Ii will hold tight with the MK2 4400s for your firmware update that lowers noise floor another 20 dB. I do like Jeff’s idea of film cap trials modding XLR interconnect.
On the DS’s filter has multiple stages, removing the C704 stops the last stage which extends the top end.
Here is a plot with all of the standard cap values from 1nF to 33nF and one with no cap at all:
The black line is the designed filter. The shallow red is with no caps.
4 Likes
?
The Mk II FPGA 173 and 179 already have the firmware in question. They are much quieter than the DS.
2 Likes
I hear this lower noise floor on the MK2 as a freedom of graininess in the music especially in the highs. The purity of the highs is incredible on the MK2. Listen to track 7 on this DSD 256 of Rembrandt Trio “Goodby If Only”. I have never heard the music so grain free and the bells ring so pure ever in my system.
2 Likes
I think this has been discussed elsewhere, but I just improved the sound of my DSD mkI by turning it off and unplugging it when I was testing different power cords. I ended up liking the original power cord, but now the DSD sounds better than it previous had after turning it off and back on again.
Thanks Ted, so, although I don’t see the dB scale, this means, except for the very left line (which is 33nF?) , the other values have no effect in the audible range compared to no cap and there’s no noticeable difference to expect from changing values, correct?
But why then does it sound so wrong with LL1948 and no cap, as some stated?
Ted,
I guess I was referring to post 207 in the comparing MK1 to MK2 thread where you stated.
“But to the ears there’s no contest between the DS and the Mk II on noise floor - it’s not subtle. A modded DS can’t get as quiet as the Mk II, there are just too many analog bottlenecks let alone the DS FPGA’s digital DSP limits.
[Edit: To be clear a better output transformer should lower the THD some, but I’ll also be lowering it in the next couple of software releases.]”
I assumed by your comments the next couple of software releases lowering THD were releases beyond 173 or 179. It seems i confused that with your comments earlier in the post, where you stated MK2 was already 20 dB quieter. I must read more closely next time and thanks for clarifying.
What is your software plan that lowers overall THD more? Is it more reduced switching operations in the FPGAs or something else? I assumed lowering THD more would also lower noise floor and inferred 20dB. Is there conditions where THD and noise floor are not interrelated?
I’m starting to think we’re doing Amir an injustice.
1 Like
THD and noise aren’t very coordinated. THD is harmonic distortion, e.g. new tones whose frequencies are a small integer multiple of old frequencies. So a little second harmonic distortion of a -10dBFS 1KHz sine, might be a -60dBFS 2kHz sine showing up too.
Noise can be correlated with the signal, but often is not. It’s just a small signal across all frequencies. Some sources of noise will add frequencies uniformly, white noise. Some add about the same amount of noise per octave, pink noise. Some add more around a given frequency, say 3kHz, but are spread from say, 1kHz to 5kHz with the peak somewhere in the middle.
I’ll lower THD by refining the sigma delta modulator. When I was tuning the new SDM I was fighting the noisy attenuators (which we’ve removed from the design.) I had to make some compromises in THD to lower the attenuator’s noise contribution. Without them I can readjust the SDM for lower THD.
7 Likes
Thanks that is best explanation i have read yet on noise and contributions. It also explains well why I hear differences in tweaks. Cannot wait to hear your software without the compromises for mK2 attenuators. Now only if I can retain what you said. I do now recall the attenuation remarks a few months back in the MK2 journey. Thanks for your patience in reteaching this information.
You asked if it was possible to tweak the top end frequency response with various capacitances, I showed how the top end would change with the original transformers for various capacitances. I was just pointing out that some smaller caps will definitely cause some peaking issues up higher, which may not be kind to your system and that we aren’t really talking about huge roll offs in most systems with most cap values. All of the peaking issues are very transformer dependent, and I don’t know what they’d look like with the Lundahls. Remember these cap values also are affected by your cable capacitance which directly adds to the cap value. (The dB scale is indicated by “(dB)” just after the signal names on the bottom, in principle different signals could have different scales.)
3 Likes
This is why we did transformer sweeps with the 15nF and without. It clearly shows what happens to the signal with both configurations.
Yes exactly…when comparing the blue line on the 2nd with the 4th LL1948 chart, correct?
The question is, could there be a cap value that provides noticeably less HF loss between 5 and 20k and anyway avoids ringing? Do you think it would make sense to try out and if so, which values could be suitable? You mentioned 12 nF…
Is there any recommendation for an alternative cap for C704?
I feel an „urge“ to go back to the Lundahls and if I do so I would just change the C704 if there is any benefit from a slightly different value (currently 15 nF is installed).
And regarding the effect of interconnect capacitance I am a little bit confused (which is my normal state of mind when leaving my „knowledge comfort zone“). But if I would try to add only 1 nF to the analog output path with a standard interconnect with let‘s say 20 pF/ft I would need an IC-length of 50 ft, right?
So I guess the influence of interconnects is nearly negligible in this case, when we talk about 15 nF down to 12 nF (or lower) for C704 for example. But, hey, I may be wrong. Just learning.
Most interconnects aren’t a problem, but some companies use geometries which have high capacitance like parallel flat conductors.
Yes, using the length of an interconnect to vary capacitance isn’t really practical.
5 Likes
Thanks Ted. Got it right.
I tested with different caps by just using a short XLR interconnect (1’) I just soldered the cap from pin 2 - 3. inside of the XLR connector. I know this is not a perfect test however it sure beats soldering caps inside the unit.
2 Likes
You will never get the distortion down lower then what the transformer performance provides.
The sweeps we provided shows that the Edcor 4400 in series secondary mode has a distortion of near 0.1% at 20hz and fairly significant phase shift and attenuation at 20k (-.6dB) (even without the 15nF capacitor) when wired in 1:2 mode.
Distortion & FR:
Phase:
1 Like
A minor clarification: unlike distortion, phase shift and attenuation are easily correctable (Not that I’m doing it or planning to do it.)
2 Likes
Thanks Ted, That’s why I only mentioned distortion in the opening statement.
Can you change FR and phase errors in the FPGA? If so that would be very handy.