Verfasste Forenbeiträge
-
Beiträge
-
Hmm, I think the region where it’s worse is more like 25 to 38MHz. I can’t edit what I wrote earlier, sorry about that.
I’m sat here not doing much, so I thought I’d try overlaying the spectra of the asymmetric and symmetric ferrite configurations (crude effort with Photoshop):
The „yellow“ is the asymmetric ferrite configuration. Generally worse than the symmetric configuration, but the region from about 20 to 30MHz is worse with the symmetric ferrites.
Interesting, huh?
Fantastic measurements! Thanks very much for your patience.
I do see a rise in noise in the 4-8MHz region (guessing at frequencies), but that appears to be the only problem with the „asymmetric“ ferrite configuration. This region is suppressed by the symmetric ferrite configuration so that it is better than the non-ferrite configuration. So there is a small problem with asymmetry here, but clearly swamped by the rest of the spectrum.
Overall the symmetric ferrite configuration looks great. I wonder how a cable covered by the cheap Topnisus ferrites compares (about 40 per metre of cable I think). I also wonder about a broader range of frequencies, upto 2GHz because sound quality problems have been reported with common-mode noise far beyond 100MHz. But in the end, the measurements already shown in this discussion demonstrate that sound quality benefits are easy to achieve with USB cable connections 🙂
I also wonder how much quieter the spectrum can be made. It still seems to me to be very noisy, all of this unwanted common-mode noise.
Changing subject slightly, this article:
https://audiowise-canada.myshopify.com/blogs/news/sound-quality-rf-energy-best-inputs
shows a Hugo 2 on battery power with various source and cabling configurations. Here we can see a low-powered PC with a USB connection can perform pretty well in terms of RF energy (as measured in the output of the DAC).
We can also see that there are differences between the two configurations that both deliver data to the DAC using a TOSLINK connection („direct from NUC“ and „from XMOS converter“). You can click the entries in the table to see a spectrum of noise from 10MHz to 4GHz.
I have been away for a while and then I didn’t turn on this computer, so that’s why it’s taken me a long time to reply.
Which TOSLINK cable do you use?
In your opinion, does the quality of the TOSLINK cable play a role in the sound?
I use a 6m KabelDirekt TOSLINK cable, which works at 192khz sample rate.
I have no personal experience of sound quality variation due to the TOSLINK cable.
That is absolutely true and the costs are low compared to „audiophile“ cables. However, with Ethernet it has been shown that the interference caused by mode conversion across the entire cable can be better combated with a very high impedance just before the receiver. I also explained this in the article on the choke cable. However, I am not sure whether this is transferable to USB.
I think you ended-up covering the entire length of all the major versions of the throttle cables, e.g. version 1 is a toroid with 8 turns and 5 clip-on ferrites filling the remaining length of the cable. V2 has solely clip-on ferrites along the entire length, each with multiple turns. In V3, 3 toroids with varying turn counts with clip-on ferrites distributed along the full length of the cable.
„Reflections“ cause „tuning“, which is what the video refers to. The reflections aren’t perfect, in general. Ideally a cable and its connectors all have uniform impedance, meaning there’s no reflections and so the cable is not tuned (not resonating).
Obviously any ferrites on a cable will make some portion of the cable have a different common mode impedance, when compared to the rest of the cable and also the connectors. So reflections will occur.
In the end, I trust the RF engineer when he says that mismatched impedances at the end of the cable are causing problems with increased common mode noise. Naturally if there’s minimal noise at the tuned frequency to amplify, then the result is innocuous…
You have the equipment to measure these effects, so it’s probably more productive to investigate directly. I can only go on listening tests 🙁
What theory do you have about how the interference can affect the audio signal?
On the one hand, these can certainly get into the analog audio signal in the audio frequency range via demodulations.
On the other hand, they can also cause interference during conversion in the DAC. John Swenson blames the noise in the mass for this.
RF noise, via intermodulation, ends up at audio frequencies as noise. There’s fundamentally no difference between DACs and other hi-fi components that have analogue signals inside them. This is why I have ferrites on the AC power cable for my amplifier.
This video discusses this topic and others:
https://youtu.be/wvfc4UYGDi0?si=nC21HPaDZkaDQx1S
Which interface is used for high-quality audio streaming?
TOSLINK provides optical isolation, but noise and jitter may even increase.
