Ben, one other thing - I don't have a spectrum analyzer, but I used the app Spectroid on a decent Android tablet (Samsung S6 Lite) to try to look at the buzz I am hearing. In contrast to other cases where I have used this app, I did not see any suggestion of a peak at either 60Hz or 120Hz, or any other specific frequency; rather, the spectrum had a pretty flat trend from low bass through 1KHz. This surprised me.
What is the sensitivity of your speakers? If around 100dB then the noise level is low. If around 85dB, then the noise level is high.
Looking at your pictures, it looks like the chassis is painted. Check to be sure that the ground connections for the PS ground lift thermistor, power transformer shield, and AC safety ground to the chassis are of negligible resistance.
You can also try measuring the noise AC voltage level at the speaker with your multimeter.
Looking at your pictures, it looks like the chassis is painted. Check to be sure that the ground connections for the PS ground lift thermistor, power transformer shield, and AC safety ground to the chassis are of negligible resistance.
You can also try measuring the noise AC voltage level at the speaker with your multimeter.
Last edited:
@northpaw I havent noticed mention so forgive me if Im covering ground already addressed. The perforations visible below your heatsink show that a considerable amount of material can be removed if you're so inclined. Wholesale removal of large sections with a dremel and adding less restrictive screen (or not) would be the most direct method with the biggest payout. Tediously drilling out each hole slightly would offer some improvement and the % increase would be easy enough to calculate. I might even find the implementation worthwhile for an irreplaceable bespoke case but I imagine you value your sanity more than a precision fix on a re-use box
Please attempt to play any of BACH's Werks for Harpsichord and Orgel if you have them. ESP. for Harpsichord as you have an ear for the piano.
The DCM Time Windows (1st gen) are speced at 89dB.
As for the grounds, I did a number of checks. While the F5m was still installed in the system, I disconnected the input RCAs (while unit cycled off, of course) from the preamp (a reworked Kenwood Basic C2), and when I restarted the F5m, 90-95% of the noise I was hearing at the speakers was quenched. I then tried hooking up the F5m directly to a SMSL DO100 DAC, and the noise returned. I take this to mean that I have a signal ground issue between the F5m and my sources, rather than an internal issue with the amp circuitry in the F5m.
To make sure it was not related to house AC ground, I temporarily lifted the AC ground to the F5m (used a 3-prong to 2-prong converter plug), which made no difference.
Moving the unit to the bench, so it is no longer connected to anything but AC power, I measure the AC on the speaker terminals. I get steady readings of 0.42mV AC and 0.25mV AC cold, and 0.55mV AC and 0.33mV AC when warmed up. These were made with a meter claiming to be a "True RMS" meter, if that makes a difference. I realize now I should have also measured AC on the speaker terminals while connected to the system sources, and I will have to update with that info later, once I return the F5m to the system.
Continuing on, to check the grounding internally in the F5m, I first reverify that the chassis post I am using as the star ground point has 0Ω relative to bare metal of chassis (threads in a screw hole). I had checked that during the build, but I had not verified that I had 0Ω between that point and the things I connected to it, so I did that now. I measure 0Ω between star ground point and IEC inlet ground, and the star ground point and thermistor lead at corner of PS PCB. I can’t verify the transformer shield connection as there is nothing exposed except for the purple wire itself.
I then checked the resistance between the star ground and the signal grounds (RCA shields, speaker black terminals) and PS grounds. All showed 10Ω, the value of the cold thermistor (that one stays cold) that sits between those grounds and the chassis, so that looks good. As a double check, I measured the resistance among just those signal and PS grounds, and all were 0Ω.
So all the internal stuff appears copacetic, and I conclude I’ve got to resolve what appears to be a difference in the ground potential on the RCA shields of the F5m and the RCA shields of my sources.
Moving on to the fan issues...
First, I was able to verify that the objectionable noise I am now hearing is the result of more suction acting on the fans once I taped up the extra air vent slots on the chassis/cover. I proved this by lifting the fans ever so slightly, and the noise level dropped. It was apparent that most of the noise was the additional load on the fans, so it does appear to be fan noise (some laboring) and maybe not so much an air rushing sound. So @Ben Mah is correct again: decreasing the resistance to flow will increase the efficiency of cooling, as well as cutting noise. Bravo, Ben. In that vein, I now verified that the 1°C reduction in heatsink temperature I saw yesterday was due to adding an additional 1" clearance under the chassis, and not the temporary experiment with a chimney above.
