Analog audio signal transmission via optical fibre (with test files)

[pma]

Analog audio signal transmission via optical fibre

Intro

Though contemporary individual SOTA audio sources (DACs) and preamplifiers and power amplifiers are at the parameters level that any improvements remain inaudible, after the components are interconnected via shielded or coaxial cables, it is not unusual that S/N parameters are degraded of 40 – 50 dB due to ground loops, especially in case of unbalanced single ended connection. One of the worst cases is a PC in class I feeding DAC in class I via non-isolated USB, and then the analog signal going to class I preamplifier or integrated amplifier. Buzz, clicks, drop-outs are not unusual, however even less evident degradation may occur as well.

USB link between the PC and the DAC should always be isolated and the isolated adapter should have the lowest possible stray capacitance. The other solution is a Toslink cable and isolated USB/SPDIF converter connected to the PC.

Sometimes we also need to have analog link signal with galvanic isolation. Signal transformers are a possible solution, but they may suffer from elevated LF distortion at higher signal level and also their stray capacitance is not negligible, allowing HF interference currents to flow in the circuit. The complete isolation solution is an optic fibre, which totally isolates signal source from the receiving circuit.
From 1980 to 2000 I designed and produced a number of fibre optic analog transmission systems that have been used in high voltage laboratories and HV circuit breakers testing plants. I thought it might be fun to try a principle of those systems to transfer audio signal.

Optical transmitter diodes are non-linear devices and their optical power is a subject of aging. Light power loss also depend on fibre length and precision of connectors used. Thus, it is impossible to get highly linear and repeatable parameters from direct analog transmission from transmitter diode – receiving diode couple. The solution is to use pulse width modulation (like class D) or pulse frequency modulation. Then, the system is immune to fibre length change or transmitter aging, if we operate within specified limits.

For the experiment with audio signal, I used Teledyne Philbrick 4705 VFC (voltage-to-frequency) converter (1974!). It transfers 0 - 10V input signal into train of impulses with frequency 0 – 1MHz. We may shift the base Fo frequency (for 0V input) and get the ability to transfer AC signals up to +/-5V. I am using base frequency Fo = 350kHz and transfer ratio 100kHz/V. For +/-2.8Vp input (2Vrms) we get range of transmission frequencies 70kHz – 630kHz. Signal from the VFC is easily coupled to the fiber optic link Broadcom HFBR1521 – HFBR2521. Output of the receiver module HFBR2521 triggers a monostable circuit with fixed output impulse width. Average value of the monostable circuit output is proportional to VFC analog input voltage. So the demodulation of the frequency modulated impulse train may be done with a simple passive low-pass filter.

VFC optical transmission sample

This is the schematics:

and the functional sample looks like this:

This is the impulse train output monitored on the optical fibre end:

Some measurements

Let's measure some audio parameters of the complete optical analog transmission link.

Frequency response is defined by the 2RC low-pass filter used:

Distortion at -6dBFS is about 0.7%, (edit - can be and was improved easily) with S/N = 80dB (similar to best tape recorders). There is no elevated high-frequency noise floor.

Noise floor

Square wave response is almost ideal:

Sine wave record:

Listening test

I have prepared a listening test, original 96/24 file and the same file recorded via the VFC optical transmission system described in this thread. The files may be downloaded from:

VFC_test.zip

drive.google.com

If you are interested, please feel free to download the files and share your foobar ABX report with us.

Below please find what the Deltawave says about the files:





____________________________________________________________________________________________________________________________________
P.S: After playing with the circuit a bit, I assume this is the limit with this VFC:

 

[pma]

Distortion can be improved quite easily, it is mostly about the output filter.

 

[yo!]

Not shabby at all, nice!
K2 @ -40 dB can add some audible vibes, some could like.

 

[AnalogSteph]

Neat concept.

So basically the linearity of this system depends on
1. the linearity of the VFC
and
2. your ability to make a filter with a linear (!) dropoff in the 70 - 630 kHz region, which is not their default behavior. That's quite the large frequency deviation you've got there.

The demodulator is basically a pulse count detector, isn't it?

I wonder whether you couldn't exponentially predistort the signal with one of these in such a way that it could be demodulated by an ordinary 6 or 12 dB/octave lowpass slope without having to jump through hoops in filter design... although it might do funny things to the noise floor.

And then there's the whole issue that you can only accommodate a single channel like that and would need a second fiber for stereo. It would be neat if you could have two FM links in parallel running over the link (or one with multiplex stereo, though its noise penalty is quite substantial - at least you'd be free to choose a higher pilot tone frequency than the usual 19 kHz).

Ah, the pitfalls of the analog age... it's fun to think about, but I think I'll stick with Toslink after all.

