AES/EBU input pulse transformer capacitors?

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

living sounds

Well-known member
Joined
Jul 26, 2006
Messages
3,931
Location
Cologne, Germany
I've got an old Protools interface from which I've salvaged the pulse transformers for my new multichannel converter.

Attached to the transformer primaries of the AES/EBU inputs are 100pf capacitors to (chassis) ground. I couldn't find a similar arrangement in any papers or datasheets.

I've looked at the input (primaries) of my converter with a scope with and without capacitors attached in an identical manner and could see the expected change in the waveform (less overshoot but slower rise time). It's hard to say which one is better, though, since the waveform doesn't looks like a clean squarewave either way on my 50MHz scope (at 44.1k).

The signal path has a proper shield connection to chassis, so the balanced connection shouldn't be picking up much RF. The trafos secondaries are terminated by a 110 ohm resistor.

Beyond RF, could there be a reason in the pulse transformer's makeup that mandates the caps? All I could measure was a primary inductance in the ballpark of similar transformers intended for AES/EBU connections.

 
living sounds said:
Beyond RF, could there be a reason in the pulse transformer's makeup that mandates the caps?
Sometimes it is better knowing his enemy. What I mean is it's better dealing with a known slew-rate than depend on the particular characteristics of a cable. As long as it does not result in exaggerated data-induced jitter, it's fine.
 
Thanks Abbey!

So, in case I do know the characteristics of the cable (I'm using proper 110 ohm wire, 3 meters max) leaving those caps out would be prefereable?
 
Is there a rise time and overshoot spec for AES? I would assume there would be.

In high frequency applications caps and inductors can be used to match the proper impedance.  I'm speculating,  but it's possible the caps are helping the input look like a true 110 ohm at the frequency of interest.

Also scopes are usually 50 ohm, so you are looking at pulses in an unmatched condition when measuring 110 ohm AES.

I would be inclined to keep the caps unless you find a good reason not too.
 
Thanks John!

These are the characteristics of the AES-1992 interface:

Transmitter characteristics
Balanced output with XLR connector
Source impedance: 110 Ω ± 20%
Balance: <-30 dB (to 6 MHz)
Launch signal amplitude:  2 - 7 Vp-p across 110Ω load
Rise and fall time: 5 to 30 ns
Jitter: <20 ns p-p

Receiver characteristics
Balanced input with XLR connector
Input impedance: 110 Ω ±20%
CMMR: Up to 7 V p-p to 20 MHz
Maximum accepted signal level: 7V p-p
Cable specification: Shielded twisted pair
Cable length: 100 to 200 m maximum
Cable equalization: Optional
 
john12ax7 said:
Did you measure the rise time or have the part number of the transformer? Could put the psraditics in a simulator and see.

No, as you said, my scope/probe isn't properly matched. The trafo is an unmarked part, so no way to look it up...
 
living sounds said:
No, as you said, my scope/probe isn't properly matched.
Even if it was matched, it would alter the measurement, because it's already loaded. You would need a high-Z active probe for that kind of measurements.
Actually, the usual way of evaluating this is to build the circuit and check signal integrity. The proof is in the pudding, not in its ingredients.  :)
 
Thanks guys!

Is there a way I can measure the AES/EBU signal with my standard scope/probe?

I've made recordings with multiple high end AD converters. Listening to these vs. the original files it's obvious the caps need to be there.
 
living sounds said:
Is there a way I can measure the AES/EBU signal with my standard scope/probe?
Not really; you can check the eye diagram, but it's quite difficult to relate it to the actual quality of transmission of the digital signal.
You need a digital signal analyser; I believe the less expensive one available is the NTI Digilizer (about €2k!).
You can rent one...
 
Thanks, abbey!

I had found a link where a guy got a clean square wave with a 100Mhz scope using a special 75 ohm adapter or something, but unfortunately I cannot locate it now.
 
