Hypex UcD OEM Kelvin connection (feedback taken at speaker terminals)

[pjug]

I had a +/- 35V power supply I wanted to use, found a pair of UcD180OEM on ebay for $25/ea so I thought I would build something. While I am waiting to receive these, I read on the data sheet that there are connections to use a Kelvin connection at the speaker terminals to extend the feedback loop.

https://www.hypex.nl/media/d2/d7/f8/1645786900/UcD180_OEM_R4.pdf

5.7 Optional remote (kelvin) feedback
When maximum control of the loudspeaker is needed it is possible to include the loudspeaker cables
in the UcD control loop thus eliminating all negative effects of long cables etcetera.
Connect both the positive feedback (pin 19) and negative feedback (pin 17) connections as close to
the loudspeaker as possible. No extra connections are needed.

I don't really care about using that, but then I thought it might be fun to try to measure at some point to see how well it works. So I think I will include connections on these pins to terminals on the amplifier case just for the heck of it. Some questions for anyone who can provide some expertise:

- Will degrade performance to run wires to the case that then have no connection? I can always just pull them off at the UcD board's header when not in use.
- Any advice on what wire type I should use or care in how they are run from the speaker back to the amp?

I don't see any UcD products actually using this. Anyone know of any?

Is this a waste of time? I think the answer is yes but I will probably mess with it anyway.

 

[SIY]

Is this a waste of time?

Unless you have massively pathological speakers (impedance dips into the tenths of ohms) or pathologically high resistance cables, then there's no advantage to this.

 

[pjug]

Unless you have massively pathological speakers (impedance dips into the tenths of ohms) or pathologically high resistance cables, then there's no advantage to this.

Yes, I don't plan to use it, just test it. Was just curious about whether it takes out those little wiggles even though I don't care about them. So just wasting some time to satisfy my curiosity.

 

[RickS]

I had a +/- 35V power supply I wanted to use, found a pair of UcD180OEM on ebay for $25/ea so I thought I would build something. While I am waiting to receive these, I read on the data sheet that there are connections to use a Kelvin connection at the speaker terminals to extend the feedback loop.

https://www.hypex.nl/media/d2/d7/f8/1645786900/UcD180_OEM_R4.pdf

5.7 Optional remote (kelvin) feedback
When maximum control of the loudspeaker is needed it is possible to include the loudspeaker cables
in the UcD control loop thus eliminating all negative effects of long cables etcetera.
Connect both the positive feedback (pin 19) and negative feedback (pin 17) connections as close to
the loudspeaker as possible. No extra connections are needed.

I don't really care about using that, but then I thought it might be fun to try to measure at some point to see how well it works. So I think I will include connections on these pins to terminals on the amplifier case just for the heck of it. Some questions for anyone who can provide some expertise:

- Will degrade performance to run wires to the case that then have no connection? I can always just pull them off at the UcD board's header when not in use.
- Any advice on what wire type I should use or care in how they are run from the speaker back to the amp?

I don't see any UcD products actually using this. Anyone know of any?

Is this a waste of time? I think the answer is yes but I will probably mess with it anyway.

Interesting as have not built any UCD amps…

Newer versions of Bruno’s design have comparable compensation but does not go past the internal outputs. If I recall correctly, the Purifi data sheet calls for it not going beyond the amp output terminations.

 

[pjug]

Interesting as have not built any UCD amps…

Newer versions of Bruno’s design have comparable compensation but does not go past the internal outputs. If I recall correctly, the Purifi data sheet calls for not going beyond the amp output terminations.

I don't see any UcD examples online where someone tried to implement this. On the surface it strikes me as asking for trouble to include haphazard long wires as part of the closed loop circuitry. But since these were cheap I think I will give it a shot and play around, while using speakers I don't care about.

 

[SIY]

I don't see any UcD examples online where someone tried to implement this. On the surface it strikes me as asking for trouble to include haphazard long wires as part of the closed loop circuitry. But since these were cheap I think I will give it a shot and play around, while using speakers I don't care about.

I'd also use an amp you don't care about. Oscillation can kill the output stage in a spectacular smokefest.

 

[pjug]

I'd also use an amp you don't care about. Oscillation can kill the output stage in a spectacular smokefest.

Yes, see first post just $25 if I play around with just one.

 

[restorer-john]

The feedback path by connecting at the loudspeaker terminals is likely to be an additional feedback path, not the only path. Kenwood's sigma drive did much the same, although it caused more than a few amplifiers to blow up.

Give it a go.

 

[pjug]

The feedback path by connecting at the loudspeaker terminals is likely to be an additional feedback path, not the only path. Kenwood's sigma drive did much the same, although it caused more than a few amplifiers to blow up.

Give it a go.

Thanks for this. An interesting read. Also thanks for encouraging giving this a go.

For wire, I have a spool of 6 conductor 24AWG with overall shield I was thinking of using. I figure I should use a bundle for the current carrying and feedback wires since that is what someone would want to do if they were actually using the Kelvin connection. So I will use 20ft of this cable, and double up on the current carrying wires for effectively ~21AWG, and 24AWG for the feedback. I don't know if the shield is good or bad but I think better to have it?

