Post by MartinT on May 3, 2023 17:04:33 GMT
Improving Your Digital Sound
I thought I would summarise some of my thinking around building a successful digital system since I started down this path three years ago. These are the things I’ve arrived at by trial and error and some considerable thought, and might just give you ideas of where to look next in order to release more performance from your system. Some are inexpensive improvements while others are decidedly not! My thoughts, my experience.
My Basic Ethos
In a digital system, noise is the biggest enemy of D to A conversion. The argument that ‘bits are bits’ and that nothing in the digital chain can affect the data, and therefore the music, completely misses the reality. It isn’t the data that noise impacts, it’s the carrier and, ultimately, the conversion at the DAC that is degraded. The more noise present on the line, the larger the uncertainty at the triggering threshold and the worse the deviation from perfect conversion into an analogue signal.
Noise in a digital system manifests as shrillness or harshness, lack of detail in the high frequencies, coarse midrange, flabby or tuneless bass and a decided lack of any real soundstage, spatial cues or note decay. Strangely, it does not manifest in the analogue domain as noise!
Mains and DC Power
Power quality to each component is critical. A regenerator is ideal to guarantee a good mains waveform; failing that, some kind of filtering is beneficial to remove the worst of the mains-borne noise from RFI and EMI. A dedicated mains spur is a good idea if practicable, and unswitched mains sockets. Power cables should be selected for their ability to filter out GigaHertz band noise and prevent further noise pickup. High quality fuses like the Synergistic Research range help to further improve power quality. I use two mains regenerators, spilt into signal and source component duties, connected to a 60A dedicated spur with its own consumer unit. The regenerators are set to sinewave output at 230V AC. I use SR fuses throughout, a mixture of Orange for most components and cables, and Purples for the core digital devices.
Stand-alone power supplies should be linear (SMPS or Switched Mode Power Supplies generally radiate too much high frequency noise and send interference back into the mains power system). An LPS with supercap capacitors, which provide a very low output impedance, tend to sound even better due to tight control of the power rails. The advice about power cables also applies to LPS. DC feed cables should also be of high quality, watch out for the poor quality ‘freebies’ supplied with some equipment. I use a Coherent QP-2 and two Zero Zone 01-A supercap PSUs with Coherent DC cables throughout. The Zero Zones are especially good value for money with excellent performance.
Supports & Weights
Vibration control is necessary in order to reduce another source of noise. This is especially important for anything that has a crystal inside, whether it be a Femto/TCXO/OCXO type. They are highly prone to vibrations due to the piezo effect. This means that routers, switches, streamers, DDC and DACs as well as clocks should all be given attention to minimise susceptibility to vibrations. Good footers like Black Ravioli work extremely well in conjunction with heavy weights placed on top. I use a mixture of Black Ravioli Big Pads and Pads with Schramm door stop weights. The key components are in turn supported on Electric Beach S1NX platforms, in turn supported on RDC Cones. The system rack is also supported on RDC Cones.
Interconnect Cables
Ethernet cables vary and the better ones are audibly superior. A decent moderately priced cable like the Supra Cat8 or AudioQuest Pearl is worth the investment. Be aware that you may cause earthing problems and inferior sound when using Cat7 or Cat8 cables with their metal shields if the connected equipment routes the socket shields to the chassis. You can test this with a meter. If in doubt, use a Cat6 cable, with only plastic connectors, in these circumstances. However, many do find that shielded ethernet cables used for audio (like the Supra Cat8) sound better. I use a Supra Cat8 ethernet cable.
USB cables also vary in sound, with the ‘freebie’ cables generally sounding thin, constrained and muddled. Look for ones that meet the high bandwidth USB spec properly and have good shielding. Even then, their sound can vary from one model to another. Keep them as short as you can get away with. The AudioQuest Carbon is a good USB cable. My current cable, the Mad Scientist Black Magic Ultra, uses graphene for shielding and it works very well indeed.
