ADS1298 Data Conversion to Volts

Mohd Aslam85

Other Parts Discussed in Thread: ADS1298ECGFE-PDK, ADS1298

I am trying to get my ADS1298ECGFE-PDK to work with my Arduino Uno. I’m trying to read some data and have had my serial port reading out a stream of binary data that looks very similar to the following structure: I am getting the 1100 in the first four bits that appear in my Serial Port terminal.

However, the only way to verify that this is the real ECG signal I’m getting is to actually plot out the voltages measured. I’m going to try to use Matlab to perform the binary conversion and plotting out on a graph in Matlab.

Screenshot from the datasheet:

You may ignore the yellow highlights. This screenshot is from the pdf that I reference and highlight for my usage.

Here is my understanding of how data format works:

The ground (or baseline) is at 0V, and the ECG signal will cary up and down that baseline. Assuming that I set my Vref to 2.4V via the CONFIG3 register, my signal should not exceed +/- 2.4V or it will be clipped at 7FFFFFh and 800000h. And based on the information on that same page, the data is in 2's complement so I will need to convert them back to its signed magnitude form.

Do let me know if I have interpreted it wrongly.

I did a bit of a search on the forum and found a post by a fellow TI Engineer, Tom Hendrick : http://e2e.ti.com/support/data_converters/precision_data_converters/f/73/t/102784.aspx

Hi Ismail, Sorry for the delay in getting back to you. The ADS1298 has a BTC output and that can complicate the SW a little in trying to convert hex to voltage. The easiest way to go about this (IMHO) is to take the raw code and convert the BTC data to straight binary. You can do this by 'XORing' the data with 0x800000. Full scale positive is 0x7FFFFF, XORing that with 0x800000 would give you 0xFFFFFF. Full scale negative is 0x800000, which yields 0x000000 when XORed. Your code to voltage input relationship is essentially [code-1 (in decimal)*LSB size]/gain. The LSB size is given in page 30 under the Format section as VREF/(2^23-1). If you need to compensate for polarity, simply take the result and subtract your reference voltage.

1) He says that i need to XOR the data with 0x800000. I don't quite understand the need for this. It seems like whatever is XORed with 0x800000 will yield the code itself. Eg if i XOR a codeword A with 0x800000 will yield codeword A itself.

2) He says that the final equation for XORed code to voltage should be [Code - 1(in decimal) * LSB size] / gain.
Just to clarify, when he said Code - 1 (in decimal) actually means the signed magnitude of the code in decimal, subtract 1 in decimal? Am I right to say that?

3) I can see the reason of dividing by the gain, but what does it mean by multiplying with LSB size? I see from the datasheet that LSB size is Vref/(2^23 - 1). Is that right?

Is there anything else I should look out for when it comes to data conversion?

Thanks in advanced for your help.

Regards,
Aslam

over 11 years ago

0 Ryan Andrews over 11 years ago

TI__Guru 68916 points

Hi Aslam,

Thanks for your question! You are on the right track, but I will add some additional explanation to help clarify things.

The individual channel inputs (INxP and INxN) are limited by the supplies. Typically, the inputs are biased to about mid-supply, so that each input can swing above and below this bias voltage, which may be 1.5V on a 3V single-supply or 0V on 5V bipolar supply (+/-2.5V). It is the differential input to the ADC that cannot exceed the reference voltage, +/- V_REF. Remember that there is a PGA in front of the ADC in the ADS1298, so the differential input must be limited to (+/- V_REF / Gain). Exceeding these limits will yield 0x7FFFFF on the positive side and 0x800000 on the negative side.

Per Tom's suggestion, the most straightforward way to convert your output codes back into input voltages is to XOR the data with 0x800000. All bits will remain the same except for the MSB. The range of your output codes will shift from [0x800000 - 0x7FFFFF] to [0x000000 - 0xFFFFFF].

