1

I want to transmit real passband analog single ended signal with bandwidth, 0 to 7.3MHz, from board to other board. distance from these board will not be more than 15cm. what's the best choice for wire interfacing these boards?

The signal is exactly pure sinewave between 0 and 7.2MHz, it's beat signal of FMCW RADAR. Thus it may become DC (0.00000001%).

Update:

The boards are FMCW radar front end and the other is processor.

Update:

signal is 2.7 p-p, I'm not sure how to determine the systems SNR yet!(Other question has been created for this issue)

update:

you can assume SNR is good :)

Mistake confess:

O O, I made a mistake, my sampling frequency is 7.2MHz and the analog bandwidth will be less than 3.6, but I think it not too different.

Community
  • 1
mohammadsdtmnd
  • 545
  • 3
  • 17
  • 1
    The crosstalk impedance and resulting noise attenuation depends on the signal + noise spectrum and impedance ratios to achieve a required SNR. When I once had dual Eddy Current probes @ 200kHz thus the noise was in-band and needed > 60 dB but wanted 80 dB SNR. I opted for balanced differential transmission with STP and CM choke to raise CM impedance while DM source impedance was < 1 Ohm to get no trace of crosstalk. – Tony Stewart EE75 Dec 10 '18 at 15:20
  • 1
    This is super easy. Use coaxial cable. If necessary, use a video op-amp to buffer the output voltage so it can drive a 50 Ohm load. At 7 MHz, layout is not that critical. – user57037 Dec 10 '18 at 17:09
  • If you need to preserve the signal to noise ratio, you may need a low-noise amplifier. Not sure if it will be difficult to find an LNA that works from DC to 7.3 MHz, but I imagine that it could be difficult. – user57037 Dec 10 '18 at 17:13
  • What SNR do you need? What voltage PP will you transmit between the PCBs? – analogsystemsrf Dec 11 '18 at 14:30
  • @analogsystemsrf question updated. – mohammadsdtmnd Dec 11 '18 at 15:26
  • Instead of adding millions of updates you should just change your question. There's no need for me to know that the first part of your question is now wrong. – pipe Dec 12 '18 at 13:29

3 Answers3

5

If you were to consider this properly, you would need to know the signal-to-noise ratio of the actual signal leaving the board and, how much that SNR could be degraded by the wires and input buffer before it becomes unacceptable. You would also need to determine what interfering signals may be present and how much the signal may be disrupted by a common ground between PCBs with the possibility of additive (but unrelated) ground currents passing down that wire.

That would be the theoretical approach and given that we don't know those details it's the only one that can be safely offered as advice. In the absense of that information I would consider a balanced-impedance transmission, screened twisted pair and a differential receiver. But you may get away with a much simpler set-up especially if the lowest frequency isn't 0 Hz as you state (you also state "passband" and that implies the possibility of an AC coupled signal.

You won't need to impedance match if the highest frequency content of the signal is 7.3 MHz however, if the signal might be a 7.3 MHz square or rectangular wave then there may be a need.