Fibre Optic. I have been astonished at how varied the sound of SFP optical adapters and fibre cables can be. My favourite SFP is the single-mode HP JD094B-C and matching ipolex cable. This combo has the advantage of supporting up to 10km lengths, so no problem with a run the length of the house! If you want the absolute best sound quality at lengths of 2-5m, go for an AOC (Active Optical Cable) which is a fibre cable with integrated SFP connectors. The best AOC I have yet tried is the Cisco SFP-10G-AOC5M (this is the 5m version), which sounds rich and dynamic with no tonal bias.
Clock cables can also affect sound quality, although generally less so than interconnects. Here, shielding matters a great deal for sinewave clocks whereas impedance is more critical for squarewave clocks. At the minimum level, a Belden 4694R or Canare LV-77S (75Ω) cable or LMR-400 (50Ω) cable will sound better than a freebie. There are debates as to whether copper or silver sounds best in a high end clock cable. Some ranges, like the Harmonic Technology and Neotech products, are available in both conductors. The best clock cables I have tried are the Harmonic Technology Digital Copper III 50Ω and the Neotech NEVD-2001 silver 75Ω.
I2S cables can also affect sound quality to about the same level as a clock cable. Here the quality and purity of the conductors and adherence to the HDMI standards (although I2S signals are different) is of utmost importance. I have found the Blue Jeans FE HDMI cable to sound very good indeed for an inexpensive product. If you want a dedicated I2S cable, the silver Tubulus Argentus is superb, but don’t expect anything other than fairly subtle improvements over the BJ.
Grounding
Grounding boxes work by sucking away ultra-high frequency noise from ground planes or mains earth. They do this through containing a mix of ingredients that enhance their ability to act like a miniature real Earth, as done by the power companies. Their reason for existence is that they are ‘quieter’ than mains earth and therefore do a better job of absorbing noise from delicate digital ground planes inside components. My strong advice, controversial as it may sound, is to have one grounding box per component. Do not share them unless they have multiple independent terminals. There are various popular models of grounding box but my favourite, by far, is the Quartz Acoustic Premium. You can use different grounding cables and they should terminate according to the equipment, with spade, phono, BNC, USB or crocodile clip to the component. In the case of SFPs, I clip to the outer metal casing of the SFP in order to provide the most effective grounding. Grounding boxes have another almost unique attribute: they can take a long time to achieve their final sound quality, sometimes as much as 2-4 weeks of continuous running.
Grounding boxes can also be highly effective when used to ground the mains earth, sucking away more noise from the core components (many components connect their ground plane to mains earth, allowing noise into the system). I use QA boxes for each main system component and a whole host of different models (11 at the last count) for grounding mains earth throughout the house using a cable with a plug end to the ground pin.
Clocks and Clock Cables
An external clock exists only as an improvement to the one inside a digital component. Clocks for audio use have a hierarchy of Femto -> TCXO -> OCXO and most quality external clocks are OCXO (Oven Controlled Crystal Oscillator) which place the crystal inside a tiny oven, kept strictly temperature controlled. This is why they are best left switched on, because their warm-up and settling time can be 12 hours or more. The most common standard is 10MHz frequency for hi-fi components, most often used for switches, reclockers, streamers, DDCs and DACs. Further to this, clocks subdivide into sinewave and squarewave clocks. There is considerable debate as to which sounds better, but it probably depends on the matching equipment and cables. The most reliable measure of ‘goodness’ of a clock signal, but not the only parameter, is phase noise, measured at 1Hz or 10Hz. A decent external clock needs to be around -100dB at 1Hz to better an internal crystal oscillator while the top models are -121dB at 1Hz or lower. Price goes up accordingly as yield drops. Some users have reported good results by adding a 10MHz low-pass filter, such as the Mini-Circuits BLP-10.7+ or BLP-10.7-75+, in order to further reduce noise but only do this with a sinewave clock.
Clock cables should be high quality and suitable for the purpose. Squarewave clock cables are sensitive to the correct impedance and should match the receiving equipment. Sinewave clock cables are sensitive to noise and should be very well shielded.