To further reduce this resistance to flow, this morning I wholesale removed the expanded metal screen that was fitted beneath the heat sink. @seventenths : this is even better than hollowing out parts of the screen with a Dremel
. I'm sure its purpose was a spider guard. I'll have to report back as to how effective its removal is.
If it is not effective enough, I plan to remove a little bit of the taping in the right places to still maintain dominant flow through the heatsinks, or I will consider running with the cover off and the fans resting directly on the heatsinks (no kids or grandkids living with us, so I can do that safely). With the F5m running on the bench right now, with cover off, no screen, fans on heatsink (running quietly), and bias pots left untouched from yesterday, the sinks have dropped to 43°C.
First, I was able to verify that the objectionable noise I am now hearing is the result of more suction acting on the fans once I taped up the extra air vent slots on the chassis/cover. I proved this by lifting the fans ever so slightly, and the noise level dropped. It was apparent that most of the noise was the additional load on the fans, so it does appear to be fan noise (some laboring) and maybe not so much an air rushing sound. So @Ben Mah is correct again: decreasing the resistance to flow will increase the efficiency of cooling, as well as cutting noise. Bravo, Ben. In that vein, I now verified that the 1°C reduction in heatsink temperature I saw yesterday was due to adding an additional 1" clearance under the chassis, and not the temporary experiment with a chimney above.
To further reduce this resistance to flow, this morning I wholesale removed the expanded metal screen that was fitted beneath the heat sink. @seventenths : this is even better than hollowing out parts of the screen with a Dremel
. I'm sure its purpose was a spider guard. I'll have to report back as to how effective its removal is.If it is not effective enough, I plan to remove a little bit of the taping in the right places to still maintain dominant flow through the heatsinks, or I will consider running with the cover off and the fans resting directly on the heatsinks (no kids or grandkids living with us, so I can do that safely). With the F5m running on the bench right now, with cover off, no screen, fans on heatsink (running quietly), and bias pots left untouched from yesterday, the sinks have dropped to 43°C.
The Circuit Ground to Chassis Connection....
These days I make that connection at the very input of the power amp. The reason is that unwanted power supply currents can flow between equipment due to the huge loop area. Mostly innocuous magnetic fields can also cause significant noise currents if the loop has not been minimized.
These days I make that connection at the very input of the power amp. The reason is that unwanted power supply currents can flow between equipment due to the huge loop area. Mostly innocuous magnetic fields can also cause significant noise currents if the loop has not been minimized.
Grounding of equipment within a system is also important.
The most important thing about ground is that all equipment in a system is at the same ground potential. To do that all equipment should be have zero voltage differential between their respective grounds. The easiest way to achieve that is to plug all equipment within a system into one wall AC outlet. Use a power bar if necessary. Also keep the power cords together as far as possible to reduce loop areas.
diyAudio member Bonsai (Andrew Russell) has information on the subject: https://hifisonix.com/wp-content/uploads/2019/02/Ground-Loops.pdf
The most important thing about ground is that all equipment in a system is at the same ground potential. To do that all equipment should be have zero voltage differential between their respective grounds. The easiest way to achieve that is to plug all equipment within a system into one wall AC outlet. Use a power bar if necessary. Also keep the power cords together as far as possible to reduce loop areas.
diyAudio member Bonsai (Andrew Russell) has information on the subject: https://hifisonix.com/wp-content/uploads/2019/02/Ground-Loops.pdf
Attachments
Thanks for your comments. I will not just change component values unless someone with more knowledge confirms it's worth a try.
Paralleling JFets is mentioned in the F5T paper and compensation is recommended from 4N/4P, I currently use 2N/2P. Regarding "Transparency", well lets just say the F5T does many things very well and better than the F4, but after using 2 F4s in balanced mode for many years I know it well, and and there is a smoothness and a level of detail I can hear in some situations that my F5T in its current configuration doesn't reach. The acoustic events are still there, but less pronounced and more in the background. Anyway, it measures and sounds very good so I assume its build correctly and my current feedback network is in line with the F5T paper.Best Regards
Sven
The measured AC seems a bit high. It was also higher when warmed up. Was the offset minimized at operating temperature?
However I do not have a F5M so I cannot say for certain what is expected. Perhaps members with F5M can measure their amps.
Last edited:
Another check is to connect only one channel of the source to the amplifier and listen for noise. Is there additional noise? Or the same as no source?