In my hi-fi I use a wi-fi connected streamer (Node) with TOSLINK. Luckily I have 2 TOSLINK inputs.
Modern DACs generally have „perfect“ jitter rejection. Chord DACs are basically as good as it gets at rejecting jitter, so it’s not something I worry about. Because I use a Hugo M Scaler, the noise problem I have is mainly caused by the scaler. Unfortunately the fact I don’t use optical between the scaler and my DAC means I have to use lots of „treatment“ to minimise RF noise.
The scaler has the two TOSLINK inputs and sends upsampled data over a pair of BNC cables to my DAC. This pair of cables is the only input I use on the DAC.
Jawed
I should have mentioned that I read the entire Ethernet Throttle thread and the blog article that introduced the major milestone. And I’ve read many of the articles on measurements of Ethernet equipment. Just trying to collect data 🙂
If the goal is to attenuate common-mode interference over the broadest possible band,
many turns could be disadvantageous.Yes. With very low cost clip-on ferrites, there’s no reason not to cover the entire length of a cable. This also solves the problems caused by mismatched impedances at the ends of the cable. I have 300 or more ferrites deployed across 6 power cables and two BNC cables.
The other is the impedance for the common mode , which is generated by ferrites or chokes, as in the video. In my opinion, the increase in interference in the video is caused by interference being reflected from the monitor. It would be interesting to take a measurement by attaching ferrites to both ends directly in front of the monitor, i.e. between the monitor and the first ferrites. In my opinion, you will then see the same increase here, as the interference is now reflected here. But that is all just speculation.
I’m unsure what you’re saying here. At the end of the video he attached a ferrite at the end of the cable closest to the monitor. Are you referring to the ferrites that are built-in to the cable and the placement of the clip-on ferrite with respect to the position of the built-in ferrite?
Anyhow, I believe you have the equipment to perform detailed measurements and listening tests with unbalanced versus balanced impedances along the length of a cable. A key reason for my efforts to post on this forum is that you have the measurement equipment, which is rare.
The solutions are really interesting and the owner seems to know what he’s doing. Thank you for the link. However, I wonder why you convert USB to SPDIF and then to optical and not just use Toslink? But here you have the timing problem again. It’s also interesting that he prefers the Mojo2 to the Dave.
TOSLINK limits sample rates to 192KHz, which is perfectly fine for streaming. Many people, including him, use software like PGGB or HQ Player to do upsampling to 768KHz (or even higher). With Chord DACs the highest sample rate accepted is 768KHz. Since this won’t go over TOSLINK, the alternative is the convoluted path. With Chord DACs dual-SPDIF enables 768KHz – I’m not aware of any other DACs that support this data rate over dual-BNCs (certainly not dCS, the company that created the dual-BNC SPDIF standard), so this convoluted path is specific to Chord DACs.
I’m mildly sceptical about his preference for Mojo 2 versus DAVE simply because the final noise shaper and PWM stage have vastly better transparency in DAVE. But there are some mathematical refinements that Rob Watts introduced with Mojo 2 that mean its volume control is better, in theory…
That’s a really cool solution – I need a CNC milling machine
He paid a company to construct the case for him, using designs that he submitted 🙂
Regarding the galvanic isolation with USB, I still think the artistic fidelity AFI + USB solution from Ralf Koschnike is top: https://kopfhoererboutique.com/products/artistic-fidelity-afi-usb-modul?variant=14147847946285
Two optical cables are used here. As far as I understand, the signal is transported on one cable and the clock on the other, so there are no timing problems.
I use this USB to TOSLINK convertor to connect a mini-PC to my hi-fi for things like YouTube:
https://www.amazon.co.uk/dp/B0B2DBGKL3
I suspect that different frequency ranges of RF noise have specific effects on sound quality. I wrote a fairly long list of the different sound quality problems caused by RF noise here:
https://www.head-fi.org/threads/chord-electronics-dave.766517/page-1113#post-16160284
Jawed
I can’t find a way to quote the messages I want to reply to, so I will simply post generally.
The USB cables I used for my experiments with multiple turns where „cheap generic types“ and were flexible enough to be turned through a toroid ferrite without much trouble.
