There's also the free version of the r8brain sample rate converting program, which also has different quality settings. No idea what quality setting has what impulse response.
Regarding the present test, it's a pity that everyone knows now that two of the three good answers are in the set {1, 5, 17}.
Regarding the present test, it's a pity that everyone knows now that two of the three good answers are in the set {1, 5, 17}.
According to the measurements in
https://www.diyaudio.com/forums/digital-line-level/373375-discrete-converter-10.html#post6697392
the transition time of a PCM1794 is of the order of a few nanoseconds. Of course the PCM1794 is not exactly the same as the PCM1792, but at least they are similar.
PGGB digital filter converted test files
Hans, I recieved a new message from the author of PGGB. He has offered to resample some test files for us via PGGB, to assist in your digital filter experiment. I immediately thought of submitting a set of files which you would choose, for the conversion. His full message is pasted below, which also includes very good pricing news for participants who may wish to purchase a permanent PGGB license.
==============================================================
Good morning Ken,
Let me know if any of the contributors in your thread need a trial license for the experiments being done, or I can create up or downsampled files using PGGB too if you provide source files. PGGB will only do the same rate-family conversion but can both upsample and downsample.
I would like to support DIY efforts, as I feel some modestly priced DACs are capable of benefiting from software based OS and can sound subjectively better than DACs that are priced much higher. In this spirit, if any of the contributors in your trhead are ultimately interested in buying a PGGB license, I would provide it at a 50% discount on the condition that the initial request comes through you (to prevent spam or abuse).
Regards,
-ZB
-------------------------------------
Author PGGB - remastro
Hans, I recieved a new message from the author of PGGB. He has offered to resample some test files for us via PGGB, to assist in your digital filter experiment. I immediately thought of submitting a set of files which you would choose, for the conversion. His full message is pasted below, which also includes very good pricing news for participants who may wish to purchase a permanent PGGB license.
==============================================================
Good morning Ken,
Let me know if any of the contributors in your thread need a trial license for the experiments being done, or I can create up or downsampled files using PGGB too if you provide source files. PGGB will only do the same rate-family conversion but can both upsample and downsample.
I would like to support DIY efforts, as I feel some modestly priced DACs are capable of benefiting from software based OS and can sound subjectively better than DACs that are priced much higher. In this spirit, if any of the contributors in your trhead are ultimately interested in buying a PGGB license, I would provide it at a 50% discount on the condition that the initial request comes through you (to prevent spam or abuse).
Regards,
-ZB
-------------------------------------
Author PGGB - remastro
Last edited:
I also did the 6-file test. I use EC Designs U192/DA96 DAC with Mori Drixo clocks. DAC connected to PP DHT amp driving Stax Omega2 electrostatic headphones.
Bach: This was much easier to decide than the other tracks. Good recording. I much preferred 21. Sounded more real to me. More natural, more dynamic and easier to "follow the music". Guessing 1 is processed then.
Round Midnight: Good recording. More difficult to tell. Going back and forth, I preferred 5, sounded a bit more real and involving. Guessing 13 is processed.
Cassidy: None of the 2 tracks sounds good IMO. Unpleasant sound, harsh voice. Preferred 17 as 2 was even more harsh to my ears. But if the processing muffles the sound like i'm guessing it does for the other tracks, then I might prefer the processing for this one? I'm guessing 17 is processed then.
So my guess: 1, 13, 17 are processed.
Bach: This was much easier to decide than the other tracks. Good recording. I much preferred 21. Sounded more real to me. More natural, more dynamic and easier to "follow the music". Guessing 1 is processed then.
Round Midnight: Good recording. More difficult to tell. Going back and forth, I preferred 5, sounded a bit more real and involving. Guessing 13 is processed.
Cassidy: None of the 2 tracks sounds good IMO. Unpleasant sound, harsh voice. Preferred 17 as 2 was even more harsh to my ears. But if the processing muffles the sound like i'm guessing it does for the other tracks, then I might prefer the processing for this one? I'm guessing 17 is processed then.
So my guess: 1, 13, 17 are processed.
Marcel,
Could be, but no single detail is given on how this was measured and how correct the circuit was built. Did it have a ground plane and was the digital gnd well separated from the digital gnd, etc, etc.
Discrete I-V converter
When not built and measured in a very controlled way, it can result in all sorts annomalies.
I gave a detailed description on what I did, and could not find any transition time below 100nsec although my gear could easily register faster transitions.