I use two AfterDark sinewave clocks in my system, rated at -107dB (1st clock) and -118dB (2nd clock).
Isolation (Air Gapping)
There are several ways of providing galvanic isolation in a digital system and each method reduces noise carried from previous stages. Your system should have one or more methods. I use these five in my system:
Other reclockers and DDCs may also incorporate a moat. It's worth checking. When providing power to fibre link media converters or a device with a moat, make sure you are not shorting the isolation by using a common power supply.
Contacts
A good contact cleaner like Mad Scientist Graphene, Furutech Nano or Deoxit should be used on all contacts from mains power connectors to DC, digital and grounding connectors. This is time-consuming and should be done on a rainy day. Great care must be taken not to over-use the cleaner and to not leave any lint in tiny contacts such as HDMI connectors. Give it time to take effect, it can take 12 – 24 hours for contact performance to maximise. I used MS Graphene as it is relatively easy to apply, can be seen as feint grey, and works exceptionally well after 12 hours.
Burn-in and Warm-up
Yes, burn-in is a real phenomenon. There are different ways of burning-in digital components but it requires patience whichever method you use. You can just leave it all switched on but that can take months to achieve full sound quality. I do, though, leave my source components switched on so that warm-up is not a problem. Few systems sound at their best from stone cold. You can run music on a loop and that cuts down the burn-in time to maybe 1,000 hours. Or, best of the lot, is to run a special burn-in tone that exercises the chain thoroughly and gets great results from around 100 hours, but makes a step-change in performance from around 300-400 hours. The most effective burn-in tone I’ve ever tried by a large margin is the Tara Labs Cascade. Don’t think about it, just run it with the volume set very low for 400 hours in stages or in one blast. You will be amazed, especially when that step-change occurs and the soundstage suddenly becomes 3D walk-in.
I thought I would summarise some of my thinking around building a successful digital system since I started down this path three years ago. These are the things I’ve arrived at by trial and error and some considerable thought, and might just give you ideas of where to look next in order to release more performance from your system. Some are inexpensive improvements while others are decidedly not! My thoughts, my experience.
My Basic Ethos
In a digital system, noise is the biggest enemy of D to A conversion. The argument that ‘bits are bits’ and that nothing in the digital chain can affect the data, and therefore the music, completely misses the reality. It isn’t the data that noise impacts, it’s the carrier and, ultimately, the conversion at the DAC that is degraded. The more noise present on the line, the larger the uncertainty at the triggering threshold and the worse the deviation from perfect conversion into an analogue signal.
Noise in a digital system manifests as shrillness or harshness, lack of detail in the high frequencies, coarse midrange, flabby or tuneless bass and a decided lack of any real soundstage, spatial cues or note decay. Strangely, it does not manifest in the analogue domain as noise!
Mains and DC Power
Power quality to each component is critical. A regenerator is ideal to guarantee a good mains waveform; failing that, some kind of filtering is beneficial to remove the worst of the mains-borne noise from RFI and EMI. A dedicated mains spur is a good idea if practicable, and unswitched mains sockets. Power cables should be selected for their ability to filter out GigaHertz band noise and prevent further noise pickup. High quality fuses like the Synergistic Research range help to further improve power quality. I use two mains regenerators, spilt into signal and source component duties, connected to a 60A dedicated spur with its own consumer unit. The regenerators are set to sinewave output at 230V AC. I use SR fuses throughout, a mixture of Orange for most components and cables, and Purples for the core digital devices.
Stand-alone power supplies should be linear (SMPS or Switched Mode Power Supplies generally radiate too much high frequency noise and send interference back into the mains power system). An LPS with supercap capacitors, which provide a very low output impedance, tend to sound even better due to tight control of the power rails. The advice about power cables also applies to LPS. DC feed cables should also be of high quality, watch out for the poor quality ‘freebies’ supplied with some equipment. I use a Coherent QP-2 and two Zero Zone 01-A supercap PSUs with Coherent DC cables throughout. The Zero Zones are especially good value for money with excellent performance.
