OPA698 circuit to get 525 mV limit

Hello Michael,
Q1. I am looking at the datasheet of the ADC(MPC3422) and I do not see a G=10. Can you please confirm how you can achieve G=10? If I look at equation 4-1 of the datasheet, then the useful operating range of the ADC in its maximum gain of 8, is from -2.048/8 to 2.048/8 (I am ignoring the 1LSB since its a small value). Also, note that if you use a single-ended input you will lose 1-bit of resolution. From this it seems that if you connect the negative side of the channel to Vss (GND) then your entire useful range is only 0V-256mV. I am assuming here that the PGA translates the common-mode voltage from whatever is at its inputs to whatever is needed at the ADC inputs (most probably 1.024V). Do you agree with all the analysis here or am I missing something?

Q2. You cannot use the OPA698 in a standard buffer config. with pin 6 shorted to pin 2 as the amplifier may become unstable or show poor settling behavior. The reason for this can be seen in the top-left figure of page 7, where Rf=25 and Rc= Open, results in 6dB peaking. Instead if you use Rf=25 and Rc=175Ohm, then you can use the amplifier in unity gain configuration safely. You could also put the amplifier in G=2 and use a resistive divider circuit as shown in Figure 6-5 of the MPC datasheet. For a single-supply application as in figure 5, page 16; the limiter voltage cannot go to 0V and 525mV as you have suggested. On page 5 of the datasheet under the voltage limiter section the lowest voltage on pins5 and 8 is 1.1V...so this sets a limit. You will have to go with a split supply config. if you want to set the limits at 525mV and GND. The function of the two 715Ohm resistors on the input is to set the common-mode input voltage of the ac-coupled signal to 2.5V. Is your signal going to be ac-coupled or dc-coupled?

I think we should resolve/agree on the understanding of the ADC before we make further design decisions on the OPA698.
-Samir

Hello Samir,

RE: Q1. I mistakenly wrote G= 10. I meant to write G = 8X.  My observation is programming the gain API value to 8 amplifies the voltage by 8 resulting in the full voltage range from 0V to 0.625V instead of 0 to 5V with 1X gain. This yields higher resolution at the low voltage level that needs to be sensed on this application. The sensor is a Coaxial Dynamics Element, which is nicknamed a "slug" and it is used for measuring RF power. When running the ADC at 8X, any value higher than 0.625V returns as 0.625 as the ADC has reached the top of its input range. Higher than 0.625V will damage the ADC:  The scenario I hope to address with the OPA698 is where the user forgets to swap slugs (sensors) and applies say 1000 Watts of RF to a slug element specified for 100 Watts [1]. The slug element simply puts out move voltage beyond the .525V FS boundary. See attached graph showing sensor's Voltage (x) with respect Power(y) applied. I should note the inherit non linearity of this sensor is addressed downstream from the ADC conversion using an Nth polynomial regression model and this outside the scope of this discussion. I hope this has clarified the application scope. 

RE: Q2.  I want to test your recommendation using a split power supply config to get limiting of 525mV. I could use some help with getting the correct supporting circuit for the OPA698.  Regarding coupling: the signal from sensor is pure DC, so it is DC coupling  A 2X Gain from the OPA698 would be great! Programmable gain options on the ADC are 1X, 2X, 4X or 8X. Is the OPA698 still a good candidate? If true, what are the resistor values for pin 3, 2, 6? I would like to use a Bourns 1K trimpot for pin 8 and 5 to benchmark a limiting range of say 0 to N volts out of pin 6.

Thank you.

Mike

Reference

[1] Sensor used with line section tap, www.rfparts.com/.../bird100h.html

Michael,
I will start by answering Q2 first since the OPA698 comes before the ADC -

Q2. For split supply operation you can just use the circuit shown in the top-left figure on Page 7 with +/-5V supplies. Note that the Rc resistor is not needed in a gain of 2 and greater. From a system noise perspective it is always better to have more gain in the early stages for best SNR, so I would say, try to keep the PGA gain at 2x and OPA698 gain at 4x. In a gain of 4x the OPA698 will have about 64MHz of BW which should be sufficient for your application. In this case I would put Rf= 402Ohm and Rg = 134Ohm. I am not recommending a gain of 8x on the OPA698 because 525mV*8 = 4.2V which is more than the specified max. of the OPA698 of +/-3.9V. Based on the 4x gain of the OPA698, the output of the OPA698 should swing from 0V to 2.1V. You can set the limiter voltage to something slightly greater than that...say 2.5V in order to maintain a linear output from the OPA698. The limiter on the low-side may not matter since the sensor cannot put a negative voltage out.

Q1. Based on my response to Q2, if you put a gain of 4x on the OPA698 and 2x on the PGA, then the limiting feature of the OPA698 will help to protect the ADC/PGA downstream from it. Also, I should mention that according to the ADC datasheet in a PGA gain of 1x the full-scale range (differential) is from -VREF to +VREF which is -2.048 to +2.048....this implies a full-scale range of only 4.096V, not 5V. The other thing I don't know about the ADC is its expected input common-mode range; for example in a gain of 2x the input range of the ADC for linear behavior is +/-1.024V. However what is the common-mode range on which this differential signal can sit without clipping. Do you know this? It is important to find this out so that we can level shift the output of the OPA698 appropriately in order to stay within the ADCs common-mode range. Another thing I wouldn't worry about destroying the ADC, since its abs max. limit is 300mV above and below its supply voltage (5V) and GND. Like I mentioned in Q2 above, the limiting voltage will protect the max. voltage on the high side and the sensors inability to put something below 0V will protect it on the low side.