Hans
O.k. That’s true, so if this is not good enough let us then try to define what would be the proper way to do, because just some undocumented image tells very little.
What do you suppose, a current conveyor instead of a transimpedance amplifier?
I could use a AD844 for that purpose.
Plus input to gnd and minus input connected to the Dac, and a 1K resistor at current source output.
The buffer behind this can steer the measuring gear such as the HP461A that supports 4nsec transitions.
Hans
Ken,
Great to notice that more people performed the test in the meantime.
When 5 have responded, I’ll disclose what was processed.
Since these 3 files served their duty, and are no longer virgin, I’ll send you 3 new files to accept the generous offer from PGGB to convert them in upsampled files at 192/24 and a second one back to 44.1/16.
We are making some progress, aren’t we
Hans
Great to notice that more people performed the test in the meantime.
When 5 have responded, I’ll disclose what was processed.
Since these 3 files served their duty, and are no longer virgin, I’ll send you 3 new files to accept the generous offer from PGGB to convert them in upsampled files at 192/24 and a second one back to 44.1/16.
We are making some progress, aren’t we
Hans
Why make it so complicated? If I were to measure the transition time of a PCM1792 output, I'd use a 47 ohm resistor to ground, a thin 50 ohm coaxial cable, a 50 ohm termination resistor and a fast oscilloscope. The cable would take care of the connection to the scope and the termination resistor would be right next to the oscilloscope input (or inside the scope, if it had a 50 ohm termination feature), so there would be no need for a probe. Semi-rigid cables are really handy for tests like this.
Ken,
Great to notice that more people performed the test in the meantime.
When 5 have responded, I’ll disclose what was processed.
Since these 3 files served their duty, and are no longer virgin, I’ll send you 3 new files to accept the generous offer from PGGB to convert them in upsampled files at 192/24 and a second one back to 44.1/16.
We are making some progress, aren’t we
Hans
Hans, I thought the software is only upsampling and down with in the same family so 44 for 192 will not be possible maybe we need to check
O.k. Marcel,
50R is what I will try, good suggestion.
There is still one small catch, the 6 mA Dac bias has to be absorbed to prevent a 300mV bias on the 50R, but that won’t be too difficult with a 2k5 resistor connected between 15 Volt and an ultra low induction 50R measuring resistor.
But in fact this is still not the ideal situation where the Dac output is supposed to look into a virtual ground.
So in addition to the 50R, I will also see if anything changes with a lower resistance value.
Maximum scope’s sensitivity with HP amp is +/- 100uV full scale, probably enough for this test.
Do you still have any comment or additional suggestion on the above procedure ?
About the 50R coax connections without probes, that’s exactly how I measured with my previous recordings.
If not, results would likekely be quite unreliable with these very short transition times.
Hans
50R is what I will try, good suggestion.
There is still one small catch, the 6 mA Dac bias has to be absorbed to prevent a 300mV bias on the 50R, but that won’t be too difficult with a 2k5 resistor connected between 15 Volt and an ultra low induction 50R measuring resistor.
But in fact this is still not the ideal situation where the Dac output is supposed to look into a virtual ground.
So in addition to the 50R, I will also see if anything changes with a lower resistance value.
Maximum scope’s sensitivity with HP amp is +/- 100uV full scale, probably enough for this test.
Do you still have any comment or additional suggestion on the above procedure ?
About the 50R coax connections without probes, that’s exactly how I measured with my previous recordings.
If not, results would likekely be quite unreliable with these very short transition times.
Hans
In fact the DAC will see a 25 ohm load: 50 ohm on the board in parallel with 50 ohm at the end of the cable. You can also use 10 ohm to ground and 40 ohm to the cable, if you are afraid that the DAC runs out of headroom with a 25 ohm load. You then have 10 ohm in parallel with 90 ohm equals 9 ohm loading the DAC.
Well that’s quite important to know and could complicate things a bit.
My Dac converts everything automatically to 192/24, so when the PGGB software can only do multiples of 44.1 it still would need an additional processing step making the use of this software less attractive for me.
I had a closer look at the software and noticed that it can indeed convert and store in .wav files, so for our 3 file test that’s ideal.
That’s a thing I didn’t see before.
Can everybody play 176.4 or should we restrict to 88.2 ?
Hans
With very low resistances, you would have to be very careful with wire inductance. A rule of thumb is 1 nH/mm, so 2 cm and 10 ohm is 2 ns, 2 mm and 1 ohm is also 2 ns. The inductance will boost high frequencies (cause a left half plane zero).
Last edited: