There can be many reasons for mains hum ingress. Usually it enters by inductive coupling or leakage. It's too low frequency to couple significantly through stray capacitances, but it can couple through EMC filter caps and such as they have large enough capacitance to allow for sufficient leakage current.
In systems with USB the EMC filter caps in the PC power supply strike me as the dominant route for mains hum in a PC-based measurement setup. The best solution I've found is to use an isolated USB interface. Those are few and far between as the isolators add cost without providing much benefit for the average consumer. There are some USB isolators out there for industrial applications, but they only seem to support "full speed" (12 Mbit/s). That's plenty for audio, but could lead to some compatibility problems. USB2/3 devices should be compatible ... "should"... Also note that some of them only isolate the power, which is useless. You need full isolation.
Another option is to run the analyzer software on a laptop on battery power. If the PC doesn't connect to ground it can't cause any current in the ground loop.
Mains hum can also enter via the cabling. So use good StarQuad cables and a differential input on the DUT. Also use the differential input on the analyzer. Beware that you don't exceed the common-mode voltage spec for the analyzer input, though.
Also keep in mind that 50/60 Hz hum may not be from the mains. I've been poking around in the lab lately and on occasion I'll get some 25 Hz and harmonics showing up, which is remarkable given that I'm on 60 Hz mains. It wouldn't surprise me one bit if that's related to wifi SSID broadcasts. The wifi bands are crowded here to say the least. #innerCityLiving.
The 25 Hz and harmonics I see are maybe 10 dB above the noise floor on the APx555, so it's not like this has any real consequences except it doesn't look pretty in the plot. Part of the reason for this is also that I have the circuit sitting 'naked' on the lab bench. Once it's in a metal chassis the RF will be shunted to ground (and attenuated heavily by the chassis).
I'll sometimes make a circuit burrito. I.e., wrap the circuit under test in ESD-safe bubble wrap and then in aluminum foil that's grounded to the test equipment ground. That usually knocks down the various RF sources to the point where they don't show in the measurement. Also, I keep my cellphone out of the lab.
Tom
In systems with USB the EMC filter caps in the PC power supply strike me as the dominant route for mains hum in a PC-based measurement setup. The best solution I've found is to use an isolated USB interface. Those are few and far between as the isolators add cost without providing much benefit for the average consumer. There are some USB isolators out there for industrial applications, but they only seem to support "full speed" (12 Mbit/s). That's plenty for audio, but could lead to some compatibility problems. USB2/3 devices should be compatible ... "should"... Also note that some of them only isolate the power, which is useless. You need full isolation.
Another option is to run the analyzer software on a laptop on battery power. If the PC doesn't connect to ground it can't cause any current in the ground loop.
Mains hum can also enter via the cabling. So use good StarQuad cables and a differential input on the DUT. Also use the differential input on the analyzer. Beware that you don't exceed the common-mode voltage spec for the analyzer input, though.
Also keep in mind that 50/60 Hz hum may not be from the mains. I've been poking around in the lab lately and on occasion I'll get some 25 Hz and harmonics showing up, which is remarkable given that I'm on 60 Hz mains. It wouldn't surprise me one bit if that's related to wifi SSID broadcasts. The wifi bands are crowded here to say the least. #innerCityLiving.
The 25 Hz and harmonics I see are maybe 10 dB above the noise floor on the APx555, so it's not like this has any real consequences except it doesn't look pretty in the plot. Part of the reason for this is also that I have the circuit sitting 'naked' on the lab bench. Once it's in a metal chassis the RF will be shunted to ground (and attenuated heavily by the chassis).
I'll sometimes make a circuit burrito. I.e., wrap the circuit under test in ESD-safe bubble wrap and then in aluminum foil that's grounded to the test equipment ground. That usually knocks down the various RF sources to the point where they don't show in the measurement. Also, I keep my cellphone out of the lab.
Tom
I took it all apart now, but it was wire from usb gnd stuck in between usb connector and the plug, and the other end connected to gnd on the input/output. It was not pretty looking or professional but it worked to remove the 60hz from my measurements.
Solution was mentioned by youtuber - phils lab - in his qa403 video review
About once a year I go through an exercise where I try new ideas on the Modulus-86 design just to see if I can squeeze a bit more performance out of it. That's one of the reasons the Modulus-86 is now in Rev. 3.0. This year I'm looking at resistors and I figured I'd share some preliminary results.
