I need to sniff the physical layer of a proprietary data line with a logic analyser, but my analyser is limited to 5V input. Unfortunately the line runs at 47.5V, which would fry everything. I considered a simple voltage divider, but I'm concerned that I might pump a lot of current through it, or affect the data on the line.

Any suggestions on a way to do this? It doesn't need to be robust - it's just a one-time hack for a single purpose.

EDIT: To answer some of the questions asked: I don't know if it's 47.5V DC offset or 47.5V p-p, as I only have a single figure given to me as part of the spec. It's an old system produced by a company that no longer exists, so getting original specs and info is practically impossible. The data rate is 2Mbps, which to me implies that it shouldn't exceed 4MHz at the physical layer.

  • 1
    \$\begingroup\$ Is the data voltage 47.5Vp-p or is the data superimposed (at say 5vp-p) on a 47.5V dc offset like in a phantom powered system? \$\endgroup\$ – Andy aka Mar 14 '14 at 16:48
  • \$\begingroup\$ @Andyaka I'm not sure, to be honest. All I have is the spec, and I'm currently without any tools that could tell me. \$\endgroup\$ – Polynomial Mar 14 '14 at 17:10
  • \$\begingroup\$ What sort of "data line" is this, 47.5v sounds a bit telecomms to me, in which case there may well be line coding etc. in the way. \$\endgroup\$ – John U Mar 14 '14 at 17:32
  • \$\begingroup\$ Reading comments / responses below, you seem to clarify that it IS telecomms, in which case PLEASE add as much data as you can to your question - equipment name/type/number will be a massive help. 2Mbits over the POTS, if it's single-pair, suggests CWSS or HDSL. \$\endgroup\$ – John U Mar 14 '14 at 17:35
  • \$\begingroup\$ @JohnU Unfortunately I don't have much information on it, and I can't reveal the name of the product (I really wish I could!) for contractual reasons. I've not even seen the device myself - there are guys on the ground working on it. I do know that it's a proprietary converter, though, which doesn't use standard DSL-based communications, but does go over standard copper POTS lines. \$\endgroup\$ – Polynomial Mar 14 '14 at 17:38

Two options.

Option 1. If data is fully swinging from 0V to 47.5V then a voltage divider using two resistors is fine. Choose values of resistors that are not too low so that excessive power is dissipated. Check also to see how much the line can be loaded with. If necessary, make a pseudo scope input from 1Mohm and 100k potential divider - this will produce about 4.3Vp-p BUT, you may need to put 22pF across the 1M and 220pF across the 100k to keep "edge" information in the data stream clean.

Option 2. If the data is 5Vp-p superimposed on a dc level of 47.5V then it's more complex - this is like "phantom power data" and to "read" the 5Vp-p data you need to remove the DC by using a capacitor - this leaves you with an AC data signal that is 5Vp-p but undulating up and down with the average mark-space ratio of the signal.

If the data is encoded in such a way that there are always data edge transitions then that's fairly easy - connect the dc-removed signal into a signal comparator that has a 0V/5V output then feed that into your analyser. However, if the data is similar to UART data, there will be long periods when the signal is 1 or 0 and this is a problem that you may never solve easily.

  • \$\begingroup\$ You mention using a cap for eliminating DC - is there a good way to choose the value of the cap? I'm pretty sure the data going across doesn't exceed 4MHz as it's a 2Mbps link. \$\endgroup\$ – Polynomial Mar 14 '14 at 17:16
  • \$\begingroup\$ It all depends on what the lowest frequency content is on the data - if the data disappears for long periods (like in a UART transmission) then no amount of capacitance is going to help. If the "no-transmission" time is a second then maybe 100uF (more of a guess than anything) might be OK. If the "no-transmission" is 1 millisecond then 1uF might be OK. It also depends on how much resistance follows the capacitor - if it is 100k ohm then you can reduce the above cap values by 100. If it is 50 ohms then multiply them by ten - without knowing the structure of the data it is hard to guess. \$\endgroup\$ – Andy aka Mar 14 '14 at 17:23
  • \$\begingroup\$ I'm honestly not sure - I have no samples of data or anything to go by. I guess I could hack together something with a cap, then use a multimeter to discover the resulting DC voltage. \$\endgroup\$ – Polynomial Mar 14 '14 at 17:28