Andy aka
  • 434,556
  • 28
  • 351
  • 777
  • Thanks for lots of information. Passband was to tell this is not complex(may be redundant!! data). I think only three way exist: 1.coax 2.ribbon 3.twisted pair 4.simply 1signal plus some gnd wire, isn't that? – mohammadsdtmnd Dec 10 '18 at 12:39
  • As I said in my answer, in the absense of details from you, the only approach I can recommend is a balanced transmission down screened twisted pair and a differential receiver. – Andy aka Dec 10 '18 at 12:50
  • 2
    Why coax is not good?, As my signal is single, diff RX and TX need time and cost! – mohammadsdtmnd Dec 10 '18 at 12:53
  • 1
    Coax isn't as good because it isn't a balanced transmission medium and, because it connects grounds at both ends, any power currents flowing through the shield/screen can induce a noise voltage onto the inner that may result in an added interference. I cannot advise anything else because there is no other information to go on (as stated in my answer). – Andy aka Dec 10 '18 at 13:10
  • Question specifies single-ended signal. Agree that differential is better for SI if implementation/cost are a non-issue. – EasyOhm Dec 11 '18 at 08:33
  • @Gonzik007 nothing wrong with a single ended signal (i.e. not a differential signal) being transmitted down one wire using a controlled series impedance and, using the other wire attached to transmit end 0 volts also through a controlled impedance. Then use a differential receiver and, it won't care that the signal is single-ended but it will see noise that has "hit" the interconnection as differential. That is why I called it a balanced-impedance transmission and not a differential transmission. Diff transmissions means lower emissions. – Andy aka Dec 11 '18 at 08:50
  • Gratefully I have some question: 1.Can you mention some part that doing this balance buffer and receiver? (If from ADI I will become more happy.) 2.How current on coax shield will induce to inner? 3.Then what's the use of coax? 4. What if I use it for prototyping and my receiver is ARM cortex M7 discovery board with single ended ADC?? 5.how you learned these & how I can (is there any rigid reference?)? – mohammadsdtmnd Dec 12 '18 at 12:29
  • Without spending a lot of time looking at your signal and then researching ADI's portfolio I can't provide a recommendation. I'm not a big fan of coax BTW. Old = have learned but forgotten where I learned it. – Andy aka Dec 12 '18 at 12:32
  • Ok just say one or two part that is now in your mind, plz. – mohammadsdtmnd Dec 12 '18 at 12:44
  • @mohammadsdtmnd there are no part numbers that spring to mind - look for integrated high speed differential amplifiers suitable for video applications - that is what springs to mind. DC accuracy may not be that good of course. Any fast op-amp should do the job too or an instrumentation amplifier with the right speed. – Andy aka Dec 13 '18 at 09:08
  • Concise answer! I have a question about balanced line, https://electronics.stackexchange.com/questions/412013/balanced-transmission-line-implementation I will be happy if you can tell your idea about that, thanks. – mohammadsdtmnd Dec 13 '18 at 11:26
3

Just go with a BNC or SMA connector and make sure to keep a 50 ohm impedance. I would suggest a coax cable.

EasyOhm
  • 4,609
  • 26
  • 35
  • 2
    If the application isn't cost sensitive this is a nice safe way to do it. But the wavelength of 7 MHz is about 42 m. Impedance control and expensive coax connectors shouldn't be needed here if cost is an issue. Ribbon cable with adequate ground connections ought to do fine. – The Photon Dec 10 '18 at 06:40
  • Signal is analog so assumed shielding is a must to avoid external aggressors. For prototyping wires with ok GND should be fine, but seems dangerous for production. – EasyOhm Dec 10 '18 at 07:21
  • @ThePhoton Gonzik007 is answered to you do you have any idea about his answer?(I asked again since he is not included you name then there will be no notification for you, Thus I notified you) – mohammadsdtmnd Dec 10 '18 at 12:05
  • Andy aka has been told in his post this: "it isn't a balanced transmission medium and, because it connects grounds at both ends, any power currents flowing through the shield/screen can induce a noise voltage onto the inner that may result in an added interference.". What's your idea? – mohammadsdtmnd Dec 11 '18 at 06:41
  • 1
    Your question specifies single-ended signal. Differential is thus not an option. Without more info and the specifications you give that's the best answer I have. I think you are relying on strangers on the internet too much for something that sounds like it needs a thorough system engineering and a mathematical assessment of the signal path. This advice is meant to steer you in the right direction, not give you a fully baked solution. Following advice on the internet without your own independent validation is a recipe for trouble. Good luck. – EasyOhm Dec 11 '18 at 08:36
  • @Gonzik007 you are confusing a differential signal transmission with a balanced transmission. A balanced transmission is impedance balanced and thus, with the correct cable (twisted/screened), a single-ended signal or a differential signal will be perfectly adequately received but, noise from external influences that hit the interconnection will be cancelled. Differential transmissions are used to reduce radiated emissions and, with the same transmission power, are no better or no worse than a single-ended transmission regards the receiving end. – Andy aka Dec 11 '18 at 08:57
  • @Andyaka You are correct that I misread balanced vs differential. Still, I assumed the transmitter is single-ended output per the question. If that is the case, I am not sure how ScTP cable helps, given that one of the signal pairs will just be GND covered by a GND foil/screen. Splitting the two GNDs to say AGND for dif pair and systemGND for the shield seems not possible as a single tie point for the aGND and systemGND 15cm away would probably result in mismatch in the two grounds, and tying off at each end seems to negate the point of splitting in the first place to use a balanced line. – EasyOhm Dec 11 '18 at 09:46
  • Read my answer [here](https://electronics.stackexchange.com/questions/295668/efficiency-of-obtaining-differential-signalling-from-a-single-ended-manufactured) - this should explain why it works. – Andy aka Dec 11 '18 at 09:57
  • Read also [this very good article](http://www.douglas-self.com/ampins/balanced/balanced.htm) written by Douglas Self. – Andy aka Dec 11 '18 at 10:03
  • I think Andy aka hate Coax, then what's coax uses?? maybe it's high frequency isolator – mohammadsdtmnd Dec 12 '18 at 12:51
0