-Samir

The ADC Pi Plus uses a voltage divider on each input with the negative pin tied to ground which on the MCP3424 sets the ADC to single ended mode.  The voltage divider consists of a 10K and 6.8K resistor which divides the input by 2.4706.

The common-mode range at 1x gain without clipping is 0 - 5.058V,
2x gain is 0 - 2.529V,
4x gain is 0 - 1.264V
8x gain is 0 - 0.632V

According to the datasheet for the MCP3424 the maximum voltage at any input is Vss-0.3 to Vdd+0.3.  The Vdd voltage on the ADC Pi Plus is 5V so the inputs on the chip should take a voltage of up to 5.3V but as a voltage divider is being used this means the maximum voltage is 13.09V.  I just tried applying 13V to an input on an ADC Pi and it still worked afterwards but I wouldn't recommend doing it for too long.  I know from experience that if you put 20V on an input by accident the chip emits blue smoke.

With the gain set to a higher value I would assume that the maximum input would be divided by that gain factor so:

2x gain is 6.54V
4x gain is 3.27V
8x gain is 1.63V

Samir, please provide link to document referencing "top-left figure on Page 7". In light of Andrew's common-mode finding, is there a way to avoid a dual power supply +/-5V design? Prior to this voltage limit protection design goal, I had been successfully used the ADC Pi Plus board's +5V reference with a unity gain op amp at +5V rail (without limit protection). I really like Samir's upfront 4X gain OPA689 config to achieve the voltage limit goal. i.e., AV = 1 + Rf/Rg, where Rf = 402 Ohms and Rg = 134; 402/134 = 3 + 1 yielding 4X gain pre-ADC input. 

Michael,

  I was referring to the datasheet of the OPA698 (http://www.ti.com/lit/ds/symlink/opa698.pdf ) for the circuit design. The issue with single supply is that you will need an amplifier that has a common-mode range that includes GND and an amplifier that can swing on the output all the way to the negative rail. This requirement comes from the fact that the lower end of the signal from the slug is 0V. The OPA698 is incapable of swinging to the negative rail. Also, I don't think a limiting amplifier like the OPA698 is necessary here, since it works on the same 5V rail as the ADC and its output swing will be limited by the amplifiers capability and its ESD protection to the positive supply.

Now, with regards to an alternate amplifier, the OPA836 may be a good choice, however it can only swing to 150mV of its negative rail on its output, which means you are going to lose some of your dynamic range. Another simple alternative is to put the amplifier in an inverting configuration and level shift the output, but I don't know if your sensor (slug) can drive a load. yet another alternative is to put the amplifier in non-inverting configuration, but you will need a 2nd amplifier to create a reference voltage for the level shift and that reference voltage is going to be a negative voltage which you don't seem to have.

Can you please let me know which buffer you used before? I may be missing something in my understanding of the circuit.

-Samir

Samir,

I am currently using the LTC1152 in a simple rail-to-rail buffer configuration.The 5V reference on the ADC Pi Plus board is connected to pin 7 and pin 4 is tied to gnd rail.This single supply operation accepts the raw mV signal from the sensor's micro-coaxial center conductor is applied to the non-inverting input pin 3 and the output pin 6 feeds back to the inverting input pin2. The outer silver coaxial mesh shield of sensor line is tied to gnd rail. This configuration provides unity gain.

For additional clarity, please notice U1 Pi 9 and 10 on the ADC Pi Plus schematic showing power supply reference, ground and 8 inputs: https://www.abelectronics.co.uk/docs/stock/raspberrypi/adcpiplus/adc-pi-plus-schematic.pdf

Your point is well taken regarding the OPA836 inability to swing down to less than 150mV above the negative rail, thereby preventing the ADC from measuring any voltages below 150mV. To achieve fault tolerant voltage limiting, I am open to using a dual power supply with 4X gain on the raw mV signal but seek a final part recommendation/configuration so that I kickoff evaluation.  

I cannot ignore the worst case user scenario that has been discussed. Downstream, are the input pins on the MCP324 that do sport protection diodes, which can sink up to 2mA of current. Likewise, out on ADC Pi board it has a 10K resistor in series with current limiting, which will tolerate input overdrive. I appreciate these safety nets but I would rather solve the problem using best TI signal conditioning practices thereby avoiding the breach of these downstream safety nets.

Thank you for all your help.

~Michael 

Hello Michael,

I understand wanting to be extra cautious in order to protect the ADC. If you can go to split supplies, then the OPA698 running on +/-5V supplies in a gain of 4V/V(12dB) with the limiting voltages tightly controlled will give you this. Another advantage of the OPA698 (apart from protection) is that in case of a fault from the sensor causing the amplifier to go open loop, the limiting circuit works to recover from the fault a lot faster than normal amplifiers which do not have this feature.

-Samir

Samir,

Yes, I have definitely favored the OPA698 throughout our discussion. Thank you for helping me with this part selection verification task. We are ordering OPA698s for eval.

Sincerely,

Michael