I think the „snags“ and other stress caused to the cable are possible reasons for the sound quality problems I heard. I also suspect that the impedance mismatch problem could have been relevant. I admit I did not go back to the multi-turn ferrite toroid to do more experiments after the initial bad experiences.
For what it’s worth I also had bad experiences with multiple turns through a single clip-on ferrite. Or with several of them arranged together.
So after many bad experiences with multiple turns through ferrites I gave up. Perhaps I should have used yet more ferrites?…
In the video that shows the problems caused by bad placement of ferrites, the key idea is that the cable itself acts as a tuned, resonating, carrier of common mode noise. The cable, ignoring the devices at either end, has a constant impedance along its length. So adding ferrites at only one end disturbs the cable electrically, producing a tuning that can make some noise worse.
The source of noise, according to the video, isn’t well documented. Was it the monitor causing the problem, or the adaptor? For the purposes of the demonstration it doesn’t matter how the common mode noise appeared on the cable, merely that with bad placement of ferrites, some of the noise became dramatically worse. Note that the measurement probe placed near the centre of the cable is measuring common mode noise, not differential noise. This noise is carried over the full length of the cable.
I hope this explains things for you Max.
In the end, 40 ferrites on a 1m cable will produce something like 4000 – 8000 ohms of impedance across a broad range of RF frequencies.
I wish I knew which frequencies are the most problematic for sound quality. From Eric’s measurements in the thread about the Ethernet Throttle, it’s clear that 10-200MHz is important. And it’s also clear that switch mode power supply noise which is typically 50 to 500KHz, is also important. Unfortunately there’s plenty of evidence that noise between 200MHz and 2GHz is also important.
Here is someone who has been doing measurements and developing RF treatment products:
https://audiowise-canada.myshopify.com/blogs/news/good-measurements-finally
There are many very interesting articles on his website related to RF measurements and treatments. The original motivation for his work comes from the connection of the Chord Hugo M Scaler to Chord DACs using dual BNC cables.
This clever chap:
https://www.head-fi.org/threads/my-choral-housed-chord-mscaler.964931/
put the Hugo M Scaler into a new case of his own design and created a fibre optic replacement connection, so that instead of using BNC cables x2 to the Chord DAVE, he uses fibre optic cables x2. His solution for fibre optics is different and superior to that developed by Audiowise.
I was directed to this site by a friend recently. I use Wi-fi for my streaming and home networking needs, so ethernet connectivity is not something I have any interest in tangling with.
I like Eric’s approach with his measurement techniques. I vaguely contemplate finding a way to doing my own measurements, but in the end it’s a low priority.
Many years ago I spent a lot of time investigating the use of clip-on ferrites with USB cables.
My original recommendations on this topic appear here:
https://www.head-fi.org/threads/does-chord-dave-system-need-a-dcc.809287/#post-13799263
which was a result of many experiments.
Along the way I learnt that it is possible to create sound quality problems:
- Toroid ferrites with multiple windings
- Some, but not enough ferrites: 8-10 ferrites
In both cases these bring some improvements but also cause some peculiar „etched“, „false detail“ effects.
Toroids with multiple windings cause a „high Q“ resonance where the strongest filtering is applied, which corresponds with „false detail“ and „fizzy stereo image“.
Clip-on ferrites, due to their air gap, introduce an unwanted resonance too. The use of more ferrites eventually seems to tame this resonance though. I ended-up recommending at least 20 ferrites on a USB cable, luckily the Topnisus ferrites are extremely cheap.
After much research and experimentation I found that it’s simplest to cover the full length of a USB cable in ferrites, all of the same type, and all close together.
When treating a cable with ferrites, it’s worth noting that the impedance at each end is relevant to the overall effect of the filtering. When this fact is ignored, some frequency ranges will be made worse by the use of ferrites:
https://youtu.be/jYR49UghNj4?si=RU3O39K47drSHEPK
Naturally, I have read Jim Brown’s (K9YC) guide to ferrites. And many many more documents and videos from that community.
The discussion of ferrites with Chord DACs goes back to here:
It transpires that discussion of ferrites on the Head-fi forum is much much older than these postings I’ve linked. Common-mode RF noise has been a well-kept secret for an annoyingly long time, but it’s getting good exposure now.
I treat my BNC cables and power cables, both alternating and direct current, with complete coverage of clip-on ferrites from end to end.
I hope you don’t mind me using English.
Jawed
