Single supply sensor measurement question

Hello,

The executive summary for this thread is: single supply measurements of signals with negative components.

I am currently working on a battery powered design which will be used to collect high impedance, generally ground referenced, sensor voltages from outside systems. The analog front end of this design would closely resemble an oscope: high impedance attenuation(10:1 and 100:1) , clipper, high input impedance buffer, pga, etc.

I would prefer to not split the battery supply in order to provide a virtual ground for purposes of measuring signals that go below ground. This is because I do not want to have to run my other regulators off this virtual ground (like the 3.3 and 1.8 required for fpga, mcu, pga, etc). Someone correct me if this is a flimsy reason. I have not yet calculated out power requirements, so it may really be a non issue.

This means adding a dc offset just after the signal hits the board. I have tried this a number of ways. To name a few:
1) Connect the ground of the sensor's system to a reference voltage on the battery powered system. I have tried this a couple different ways, all of which work fine but the problem is what happens if the sensor system's ground was the same ground as the computer which may be attached via jtag programmer or usb. The reference voltage would get shorted. Depedning on the attenuator arrangement it either causes damage or simply ruins the measurements.

2) Use a regular voltage divider as the attenuator, but instead of connecting it to ground connect it to a reference voltage instead. The problem with this is that the reference voltage will be attenuated by the attenuator, which means I need a seperate reference for each attenuation mode (10:1 vs 100:1).

3) Traditional summing opamp, the problem is that I dont know of a configuration that will allow me to maintain the high input impedance of the input buffer.


Currently my favorite option is to use a differential attenuator, connect one output of the attenuator to a reference voltage, send this differential signal to an instrumentation amp which then allows me to adjust the output common mode down to levels useful with a 3.3v pga. This doesnt resolve the ground path issue, and it also means I need to double the number of relays used to switch attenuators.

If I were to rank system requirements I would say: accuracy/robusteness, complexity, then cost.

Any help would be appreciated.
Thanks,
Jay

What's the purpose of the attenuators? Are some of the sensor signals larger than the dynamic range of your system or are you trying to amplify low level signals?

Correct, some signals (but not all) will be larger than the dynamic range of the system. +-1V, +-50V, +-100V, etc.

q2418130103p said:

I would prefer to not split the battery supply in order to provide a virtual ground for purposes of measuring signals that go below ground. This is because I do not want to have to run my other regulators off this virtual ground (like the 3.3 and 1.8 required for fpga, mcu, pga, etc). Someone correct me if this is a flimsy reason. I have not yet calculated out power requirements, so it may really be a non issue.

Click to expand...

This is definitely not a flimsy reason. If you placed two battery cells in series and drew power for most of your devices from only one of them, then you would have some challenging battery charge issues to deal with.

What you really want here is the ability to generate a negative voltage so the 1st stage amplifier can have its inputs below GND, thus allowing your signal to be GND-referenced. You can easily generate a negative rail using a charge pump, such as this one. This charge pump can also power subsequent amplifiers, if you choose to keep the signal GND-referenced for a while.

The package of your device should form a metallic shield, attached to the device's GND to avoid interference. Your device's GND will be attached to the GND of the device under test (DUT).

When plugging the device into a computer, you suggest that its GND might also be attached to the computer GND. Here we happen upon an interesting situation: Computer GND is attached through a USB cable, to your device, then through another cable to the DUT's GND, which is then attached back to Computer GND through the wiring of the building. This is often called a "ground loop," and it can wreak havoc in sensitive measurements. Why? Well, the GND of one power strip in your house will be at a slightly different potential than the GND of another (especially if somebody turns on a high-power device on one power strip, such as a hair dryer). This difference in GND potentials oscillates at the line frequency. This difference in GND potentials will also appear across the wires between the computer and the DUT; since your device's GND lies midway between the computer GND to the DUT GND, your device's GND (and thus, the inputs to your amplifiers) will be an incorrect representation of the actual signal's GND. The measurement will be thrown off.

To avoid this, you should isolate your device from the computer. There are also several options here. One is to get USB isolation for your USB port, and if you really want to take measurements with the JTAG attached, **broken link removed**. Another option is to convert the signal, and then transfer it between two sections of your PCB which are riding on the two different GNDs using an isolated I2C transceiver (or, if data will only flow one direction, you can use use an optocoupler). (You can isolate other protocols... SPI, UART, etc.) This option means that you also need isolated power delivery. USB and JTAG provide their own power anyway, so this might not be a problem. Another option, one that is often used for multimeters, is to build your own transceiver out of IR LEDs and phototransistors.

Hope this helps!

I don't see any easy way to do what you need with a single battery supply.