Could you clarify your listening test and the caps statement? I thought there were transformers for digital data to feed a DAC? If you are hearing errors it would be jitter and / or error related.

Building a test pcb would be the best way to test these and would open up some options.  I somewhat disagree with Abbey on the usefulness of eye diagrams , as you gain a lot of info from the size and quality of the eye opening.
 
john12ax7 said:
Could you clarify your listening test and the caps statement? I thought there were transformers for digital data to feed a DAC? If you are hearing errors it would be jitter and / or error related.

Building a test pcb would be the best way to test these and would open up some options.  I somewhat disagree with Abbey on the usefulness of eye diagrams , as you gain a lot of info from the size and quality of the eye opening.

Yes, the DAC input is sensitive to the quality of the digital signal; jitter somehow transcends the FIFO, it's complicated and not all clear, but there is an obvious difference between the recordings in favor of the caps.

What would a test PCB require?
 
You can view the transformer as a 4 terminal (port)  device and then each port is 50 ohm.  This would be close to 110 ohm differential.

Then you can feed the transformer either differential or single ended depending on your available equipment.  The PCB would have appropriate AES XLR and/or BNC i/o.

Edit : Started writing more but I think drawing a diagram would be more useful.
 
You mean just send a high frequency squarewave through and look what comes out at the other side on a scope? I haven't got a signal generator with sufficient bandwith, I think.
 
This would be a universal type test pcb.  The jumpers allow choice of BNC or XLR i/o. Use a ground plane,  SMT parts,  all traces treat as 50 ohm lines (there are calculators for this).  The BNC and any coax cables should be 50 ohm as well.

If you don't have a good signal generator, use an AES signal.  Then you could hook up a scope,  or do a big long digital loopback test in a DAW,  things should be bit for bit identical.

With a scope you could run differential by using two channels,  or single ended,  by adding a 50 ohm load to the other BNC.

With an eye diagram you can do a comparison and estimate jitter and error rate.  One caveat though is that I've never done this in a home lab type environment. The equipment most of us use for home use may not be sufficient to detect minor differences.
 

Attachments

  • IMG_20201104_183231.jpg
    IMG_20201104_183231.jpg
    692.5 KB · Views: 10
john12ax7 said:
With an eye diagram you can do a comparison and estimate jitter and error rate.
Can you really? You can certainly make a comparison of the quality of transmission of the HF signal, but I very much doubt you can predict how the receiver will react.
I went to a seminar with Bernhard Weingartner (Neutrik) where we had the possibility to introduce various amounts of jitter and it showed that serious deterioration of the audio signal came suddenly when the peak value of jitter resulted in sample displacement. Below that, the audio quality was affected in various amounts that were mainly related to the PLL's performance.
 
john12ax7 said:
This would be a universal type test pcb.  The jumpers allow choice of BNC or XLR i/o. Use a ground plane,  SMT parts,  all traces treat as 50 ohm lines (there are calculators for this).  The BNC and any coax cables should be 50 ohm as well.

If you don't have a good signal generator, use an AES signal.  Then you could hook up a scope,  or do a big long digital loopback test in a DAW,  things should be bit for bit identical.

With a scope you could run differential by using two channels,  or single ended,  by adding a 50 ohm load to the other BNC.

With an eye diagram you can do a comparison and estimate jitter and error rate.  One caveat though is that I've never done this in a home lab type environment. The equipment most of us use for home use may not be sufficient to detect minor differences.

Thanks, that's just the trafo and the caps... So the idea is to test the trafo without any load but the scope?
 
abbey road d enfer said:
Can you really? You can certainly make a comparison of the quality of transmission of the HF signal, but I very much doubt you can predict how the receiver will react.

Yes the receiver can recover and correct for some of it.  But that s an unknown and dependent on receiver design.  If you prefer it be stated that the eye opening is proportional to error rate given the receiver takes no corrective action, I'm ok with that. Or perhaps potential error rate is a better description.
 

Latest posts

Back
Top