Connector Quandary? Data Acquisition device at 500V

[max_torque]

I'm building a small, multichannel ( 8 channel ) analogue data acquisition device, using a 20b ADc with built in PGA, stuffing data into an ARM, that then buffers and sends that data onwards. So far, so normal. The input range of the device is +-10V, again, normal ,however, i'm going to include a significant common mode handling capability, although the channels are not totally galvanically isolated from each other.  Using the INA149 unity gain buffer, i can measure that +-10V on top of up to 275V, and that buffer will survive up to 500V for short periods.  That basically gives the system a built in level of robustness and the ability to handle differential measurements across a system with a range of voltages or ground potentials.

My question is what to use for the input connectors?  The device will be fairly small, so i want small form factor connectors, and those connectors want to be reasonably robust, but also because of the potential for, er high potentials on the contacts (  ) i guess they need to be suitably insulating, ie not present conductive surfaces to the end user.  Classical instruments use standard metal BNC sockets, but the outside is always mains earth referenced, whereas mine will be (effectively) floating.

Sample rates won't be that high, perhaps 400 ksps across 8 channels, so 50 ksps per channel or so, with an analogue bandwidth of something like 10 to 20 Khz i think.

Can anyone recommend a nice connector to use?  I'm wondering about something from LEMO, ideally with a direct to Pcb footprint to make assy easy, obviously cost of that connector will be high, but they make nice small ones ie:




Are their cheaper alternatives anyone has used?  Actually wondering about a nice gold plated dsub type, although i don't know the bandwidth of those sorts of connectors (probably ok at just 30Khz)


Ideas anyone??? 

[reboots]

Actually wondering about a nice gold plated dsub type, although i don't know the bandwidth of those sorts of connectors (probably ok at just 30Khz)

D-sub seems fine for your application. Common applications have included Ethernet (10MHz AUI) and VGA (150MHz+). Check the datasheet for your selected part, since official voltage ratings do vary. Amphenol DE09P064TXLF is rated for 1000Vrms.

https://www.digikey.com/product-detail/en/amphenol-icc-fci/DE09P064TXLF/609-1524-ND/1001838

I think it would be good practice to put the male connector (exposed pins) on the DAQ side and female connector (shrouded pins) on the cable side, to prevent a shock hazard if the cable is disconnected while high voltage is applied.

[jbb]

Hold on...

The INA149 has limited Common Mode Rejection Ratio. It’s 90dB or higher; at 275V input, that’s 8.7mV.  For a range of +-10V, that’s 1/2300 of the ADC input, so your useful accuracy would be about 11b.

You need to think carefully about whether this is good enough. Can you filter it out? Can you calibrate it out? I’m guessing from the 29b ADC you mention that you want quite some precision and accuracy.

Maybe you would do better to float the ADC etc. (use isolated power supply and communications) so that it can float up to the measurement potential? This is how good bench meters do it.

PS: sorry for threadjacking.

[max_torque]

The Common mode rejection is mainly for "device robustness" and not really in place to allow continuous high precision measurements of small differential signals on top of large Common mode ones.   The 20bit adc is mainly because that adc comes with a nice PGA and front end multiplexer, which in conjunction with the INA149 means a simple "front end".  Most of my applications for this device will have CM voltages of less than 50V.

 For the applications i have in mind, mV levels of absolute accuracy are plenty good enough.  Ideally, i would use a fully galanvically isolated front end for each individual channel, but this rapidly gets complex and really pretty un-necessary, when the INA149 provides enough impedance to provide a decent level of robustness for the device itself. The unit includes full galvanic isolation in the datastream after the acquisition & processing stages to furnish a suitable level of user safety.

In effect the INA149 is a nice easy "half way house". (im not making lots of these devices, so the highish cost of that amplifier is not really an issue)

This does lead me onto one interesting point, in that the high frequency CMRR of the INA149 is somewhat limited, and my device may be used in situations where there are high frequency common mode content (up to around 1Mhz) and as such, i wondered if the high frequency CMRR could be boosted with an external common mode choke?  If this a viable option do you think?  I guess to some extent it will depend on the choke chosen and how matched any of those parts actually are to each other?

[max_torque]

PS. no need for apologies for the "hijack" as it was very relevant (and interesting too!  )

I actually have a previous version of this device that has a hand-rolled front end, that has worked very well in the end application, but is a PITA to source and trim, requiring precision resistors and trimmers and an "end of line"  (well end of my bench anyway....) calibration process.  The idea with the use of the INA device was to try to build a bit more of the front end with pre-trimmed OTS devices to make building them easier (at a small loss of precision that is un-important)

[capt bullshot]

This does lead me onto one interesting point, in that the high frequency CMRR of the INA149 is somewhat limited, and my device may be used in situations where there are high frequency common mode content (up to around 1Mhz) and as such, i wondered if the high frequency CMRR could be boosted with an external common mode choke?  If this a viable option do you think?  I guess to some extent it will depend on the choke chosen and how matched any of those parts actually are to each other?

Yes, a common mode choke can boost CMRR at higher frequencies, but is highly dependent on the sink impedance (the input impedance of the INA149). Imagine a voltage divider consisting of the common mode choke's inductance and the the input resistance of the INA and do the calculations. It's quite difficult for an inductor to reach a high enough impedance in comparison to the high resistance of that INA, a common mode choke works way much better if it is terminated by a rather low common mode resistance.