Ok. I was curious since the normal approach is to drop a via to ground plane from the GND pin. Using traces can lead to problems at least according to this source (see Figure 6):
https://www.analog.com/media/en/training-seminars/tutorials/MT-101.pdf
To make it complete you could add the MCKs used in I2SoverUSB as they are used to generate the input signals to Marcel's digital input circuitry. I2SoverUSB MCK can also be external which is another option. To make it more apples to apples comparison similar clock should be used for both Marcel's dac and FIFO.
It seems like a pretty forlorn hope Acko, most of the 'tests' of Marcel's RTZ decoder seem to be subjective and with expectation bias - not a good basis for informed decision making.
To illustrate the point, I am genuinely curious about Andrea's DAC but it would be a significant investment (especially in comparison with the RTZ decoder) and the opinion-based subjectivity I'm seeing on this thread is actually giving me significant reservations about making such an investment.
nautibouy, I think it sounds a lot like when you describe your building adventures... sure you are not just self-suggesting in order to avoid the costs... you know you want one ;-)
//
//
Hi @TNT - My build descriptions are always subjective and not presented as otherwise, and certainly not presented as definitive. The problem I have is that I have experience of just how good the RTZ is, not quite as good as the ValveDAC in my (subjective) opinion but certainly in the top handful of the broad range of DACs I've heard and incredible value for money. I also have plenty of experience of being tricked by my own expectations - I can't get rid of that but I do have awareness of it - and all of my experience is that 'improvements', beyond the basics of a good quality buiild, are small and often subtle, and are frequently just differences rather than better or worse. So, knowing that the RTZ is so much more capable than just having a rather patronising "...certain 'charm'" my reservations are about the actual bang for buck Andrea's DAC would actually yield and I don't feel that I have a reliable opinon/experience to make an informed decision about such a purchase.
Last edited:
Perhaps at some point such information can be folded back into the opening post.
When Mark's experiments have converged, I can put a link to the latest status in the opening post, under Variants.
As a data point of possible interest, here's a pic of the output array for a Marantz high end SACD player:
One may notice how the bypassing is done, how each output switch IC outputs a mix of inverting and non-inverting signals. In some ways it has some similarity to the scheme Marcel uses.
The output of the stage shown above goes to a number of op amp-ish and other circuitry for filtering, etc. The next stage looks like:
As can be seen, its all discrete. Also, there happens to be some differential summing towards the end on the right.
Was it designed by FFT measurements? IMHO probably not.
One may notice how the bypassing is done, how each output switch IC outputs a mix of inverting and non-inverting signals. In some ways it has some similarity to the scheme Marcel uses.
The output of the stage shown above goes to a number of op amp-ish and other circuitry for filtering, etc. The next stage looks like:
As can be seen, its all discrete. Also, there happens to be some differential summing towards the end on the right.
Was it designed by FFT measurements? IMHO probably not.
Last edited:
You are just guessing. IMHO that circuit was both extensively simulated and measured.
BTW are you implying that Marcel designed the output stage for FFT measurements? IMHO it measures well because it was designed well.
It's an interesting circuit diagram, also their supply decoupling and the fact that their (non RTZ ?) Firdac uses seven
resistors per channel.
When this is from the SA-1, the introduction was in 2000, so the design took place in the ninetieth, 25 years or more ago.
Hans
Last edited:
BTW, got Marcel's dac working with DSD256 this morning. Took awhile to solder on more u.fl connectors for scope measurements. Got the phasing between MCLK and the DSD DATA signals correct for MCLK to work as BCLK. Sounds good. Stilling warming up though. Probably more to say after I get some other people over to listen.
Then why did they use ferrite beads? According to you they should be a no-no for good sound 😉
They found some they like for their circuit.
EDIT: Schematics were posted for people to look at, is all. Don't want to get into a critique type of discussion.
EDIT: Schematics were posted for people to look at, is all. Don't want to get into a critique type of discussion.
Interesting that they chose a Sallen and Key filter, rather than an MFB filter or something else that more or less behaves as a virtual ground for low and high frequencies.
Notice dac resistors called out in BOM are Susumu RR series, which is not recommended for new designs. Reason I bring this topic up is because I think I hear something a lot like a little bit of resistor sound. Has that some of that particular blur to it.
Anyway here is the current selection guide from Susumu:
As can be seen they recommend RS series for hi-fi audio.
When I was trying different resistors in Andrea's DSD dac, one of the types I tried was Susumu RG. IMHO they were the worst sounding resistors I tried. Don't know what else could account for the sound except Excess Noise, most of which is probably current noise. Best resistors were the ones Andrea used, IIRC they were Yageo. We never tried SMD foil resistors because of the expense. However, since this dac uses fewer resistors than Andrea's dac maybe it would be less costly to try enough of them to work for SE operation. About $8/each in moderate quantities. Makes the Rubycon caps look cheap.
Anyway here is the current selection guide from Susumu:
As can be seen they recommend RS series for hi-fi audio.
When I was trying different resistors in Andrea's DSD dac, one of the types I tried was Susumu RG. IMHO they were the worst sounding resistors I tried. Don't know what else could account for the sound except Excess Noise, most of which is probably current noise. Best resistors were the ones Andrea used, IIRC they were Yageo. We never tried SMD foil resistors because of the expense. However, since this dac uses fewer resistors than Andrea's dac maybe it would be less costly to try enough of them to work for SE operation. About $8/each in moderate quantities. Makes the Rubycon caps look cheap.
Today I've upgraded my music server with the latest version of Audiolinux and HQPlayer 5 today (and I'm using HQP embedded instead of HQP desktop) and it does seem that the CPU loading is a bit reduced so I will see how it does with DSD512 with my preferred HQP configuration.
I've just received back my reflashed PCM2DSD and did a quick comparison and, good though the PCM2DSD is, it falls a little short of HQP resampling IMO.
I've just received back my reflashed PCM2DSD and did a quick comparison and, good though the PCM2DSD is, it falls a little short of HQP resampling IMO.