Transformers have many great properties that can be useful for audio applications. Most useful is the galvanic separation of input and output, which helps to eliminate ground noise issues. Many great articles are available how to properly ground audio equipment and following good advice helps to minimize the problem. Ironically ground noise issues are even common when using professional balanced interconnects since equipment design is often subpar (the so called pin 1 problem). In order to enjoy absolute freedom from ground hum issues, audio signal input transformers are the complete cure. Transformers offer many more great features like unmatched CMRR, low THD and super low IMD. Mayor downside is high cost and high low frequency distortion in case the driving impedance is high. An OPamp output would be well suitable to drive an audio input transformer. Another downside is that the sink impedance needs to be high (>10kOhm) and capacitive loading should be avoided (even a longer coax cable may be too much).
- DC blocking capacitors to avoid core saturation by DC, which would increase second harmonic distortion
- Optional film capacitor to bypass the electrolytic DC blocking capacitors
- Option to mount three different transformer types:
- Type1690 from Lundahl
- Type NTL-1 from Neutrik
- Type A187A12C from OEP
R7 and C4 need to be adjusted for each transformer for correct square wave response. Pin 1 of the XLR connector needs to be connected to chassis ground, not audio ground. This is true regardless of whether transformers or any other circuitry is used as input stage. Pin 1 is safety ground, not audio ground.
For the input I chose XLR connectors since they offer differential signal support. This is by far the best way to transmit any signal with best noise immunity. In case the feeding signal is single ended, an adapter cable from XLR to RCA / Cinch connecting the signal to pin 3 and ground to pin 2 could be used. The output is available on SMA sockets. The reason I chose SMA connectors is that the interconnects are well shielded, the connection can be established easily without any tools and cables configured with connectors are available in almost any length. Each channel has two sockets in case the audio signal needs to be routed to two modules at the same time (I have plans for this already). The option to mount three different transformers provides a variety of both sonic and economic options. The most economic option would be the OEP transformer, which sells for ~35&Euro;, next is the Neutrik for ~50&Euro; and most expensive is the Lundahl for ~100&Euro;. Resitor R7 and capacitor C4 need to be tuned for correct damping of the transformer.
Here is a visual comparison of the two more affordable transformer options: OEP model A187A12C on the left and Neutrik NTL-1 on the right. My guess is that the OEP performs better since it is larger, but this may be a fallacy. I would have kept the OEP for doing measurements, but it was shipped by RS Components in such a lousy packaging that I seriously doubt the parts survived transport without damage. It probably doesn't make much sense to measure broken parts. The Neutrik transformer shipped by Mouser was properly padded for transport and I have way more faith that it makes sense to continue with those instead.
Neutrik does not provide any application support for those transformers. So I had to figure out compensation trial and error. Measurement setup was using the square wave output of my oscilloscope and the output was loaded by 50cm coax cable and 22pF of the oscilloscope input.
The experiment with the Neutrik transformer shows how sensitive an audio input transformer is to secondary loading. Extra care is required in case the amplifier features any input filter that loads the transformer. This will add back some overshoot to the square wave. With the Neutrik NTL-1, I settled for a compensation network made of 7k15 and 1nF. In case the amplifier input network is designed like in the article mentioned, (two times 1k1 and one 17k8 resistor), the input resistance is 20kΩ. Using 100pF capacitors either in the first or second order low pass filter sharpens the square wave corner a bit. Two times 100pF are perfectly acceptable. This would set the filter roll-off frequency to 350kHz and reduces phase shift to slightly over 2° at 20kHz. Considering that the input transformer mostly suppresses common mode interference, this sounds acceptable given the amplifier slew rate is sufficient. Adding higher capacitance to the low pass filter results in overshoot at the square wave corner regardless where the higher capacitance is located inside the filter.