One other option is to use a small DC-DC converter to generate a plus and minus voltage for the input buffer amp power.

I propose you use a dual battery operated measurement device with 200dB CMRR when you are using high impedance signal sources and high impedance loads (OA) you can use large divider ratio ( 10~30:1) then amplify with a single resistor to choose any gain you want 1~1000.

Why?

Because E-field from AC mains will be high; and ESD events can be 10~25KV E-fields from just connecting the cables to the sensor..
Just touch any 10MΩ*probe and you can get 80Vac from a 120Vac line. You want possibly 100:1 to 1000:1 SNR on all scales from +/-1 to +/- 100V..


How?

1. First choose a 3 OA configured IA (Instrumentation Amp) with 130 to 140db CMMR over range.

2. Choose signal bandwidth.. unless just DC and add active LPF to differential IA chip feedback and plan on using some differential LPF

3. Choose a CM choke that will provide 60~70 dB CM rejection over 50/60Hz and up to LPF cutoff.. If not go for 40dB... THis will be the highest permeability ferrite avail. often found on USB cords, mic cords, charger DC cords and come in any sizes & shapes from wound torroid.. dual donut hole to clam-shell donut torroids. (best) All your wires can be small wire gauge or even Litz wire or stranded 24 AWG to 26AWG in twisted pair shielded and run through or around donut CM choke several turns or use thin wires without jacket to wrap around CM choke then seal with PolyUrethane potting or Epoxy. or for a prototype PL400 sub-floor adhesive that will turn rock hard and is cheap. Then run cable jacket over twisted pair bundled.. or consider using subminiature Coax if you opt out of the differential design ( Cable suppliers use clamshell types and then mold in hard PU material or hard PVC to protect brittle ferrite.

BTW: You cant use ethernet magnetics because they don't work well at 60Hz as well as they work at 1:10:100:1000 MHz.. and you need CMRR for 60Hz.

4. Then choose dual batteries such as 3V Lithium $1.5 ea CR123 socket $1.5/pc by the dozen at Home Depot.

5. Then get Battery receptacles for CR123.

6. Then design gain and LPF active filters using your 200 dB CMRR precision Instrumentation Amplifiers ( commonly used for low EEG signals on noisy body with galvanic skin DC offsets. with 50uV and > 40dB SNR

7. Look here for a wide selection of I.A.'s https://www.linear.com/products/instrumentation_amplifiers

8. Look here for Lithium batteries 1500 mA-hr https://www.homedepot.com/buy/tools...-lithium-3-volt-batteries-12-pack-188349.html

9. Don't forget to prevent battery ejection lids.

10. Consider putting it in a tinned brass shield soldered onto PWB and then mount inside a plastic box preferably conductive ESD/EOS safe plastic...

11. Ensure you put in TVS to protect the IC's on every signal input and choose your termination resistance to be whatever you want 10MΩ probe like with tuned cap to offset cable capacitance or whatever going into 2MΩ input load for 50:1 divider.

12. Use good plastic low ESR polyester caps across battery (1~100mΩ) that are not microphonic like ceramic... or cheap alum. high temp. 2nd best.

13. Dont forget switch and power indicator using 3mA of a 20Cd white or Blue LED that runs off 3.2V .. 20mA will blind you... so 3 Vdrop over 3mA will plenty bright. with 1KΩ.. To get fancy make a low voltage indicator using Schmitt trigger inverter which changes pulse repetition rate with battery voltage... any inverter cap. to ground and 1~10MΩ feedback resistor on low leakage cap. say 1Hz blink.. use pull-up input bias resistor for 10% duty cycle on full charge and 1% on end of battery life 1 to 10Hz .... or better yet use two blue LEDs in series on 6V which is ~6.5V new and ~5.5 dead so LEDS go out when battery is dead and use 470Ω i series with 2 LEDs. 2~3mA new. battery 0.5mA dead... Close enough for 1st estimate.
You got 1500mA-hr...

14. Then test your design with CM noise pulse and differential signals with zero input. for SNR and all the other tests.. Vbat noise sensitivity. drive a signal to see PSRR. etc etc... calc battery life... .. write user manual...and theory of operation spec.

{Farmout design layout so a good analog PWB designer.} or do yourself.

15. When happy.. go for a soda.

https://www.linear.com/product/LT1167

OK maybe the 200dB is ambitious.. but I hope you got my point.. Overdesign and hope it works at worst case.
10 kV CM /1uV Diff Mode = 200dB I guess that is overkill...if you just measuring big signals such as +/1V and +/-100V but then my assumption may have been wrong that you needed 10 bit resolution and accuracy.
Comments?
Total cost.. cheap... Your time... expensive...

High impedance Scope like accuracy on long probes with high impedance source.

