16x Digital interpolation filter - drive PCM56, PCM58, AD1865 and so on up to 768 kHz

Thanks Abrax, I guess thats the point of DIY, learn about what you like.
I'm using a JLSound usb card, Ians PCM to i2S board, a couple of TDA1541A in balanced out mode.
The dac sounds pretty good with a simple resistor i/v triode gain stage, but when compared to sabre, its missing some top end. The actual i/v Im going to use is Thorstems e180F. He uses a simple PP 0.001uF cathode bypass cap and I think it gives about a db lift at 20Khz.
Another option is to use foobar2K equaliser.
 
I guess I'm trying to the originality and / or novelty of the 3lite's original project. FPGA digital filters -- for consumer dacs like Wadia, etc -- have been around since the late 1980s.
That said, not sure I've seen a dedicated-device project appear yet on DIYA . . or for that matter, Ali or ebay. Which is a bit strange because the ways and means have been around for decades.

If anyone REALLY wants to impress with an FPGA project, two ideas:
(1) Digital filter: Rob Watts/Chord knock-off (1 million taps)
(2) FPGA-based d-to-a: Chord, PS Audio, etc.
 
Last edited:
I guess I'm trying to the originality and / or novelty of the 3lite's original project. FPGA digital filters -- for consumer dacs like Wadia, etc -- have been around since the late 1980s.
That said, not sure I've seen a dedicated-device project appear yet on DIYA . . or for that matter, Ali or ebay. Which is a bit strange because the ways and means have been around for decades.

If anyone REALLY wants to impress with an FPGA project, two ideas:
(1) Digital filter: Rob Watts/Chord knock-off (1 million taps)
(2) FPGA-based d-to-a: Chord, PS Audio, etc.

Since you have no idea what you are talking about or looking at, I can see why you don't recognize the value of this project.

BTW - it takes very little effort to make a stupidly long digital filter with massive latency like Chord.

Their DAC is not necessarily what I would call FPGA based, either. It's just another take on a switching DAC, but the front-end is realized in an FPGA instead of an ASIC like the ESS Sabre or a DSP like the (better) effort from Bruno Putzeys. The ESS Sabre or AK449x prototypes could have been called FPGA DACs too probably by your logic.

You show me a FPGA dac from before 1990 and I'll show you a break-dancing unicorn

He'll discover the difference between an FPGA and a Motorola DSP one day.
 
Last edited:
FPGA digital filters -- for consumer dacs like Wadia, etc -- have been around since the late 1980s.

As comes to Wadia Computer:

. the oversampling was last done with a SPLINE algo

. also the spline fixes the overshooting issue as be the Dirac impulse

. on auto range, some dB's lost on 20kHz range, but the sound is far better than the SDM DAC. Listen to piano & pianissimo classic music, as my used PCM63 & SM5843 FIR could not perform

. It required a DSP, as on Wadia 27 & 2.xxMHz SR with several PCM1704

. No real I/V, while output went the a single resistor

. The Wadia 27xi had a up-link option to the CD player to run all in sync

so, since Wadia left, no spline "oversampling" seen on sky as it goes in hand with the obsolete PCM1704.. I do not much about the later implementations after Wadia 27, while the newer gear prices xx'xxx $ are :eek:

Hp
 
I take it you're running your TDA1541 at 44.1kHz rate? In which case yes, you'll get 'NOS droop' leading to an attenuation around 3dB at 20kHz.


Always thought that a simple sinX/X filter is all it takes to fix this.

Also thought that the true NOS devotee accepted this as a feature rather than a bug and would not bother fixing it. Wouldn't the presence of high frequencies destroy the NOS magic? :cool:
 
Last edited:
@3Lite, I've performed only a quick search through this thread for an answer to the following question, but didn't spot one. Forgive me, if I've overlooked it.

Q: Is your design an equiripple, or an windowed-SINC or something else? I've always noted that the long out of production PMD200 and PMD100 appeared to utilize windowed-SINC interpolation kernels, rather than the typical equiripple design. Thanks.
 
so, since Wadia left, no spline "oversampling" seen on sky as it goes in hand with the obsolete PCM1704.. I do not much about the later implementations after Wadia 27, while the newer gear prices xx'xxx $ are :eek:
I think you guys should get out and learn a little about the world outside diyAudio.

From the CH Precision C1 User Manual Rev 2.2, 12/6/2019:
The 4.8GFLOPS DPS engine for each channel performs several tasks, including synchronous spline-based PEtER upsampling (to 705.6/768 kHz ...

The digital conversion for each channel is handled by four Burr-Brown PCM1704 R2R chips.
CH Precision is not the only one producing statement products based on the 'obsolete' PCM1704.
 
Unfortunately this project seems to be dead: I ordered the FIR board from them almost one year ago, they responded the assembly would take 4 weeks, never heard again from them...

I do very much apologize as I am rarely available on forums nowadays due to being busy with other work. I did notice my last login here was almost 1,5 year ago... Also, since those digital filters are hand assembled (due to required extensive tests of the quality) they had to be ordered for assembly by a person I can trust. The same goes to NC2k buffers we sent for assembly. That being the case, a friend of mine (Hermes5) will take care of your questions and orders via email (KuSy Audio - Unconventional And Perfect Hi-End Solutions).
 
Last edited: