I adjusted the default gain a little in 1.19, just set the volume a little lower.
If it clips internally, does volume really help? Taking the volume is last in the chain.
Could you describe the different "gainstages" and their order please?
//
Internally the dams have extra guard bits in all digital processing, and the digital volume control is last, so adjusting the digital volume control should do fine....
I don't know what filters they were using, assuming stock, they were probably not all running a full output level, but a thing to check....
One of the new filters required me to lower the gain in all the filters so they match each other, that's why I then added gain, actually added in the volume control....
I have problems with 1.19 using stock filters at volume 0db since I control volume externally. But the clicks are only percaptable at low volume passages of the music, so probably not clipping.
They sound like sample spikes. Something I heard when converters were not in proper sync in a studio environment. But lower volume and not as often and regular.
Last edited:
The filters have not been designed by me, as I explained they are from Soekris and Paul that I liked. I calculated from the 44 and 48 the rest of sampling frequencies and I adjust the gain to the maximum. Because Paul designed the rest of frequencies attenuating all to 20 Khz. I with my bat ear wanted them to come higher in 96 Khz. frequency etc.
I adjusted all the filters with many decimals so that they did not saturate, but they were as much as possible of gain in order to take advantage of the dynamic margin and that the 4 had identical sound volume. For new versions with gain above 0 they can saturate. To me the gain above 0 seems an error precisely because of that. It would be necessary to retouch the gain in the TXT so that it had the maximum signal at the maximum output volume. I was even undecided as to whether it was convenient that at maximum volume it was convenient for them to be at -3 dB because oversampling have peaks above digital 0 dB. The most critical case is a square wave and I verified with oscilloscope that the ringing of this square wave manufactured with peak-to-peak values did not saturate the ringing produced by oversampling.
There was a failure in the F6 to 192000. Found by Jimmy2002 in September 2016. I have not bothered to change it in my DAC or in publishing it.
The error is at line 6473.
dam1021,192000,4,6,33,1.999999995
It should be
dam1021,192000,2,6,33,1.999999995
This line is corrected in post 1379. And obviously in the skr file from post 1342, as they appear identical.
I did not remember if I had corrected it. Jimmy discovered it and told me in private.
Correction: The latest version is adjusted so that the file created by MKROM has the maximum values of impulsive response. Displaying one by one the binary result in the skr file so that it does not saturate that impulsive response. So take advantage of the maximum precision in the coefficients. And for the DAC not to saturate, it is recommended to play the music at -2 or -3 dB of volume. Especially the very compressed music (all the today music) can saturate and sound very hard.
Correction: The latest version is adjusted so that the file created by MKROM has the maximum values of impulsive response. Displaying one by one the binary result in the skr file so that it does not saturate that impulsive response. So take advantage of the maximum precision in the coefficients. And for the DAC not to saturate, it is recommended to play the music at -2 or -3 dB of volume. Especially the very compressed music (all the today music) can saturate and sound very hard.
Just read everything from page 50, together with the posts at the beginning I read previously, and I now realize how much Paul has contributed...
Sad that there's no way to get in touch with him. It may be close to trivial for him to make new 4k filters based on the same principles as the party pak. Read a bit about how Chord Dave has 164,000 taps... Wonder why Soren didn't use better FPGA chips in his recent products... We have LX16, Dave is LX75, top of the line is LX150. The number of logic cells is proportional to the product number.
Very impressive that Soren managed to upgrade the product family to 4k taps though...
Sad that there's no way to get in touch with him. It may be close to trivial for him to make new 4k filters based on the same principles as the party pak. Read a bit about how Chord Dave has 164,000 taps... Wonder why Soren didn't use better FPGA chips in his recent products... We have LX16, Dave is LX75, top of the line is LX150. The number of logic cells is proportional to the product number.
Very impressive that Soren managed to upgrade the product family to 4k taps though...
Last edited:
That’s somewhat true yes. But it’s complicated
If we don’t have the data there isn’t much point to begin with I think... my listening skills aren’t that good anyway.
Although it might be interesting to find out why the balance on Soren 4K filters are off by so much. Even the 1.19 has a lot more bass, let alone the distorting 1.20. I doubt Soren messed with the freq response
