HDD to Case Bonding - Uptick in SQ

[One and a half]

@sandyk and I have differences now again with perception on files, but we do have an agreement on a simple method inside a PC which provides quite an audible uptick in SQ.

There's less noise generated between the hard drives, mobo and the PC's PSU case, the SQ is clearer, lifting the soundstage. The effort is minimal with the result quite a surprise!

 

The Music Server I use is a standard Supermicro case for audio serving only. There's a JCAT NET Femto wired directly to a Lumin U1. The PC has a standard ATX supply, Noctua Fan, and plenty of other bog standard type features found on typical PCs. The only LPS is fed to the JCAT,. The rest of the voltages are sourced from the ATX. For those that are building their split box Server, ithis method provides for simialr SQ enhancement, I would recommend to use at least a 4mm or 10AWG wire to bond the two cases together. Keeping the chassis at lowest potential difference minimises voltage across the cases and less noise transmission.

 

While poking about, I noticed a voltage in mVaC across the case of the SDD or 10TB HDD and the case. If this voltage is measured by a multimeter, then it's mains sourced. Look, no doubt if a very high impedance probes are placed on a scope, the most likely culrpit is the PSU but let's see how to 'shunt' this noise away.

 

The Supermicro case has insulated carriers for the hard drives, so their cases float above the chassis ground at various potentials, so the noise travels from the drive to the PSU with antenna effects making a general mess of things. To this end, I found a spare screw on each drive and made a short loop of braided wire to a single chassis point. I had a 4mm wire to hand and used this to bond the rear of the case and from there to the PSU. The JCAT Femto frame was also bonded.

 

End result is, the voltage developed across the 0V from the PSU and the cases are now 0.0000V with a Fluke 289. The SQ has a nice lift in reducing the gunge over Ethernet. If you have the JCAT USB card, the results will be similar. If you have a PC fed by a linear supply, the shunting of noise would also yield an improvement. For those PCs that have their mounting screws directly to the case, I would suggest to still wire in the bonding wires as well as using a star type washer which bites into the metal at one point only.

 

For those with an EtheRegen, I would bond the chassis ground to the ground terminal on the Etheregen. I will buy one of these one day, darn Covid is reducing the amount of available cash.

 

This post details where the wires went for the curious.

 

PSU 0V to frame

 

 

Drive to chassis bonding

 

 

Single point ground for the mobo

 

 

JCAT NET Femto and rear case bonding (stud attached with star washer). The blue wire of the JCAT is to the Lumin U1 25m away, the grey is from a TPLINK 24 way switch via a Baaske filter.

 

 

Measure Drive chassis to frame

 

 

[sandyk]

17 hours ago, One and a half said:

@sandyk and I have differences now again with perception on files, but we do have an agreement on a simple method inside a PC which provides quite an audible uptick in SQ.

There's less noise generated between the hard drives, mobo and the PC's PSU case, the SQ is clearer, lifting the soundstage. The effort is minimal with the result quite a surprise!

It's not only A.C.  I read more than +20mV from one SSD to 0 Volts, and around +40mV for an HDD before improving the Earthing.

 My PC used rotating plastic clips to lock them into position, and I also reverted to screwing them in via the existing screw holes in the metalwork.

[One and a half]

1 hour ago, sandyk said:

It's not only A.C.  I read more than +20mV from one SSD to 0 Volts, and around +40mV for an HDD before improving the Earthing.

 My PC used rotating plastic clips to lock them into position, and I also reverted to screwing them in via the existing screw holes in the metalwork.

Of consideration is AC induced hash, but DC , that's an offset built in.

[One and a half]

The venerable HP-Z800 with dual Xeons finally went belly up, mobo showed no life. Time for a different computer. Local dealer had a slightly older machine for sale at a good price, and put this to use in the office for audio playback and audio file processing.

.

For audio system in the office, there's an Accuphase E-450 integrated with internal DAC-50, KEF LS50 , the purpose is just to listen to music while I work.

 

Something was lacking though, the KEFs are not really desktop speakers thought they were muffled just sitting on the desk with acoustic foam, but playing files from the computer was not as involving as from a Yamaha CD-S2100.

 

So, let's try some different bonding techniques. The office PC case is pop riveted together, structurally OK, but not for keeping all components at the same potential.