Yes, I have everything plugged into the same power strip. Earlier I had the dac on a different strip, but I have corrected that and found no change in the noise issue. I am aware that a different AC ground can show up on the RCA shields and the mismatch will propagate via that to the preamp, etc., so it was good to correct that. As for AC cord management, to group the power cords together to minimize loop area is something I have to work on. Currently it is a mess, but this remarkable new amp is a good impetus to cull some less-used items out of the stack I have let grow.
That 0.55mV AC and 0.33mV AC when warmed up with no RCA input connection compares to values of 0.42mV and 0.28mV AC warmed up with the RCA inputs connected to the system.
Yes, the DC offset at operating temp was minimized (always well under 10mV, and presently at 2.6 and 1.3mV); with the 1-turn pots, it isn't possible to precisely set the offset or bias, at least by me, because seemingly just holding a screwdriver in the slot of the adjust screws and wishing it to turn it will produce a change. It is more of a do the slightest jog you can manage and see where it lands, with a fair amount of back and forth between the two pots. But as far as the DC offset goes, it seems pretty stable once set.
I tried disconnecting the RCA input on one channel, and the buzz remained on both. I then disconnected just the other channel, with the same result. Smells to me like a system ground or PS issue. I also disconnected both RCA inputs, and in contrast to what I reported in post #1086, it now made no difference in the noise. I suspect I am moving too fast and may be making some mistakes, so I've got to slow down and be sure I am being rigorously methodical. I was also juggling grandkid babysitting today, which can't help...
On a more positive note, I feel I am now at a good place with the fans. Removing the spider screen from under the heatsink, removing a little bit of the tape I had added to the air vents, and providing an additional inch of clearance under the chassis appears to have both reduced fan noise and improved cooling. I was able to up the bias to 530mV R and 539mV L, with the sinks at 50°C, giving an average bias of 1.14A, all on the bench. Moving the F5m back to the system and letting it settle for a few hours recovered the same 50°C sink temp, so the bias number should hold. I also was able to lower the physical height where I place the F5m, so that it is no longer line-of-sight to the listening position. I can't seem to hear it now, but I will know for sure overnight, when the room background noise drops to 0.
I want to thank Ben and all the others who have offered so many helpful comments. I greatly appreciate it. I'm not out of the woods on the grounding yet, but I've got to do more reading and thinking first on how to optimize those things. That said, the buzz isn't audible to me from a foot away, so I do feel it is a problem I can pursue as time allows.
As part of that, I want to place the F5m in the main system I have in the living room, to see if the noise issue presents in the same way there. That system has a Nakamichi PA-5, a McCormack TLC-1 Deluxe Preamp, an Adcom GDA-700 Dac, and some DIY powerline filtering (Destroyer OS's Fo-Felix and a DC-blocker I made from Rod Elliott's design), and is very quiet outputting to ADS L-730 speakers (92 dB sensitivity). Putting the F5m in place of the Nak PA-5 in that system will be a good test.
I powered up ch #2 this morning for one last check before hauling the whole unhoused rig upstairs for a first listen outside of my crappy test speaker. Reading this thread as it warmed up, I reached over and switched the output offset meter to mVac. It was right there with your reported readings at first but climbed as high as 18mV @3-4kHz. My test rig PS is a basic CRC without any caps across the primary, secondary or bridges and after 20 minutes or so of reading on the subject and ready to start playing around with it, the amp had plateaued at operating temp and was sitting at under 0.15mVac/0.000Hz . Nothing I could power cycle in the shop had any effect whatsoever, including a 450W LED fixture less than 4' away. Curiosity has me letting it all cool down to room temp to see what happens. Just before powering it down I held my phone next to the board and it did indeed induce a small amount of ac but is clearly not associated. Beyond 2"-3" there was nothing.
I guess everyone is getting along just fine with those single turn pots eh?
I've come to realize how much I enjoy the process these days and an evening of watching the interplay between the two pots and tweaking the offset beyond any meaningful level is just fine. I can't help but think that a single turn pot would feel like rolling dice and eventually getting lucky. Maybe you boys with better fine motor control have a different experience.
I guess everyone is getting along just fine with those single turn pots eh?
I've come to realize how much I enjoy the process these days and an evening of watching the interplay between the two pots and tweaking the offset beyond any meaningful level is just fine. I can't help but think that a single turn pot would feel like rolling dice and eventually getting lucky. Maybe you boys with better fine motor control have a different experience.