JCAT Net Card Femto

[austinpop]

I posted a system review on another thread which featured the JCAT Net Card Femto, so posting a link here for those interested:

 

 

[austinpop]

On 2/23/2020 at 2:22 AM, Patatorz said:

Thanks Marcin, I’m checking with Piero from AudioLinux if he has some ideas

 

Please let us know what you learn. It seems like this can be done with a sequence of "ethtool" commands, but there may be a cleaner way using a config file somewhere.

[austinpop]

4 hours ago, bodiebill said:

Last week I received a JCAT net card FEMTO (JCF) and it lead to some experimenting that I would like to share here.

 

For my dual Linux PC setup I had to make a choice whether to first put it in the server with HQPlayer embedded or in the NAA endpoint. I started with the latter as intuition told me that the card would have the most effect downstream. As shown below I was wrong (again 🤔). I had to use a riser cable as my Streacom does not allow for direct plugin of the card.

 

Out of the box I immediately heard more detail with the card in the endpoint. However the sound was not yet listenable: classical violins where strident and harsh, and it lacked the holographic experience that I am looking for. This was to be expected (as others have written here), so I first let the card burn in for 4 days. The sound improved, but I was not happy yet. Some of the stridency was still there. For comparison I used a Beethoven violin sonata (opus 12-3, 2nd movement with Haebler and Szeryng) because it tests the soul versus stridency of the violin and credibility of the (to me) familiar piano sound as recorded by Philips.

 

From the start I noticed that a good external power supply really makes a difference for the JCF, as expected, so for the comparisons below that remained a constant. For now I am using:
iFi power 9VDC => DLS-HPULN LT3045 => 7.7VDC => (Chinese) LT3045 box => 5VDC.

 

In the comparison table below, I is the initial setup with JCF in the endpoint. Time to try the card in the server in several setups, namely II to IX. For this I replaced an Intel dual fiber card by the JCF. I experimented with and without fiber and same for bridging.

 

Bridging to me appears beneficial (notwithstanding Miska's advice against it for HQP-NAA). In Audiolinux I SW-bridged all 4 available server NIC's which includes the 2 JCF NIC's. It appeared that SQ improved as follows depending on which 2 NIC's were used:
stock in & stock out << JCF in & stock out == JCF in & JCF out << stock in & JCF out

so I am using the latter now.

 

For now I have decided to keep setup IX -- although SQ-wise hard to distinguish from copper-only setup VI -- as it saves me the work of changing a long run of hidden fiber cables to CAT6 cables, and gives me the piece of mind of having some isolation between server and endpoint.

 

Changing the net cards in the server also changed the fingerprint for the HQPe software, so I will have to ask Miska for a new license key after having verified that everything is stable.

 

Reactions / similar experiences / suggestions welcome!

 

Great report, and kudos for the experimentation!

 

Did you try a configuration - let’s call it X - which is the same as IX, except you bridged both JCF ports, while keeping the 2-switch chain between server and endpoint?