The resistor that's most likely to affect performance is the resistor from the output to the inverting input of the global feedback network. Same as in any other power amp. In Modulus-86 Rev. 3.0, that's R18.
The Dale RN55 "military grade" metal film resistors get a fair amount of press here. The "civilian grade" CMF55 is identical to the RN55 and I decided to give that a whirl. After all, the CMF55 has a pretty detailed spec sheet and seems like an overall good product. I decided to compare it against a TE YR1B-series resistor of the same resistance. The YR1B is roughly half the cost of the CMF55. Both resistors were 20 kΩ, ±0.1 % tolerance.
Here's an FFT of 50 W into 8 Ω at 1 kHz with the TE YR1B-series resistor.
And here is the same measurement with the Vishay/Dale CMF55 (= RN55) resistor:
Seems pretty obvious that the extra money doesn't buy you better performance.
These results are repeatable, by the way. I tried various combinations of resistors. The result is the same: The CMF55 is about 10 dB worse than the plain vanilla TE YR1B.
I think I'll try Holco (now owned by TE) next. Just because that's another brand of boutique resistors that seems well-regarded.
Tom
The resistor that's most likely to affect performance is the resistor from the output to the inverting input of the global feedback network. Same as in any other power amp. In Modulus-86 Rev. 3.0, that's R18.
The Dale RN55 "military grade" metal film resistors get a fair amount of press here. The "civilian grade" CMF55 is identical to the RN55 and I decided to give that a whirl. After all, the CMF55 has a pretty detailed spec sheet and seems like an overall good product. I decided to compare it against a TE YR1B-series resistor of the same resistance. The YR1B is roughly half the cost of the CMF55. Both resistors were 20 kΩ, ±0.1 % tolerance.
Here's an FFT of 50 W into 8 Ω at 1 kHz with the TE YR1B-series resistor.
And here is the same measurement with the Vishay/Dale CMF55 (= RN55) resistor:
Seems pretty obvious that the extra money doesn't buy you better performance.
These results are repeatable, by the way. I tried various combinations of resistors. The result is the same: The CMF55 is about 10 dB worse than the plain vanilla TE YR1B.
I think I'll try Holco (now owned by TE) next. Just because that's another brand of boutique resistors that seems well-regarded.
Tom
Hi Tom,
Your effort here begs a question you'd probably prefer to avoid -- SM resistors. If a few in critical locations could achieve predictable, consistent improvements, it might work if you were to space the through-hole pads closely such that they could work both ways. TH resistors would stand vertically and the SM ones (larger sizes) would lie flat. For other devices flexible-use layouts could extend DIY board design lives. It could well be that there are bigger compromises than I realize, but this might be a path forward for some situations.
Skip
Your effort here begs a question you'd probably prefer to avoid -- SM resistors. If a few in critical locations could achieve predictable, consistent improvements, it might work if you were to space the through-hole pads closely such that they could work both ways. TH resistors would stand vertically and the SM ones (larger sizes) would lie flat. For other devices flexible-use layouts could extend DIY board design lives. It could well be that there are bigger compromises than I realize, but this might be a path forward for some situations.
Skip
Skip PACK, I think he really wants to keep modulus86 thru hole only, and I appreciate that.
Tom, are you talking about 20k cmf55/rn55 1% 25-50 ppm vs 20k TE YR1B 0.1% 15ppm ?
Funny you mentioned the Holco H8 resistors, I have 2 of your modulus86 v3 boards I never started populating, for each "important"(r4/17 r7/18) resistors I actually have 3 kinds, cmf/rn55, h8 and TE YR1B ones.
I was also considering the ptf55 or mox700, 0.1% 15 ppm or less.
For electrolytic capacitors it was a little bit easier, I am very in love with the panasonic FM, great specs and good looks.
Tom, are you talking about 20k cmf55/rn55 1% 25-50 ppm vs 20k TE YR1B 0.1% 15ppm ?
Funny you mentioned the Holco H8 resistors, I have 2 of your modulus86 v3 boards I never started populating, for each "important"(r4/17 r7/18) resistors I actually have 3 kinds, cmf/rn55, h8 and TE YR1B ones.
I was also considering the ptf55 or mox700, 0.1% 15 ppm or less.
For electrolytic capacitors it was a little bit easier, I am very in love with the panasonic FM, great specs and good looks.
Edit: I can see now from the graph that the cmf55 are actually 0.1% but 25ppm vs 15ppm for YR1B resistors
Last edited:
Do you have one in 20 kΩ with 0.4" pin spacing that I can borrow? I'm not super keen on buying custom resistors as it would be completely unrealistic to expect my customers to do so.
One could do that. I'd probably just plunk down the SMD footprint on top of the PTH footprint in that case. I might have to fit a 1206 there.
I would indeed prefer the Modulus-86 to be all-PTH. No surface mount. In fact I'm not even all that keen on having a hybrid approach. I would only offer the SMD footprints if the SMD resistors provided better performance. And if I were to offer both footprints, the first support question would be, "would you solder the SMD resistors for me?" The answer will have to be no unless I get my assembly house to do the work. I've been down that path before and I just don't enjoy it.
That said, there are some more exciting options available in SMD, including Vishay's Z foil (or Z1 foil).
Tom
Heh. That's pretty funny. I'd rather deal with RN55 than with MELF. I've launched a few MELFs from my tweezers over the years.
I think it was in this thread that someone mentioned that they smacked a radial, dipped ceramic cap around a bit and concluded it was an 0603 with pins under the enamel.
A few years back a friend of mine sent me these:
Wouldn't you know it? They're all from TxCC (Texas Components Corporation). Some have incredibly tight tolerances (±0.001 %). Not that the absolute tolerance matters much for the MOD86, but it does indicate that these are pretty fancy resistors (as does the HC-49 package). There's a 20 kΩ and two 10 kΩ in the mix, so I can test with those.
Now, I don't know if these qualify as "naked". If you still see value in measuring the performance with a current generation TxCC Naked Foil resistor, toss me a PM and we can exchange address info. I'm happy to ship the resistors to you after testing if you would like as I won't have much use for them after.
Tom
I think it was in this thread that someone mentioned that they smacked a radial, dipped ceramic cap around a bit and concluded it was an 0603 with pins under the enamel.
That is an incredibly generous offer. Thank you. I'm happy to take the measurements, but see below first.
A few years back a friend of mine sent me these:
Wouldn't you know it? They're all from TxCC (Texas Components Corporation). Some have incredibly tight tolerances (±0.001 %). Not that the absolute tolerance matters much for the MOD86, but it does indicate that these are pretty fancy resistors (as does the HC-49 package). There's a 20 kΩ and two 10 kΩ in the mix, so I can test with those.
Now, I don't know if these qualify as "naked". If you still see value in measuring the performance with a current generation TxCC Naked Foil resistor, toss me a PM and we can exchange address info. I'm happy to ship the resistors to you after testing if you would like as I won't have much use for them after.
Tom
Last edited:
More measurements. Note that I've also now swapped the local feedback resistor around the LM3886. That's R2. It's 10 kΩ and I swapped it for 2x 4.99kOhm to see if it would make any difference. It didn't.
Here are the measurements with the fancy TxCC resistors. I had two 20 kΩ resistors. One with ±0.01% tolerance (boo!) and one with ±0.001% tolerance (yay!) in an HC-49 package (extra fancy!).
But the reference plot with $0.44/each (QTY 1) TE YR1B20KCC (20 kΩ, ±0.1 %, 15 ppm/ºC):
And the TxCC 20 kΩ ±0.01% (K1048S102K):
And the fanciest of them all, the TxCC 20 kΩ, ±0.001 %, K0624VHP202:
It looks like H3 is a smidge lower with the TxCC resistors but not dramatically so. The difference is 2.1 dB.
Tom
Here are the measurements with the fancy TxCC resistors. I had two 20 kΩ resistors. One with ±0.01% tolerance (boo!) and one with ±0.001% tolerance (yay!) in an HC-49 package (extra fancy!).
But the reference plot with $0.44/each (QTY 1) TE YR1B20KCC (20 kΩ, ±0.1 %, 15 ppm/ºC):
And the TxCC 20 kΩ ±0.01% (K1048S102K):
And the fanciest of them all, the TxCC 20 kΩ, ±0.001 %, K0624VHP202:
It looks like H3 is a smidge lower with the TxCC resistors but not dramatically so. The difference is 2.1 dB.
Tom
Last edited:
These objective results give me incredibly greater confidence that I am not missing out from boutique resistors. In a large sense, the numbers don’t lie. My only thought if there are other measurements-verified via well thought out listening tests-that are correlated with better perceived sound quality. I am not an EE, so I cannot tun much less interpret scientific/engineering journal articles, so someone else may wish to explore that, perhaps via a grant?