this project this upcoming week, though.
Tom
😀😀😀
Just talked with a chip company executive today, trying to convince me to get into wireless audio. In addition to looking at all the discussion here really amuses me how history repeats itself and also how people stick to different ends of ideas.
😀😀😀
But I was asked to give a presentation to the chip engineers about what we look for in audio applications. The good thing with these is that most of these really focus on the issues and see how the need and supply match up. I never envisioned face to face discussions like these could be more productive than forums, but sometimes these sessions really are.
Just talked with a chip company executive today, trying to convince me to get into wireless audio. In addition to looking at all the discussion here really amuses me how history repeats itself and also how people stick to different ends of ideas.
😀😀😀
But I was asked to give a presentation to the chip engineers about what we look for in audio applications. The good thing with these is that most of these really focus on the issues and see how the need and supply match up. I never envisioned face to face discussions like these could be more productive than forums, but sometimes these sessions really are.