Supports & Weights
Vibration control is necessary in order to reduce another source of noise. This is especially important for anything that has a crystal inside, whether it be a Femto/TCXO/OCXO type. They are highly prone to vibrations due to the piezo effect. This means that routers, switches, streamers, DDC and DACs as well as clocks should all be given attention to minimise susceptibility to vibrations. Good footers like Black Ravioli work extremely well in conjunction with heavy weights placed on top. I use a mixture of Black Ravioli Big Pads and Pads with Schramm door stop weights. The key components are in turn supported on Electric Beach S1NX platforms, in turn supported on RDC Cones. The system rack is also supported on RDC Cones.
Interconnect Cables
Ethernet cables vary and the better ones are audibly superior. A decent moderately priced cable like the Supra Cat8 or AudioQuest Pearl is worth the investment. Be aware that you may cause earthing problems and inferior sound when using Cat7 or Cat8 cables with their metal shields if the connected equipment routes the socket shields to the chassis. You can test this with a meter. If in doubt, use a Cat6 cable, with only plastic connectors, in these circumstances. However, many do find that shielded ethernet cables used for audio (like the Supra Cat8) sound better. I use a Supra Cat8 ethernet cable.
USB cables also vary in sound, with the ‘freebie’ cables generally sounding thin, constrained and muddled. Look for ones that meet the high bandwidth USB spec properly and have good shielding. Even then, their sound can vary from one model to another. Keep them as short as you can get away with. The AudioQuest Carbon is a good USB cable. My current cable, the Mad Scientist Black Magic Ultra, uses graphene for shielding and it works very well indeed.
Fibre Optic. I have been astonished at how varied the sound of SFP optical adapters and fibre cables can be. My favourite SFP is the single-mode HP JD094B-C and matching ipolex cable. This combo has the advantage of supporting up to 10km lengths, so no problem with a run the length of the house! If you want the absolute best sound quality at lengths of 2-5m, go for an AOC (Active Optical Cable) which is a fibre cable with integrated SFP connectors. The best AOC I have yet tried is the Cisco SFP-10G-AOC5M (this is the 5m version), which sounds rich and dynamic with no tonal bias.
Clock cables can also affect sound quality, although generally less so than interconnects. Here, shielding matters a great deal for sinewave clocks whereas impedance is more critical for squarewave clocks. At the minimum level, a Belden 4694R or Canare LV-77S (75Ω) cable or LMR-400 (50Ω) cable will sound better than a freebie. There are debates as to whether copper or silver sounds best in a high end clock cable. Some ranges, like the Harmonic Technology and Neotech products, are available in both conductors. The best clock cables I have tried are the Harmonic Technology Digital Copper III 50Ω and the Neotech NEVD-2001 silver 75Ω.
I2S cables can also affect sound quality to about the same level as a clock cable. Here the quality and purity of the conductors and adherence to the HDMI standards (although I2S signals are different) is of utmost importance. I have found the Blue Jeans FE HDMI cable to sound very good indeed for an inexpensive product. If you want a dedicated I2S cable, the silver Tubulus Argentus is superb, but don’t expect anything other than fairly subtle improvements over the BJ.
Grounding
Grounding boxes work by sucking away ultra-high frequency noise from ground planes or mains earth. They do this through containing a mix of ingredients that enhance their ability to act like a miniature real Earth, as done by the power companies. Their reason for existence is that they are ‘quieter’ than mains earth and therefore do a better job of absorbing noise from delicate digital ground planes inside components. My strong advice, controversial as it may sound, is to have one grounding box per component. Do not share them unless they have multiple independent terminals. There are various popular models of grounding box but my favourite, by far, is the Quartz Acoustic Premium. You can use different grounding cables and they should terminate according to the equipment, with spade, phono, BNC, USB or crocodile clip to the component. In the case of SFPs, I clip to the outer metal casing of the SFP in order to provide the most effective grounding. Grounding boxes have another almost unique attribute: they can take a long time to achieve their final sound quality, sometimes as much as 2-4 weeks of continuous running.
