I'm looking to transmit TDM signals (digital audio) over long-distance cables (around 5 meters). I did some tests with an oscilloscope and the signal is getting too much distortion. I was wondering if I could use a re-timer to rectify the signal. Is it even possible to get the original signal without loss? What is the best way to minimise signal reflection, and will termination resistors be effective in this situation? I came across this chip: DS250DF410; but it only seems to deal with higher frequencies:

Each channel of the DS250DF410 independently locks to serial data rates in a continuous range from 20.2752 Gbps to 25.8 Gbps or to any supported sub-rate (÷2 and ÷4), including key data rates such as 10.3125 Gbps and 12.5 Gbps).

Are there any other re-timers you would recommend for this case?

The TDM clock is 11.3MHz. See below the images of the original data signal and the deteriorated signal. original deteriorated

Thanks for your help, any other ideas will be much appreciated.

  • 1
    \$\begingroup\$ How did you terminate that signal? The distorted edges look like the reflections are not well managed. \$\endgroup\$ Commented Nov 17, 2023 at 15:28
  • \$\begingroup\$ there is no termination, this is the signal as it comes out of the cable at 5M. \$\endgroup\$
    – Emerson
    Commented Nov 17, 2023 at 15:35
  • 1
    \$\begingroup\$ You have to terminate this signal. \$\endgroup\$ Commented Nov 17, 2023 at 16:22
  • 1
    \$\begingroup\$ The issue is that you didn't consider the impedance of your line and terminate properly. It isn't clear why you think a retimer would help, but since the underlying problem is your wiring, you're going in the wrong direction trying to add something on the next board after you've already screwed up the signal. \$\endgroup\$ Commented Nov 17, 2023 at 16:31
  • 1
    \$\begingroup\$ A retimer would be useful if the signal had clean edges with a bit of jitter or random delay to these edges. Not the case here. \$\endgroup\$
    – bobflux
    Commented Nov 17, 2023 at 17:12

2 Answers 2


TDM has 3 signals: Frame sync, clock, data.


These need to be aligned in time properly. The source emits sync and data on the falling edge of clock, and the receiver samples sync and data on the rising edge of clock.

Easy mode: source-synchronous, with clock and data generated in the same device, then both are sent through the cable. They will arrive at the receiver with correct alignment.

Hard mode: clock is generated at one end of the cable, but the chip transmitting data is at the other end. There will be a roundtrip delay added to the data (and maybe sync, depending on who generates it). If the cable is long enough this can screw up the timing.

Back on topic

The signal on your scope looks like it was sent through a cable with the wrong termination impedance at the source which means it is not usable.

Here's a sim with a 100 ohms transmission line (like 2 wires next to each other in a ribbon cable) directly connected to the output of a logic gate with a bit too much drive strength. The 22 ohm resistance approximates the driving chip's output impedance.

enter image description here

Some of the wiggles are close to logic level "0.5" so you can get double clocking as the input circuit in the receiver interprets it as two clock edges instead of one. In other words it's not working. It's a bad idea to use a circuit at the receiving end to make sense of this signal, the solution is to make sure it arrives in good shape instead.

So you need a clean signal, which means terminating the source of the transmission line at its characteristic impedance:

enter image description here

Termination at the receiver is not mandatory. If it is absent, then signal will fully reflect at the receiver, go back through the cable, and be absorbed by the termination resistor at the source.

Termination at both ends works better if the termination resistance does not accurately match transmission line impedance. Here's an example with 80 ohms instead of 100. Termination at both ends delivers a cleaner output, but of course you lose half the amplitude, which is inconvenient with standard logic levels.

enter image description here

Over 5m with ~5ns edges it means you need a transmission line of known impedance so you can terminate it. Shielding and/or differential is a plus if you don't want to make a wideband radio transmitter. So either a differential signal over twisted pair, or a single ended signal over coax. You could even use ribbon cable, after all PATA did exactly that, but it was inside a shielded computer case so hopefully emissions didn't get out.

Coax is relatively simple, but you need three with assorted connectors, and that's cumbersome. Also propagation delays have to match so the cables should be the same, and you don't have extra pins for power supply and other stuff. Although if you want 3 coax with connectors and extras you can use a VGA cable. Another drawback of single-ended signals is you have to be more careful about EMI and crosstalk through the shared ground.

Fortunately, there's a simple solution: you can either use Cat6 cable for a cheaper option, or an off the shelf HDMI cable which has the advantage of coming readymade with connectors on both ends, a good shield, and plenty of useful extra pins.

Then you can stick a LVCMOS-to-LVDS chip on the sending end, its matching receiver at the other end, decoupling caps, ground plane, and it should work fine.

LVDS allows terminating at both ends without issues about losing half of the logic level 1 voltage, since the receiver is a differential comparator. Just check if the driver and receiver you buy have integrated resistors of the correct value for your cable, or if you have to add them.

  • 1
    \$\begingroup\$ Thanks for your amazing answer. This is very useful, and the LVCMOS-to-LVDS chip is perfect for what I have in mind. I will test it soon. Thanks! \$\endgroup\$
    – Emerson
    Commented Nov 17, 2023 at 17:42
  • 1
    \$\begingroup\$ I would even recommend RS-422 over LVDS as the bitrate isn't terrifically high; which gives far more CMR than LVDS does. \$\endgroup\$ Commented Nov 17, 2023 at 17:58

Your cable is probably 50 or 60 ohms single ended impedance.

This porch you see on the rising and falling edges of the signal indicates that whatever is driving the cable has an output impedance close to that of the cable.

enter image description here

What happens is the driver's output impedance forms a resistive divider with the cable impedance. The signal doesn't reach full strength until the reflection from the open, unterminated end of the cables comes back. This is typical of a source-terminated interface.

It's hard to tell from the 'oscope pics you included, but the duration of this "porch" would be consistent with the out and back flight time of the signal (2 ns/ft x ~32 ft) = 64 ns.

To fix this, you need to do two things.

  1. you need to terminate the cable properly at the destination and
  2. you need a driver strong enough to drive the cable impedance.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.