need help on sigma delta ADC?

[7rots51]

ad7730 site:edaboard.com

Hello

I have worked with analog device AD77 series(like as AD7714,15,19,30)

But I have problem with them in noisy environment and in sensors that connected to it with long cables ,Sometimes they hang or lock, if I power off the card then power on it ,it works,I do not know what to do with them,
They are good ADCs ,because they have low pass filter and PGA in them and I can connect the sensors directly to them,but it seems they are not reliable.
please help me on this subject and share your experience!

bye

 

[Bus Master]

ad7730,latched

Hi there,

I doubt it's a matter of ADC's reliability. These stuff are completely tested upon menufacturing. Are you sure your system is functioning correctly, you do NOT have race conditions, glitches, power supply filteration problems. Are you using a watch dog? if not, enable it (if you have a built in one in your microcontroller) and check if you still have the fault.

Hope this helps.

Yours,

 

[7rots51]

pga in sigma delta adc

hello

I have WDT in circuit and the supply is provided by a DC/DC converter that its output is a regulated 5VDC,I used 100n and 10 uf TAN capacitors for power pins of ADC and the supply filtering is good.

But in noisy field I think it is difficult to use sigma delta ,and the ADC lost its sequence.

Only with power off and then on after several seconds everything becomes ok.

Regards

 

[zoovy]

sigma delta adc diode temperature

hello

I have WDT in circuit and the supply is provided by a DC/DC converter that its output is a regulated 5VDC,I used 100n and 10 uf TAN capacitors for power pins of ADC and the supply filtering is good.

But in noisy field I think it is difficult to use sigma delta ,and the ADC lost its sequence.

Only with power off and then on after several seconds everything becomes ok.

Regards

7rots51,
Can you give me alittle more explanation in defining the problem?
I want you to break down the described problem.
Here are some questions. I am sure you are knowledgable. But sometimes seemingly silly questions might get you thinking something you didn't look for.
1) Is the latchup really the ADC? How did you determine that?
2) When it was latched up, could you read/write the status/command etc. registers and get some meaningful information?
3) Can you spare one of the ADC inputs to some reference voltage so that you can test your theory that the ADC is latched up?(switch to that port do conversion and check result type of verification)
4) Which side has long wires (ADC to uP or ADC to the sensor?)
5) If the long wires are on the ADC<->sensor side, do you have protection diodes? (Preferrably clipping the voltage below the supply voltages and/or within the range of the input.)
6) Is there any signal that is floating on the ADC?
7) Last, let us know about the results!!!

Zoovy

 

[7rots51]

ad7714 filter high register

Hi zoovy

this is answer to your questions:
1: before each conversion for example for AD7714 ,I read filter high register if I readback the programmed value I know that the communication between ADC and Micro is OK,else the ADC does not respond or respond with error,then micro waits and WDT reset the micro and also the ADC ,at begining of micro program ADC is initialized and self calibrated and then set to get samples

2: yes ,because the complex chips on the board is a micro and the ADC other components are some TTL logic and DC/DC and optocouplers and RS485 line driver.
when WDT resets the micro the micro again performs the initial of program until reach to test routine of ADC ,and because ADC does not respond it wait here and WDT timeout will occur again and again.

3. when it hangs the sequence of communication will be lost and the test routine says there is an error in ADC.
If I turn the power off then I turn on the power of the card everything becomes ok.But WDT reset on ADC and Micro has no effect

4: sensor to ADC is long

5: I used four diode for clamping (two at each leg)and two resistor for current limiting on input of ADC and I used three capacitor for input filter.
the input for example is from a THC sensor and I used two 100n capacitor for common mode noise rejection and one 1uf cap for differential noise rejection.

6: one leg of input sensor is biased to middle of ADC suply and the common mode voltage is in the range.

7.all you said I had done but the problem exist!!
I do not know why some cards does not work properly.
BTW: some cards work ok in safe area but some may crash in noisy area after some days!

the situation is industerial(the place that I used these cards)

bye

 

[mystery]

adi adc latchup

What is the difference between sigma delta adc's and slope adc's ?

 

[techie]

AD7730

I beleive you are talking about AD7730. I have used this with exactly the same problem as you have described. In fact, one of my products was scrapped due to this problem. The problem was that whenever there was a some switching noise in the system from relays, the AD7730 just hangs up, gets reset or loses values of it configuration registers. The only way to recover was either a system reset or somehow determining this status and reinitializing the AD7730. I did try both options but this meant loss of precious real time. As I was using the chip in a batching controller, this was entirely unacceptable. No amount of filtering, shielding, etc etc could get me out of the trouble. So I had to convert the design on some other chip after all the time-to-market loss and customer annoyance.

 

[7rots51]

hi techie

I have the problems that you have but my system does not work unless I switch the power off and on,this is funny,how can I solve this problem?

I want to use other ADCs for future use ,But I do not know TI or intersil or maxim sigma delta ADCs have these problems or not??
Please if some one worked with them respond.