 

[pma]

1. Yes, linearity of the VFC converter is the key factor to define the limit of distortion.

2. Fo - carrier frequency may be shifted higher (to 500 kHz with this VFC) and there is also no need to use the whole range 70-630 kHz, this is easy to set by the input divider, reasonable is not to go below 200kHz. The demodulator is nothing but a lowpass filter, to get the average value of the pulse train with variable frequency but constant width. The lowpass filter is normally more sophisticated than a mere 2RC, Bessel and Butterworth active lowpass of 6th order were used in the real instruments. You can see the filter opamps it in the photo of the receiver board in the post #1. The board is the spare from those old systems I built.

Several 8 channel units were made in the nineties of the past century, for HV and high current testing labs. With different VFCs of course and higher BW, up to DC - 400 kHz. I thought it would be just fun to try it for audio, and it is clear that with some small effort it would work.
--------------
P.S.: I am getting this at the moment and I assume this is the limit with the Teledyne Philbrick 4705 VFC,1974 circuit.

New frequency response curve (DC - 70kHz):

 

[kchap]

Very interesting. In the 1980s, Siemens (and probably other Telco vendors) made short haul optical systems that frequency modulated PAL video on to a 20Mhz carrier which, in turn switched a 1300 nm laser. They were marketed as VOLTE, Video Optical Line Terminal Equipment.

As another approach, have you thought of driving a 1 Gb/s SPF with the a standard AES,EBU,S/PDIF signal and using another SFP as the receiver? That would solve the distance and bandwidth problems associated with the plastic fibre.

 

[pma]

Very interesting. In the 1980s, Siemens (and probably other Telco vendors) made short haul optical systems that frequency modulated PAL video on to a 20Mhz carrier which, in turn switched a 1300 nm laser. They were marketed as VOLTE, Video Optical Line Terminal Equipment.

As another approach, have you thought of driving a 1 Gb/s SPF with the a standard AES,EBU,S/PDIF signal and using another SFP as the receiver? That would solve the distance and bandwidth problems associated with the plastic fibre.

Thank you for interesting info, of course there are ways to do it much more sophisticated.

Interestingly enough, I was testing 192/24 SPDIF optical link recently, and this is one of the screen shots of the optical impulses measured at the end of the optical fibre with a fast opto-to-voltage analog converter.

Bitrate is up to 12.3 Mbit/s.

BTW, please do not forget that fast opto-electronic modules often have low bitrate limit. For example, 125 Mb/s modules usually work only at 20Mb/s - 125Mb/s range. So, they are unusable for SPDIF then.

 

[kchap]

Thank you for interesting info, of course there are ways to do it much more sophisticated.

Interestingly enough, I was testing 192/24 SPDIF optical link recently, and this is one of the screen shots of the optical impulses measured at the end of the optical fibre with a fast opto-to-voltage analog converter.

View attachment 399911

Bitrate is up to 12.3 Mbit/s.

It's a pity that there is not there is not enough consumer interest in taking the the existing standards with the addition of better receivers and glass fibre to achieve longer distances and higher bit rates.

BTW, please do not forget that fast opto-electronic modules often have low bitrate limit. For example, 125 Mb/s modules usually work only at 20Mb/s - 125Mb/s range. So, they are unusable for SPDIF then.

I thought that might be the case when I suggested the use of an SFP. Oh well!

 

[restorer-john]

It's a pity that there is not there is not enough consumer interest in taking the the existing standards with the addition of better receivers and glass fibre to achieve longer distances and higher bit rates.

Why would consumers care? They have simple needs and certainly aren't trying to send pristine audio/video over vast distances in the home.

This is fringe stuff. Fun for sure, but until you can convince me a "consumer" needs his source any more than a few feet away from the receiver of that source, what's the point?

Just because I can send SPDIF at 0.5V through a potato, a strawberry, or a piece of meat, and get bit perfect audio means nothing. (yes, I did it first many years ago)

 

[pma]

Regarding longer distance transfer, I use optical SPDIF 10m cable between my workroom (with PC) an my sitting room. PC sends signal via USB/optical SPDIF converter. I always use optical fibre to connect my PC with the main system DAC.

In fact, any direct connection between non-isolated USB from class I PC and class I DAC like most Toppings and SMSL is a disaster. It degrades the signal purity and in worse cases we have clicks, drop-outs or whistles. This cannot be seen from the reviews posted here in ASR, unfortunately. In fact the S/N and signal mess is completely defined by the interconnection used and loop currents over signal ground and USB ground. Who cares, manufacturers need pristine measurements under clinical single-component conditions.

 

[pma]

I made a fully working analog audio optical link with impulse FM modulation as described here above.

It uses active filter Butterworth 5th order as a demodulator and carrier frequency and FM products are suppressed of 96dB and more related to full scale output audio level. The spectrum is absolutely clean up to 60kHz. FM products start to be visible above 100kHz with FS signal, however suppressed as mentioned above.
Sound source is completely isolated from the audio system via optical fibre, so no loop troubles. And no SW, no worries about SW updates or OS changes. Old school solution .

Quite large instrument cases used just what I found in my stock.