Now that the data is in this range, we can convert the codes to volts directly using their decimal equivalent:

(Code (decimal) - 1) x LSB / Gain = Volts

The LSB size is the voltage weight of the least-significant bit. This change in voltage at the input produces a change by 1 code at the output.

I hope this helps answer your question.

Best Regards,

0 Mohd Aslam85 over 11 years ago in reply to Ryan Andrews

Prodigy 160 points

Hi, I greatly appreciate your reply and thorough explanation!

I have another question, but this is more related towards the data acquisition.

As I had mentioned earlier, I connected my Arduino to the ADS1298ECG-FE, along with an ECG simulator that generates ECG signals.

I'm using a rather trivial method to store the data. What I do is I use the Coolterm to save the data by doing a Serial.print of the byes onto the Coolterm terminal . Setting the baud rate as 115200, I print the data in the serial terminal and copy into an excel file. I understand this is rather primitive but I guess this is the only quick way to check if the data is correct, before I move on to making more advanced settings and configuration of the ADS1298.

Right now, I'm just trying to read off one channel. As such, I am expecting a total of 48 bits (24 bits of status word + 24 bits data/channel). I am trying print the data as such

Example:
11000000
00000000
00000000
11010101
10101010
00101010

The first three bytes in blue belong to the status word while the next three bytes in green belong to the Channel 1 data. The reason I have to print them vertically, byte-by-byte, is because if I try to print each cycle of data (48 bits) horizontally, I get a lot of gibberish and inconsistencies. So this is the best I can accomodate for now.

However, here comes the problem, in some cases, the bytes seem inconsistent. Sometimes it doesn't print a full byte. Instead of 8 bits, I get lesser.

11000000
0
0
11111011
11001010
11010100

11000000
0
0
1011
10100
100011

11000000
0
0
111
11000111
100101

The ones in green are what appears to be correct while the ones in red seem wrong.

1) I suspect the Coolterm only prints the significant bits. If the entire byte is 0, it only prints a single 0 as opposed to 00000000. That is why the bits get truncated.

2) Maybe it is the data rate coming from the ADS1298ECG-FE and the Coolterm baud rate not in sync?

I'm at my wits end and it has been quite frustrating. Is there something I should be doing?

Any advice would definitely be useful and appreciated.

Once again, thanks in advanced.

Best regards,
Aslam

0 Arun Joe over 9 years ago in reply to Ryan Andrews

Expert 1780 points

Hi, Why is that -1 there in the formula, (Code (decimal) - 1) x LSB / Gain = Volts?

0 Brian Pisani over 9 years ago in reply to Arun Joe

TI__Mastermind 38745 points

Hello Arun,

It is simply to correct for the single LSB error that occurs when you XOR the original value with 0x800000. For example, if your data was exactly 0x00000, meaning the input voltage was exactly zero volts, and you used a gain of 1, you could would XOR 0x800000 to get 0x800000. If you don't subtract 1, you get 8388608 x 0.2862 uV = 2.4000002862. Then subtract Vref into a signed data type and you get 0.2862 uV - which is exactly 1 bit. If you subtract the bit, you get exactly zero.

That being said, the noise of this device, and therefore the effective resolution, is greater that a single bit. Therefore if you skip the single bit subtraction step, you would never be able to actually measure the difference.

Regards,
Brian

0 Arun Joe over 9 years ago in reply to Brian Pisani

Expert 1780 points

Hi Brian, 8388608 x 0.2862 is 2400819.6096 and subtracting Vref will yield 819.6096. Please explain hoe can I use it in my code, when done as per the formula I'm getting 819.

0 Brian Pisani over 9 years ago in reply to Arun Joe

TI__Mastermind 38745 points

Hello Arun,

It should be 8388606 x 0.2862 uV. The voltage 0.2862 is measured in microvolts which is 1/1,000,000 V. If you correct for the units then you will get the same results as I did above.

To use this is your code, simply apply the forumla to every sample.

Brian

Data converters

Data converters forum

ADS1298 Data Conversion to Volts