The main advantage of sigma delta is its PGA and low pass filter and burnout detection,these features enable us to connect sensors directly to ADC without large amount of signal conditioning circuits and amplifiers,But these faults are dangerous for industerial projects ,because they must work 24 hour/day,7day/week.

please give comments on sigma delta problems and how to solvw the problems! :?:

 

[zoovy]

more questions

7rots51
Here are some more questions
8) The device seems to be quite complicated. To eliminate any software problems can you provide reset from the micro as a software control? That way we can eliminate software mistakes if any. So you initialization routine will reset the IC, initialize then run it.
9) Have you checked your power supply pins? More than anything most of these sensitive devices fail due to power supply problems, ripples, spikes etc. I'd suggest 0.1 ceramic caps directly on the pins of the IC. One per power supply pin.
10) Does your watchdog reset the ADC? (I think so but I thought I should ask)
11)Looks like the part has some test pins which tell you it is ready or not. If you can break your reboot of your micro and hand verify it (when it locks)you might get some additional info.
12) The device seems a full cmos. Is your processor CMOS I/O?
13) I forgot to ask. How embedded is this processor? Does it have a serial port for monitoring/controlling it? Can you jump to a monitor program to hand-manage the device?

Don't forget. Assumption is the most expensive quick decision. I may sound paranoid but somebody said "Expect the unexpected".

Let me tell you a quick story that happened today. A new batch of boards came from our outsourced build-house. This is a quite complicated board with ATM, E1, cpu, SDRAM, AAL processors, Ethernet etc. Everything works on these boards except the E1/T1. Once in a while the system crashes too.

Problem:
The system reference is a 50MHz clock. This happens to be one of the Just-In-Time programmable oscillators.

In case you don't know... The supplier has a special gadget that programs the frequency whatever you like when you order them. Internally it has a PLL and it generates any frequency you like.

Turns out these oscillators are not so stable for some reason. The internal pll is not programmed to correct values. When anything disturbs the device (you blow on in so it cools down or warms up) it oscillates wildly for a while then stabilizes. During this time the instantaneous frequency moves wildly between 30 MHz to 70 MHz. Well the processor is rated for 50 MHz. Guess what happens when the instantaneous frequency goes to 70 at temperature extremes.




Good luck in your debugging.


Zoovy

 

[7rots51]

hi zoovy

I have done everything that you said ,my problem is not n my laboratory ,but my problem is in field ,some cards may have this problem after several dayswhen they work,because my application is industerial and noisy environment(I used sheilded with earth for sensor cables)some thing are unpredictable!

I think my mistake is in using sigma delta ,I think sigma delta is good for sound and audio system design ,they are not reliable for harsh and industerial systems ,I think I must ignore some feature of sigma delta and use other type of ADCs that are 16 bit or more and use some PGIA or IA in front of them for sygnal conditioning and some filtering in analog section and some in software.

I think the architecture of sigma delta is not good for operating in harsh and complicated environment,they are not reliable at all.

Regards

 

[zoovy]

one more quote

hi zoovy

I have done everything that you said ,my problem is not n my laboratory ,but my problem is in field ,some cards may have this problem after several dayswhen they work,because my application is industerial and noisy environment(I used sheilded with earth for sensor cables)some thing are unpredictable!

Regards

I am running into the danger of pestering you with my suggestions but I will risk it one more time. Because I hate to give up!

The Analog Devices makes protection devices for their ADCs. Do you think it would be useful if we were to use one of them to find out wheter the problem is happening because of the long wires to the sensor or power-supply design problems? They are usually expensive ($50 > XX >$100).


Below is the link to the devices.
**broken link removed**


Zoovy

 

[zoovy]

is there a layout problem?

7rots51,

I just remembered a problem in one of our designs. The layout had a problem which the input wire passed very close to the reference input/output of the ADC. The High impedance input got the glitches and created chaos.

Zoovy

 

[7rots51]

hi zoovy

5B series of analog device is too expensive for multi channel design.
I want to examine the TI ADC both sigma delta and SAR (sampling ADC) ,if the result was good ,I will use them in my future design,it seems that the price of TI is lower than ADI ADCs.

I need to work on pcb layout design(for high resolution systems) and on ADC selection.

zoovy thanks for your helps if you have any sch or pdf about this topic please send the link.

bye

 

[techie]

try the crystal semi cs5520 series. i have seen these also in some designs though never used myslelf

 

[Wowee]

Hi,7rots51
I'v designed a weigh scale based on AD7714 .
When I apply a known voltage to the AIN1-AIN2 of AD7714(be set to

bipolar,gain128) ,there were always some wrong codes among other

approximately right results.
When I changed the input voltage,also got right results with some wrong

codes.But the wrong codes didn't change ,fixed at
259XXXH,58DXXXH. I was confused with this.Could you encounter that in

your debugging AD7714?
Regards
Wowee

 

[7rots51]

hi wowee
For weigh scale application I recomment you AD7730.
I did not counter the problem hat you have ,(my application is unipolar)

Totally,I fear to use analog devices sigma delta for my future products ,I think It is better to test TI sigma delta ADCs (their price is also lower than AD ICs),for application that we can use sampling ADC(SAR)it is better to use SAR.
for very fast sampling flash ADC is best.
bye