Here's a simple inverting circuit that should work, provided you can safely connect the grounds together. It presents a very light load (120K) to the signal, and should work okay up to maybe 50 or 100kHz. If you need higher frequency response, shunt R1 with a few pF.


simulate this circuit – Schematic created using CircuitLab

This assumes the input signal swings from 0 to +47.5V so sets the transition at around 20V. As Andy says, if your signal does not swing 0/47.5 then other means must be used.

Note that it is inverting, so output is low for 47.5 in and high for 0V in.

  • \$\begingroup\$ The data rate is 2Mbps, so I'd expect 2-4MHz rather than 100KHz. \$\endgroup\$ – Polynomial Mar 14 '14 at 17:19
  • \$\begingroup\$ It's possible to make that work, but I seriously doubt it's swinging 47.5V at megahertz, at a minimum that would be a serious EMI source. What do you know about the levels? If not much is the answer, an oscilloscope will tell you all you need to know. \$\endgroup\$ – Spehro Pefhany Mar 14 '14 at 17:23
  • \$\begingroup\$ Unfortunately the device in question isn't in a place I can take an oscilloscope. I'm actually doing this by-proxy (I don't have access to the device myself) and the guys on the ground have limited experience and tools. You may well be right about it being a DC offset, though. \$\endgroup\$ – Polynomial Mar 14 '14 at 17:26
  • \$\begingroup\$ It's not some kind of PoE setup is it? \$\endgroup\$ – Spehro Pefhany Mar 14 '14 at 17:28
  • 1
    \$\begingroup\$ The 48V (-48V?) made me think of phone systems. It's probably a DC offset. \$\endgroup\$ – Spehro Pefhany Mar 14 '14 at 17:31

Even though this is most likely a DC offset scenario, you can eliminate the peak/peak possibility for your sanity. Take an appropriately sized resistor, and an LED, and connect this between the signal and its reference (ground). If the LED is constantly lit, it is an offset scenario. If the LED flickers, it is a peak/peak scenario. (If you are swinging 48 volts then at some point that signal will be below ground and the LED will be off.) Hopefully your ground team can manage this much.

So, let's assume this is a DC offset scenario where the voltage could swing significantly around 47.5V. I propose using a level shifter circuit, using two resistors and an NMOS, to bring the voltage down to a 5V range. You can adjust the range by adjusting the resistor ratio. I played with this a bit in LTSpiceIV and the idea seems to hold up.


However there will still be some DC offset present, and the swing won't be optimal for a TTL UART bus, so you'll need to condition further. A single instrumentation amplifier should be able to difference the signal with a DC level, and gain it properly to 0-5 V.


Also the level shifter might give a spiky output - use a decoupling cap before the input to the instrumentation amplifier to mitigate this.

  • \$\begingroup\$ If the data is continual and above (say) 100bps, the LED will appear to be lit continuously. If the data is fully 48Vp-p then the brightness on the LED will approach half the other scenario but without both types of scenario being available how would the OP tell? Why would the signal be below ground? Also, at 2Mbps, an instrumentation amp is likely to be a poor candidate for sanitizing the data. Please consider what I've said. \$\endgroup\$ – Andy aka Mar 14 '14 at 21:32
  • \$\begingroup\$ A parts search on Digikey yields plenty of AD in-amps with sufficient -3dB for this application. Good point re: LEDs. Easier for OP to skip the sanity check. \$\endgroup\$ – neelfirst Mar 19 '14 at 23:26

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

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