[edit: having to use 20ft ~21AWG for the speaker wire to make it easy to see the wiggles in frequency response shows how dumb it is to spend time doing this]

 

[MaxwellsEq]

I had a Deltec DPA 50S which deployed a non-optional Kelvin-sensing cable. It never failed, but I was always a bit nervous about it.

Internal photo

 

[pjug]

I had a Deltec DPA 50S which deployed a non-optional Kelvin-sensing cable. It never failed, but I was always a bit nervous about it.

Interesting! What was the construction of the cabling?

 

[MaxwellsEq]

Interesting! What was the construction of the cabling?

Two sets of XLR connectors per channel, one male and one female. They shipped with premade "Slink" cables and highly recommended not using anything else! The speaker ends were connected together into a single pair of banana plugs. The normal feedback connector was run in a meshed sleeve. I never knew why, but perhaps they were nervous about people twisting the cables together. Doing something like this places a lot of trust in the end user!

 

[pma]

I would finish at taking feedback from amplifier output binding posts. Would not go to the end of the speaker cable. Tricky and dangerous.

 

[pjug]

Two sets of XLR connectors per channel, one male and one female. They shipped with premade "Slink" cables and highly recommended not using anything else! The speaker ends were connected together into a single pair of banana plugs. The normal feedback connector was run in a meshed sleeve. I never knew why, but perhaps they were nervous about people twisting the cables together. Doing something like this places a lot of trust in the end user!

Two sets of connectors coming out of a single bundled cable, or two cables per channel?

 

[pjug]

I would finish at taking feedback from amplifier output binding posts. Would not go to the end of the speaker cable. Tricky and dangerous.

Just danger of damaging equipment or dangerous to myself?

 

[pjug]

Doing something like this places a lot of trust in the end user!

I was really surprised when I noticed Hypex would include this. I think it is only on the UcD OEM versions. I wonder if some OEM pushed them to include it?

 

[EERecordist]

I spent a lot of time looking at subwoofers with feedback. There is one in the review database from Rhythmik. Velodyne, Mackie, and others have used the technique to reduce distortion. They used accelerometers on the cone or measured electronically. (Velodyne made great products. They discontinued audio to focus on LIDAR for self-driving cars.)

I agree with all the suggestions on the risk to the amp and the speaker. The sense leads are part of a powered feedback loop. The amplifier will provide power up to its maximum power within the feedback loop. If you have any noise or a broken connection the amplifier is going to provide power to zero error in the feedback loop. The amount of error between the speaker terminals and the amplifier terminals is essentially 0 under normal operation.

Perhaps you can have a discussion with Hypex on how they manage sudden loss of the feedback loop?

Your amplifier is like the great machine in the classic science fiction film Forbidden Planet. In that film the great machine had the capability to provide relatively infinite power including the power to destroy itself.

 

[DVDdoug]

I also wouldn't expect any actual improvement because (normal-length) speaker wires have very low resistance.

While I am waiting to receive these, I read on the data sheet that there are connections to use a Kelvin connection at the speaker terminals to extend the feedback loop.

I'd also use an amp you don't care about. Oscillation can kill the output stage in a spectacular smokefest.

Without knowing anything about the circuit... Since the manufacturer says it's OK I'd assume it's safe. I also assume that it has low open-loop gain, or that there is some internal negative feedback so it's not fully open-loop without the external-user feedback.

For wire, I have a spool of 6 conductor 24AWG with overall shield I was thinking of using. I figure I should use a bundle for the current carrying and feedback wires since that is what someone would want to do if they were actually using the Kelvin connection. So I will use 20ft of this cable, and double up on the current carrying wires for effectively ~21AWG, and 24AWG for the feedback. I don't know if the shield is good or bad but I think better to have it?

Since speakers require voltage and current, heavier wires might be better than the feedback.

Normally speaker wires don't need to be shielded. Shielding adds capacitance which could be "dangerous" with feedback, but I'd guess it's insignificant.

They used accelerometers on the cone or measured electronically.

THAT makes sense, at least theoretically, and the manufacturer of an active speaker can control EVERYTHING to keep it stable. In the real world, speaker design involves lots of trade-offs and don't know of those speakers perform better than everything else. I kinda' doubt it...

 

[pjug]

Normally speaker wires don't need to be shielded. Shielding adds capacitance which could be "dangerous" with feedback, but I'd guess it's insignificant.

So I am worse off with shield on the feedback lines? The 6X 24AWG with shield just happens to be what I have but I could wrap my own cable bundle of just about anything if someone wants to suggest what to do. If the current carrying wires are not fairly thin then it won't be easy to see a correction to the frequency response. Even with 20 ft of 21 AWG it should be less than 1 dB, looking on VituixCAD.

For thick cable I have this: https://www.amazon.com/CircleCord-Extension-14-30P-14-30R-Charging/dp/B08XBP5XDZ
Just kidding but it might make quite a star quad cable.

 

[restorer-john]

The sense cabling can be anything. Cheap two core speaker wire you have scavenged is fine. It's only voltage feedback at the end of the day. The impedances are low enough that shielding should be unnecessary.

The whole concept of being able to remove the effects of speaker cable (particularly long lengths) from an amplifier/loudspeaker system is cool. I remember experimenting with that and negative impedance back in the 80s.