ADS1225 conversion time out of spec

Additional information: I have tried to check the power supply start-up. The 3V supply takes about 2.5ms to settle. It is a smooth capacitive charging curve, without glitches or overshoot. Please leave a comment for me if something else is missing.

Best regards,
Tobias

Hi Tobias,

Welcome to the forum!  The first thing I would do is remove the inductance (probably you have a ferrite) on the analog supply.  I would replace it with a 10 ohm resistor if you want to add some filtering.  The problem with adding any inductance on the supply lines is the restriction of startup current which is a fast rise time response and is choked by the inductance.

How are you determing the conversion timing?  Are you using an oscilloscope?  Can you send me some screen shots?

Best regards,

Bob B

Hi Bob,

thanks for helping. Indeed I am using a scope, so please find two pictures attached. I have removed the two ferrites and placed a 10R only for the AVDD, so the power supply is now connected directly. As you can see in the first picture, start-up is now slightly a bit faster. In the second picture you see the data line with the data ready signal, which appears every 17ms and has a pulse-width of about 60us.

Best regards,
Tobias

I checked also the initial self-calibration, which is performed after pulling the start pin high. In the attached picture the yellow chart (ch1) is the clk to ADS1225 and in blue (ch2) you see the data from ADS1225. As far as I can see the calibration takes more than 400ms, which is also much more than the 313ms stated in the data sheet.

Any further suggestions?

Best regards,
Tobias

Hi Tobias,

It appears that something has affected the internal programming of the device.  Have you tried removing and installing a different device?  Do you have more than one board that is exhibiting the same behavior?  Did you get the device as a sample or from another means of distribution?  Are you able to read and tell me the marking on top of the device?

Thanks,

Bob B

Hi Bob,

the QFN is not so easy to solder by hand, so I prefer to keep it in place. Yes, we have several boards with the same effect. In fact we produced some prototypes for this project here in house and purchased the components via a distributor like Digi-Key. The top mark on the one part on my desk (difficult to read with naked eye) is: ADS, 1225, T1 08K, C3KV.

In the meantime I have found out, that it works with a higher AVDD. So I applied externally 5V to AVDD and kept the rest as is, suddenly everything is fine. I will continue investigating from this point tomorrow.

Best regards,
Tobias

Hi Tobias,

I think there may be an issue with reset timing between the digital and analog portions.  I know that added inductance on the line is a problem, and the resistance may be a problem as well as this will delay the POR.  With the 3V supply at AVDD try connecting 3V directly to the pin bypassing the 10 ohm resistor to see if that makes a difference.

From documentation I've seen this may be the problem.  Unfortunately the desiger of this part is out of office this week so I can't get a definitive answer at this time.

Best regards,

Bob B

Hi Bob,

I have replaced the resistor with a solder bridge and I had also removed some of the bigger capacitors from the 3V (analog) path, to remove capacitive load. This didn't solve the problem so far. The start-up of the power supply is still looking the same, i.e. doesn't become faster than 2ms, but should be correct according to the data sheet of the regulator. The PCB trace of AVDD to supply is about 2mm longer than DVDD, but it cannot be that sensitive. Also I could not see any inductive ringing - the supply line is really quiet.

Let's try again with a higher voltage, even if this is not feasible for the system. One other question: Can we still trust the data, after the conversion time was longer than expected? We could match the algo, to adapt to the lower speed and throw a notification message.

Best regards,
Tobias

I could not reproduce the success of yesterday, using 5V as AVDD. So it doesn't work with 5V either. I am almost sure that the issue has something to do with the POR or is at least supply related, but how? What else could have an influence on the internal oscillator? I don't want to give up, but the debugging takes too long already, so resignation is maybe the best choice. Any other suggestions?

Best regards,
Tobias

Hi Tobias,

This problem is most likely a startup issue.  About the only thing to further investigate is attempting to shorten the ramp time of the supply.  I would remove the ferrite in series with the supply input of the LDO that is generating the 3.0V supply.  This isn't necessary as the LDO will do a good job of filtering by design. 

One further thing to try is to pull the START pin low instead of high.

I will have to do some more investigating, but I think the only negative result  with the startup is the internal clock frequency trim.  I will have to do some more checking to be sure. The problem is the oscillator can vary quite a bit by process variation and that is why it is trimmed. The best method for collecting data is to use an interrupt driven system if trying to read a continuous conversion result.

Best regards,

Bob B

Hi Bob,

as I said before, I have removed all ferrites and to increase the ramp time I removed some capacitors on the supply line. But it seems that the ramp time is limited by the LDO, so it will not be much better than 2ms. I could eventually replace the LDO, but then it should be a pin compatible type.

