Harveysouth
New Head-Fier
2 GHz Noise
Before I added the RF isolation treatment on the M scaler outputs, HF GHz clip on ferrites sounded considerably better - warmer sound quality, with better depth and focus. I understand why it works as the problem is down to RF currents (peak problem area is around 2 GHz) flowing in a loop from the mains>Hugo M scaler FPGA ground>galvanic isolated OP>BNC cable> DAC ground plane> back to the mains. It's when the current flowing through the DAC ground plane is where the SQ problems are happening. If we can reduce the value of the currents by increasing the loop impedance then we get less voltage drops across the DAC ground plane and then less correlated noise and lower noise floor modulation. The lower noise floor modulation improves warmth; the reduction in correlated noise (that's noise/distortion that is in the audio bandwidth and is linked to the actual signal) degrades small signal amplitude accuracy, and that degrades the perception of depth.
It's all very straightforward and well understood - double the loop impedance at 2GHz and currents will half, and we get a corresponding improvement in sound quality - better depth and a warmer sound.
It doesn't matter too where you increase the loop impedance either; you can do it at the BNC output drivers, or through the cable, or via the PSU on the M scaler, or via the PSU on the DAC - it's the total loop impedance that counts. If the loop impedance is infinite, no current can flow, audio nirvana results. Problem is, at 2GHz even 2pF of capacitance (which is a very small capacitance) is very significant - it is 40 ohms, so will allow current to flow. Moreover, the ear/brain is ultra sensitive to this problem, so the faintest trace of an error is audible.
So why did adding ferrites improve the SQ? Ferrites work by increasing the common mode impedance, but keeping the differential impedance the same. The wanted signal is differential, the unwanted current is common mode; so adding ferrites increases the loop impedance, without affecting the wanted signal; if you double the number of ferrites, you half the loop currents, thus halving the SQ problem.
The treatments I put into the M scaler was extensive and aimed at increasing the loop impedance, by improving the galvanic isolation, improving the drivers and the driver PSU, and adding RF filters and chip ferrites into the BNC output ground. In terms of sound quality, I gained about an order of magnitude improvement in SQ - it went from a big problem to a small one. When I then added clip on ferrites it sounded worse - and I was not expecting that. I am not sure of the reason exactly - I suspect a resonance between the internal M scaler ferrites and the clip on ferrites air gap, making the loop impedance lower overall. But solid core ferrites do still improve the SQ - but I should add that it's not a huge difference. When I travel, I don't bother, as it's simply too small a difference and I stick to stock. Compared to the SQ that the M scaler adds, we are talking about the surface finish of the icing on top of the cake - it's the iced cake that's important.
But remember it's the overall loop impedance that's important, and the PSUs come into this too. If you replace the supplied PSU with linears, apart from breaking the warranty, will degrade the sound quality. This is because linear PSUs have no RF filters - the supplied one comes with input and output RF filters. Also the supplied one has been very carefully selected for no grounding, and very low inter-winding capacitance - many times smaller than toroidal transformers. The lower inter-winding capacitance, together with the RF filters, increase the loop impedance and improve the sound quality. By using a linear will increase noise floor modulation (making it sound brighter, fooling the audiophile into thinking it's more transparent) and increase correlated noise thereby degrading depth (and then the hapless audiophile will then convince themselves into thinking it's better because perceived width is artificially wider due to the poorer depth). Confirmation bias and placebo is a very real problem with listening tests!
But if you want to almost completely isolate the M scaler from the DAC then use a battery power bank (so long as you site it correctly) - then the loop impedance becomes very high, and then the BNC cables no longer have an impact on SQ as no current can now flow as there is no longer a loop. Is this something I do in practice - no - the difference isn't large enough to bother with charging batteries. At the end of the day I just want to simply enjoy my music more, and I don't need to go to silly lengths to do that.
I have used a reply to a post by Rob Watts from November 2019 to explain that the information about the 2 GHz noise originally came from the designer of the DAVE / M-Scaler. The post clearly states that the peak problem area for the noise is at 2 GHz.
This noise is at a low level and escapes the internal counter-measures in the M-Scaler that successfully block noise at lower frequencies. It is the presence of this noise on the shield of the coax cable, not on the central conductor, that is the point of concern. The screen is connected to ground at both the M-Scaler and DAVE ends. The grounds of the M-Scaler and DAVE are also joined through the mains Earth wiring. There is therefore a loop in which this radio frequency noise voltage can cause current to flow. This is, I believe, a well established, textbook mechanism for noise transmission in coax cables
As I understand it, Rob believes that noise floor modulation in the DAVE caused by this ground - born noise is responsible for a small degradation in sound quality, because it impinges upon The D to A conversion process.
For the home user of DAVE / M-Scaler there are a number of possible responses to learning of all this, including:
1. Making no changes to one's audio system around the DAVE / M-Scaler. They are great sounding components anyway. This is what I did for over 3 years. I just made sure I was using cables professionally constructed of well-screened 75 ohm coax cables terminated with 75 ohm characteristic impedance BNC connectors to link the two devices.
2. Breaking the ground loop by running the M-Scaler from a battery or Power-bank to gain a small uplift in sound quality. This approach requires one to make provision for either manually or automatically charging the battery in the listening room.
3. Breaking the ground loop by using twin optical links, between the M-Scaler and DAVE. This involves adding active circuitry between the components and in one becoming an early - adopter of technology (384KHz optical transmission of digital audio signals) that is not standard in domestic audio equipment.
4. Adding ferrite cable cores to the coax cables to increase the impedance of the shield. This has the disadvantage of not fully breaking the ground loop. It does, however attenuate the ground currents. If these can be sufficiently attenuated, one hopes that they will fall below a threshold value at which no further improvement in sound quality can be heard.
I admire the ingenuity of those who have implemented methods 2 and 3. They should be getting the full uplift in sound quality that is possible. Personally, I have not been inclined to go down these routes.
In the end, I succumbed to curiosity and tried method 4, adding ferrite cable cores, as outlined in my earlier post. It does unfortunately require one to make one's own cables in order to get the ferrites in position between the BNC connectors. This method only uses passive means to attenuate the noise, and there is no added circuitry or equipment in the listening room. The new cables just disappear back in behind the audio rack and one is left to relax and enjoy listening to music.