 

The before measurements on the case to HDD and expansion board frames were several hundred mV due to the pop rivets and using painted black screws for fixings.

 

The after measurements were consistently lower than 1mV ac between frame, PSU, Hard Drive cases, JCAT USB, other expansion cards. Like with the Music Server, there's now a lift in soundstage, and revealing of details more familiar with my main system. Resistance values were worst at 3 milli ohms.

 

Photos

 

 

All cable bonds are 4mm (10 AWG) or 2.5mm (12 AWG). The busbar is bolted to the frame using standard brass motherboard spacers. Expansion and motherboard are directly wired.

 

 

HDD Cages are bonded, some drives are located with plastic supports, but all are wired separately. Anti-locking external washers were used for the drives fasteners to make sure the connections don't come adrift, while biting into the paint to form a connection to the metal. The coating on the Cooler Master paint is mainly  powder coated black, but underneath is conductive metal. I'm not going to water cool this PC, so the holes are handy.

 

 

 

Close up of the external washers on the PSU. 

 

[sandyk]

3 hours ago, One and a half said:

So, let's try some different bonding techniques. The office PC case is pop riveted together, structurally OK, but not for keeping all components at the same potential.

 

The before measurements on the case to HDD and expansion board frames were several hundred mV due to the pop rivets and using painted black screws for fixings.

 

The after measurements were consistently lower than 1mV ac between frame, PSU, Hard Drive cases, JCAT USB, other expansion cards. Like with the Music Server, there's now a lift in soundstage, and revealing of details more familiar with my main system. Resistance values were worst at 3 milli ohms.

Gary

 I achieved similar results with my Desktop PC, although my initial readings were not as high as your readings .

 Eventually, it dawned on me that the top half of the metal drive bay where my LG BR writer is, with the SSDs and HDDs in the bottom half, was also pop riveted to the bottom section. A short bypass link between them gave a further obvious improvement over just a short heavy duty strap to the bottom of the case, however connecting a short braided strap between the top half and chassis near a motherboard mounting hole, appeared to markedly improve the lower LF area, with increased dynamics as well, even with a compressed  thunder clap on a Telarc recording.

 

Alex  

[One and a half]

When measuring 'the before ', had the brainstorm to measure the frequency of the noise. Turned out frequencies were multiples of 60Hz, eg 180Hz, 360Hz. There was an odd one at 100Hz. The Fluke 289 can measure to just under 1MHz. I suppose with some more poking about others could be measured, but just concentrated on the hard drives and the expansion port frames.

 

Up to a certain frequency, is it 100Khz (?), plain wire is effective, above that ,braids are more effective as far as impedance of the cable is concerned. I tried to look for the reel of small 5mm braided wire in the garage, but a recent re-organisation &  poor memory recall didn't help in locating the reel. It will turn up. The results with the plain wire aren't bad at all and can be run anywhere, whereas the braid needs to be short and near hard drives needs insulation.

Still sticking with a star point system, rather than bunny hop jumpers, bit of a trade off between wire length ad jumpers, but decided to use short cables from the hard drives to the HDD frame, and from there to the star point.

 

In searching for PC cases that have a welded rather than pop rivet construction is met with some difficulty.  The old HP Z-800 had spot welded frames, but its non AT design means standard ATX, EATX  motherboards don't fit, so really can't be used unless a lot of work in modifying the power supply takes place. 

[One and a half]

12 hours ago, sandyk said:

<snip> appeared to markedly improve the lower LF area, with increased dynamics as well, even with a compressed  thunder clap on a Telarc recording.

 

Alex  

Given the LF noise that wanders uncontrolled in the wrong places, it would make sense that LF response would improve. In listening to Deep House music at the moment, the bass is nicely tight, before was quite woolly.

[sandyk]

On 10/8/2020 at 6:55 PM, One and a half said:

Close up of the external washers on the PSU. 

 Gary

 I have now also done something similar to that, but in my PC's case the earth resistance problems were much less to start with than in your PC's case .

 

Alex

[JonD]

An older thread, but an interesting one. 

 

I have moved my music out of my PC and onto a separate NAS drive. I was wondering if anyone has tried this, or similar, grounding on a NAS drive and found there to be a SQ improvement?

 

 

[lpost]

I'm curious, those that found audible SQ improvements, is your PC connected directly to DAC (USB, I2S, COAX) or network connected?