Here is a simulation of the results of moving 1volt signal over a 15cm trace that is 1/16th inch above the Ground plane. I assume the source and load resistances are 1,000 ohms; high impedance nodes (your 15cm trace) are very vulnerable to Electric Field Interference EFI; the Loop Area (that 15cm distance and the 1/16" height) are very vulnerable to Magnetic Field Interference HFI.

Also, currents flowing in the Ground impedance will inject errors.

Here is the resultant display (with Gargoyles enabled at the top right--- HFI, EFI, GPI):

enter image description here

The Signal Noise Ratio is only 19.6dB as shown in the top right, computed at 7MHz Frequency Of Interest. Given the assumptions about the various interferers, simply call that 20dB +-5dB. Look down the right column, and find the Aggressors (RSS combined) are 18 milliVoltsRMS; notice the Signal is only 175mVRMS because of the 6dB division and the 1volt source PP.

What are the Gargoyles (the interferers, the aggressors)?

HFI magnetic field is a switching regulator, located only 10mm from the 15cm trace, with slewrate of 10 Million amps per second.

EFI electric field is 3 fold; the dominant interferer is a microcontroller at 10mm from the 15cm trace, with slewrate of 2.5GigaVolt per second. There are secondary EFI sources, as you can see in the table of Analysis Details (that Text display), from 60Hz power wiring that are by default located only 1mm (1 millimeter) away from the 15cm trace.

GPI ground interference is 0.1 amps flowing thru 0.038 ohms of Ground/RETURN path.

The trace is 15cm long, 1mm wide, placed 1/16" above the Ground. The return path is (in this version of the tool) modeled as the same resistance as the signal, to emphasize the danger to SNR if you have Ground currents.

To define the inter-stage trace (15cm long), enter the Wiring Wizard (click on TRC between the stages) and edit the default 10mm length to be 150mm.

enter image description here

And here is the built-in (it is editable) HFI magnetic field aggressor table. Notice only the switching regulator is selected.

enter image description here

analogsystemsrf
  • 33,703
  • 2
  • 18
  • 46
  • It's probably worth mentioning that if the source impedance is less than 1000 ohms, the results will get much better. – The Photon Dec 10 '18 at 15:55
  • Electric Field interference will decrease (because the same displacement current, times a lower NODE impedance, injects a smaller trash voltage. The HFI and GPI do not decrease. And the HFI dominates the Aggressor output, thus the 19dB SNR will only minimally improve. ----- I unchecked the EFI, and reran the sim (took about 1/2 second); SNR only improved to 20.5dB. – analogsystemsrf Dec 11 '18 at 03:03
  • The magnetic field trash dominates; the 15cm by 1/16" loop area is the problem. Or the closeness of the one HFI built-in source I left active: the switching power supply, located only 1cm from the signal path. – analogsystemsrf Dec 11 '18 at 03:07
  • Excellent comprehensive analysis, god bless you. But can you also analysis them for these: 1.coax 2.ribbon 3.twisted pair 4.simply 1signal plus some gnd wire. To finally have an overview on them for final decision? – mohammadsdtmnd Dec 11 '18 at 07:15
  • What are your interferers? If you keep the switching power supplies **AWAY** from the signal trace, the SNR greatly improves. The random-noise-floor is about -86dBc with1Kohm Rsource and Rload. As people have stated, at this frequency the Rload is optional, thus the 2:1 voltage division goes away. To compare, the Ribbon and the Isignal+Gnd are likely equivalent, assuming 1/16th inch spacing. TwistedPair should have some benefit in reducing HFI (I assume 20dB benefit in the tool; in/out wiring really matters). And as Andy mentions, there is no magic regarding the use of Coax. – analogsystemsrf Dec 11 '18 at 14:51