In my Cirrus Logic CS8416 SPDIF to I2S board design , I want to include a SPDIF transformer to avoid ground loop problem . Although there’re a lot of others ready made digital transformers out there, I plan to build my own because I have a lot of toroid rings from my previous projects.
Not all toroids are suitable for SPDIF signals, I need to choose one with good bandwidth, at least from couple of Khz up to a few Mhz, Other than downloading data sheet from various manufacturer, I just try to wind an actual transformer and measure its response because most of my toroids are actually from unknown manufacturer.
SPDIF signal from 24 bit 192 Khz are actually operate at around 18Mhz, I try to experiment with a ring core which is slightly narrower response and see how it can reduce spikes on those overshoot and hence reduced higher frequency noise.
I tried wide band EMI suppressor ring core, which can be found on power cord feed into industrial or some computer switching power supply. They wind a few turns of wires on EMI suppressor ring core before connect into the PCB. The suitable size are around 20mm to 25 mm outer diameter, any smaller will work too but harder to work with , and need a few extra turns of winding for correct inductance.
I wind around 10~16 turns for 25mm OD core, and aiming for inductance around 0.8~1.5mH. In fact , if the inductance measured greater than 0.2mH ++, then it should work fine.Copper wires in between 0.4mm~0.7mm in diameter can be used, thinner are better for high frequency response I believe, I even tried Litz wire but I think single core enameled copper wires are good enough for the job. This EMI suppressor version works well as isolator to prevent ground loop.
I found some info regarding ring core SPDIF transformer manufactured from a big brand company, and they used nickel permalloy tape core for their SPDIF transformer product . I do have thin tape wound nickel permalloy ring core left over from my previous MC step up transformer projects. I can’t believe nickel tape wound core can operate at above 10Mhz , I wound one and compare with my EMI suppressor version, there’s slight different in listening test, the nickel core sound a little bit darker background but I only test it on 16bit 44.1 Khz media file , operating in more than 10th of Mhz are very challenging for nickel permalloy , at least in my diy version. The measured frequency response of nickel permalloy core are not as good but I can see the benefit of using his core because it filtered out those sharp peaks from ringing . The manufacturer make used of the hysteresis characteristic to act as a spike filter. I will try to experiment more, maybe thats a secret behind the big brand product.
Also , I tried N30 material from EPCOS, it works ,and sound slightly different than EMI suppressor core. The frequency response from their data sheet shows optimum frequency up to 400Khz. Measured waveform are not as good too. It filtered out some spikes though.
Winding with two wires twist together will induce a lot of ringing , and traditional text book style winding are much less distortion.
I need more time to experiment with others core but I’ll stick with EMI suppressor core as a reference . Until I found a better design , I will share with you guys again.