 

It's easy enough to ground all the components and case parts (relatively) just wondering in what scenario you realized results.

[One and a half]

18 hours ago, lpost said:

I'm curious, those that found audible SQ improvements, is your PC connected directly to DAC (USB, I2S, COAX) or network connected?

 

It's easy enough to ground all the components and case parts (relatively) just wondering in what scenario you realized results.

The transmission is Ethernet to a Lumin U1. Have also tried with USB Jcat card with similar results on a different daily driver PC.

 

Yes the bonding is easy to figure out and don't rely on the pop rivets.

[One and a half]

Update 21-11-28

Due to a Windows 10 upgrade impossible to fix Page_Fault blue screen on the main file server, I had to use the Music -Server to add in the 3 extra 6TB drives.

Just to get the music server going again, I suspended the wiring to the frame to each hard drive. The sound became worse, as expected. I thought to keep the wired network separate from the regular network, and used a disused Apple Extreme router via the JCAT NICs to the Lumin player. 

That worked out quite well, better than expected, the background common mode noise was reduced quite a lot with the DXE filters. I did miss a slight loss of stage height, once I have some time, I will re-bond the HDD frames to the main earth point and take some before and after measurements of the voltages. 

[One and a half]

A bit of work later and all drives are bonded, some better than others...

 

The main drama are the plastic cages that hold the drives in place, mainly to stop vibration from the drives to the case, but electrically isolating each one. There is a screw that can electrically attach the drive frame to the case, I found 3 of them in a screw bag that are left over screws from 20 odd years with various cases and motherboard hardware, but the fourth I could not find. They are fractionally longer than standard HDD screws and allow for the plastic HDD frame thickness. 

I saw these after the wires were bonded to the HDD Frames, but more the merrier.

 

The SSDs are on a separate painted metal 3.5in to 2.5in adapter, and the screws used there don't have the outside star washers, so they don't bite into the paint, so the figures for them are not so good. Have run out of M3 star washers, will purchase some more and use round head screws instead of countersunk.

 

The bonding cables are bare braided plated copper about 3mm across x 0.5mm thick, covered in heat shrink to avoid shorts.

 

Here's a table of the measured voltages, as soon as the case is bonded to earth (ground), the DC volts drop dramatically. The AC mV remain the same, though, tried to measure the frequency, but the meter couldn't lock on to the waveform. DC figures are very good, with the exception of Drive 1, must be that missing screw to the case.

 

 

SQ

Immediate effect of stage height returning to a position about 30cm above the tweeter line. Vocal especially pure and tightly focused image. Bass improved, since there's no anti bass frequencies any more from the noise. The image is now spread right across the two speakers with precise positioning and that used phase PRAT is really cool. Very happy with the result.

 

 

 

 

 

 

[One and a half]

Sorry for the twin excel sheets, the lower in the post should go, please ignore.

 

Bought some more star washers, added to the SSD's and the measurements dropped a little more. There's a differential of 77mV ac between 0V  (the 5V and 12V circuit power to HDD connector) that needs more tracing. The values are very small, if the DMM is set to auto measure, the values are all 0.0000 (shrug). 

The music drive, since I couldn't find the 6-32 x 3/8 UNC screw, remains at a higher potential than the other drives (2 to 4) in the same cage.

This server is mimicking a NAS, so the extra drives aren't normally there, plus other activity in the background is also not there either. Despite the extra activity, the SQ is actually very good. 

The WD Green SSD is not going to cooperate very well, despite improving the Samsung that sits directly underneath the WD by 1-2mm, the case voltages won't come down. Next maintenance time, will remove the WD, and see if the PC can boot off a M.2 drive with a PCIE adapter. The mobo built in 2018 may not boot from an NVME drive, another drama.

The sound improved a little more height to the sound stage, but bad recordings (like rock from the 70's) come out really bad, just no life, which is fairly accurate conversion from LP to digital conversions at the time. Good recordings do really shine, like Carmen Gomes, the hi res remaster of Court and Spark, just about any DSD.

 

 

Latest measurements with additional star washers on SSDs.

[christian u]

Very interesting. I have some investigation to do in my system. Thanks for posting!

(Came upon your post when I put Carmen Gomes in the search engine. New analog recording on the way it seems.)