Ref my selection of high CMRR I.A's

Product Channels (#) Vos (uV) Vos TC (uV/C) CMRR (dB) Vs Min (V) Vs Max (V) Comments Packages
LT1167 1 15 0.05 140 4.6 40 Precision, Low Bias Current IA DIP-8,SO-8 $6.45 ea
LT1167-1 1 15 0.05 140 4.6 40 Precision, Low Bias Current IA, Controlled Input Current at High CM Volt and Hi Z SO-8
LT1168 1 15 0.05 140 4.6 40 Precision IA, Low Bias Current, Low Power DIP-8,SO-8
LT1920 1 30 1 140 4.6 40 Resistor Programmable IA DIP-8,SO-8

Referring to the original question. A small DC/DC converter (either inductor based or charge pump) is usually the most flexible and least coast/best performance solution.

The DC-DC convertor which is isolated in my Laptop charger is the biggest source of noise on my external mic. Unless I ground which defeats the safety isolation due to very high CM noise levels on DC cable.. Once it is disconnected, it's silent and truly floating without all those ingress caps across the transformer. About 50% of al laptops have at least some faint hum when external 3rd party mics are used. ( poor CMRR of audio mic pre-amp and amp in PC.. )

best I can get is 26 dB SNR ungrounded and 46 dB SNR grounded. Using spectrum analyzer in Cool Edit Pro 2 (now Audacity) on Gateway 5 yr old lappie using XP Pro. Same in Linux...

But a DC-DC with isolation AND a good CM Choke is cheap, but still dirty since the the CMRR of the OpAmp converts unrejected CM into a differential signal of 50/60Hz hum but in the case of the DC-DC SMPS it make be supersonic and easily filtered.... ahh but then you still have high impedance probe ingress....

The DC-DC convertor which is isolated in my Laptop charger is the biggest source of noise on my external mic.

Click to expand...

Yes, it's generating common mode noise because it's isolated. Without additional grounding, the common mode voltage of the DC output is centered between neutral and phase by the Y-capacitors with some added switching noise.

For the present problem, we would consider a non-isolated inverting converter, which can be perfectly filtered if necessary.

Wow, I did not expect this number of detailed responses. I read them all thoroughly. Thanks everyone. I am going to try and respond to as many bits as I can.

1) DC-DC / Charge Pumps: I initially wanted to find a way to do the entire design without a switching regulator simply because I didn't want to waste the cleanliness of having a battery but this is basically lazyness since, as mentioned earlier, it can be filtered just fine. It seems like it may be more trouble then it is worth to avoid dual supplies, so I have given in.

2) Isolated pc connections: Thanks for those links. I dont really have an intention of having a pc connected while measuring anything, but I want to reserve the ability to do it without damaging anything or creating completely worthless measurements. But since I will now keep a normal ground reference the possibilty of damage is greatly reduced, obviously the ground loop will not be removed but I can always lift the ground from the sensor(s) while I am connected to a pc.

3) Using two batteries: I like this idea because it gives the best of both worlds, dual supply and clean battery lines. But I dont like the idea of being able to drain down one battery before the other, which will likely happen becuase there are far more single supply devices in the design than dual supply. Any thoughts on that?

4) CM choke: Are you refering to the cables used to connect to the sensor? What is wrong with using a board mounted CM choke? I have already decided to opt out of differential, I will likely be using regular coax or submin coax as you suggested.

4) General configuration (for clarification): My intended (and now somewhat modified) configuration was (is) as follows, in order of signal flow.
- 1 MOhm||10pF input impedance, relay selectacble 10:1 or 100:1 single ended passive attenuator
- Dual supplied unity gain buffer/level shifter (instrumentation amp or otherwise, several MOhm input impendace or higher, 0.1dB flatness DC-100MHz, low offset)
- PGA (SPI programmable, differential in and out, probably single supply)
- Filter (either here, before the pga, or as part of the pga)
- ADC (10 bit, differential input, likely with a slectacble +-1 or +-2 Vp2p full scale range, single supply, 100MSPS)
- FPGA||MCU||Bluetooth transceiver plus other misc digital components
I have this diagramed out as functional blocks if anyone cares to see that.


5) High CMMR IA/OA: I would love to find a unit with 140dB CMRR or better, but finding one that has the bandwidth seem difficult. A number like 120dB may be practical, even that seems hard to find. Something like the OPA656 only has 80dB CMRR and an offset of 250uV

6) Enclosures: I hadnt given much though to enclosures yet, since I hadnt gotten an idea of how large the board will be. But I will remeber your suggestions. The design will incorprate an RF transmitter, so I will have to keep that in mind when slecting the enclosure.