For the start pin, this is something I have tried already. I am using the following sequence: all pins low -> enable power supply -> after supply settle, set start pin high (and keep it high) -> when first conversion is done (data ready signal) read data and initiate a self-calibration -> then let it run free, by keeping the start pin high.

I will play a bit with the input buffer and try the internal temperature sensor, to see if the data is valid.

Best regards,
Tobias

I have made a test, starting the ADC in low-speed mode and changing to high-speed, after calibration. Only result I can provide is that calibration in low-speed mode takes 500ms and sampling time is 100ms. After switching to high-speed it behaves like before.

In addition I used the internal temperature sensing diodes, which should bring up 106mV at 25°C. I have 26°C currently, so it should be close to that, but reading the data I get only 54mV. Data is 0x0B1C51 and I calculate the voltage Vx = Vref * data / (2^24 - 1).

Am I doing wrong? Nothing is working?

Best regards,
Tobias

Hi Tobias,

Your calculation is incorrect for LSB size.  Full-scale range is +/-VREF/(2^24-1), so you need to multiply your result times 2 and you get 108mV.

As far as the data rate timing, it is tracking at the same decimation ratio between high speed and high resolution modes.  The problem is still the oscillator trim that gets resolved at power on.  I will attempt to slow the ramp time on my system to see if I can find out want may be happening with your system.

Best regards,

Bob B

Hi Tobias,

There is one other thing that I should also mention.  As I have only a partial schematic it wasn't obvious, but if the micro is powered from 3.3V and is active, you need to be very careful not to have a voltage applied from the digital side (or analog side inputs either) as this exceeds the absolute maximum specifications.  Leakage current can pass through the inputs via the ESD protection circuit and effect the device startup if the voltage is applied before the power has fully ramped to the ADS1225.

Best regards,

Bob B

Hi Bob,

thanks for your answers. Ah for sure it must be +/-VREF, it was so obvious. ;-) Okay, so the measurement is correct, that is an important point for me. Yes indeed the uC is 3V3 powered, but the interface is an open drain, which has 3V pull-ups. You can see the pull-ups also in my schematics print. So there should be no leakage current, just common ground. I have tried to disconnect the ADS from the microcontroller, by removing the 100R resistors and just pulling the inputs to defined level.

Btw. have a nice weekend!

Best regards,
Tobias

Hi Tobias,

It appears that you have the basic connections as they should be so that brings us back to the power up ramp time.  According to the TPS78330 datasheet on page 3 of the electrical specification table it shows that the LDO should be at 90% of VOUT typically in 500uS using a 1uF load.  It would appear that something is greatly loading your system according to the graph of the power up ramp you gave in an earlier post.  If 5 tau takes 2.5ms and tau is 0.5ms, then with a capacitive loading of 4.7uF that leaves about 100 ohms of series load.

I would review the PCB layout for any possible restrictions on the supply output or inadequate grounding.

Best regards,

Bob B

Hi Bob,

thanks for this hint. I re-checked the TPS78330 and the data-sheet. First of all, the grounding on my PCB is done with a dedicated layer (4-layer structure) so GND connection is really low impedance. Now I had placed a second LDO, as a dedicated supply for the ADS1225. I have found two deviations:
1. in the data-sheet power-up is started at 10%, while I have started at 0% Vout.
2. Data-sheet recommends Vin = Vout(nom) + 0.5V and I am using only a 3V3 input, to protect the LDO against Vin >5V. But in the current application battery voltage will not exceed 4.2V, so I could change the input.

I will try it like this and keep you updated. Attached you can see the power-up of the dedicated 3V LDO supply, with a (90%) trigger level of 2.68V.

Best regards,
Tobias

Here is the result of testing with Vin = 4V at the TPS78330 LDO. I also tried 5V input, but this doesn't help. The input power was always well settled before enabling the LDO. So if it cannot go faster, I should consider using a different regulator.

Best regards,
Tobias

Hi Bob,

first of all, sorry for letting you wait for my answer. I took a new / naked PCB and soldered it on my own, with a calibrated IR reflow profile. Then I used the same SW version for testing as before, i.e. self-calibration and continuously fast conversion mode. Now the ADS1225 finishes calibration after 250ms and each conversion takes 11ms! No way to get it out of spec anymore and it even works with the ferrites in place.

So now I have to talk to our lab who soldered the 5 devices before. I am almost sure that their profile was not good and destroyed the ICs. Thanks a lot for your great support! I think we can close the issue like that.

Best regards,
Tobias