I use poly-sync-ext2, to PCM 176/192, depending on the source original rate (using the auto family rate option) and NS9 as dither. Best configuration for me.
I chatted with Jussi about this. He was not aware that the DirectStream DAC supported up to 24/352.8. He told me that he would use HQPlayer to upsample to the highest PCM rate the DAC supports. That means you should use your favorite filter to upsample to 24/352.8 and use the NS5 noise shaper.
I’ve seen several DS/DSJ owners say they prefer 176/192 to 352, saying the lower rate has more weight and better bass. I’ve tried both but don’t hear a dramatic difference either way…
It seems a bit illogical. Let’s have HQPlayer upsample to a higher rate PCM. Then let’s have the DAC upsample to the highest rate PCM. Then let the DAC upsample 20x DSD.
Either you like the DAC’s PCM upsampling or you don’t. Mixing the two, if you don’t have to, makes little sense.
If it made sense, it would be a different hobby
I’ve never used HQP and I am in the same camp as speed-racer: you own a top-of-the-line DAC - send it the bitstream in its native format and let the DAC do its thing. However, HQP claims to be the best SRC tool in the business, so maybe it is better at this than the DAC itself? Or, it seems like it has a myriad of options to allow you to dial-in the sound to your liking during the SRC process - something the DAC doesn’t have. Where does HQP provide value? I thought I saw some posts from a couple years ago where Ted liked HQP but I’m wondering if this has changed given all the PCM upsampling improvements in Snowmass?
I actually prefer HQPlayer’s PCM upsampling to Ted’s. Not by much as Ted’s is excellent. I am using the poly-sinc-xtr-lp filter with the NS5 noise shaper and upsampling all PCM to 24/352.8.
Having said that, I would be quite content using Ted’s upsampling. The reason I bought HQPlayer Embedded was to use it with Roon. For some reason, using HQPlayer and an ultraRendu acting as an NAA sounds better than using RAAT to an ultraRendu acting as a Roon endpoint.
My understanding is that DSD DAC does not upsample PCM to PCM. It only upsamples PCM to DSD. But I might be wrong.
Anyway, I also understand that the main advantage of upsampling is to move the noise to subsonic and ultrasonic levels, and therefore to alleviate the “brick wall effect” of the filter that limits the frequency to about 22khz.
And, for this purpose, 176khz and/or 192khz seems to do the trick. I hear no difference when upsampling 176 to 352. But I do feel differences when mixing up rate families (44.1khz and 48khz). So I just kept my HQPlayer limited to 192khz and respecting the rate family.
Having said that, right now I am preference the SQ of the Bridge II, over HQPlayer directly to USB, or HQPlayer to microRendu. In these latest configurations, the sound appears darker.
IT is interesting you prefer HQPlayer’s PCM upsampling to that in the DS, but with the ability to choose your preferred filters, noise shaping, dithering, etc. it certainly seems possible to find an upsampling solution which matches one’s taste.
Did it take a great deal of effort to go through the many options/combinations to find something you liked better, or was it better right away?
HQPlayer has so many combinations, not sure how to do the math, but it must be many hundreds of potential combinations.
The settings speed-racer mentioned above are a good place to start - they are close to what the HQPLayer’s author recommends and close to what I’d recommend.
I have an I2S connection to the DS, the card is limited to 24/192 PCM. With Redcloud I liked having HQPe upsample to the max 192. With Snowmass v1 I prefer HQPe to be in the path (taking Roon out of the audio path) but in bitperfect NOS.
I’ve tried numerous times but have never been able to detect any difference with the various noise shaper dither options but readily hear the difference in filters.
There really aren’t that many options once you narrow down to a couple of preferred filters. I like the newer poly-sinc-ext2, sinc-M and closed-form-M when I do chose to use one.
Like I said, @Elk, Ted’s PCM upsampling is excellent and I would be quite happy if it were all I could use. But, HQPlayer gave me more filter and noise shaper/dither options to try…so I did.
The upsampling filter I ended up liking the best, poly-sinc-xtr-lp, makes sense. You see, I really like the sound the Chord Dave filter produces and the poly-sinc-xtr filters are similar to the Dave filter. I also like the poly-sinc-short-lp filter. I am pretty sure that Jussi like the poly-sinc-short-mp for rock and pop music and poly-sinc-short-lp for Classical. Regardless, he likes upsampling with his filters as high the DAC input will accept.