Grounding boxes can also be highly effective when used to ground the mains earth, sucking away more noise from the core components (many components connect their ground plane to mains earth, allowing noise into the system). I use QA boxes for each main system component and a whole host of different models (11 at the last count) for grounding mains earth throughout the house using a cable with a plug end to the ground pin.
Clocks and Clock Cables
An external clock exists only as an improvement to the one inside a digital component. Clocks for audio use have a hierarchy of Femto -> TCXO -> OCXO and most quality external clocks are OCXO (Oven Controlled Crystal Oscillator) which place the crystal inside a tiny oven, kept strictly temperature controlled. This is why they are best left switched on, because their warm-up and settling time can be 12 hours or more. The most common standard is 10MHz frequency for hi-fi components, most often used for switches, reclockers, streamers, DDCs and DACs. Further to this, clocks subdivide into sinewave and squarewave clocks. There is considerable debate as to which sounds better, but it probably depends on the matching equipment and cables. The most reliable measure of ‘goodness’ of a clock signal, but not the only parameter, is phase noise, measured at 1Hz or 10Hz. A decent external clock needs to be around -100dB at 1Hz to better an internal crystal oscillator while the top models are -121dB at 1Hz or lower. Price goes up accordingly as yield drops. Some users have reported good results by adding a 10MHz low-pass filter, such as the Mini-Circuits BLP-10.7+ or BLP-10.7-75+, in order to further reduce noise but only do this with a sinewave clock.
Clock cables should be high quality and suitable for the purpose. Squarewave clock cables are sensitive to the correct impedance and should match the receiving equipment. Sinewave clock cables are sensitive to noise and should be very well shielded.
I use two AfterDark sinewave clocks in my system, rated at -107dB (1st clock) and -118dB (2nd clock).
Isolation (Air Gapping)
There are several ways of providing galvanic isolation in a digital system and each method reduces noise carried from previous stages. Your system should have one or more methods. I use these five in my system:
| Method | How |
| 5G | is mobile data delivered by radio over the airwaves |
| DXE | Ethernet isolation modules with transformers |
| EtherREGEN | Incorporates a ‘moat’ between input and output |
| Fibre Optic | Optical fibre link with SFP transceivers at either end |
| U18 DDC | Incorporates a ‘moat’ between input and output |
Other reclockers and DDCs may also incorporate a moat. It's worth checking. When providing power to fibre link media converters or a device with a moat, make sure you are not shorting the isolation by using a common power supply.
Contacts
A good contact cleaner like Mad Scientist Graphene, Furutech Nano or Deoxit should be used on all contacts from mains power connectors to DC, digital and grounding connectors. This is time-consuming and should be done on a rainy day. Great care must be taken not to over-use the cleaner and to not leave any lint in tiny contacts such as HDMI connectors. Give it time to take effect, it can take 12 – 24 hours for contact performance to maximise. I used MS Graphene as it is relatively easy to apply, can be seen as feint grey, and works exceptionally well after 12 hours.
Burn-in and Warm-up
Yes, burn-in is a real phenomenon. There are different ways of burning-in digital components but it requires patience whichever method you use. You can just leave it all switched on but that can take months to achieve full sound quality. I do, though, leave my source components switched on so that warm-up is not a problem. Few systems sound at their best from stone cold. You can run music on a loop and that cuts down the burn-in time to maybe 1,000 hours. Or, best of the lot, is to run a special burn-in tone that exercises the chain thoroughly and gets great results from around 100 hours, but makes a step-change in performance from around 300-400 hours. The most effective burn-in tone I’ve ever tried by a large margin is the Tara Labs Cascade. Don’t think about it, just run it with the volume set very low for 400 hours in stages or in one blast. You will be amazed, especially when that step-change occurs and the soundstage suddenly becomes 3D walk-in.
