4 percent gain error with ADS1602

Michael,


I just have a a few questions:

Can you start by measuring the gain error with a DC source on the input (instead of an AC source). Of course, you'll need to wait for the digital filter to settle. This may require you to throw away data at the start of the measurement. But it is important to only collect settled data.

The 33220A is a function generator, so I'm assuming you're using it as an input source. If you're using some other meter (such as a 34401A), then it will have sufficient accuracy. However, if you're using the 33220A as the source, the error is 1% of full scale (with an absolute error on top). Let me know how you are making the measurements. Just remember, you will need to measure back the both the input and the reference voltages.

If you are already using a DC source, what are you using?

Are you using the buffer circuits shown in the datasheet (Figures 41 and 42) with the OPA2822 or the THS4503? They may impart some small gain error. This includes the reference buffer of Figure 45. Granted, this is likely not going to be the 4%, but I still thought I'd ask.

However, going without the buffers may also be a problem because the input impedance of the analog inputs and reference inputs are rather low. Any series impedance may be a source of gain error. Looking at the diagrams for Figure 39 and Figure 41, the input impedance may be several kOhms, while the reference input impedance may be under 200Ohms.

You can test if the input impedance and reference input impedance are a factor by slowing down the clock. If the gain error drops as the clock is dropped then this is likely the problem.

Do you have the recommended bypass capacitors (for the reference, supplies, and any recommended for external circuitry)?


Joseph Wu

Michael,


First congratulations on your new son. If I have any important advice for you, I would recommend that you catch up on your sleep while you can!

As for the ADS1602, I need much more information to help you debug your problem. Many of the questions I'd asked in the previous post were posed to get some information about your particular system and see if there are any problems in calculation or in the method of using the part. Gain error listed in the datasheet is typically 0.25% and while there is no max, I would expect that the typical number is representative of the value and it is tested in our final test.

When you have a chance, look through my questions and answer them specifically, I might have more insight once I get some of those answers.

In the meantime, since it looks like the gain error is the same for DC and AC signals, let's concentrate on the DC measurement. This is typically a more stable and accurate measurement. Since the Agilent 33220A has an accurate meter similar to the 34401A, can you measure the reference voltage with the meter and the input signal with the meter to get the expected code? Comparing this with the output code average gives us an exact value for the gain error.

Remember to measure the inputs of the ADC directly at the inputs. The input impedance of the ADC is rather low (several kOhms with a differential and common-mode component) and any series resistance might have some impact on the measurement. I recommend this also for the reference. The reference input may have a low impedance. These low input impedances come from the sampling nature of the ADC. The inputs are sampled and discharged through the ADC's sampling capacitors.

I'd also like to know what you are using for your reference and for the inputs. It's important to know if they are stable, low noise, and have no periodic noise or oscillations.

Let me know if you have any further questions. Remember - it helps to be descriptive and I'll know more about how to solve your problem that way.


Joseph Wu

Michael,

In your schematic, you have several OPA1612s (U304, U305) listed in the circuit. Are these really the OPA2822s? Also, are they running at +5V single supply?

I think the supplies used for the OPA2822 should be higher and bipolar. If this is not the case, the inputs of +1V and +4V would violate the input common mode range. With a +5V supply, the least positive input voltage is 1.5V and the most positive input voltage is 3.5V (see the OPA2822 datasheet on page 4). You can see the setup in the ADS1602EVM user guide in its schematic as well.

Let me know. I'm still looking through the schematic for issues. I've seen a few small things, but this is the one that first struck me.

Joseph Wu

Michael,


I don't think the TLC4502 is that much better, it's input common-mode range goes all the way to the negative supply, but the highest the input common-mode range goes is to 2.7V on a 5V single supply.

However, let's just go through some of the comments that I have for the schematic.

1. As I'd mentioned before, the input impedance on the ADS1602 is rather low. If you an input driving the device, then the 51Ohm resistors in series with the inputs (R300 and R301) create a resistor divider to the inputs. If you measure the input at the pins as you mention, this will help you get an accurate gain error. However, it's something you'll need to consider in your system.

2. The input impedance is also a factor for the reference inputs as well. In the reference buffer you construct, you have the same series resistance in R310, R312, and R314. This also is a factor in the gain error as well. If you are measuring the reference, then you'll also need to measure at the pins to eliminate their effect.

3. For U304 and U305, you really should use OPA2822s. There are a couple of reasons for this. First, they were likely proven out by the development group as input buffers through testing. The bandwidth is 400MHz, which is necessary because the ADS1602 samples the input at a rate of 40MHz, or every 25ns. You need a large amount of bandwidth to get the input sampling to settle correctly. The OPA1632 and TLC4502 are likely not fast enough. Additionally, the capacitors used to supply charge in the sampling of the reference (C306, C309, C311) are very large and can make some op-amps unstable. While the OPA2822 might generally have a hard time driving a capacitive load, it has been tested with the ADS1602EVM. Again you should have a larger supply from both the negative and positive sides or the common-mode range comes up short.

4. What are the values and accuracy of the resistors used to setup the reference voltages? This would be the resistors connected to the REF5050.

One other thing that I would recommend is that there is an ADS1602EVM and the user guide has extra comments about the use of the device and schematics that show some of how the device was tested and can be used. You can find the link below.

http://focus.ti.com/docs/toolsw/folders/print/ads1602evm.html

Also, I'll need some more information to help debug this. For example I'll need the output codes when certain certain voltages are applied. For example:

What is the output code from the ADS1602 in operation, and with what inputs? So give me the code, the exact voltage measurement between pins 4 and 5 of the device, the measurement between V_IN+ and V_IN-, the measurement of +4V_REF, +2V5_REF, and +1V_REF, and the measurements of the reference inputs of the ADS1602 at the pins of VREFP, VMID, and VREFN. This gives a more accurate view of the system. ADC behavior, and any circuit irregularities. Make sure the voltages are precise, this can be used to calculate the input currents into the device as well.

Joseph Wu