Please note that we are not talking about huge differences here. The difference between Ted’s filter and my favorite HQPlayer filter is subtle and it took a lot of hours of listening to figure out which one was most pleasing for long listening sessions.
This makes perfect sense. Thanks!
Does anything out there do ethernet -> I2S in one box? Seems like having all these boxes strung along to finally input into the dac just over-complicates things.
Well, the Bridge II does that. Now a box with a great PS that accepted an optical fiber input to I2S output I would kill for.
Here’s another take on optical…
Unfortunately, it only has USB output. I wish they’d make the same thing with I2S built in.
Yes, but I think the biggest issue that everybody is facing, is the noise that Ethernet introduces into a HiFi system. And when you go for streaming, there is an Ethernet network somewhere in the chain.
To what I have read so far is, that the usual boards and chipsets available on the market are not solving this issue very well. And to what I read here in the forum is, that Bridge II is not the best input source on the DirectStream DAC’s. Well, Bridge II is a 3rd pary product and therefore not 100% under control of PS Audio. They might solve this issue in a future version (Bridge III?) or with their Octave product. Who knows…
So I2S might look superior to USB, but on what level, if the used Ethernet chipset in the chain is of bad quality?
What the guys at Sonore did understand very early, is that streaming is dependend on Ethernet. And if you want to solve the noise issue, you have to work very hard to reduce noise on your Ethernet solution. That’s what all the Rendu’s did very well from the beginning. Sonore has a deep level of understanding of building their audio optimized Ethernet solutions. The result is a very clean USB signal. If that is not enough, you can add the Sonore ultraDigital, a USB to I2S converter.
Now with the new opticalRendu, Sonore went to the extreme:
The opticalRendu utilizes an SFP fiber optic transceiver at its input to provide 100% galvanic isolation from the network and USB-audio output. The opticalRendu has linear power regulation, CPU circuit design with femto clocking, USB circuit design with femto clocking, and a network circuit design with femto clocking.
And on the Ethernet design:
Quote John Swenson:
In the opticalRendu I re-designed the Ethernet circuit. I’m not at liberty to discuss the details of this circuit. What I can say is that this circuit uses a ton of very high quality voltage regulation and very low phase noise clocking. The result is that the processor is fed a signal that is cleaner than any external Ethernet connection you can buy today.
This circuit is the primary new addition to the opticalRendu and is where the additional oscillator is used. Using an optical network connection provides a remarkable synergy with this circuit allowing it to produce such a clean output to the CP.
I really think, that the know how in the area of networking, which Sonore has built-up over all these years, is very difficult to top. In my opinion, the opticalRendu is by far the best (digital-to-digital) streaming client available on the market today, the best conversion from Ethernet to USB Audio. I also suggest to use the opticalRendu without any further optimization behind the USB output (which means: no additional ultraDigital), as the USB signal is the cleanest you can get.
I would agree that electrical noise is now the biggest issue affecting SQ. In the case of the DS DACs, you can verify bit perfect transmission (with PS Audio supplies filed). Since they do not reference prior clocks, they are virtually immune to external clock phase noise. That leaves electrical noise as the primary issue.
There are several solutions available that place an optical break in the Ethernet feed. If placed properly, this means the final “digital conversion” and the transmission of that data to the DAC are the critical components. While I think the Bridge II fed such a signal, sounds very good, I would guess that performing the final digital conversion inside the DAC with the same power supply is not the optimum solution noise-wise (although one should not discount the benefit of a short, direct I2S connection from the Bridge to the DAC).
I would also agree that the OpticalRendu seems to be an intriguing solution as they have clearly tried to minimize noise post optical conversion.
Unfortunately, the clean USB output must be connected with a USB cable. Even if we assume that what happens after the DAC connection is equivalent, an HDMI cable is technically superior to USB with respect to noise resistance. Couple that with the near universal personal preference (at least on this forum) of the I2S connection, and I remain disappointed that Sonore doesn’t offer an I2S connection option. The UltraDigital unit adds an additional power supply and processing to the chain which is somewhat antithetical to our purpose.
All this said, I will likely compare an opticalRendu (with an expensive USB cable) to an Optical -> Matrix, to Optical -> Bridge. And as long as I’m at it an Optical -> Aries G2.