Power supply dependancy in analog cuircuits

[KrDe_284951]

Hi,

   

I'm designing a low cost conductivity sensor device based on the Cy8C3446-LTI073, according to the schematics in attachment. The method of measurment in the temperature over pins labeled Ex1/Ex2 is a ratiometric method, is highly power supply independant and works perfectly in previous designs with the same chip layed out on the same pins.

   

We now added the circuitry for EC measurements (using the IDAC, comparators, PGA,...), causing the analog routing to be changed in such a matter that all analog properties in the device suffer from severe inaccuracy due to fluctations in the power supply. Even the ratiometric temperature measurement shows a severe power supply dependancy.

   

My question are mainly what is causing the problem and how can be dealt with them?

   

(BTW this is a high volume product)

[odissey1 · Accepted solution]

        If you believe that analog routing change is the cause of noise pickup, you may try to lock original routing in Analog view. To confirm that routing is an issue, delete late addition components to see if noise disappear. If not, that might be a PCB layout issue, not PSoC.   

[KrDe_284951]

Thanks for your reply!

   

It seems to be an internal routing problem indeed. I've deleted all analog parts from the design except those required for the temperature measurement, rebuilt the design, checked the power supply independence (which was ok), locked the analog design and subsequently added the other parts again (without any other hardware modification what so ever). The temperature measurement now remains stable under various power conditions. But as for the rest of the design, the power supply dependency has become far worse. Moreover, the routing of the IDAC is now in such a way that the load (in our region on interest) of the corresponding hardware pin  is of little or no influence on the measured result. Any suggestions as to solve this issue?  (A SOC is a very nice concept, but is apparently very limited in its real use, maybe I'd better stick to the old fashioned opamps, current sources, etc,...which do not suffer from the obscure ways of internal routing schemes)

[odissey1]

KrisD,  since this is commercial project, I suggest to send your design to Cypress for review (File-Create Workspace bundle->Minimal). The application engineers at Cypress are very helpful and fast-responding. If you have PCB designed already, provide it as well, Cypress can help with that also.

   

The schematics shown does not seem to be overcrowded. In PSoC analog and digital domains are well-separated, having own power inputs, grounds and pin locations. Any noise on the digital pin Ex0, which has frequency above the cut-off frequency of the correlated double sampling will affect RTD measurement. I  suspect that issue is the noise on digital pin Ex0, which provids "analog" power to the RTD (please check this with scope).  Try to bypass it with some large (~10uF) capacitor and see the effect.

   

Try also to change pin Ex0 to analog and power it using OpAmp+Vref or IDAC.   

[KrDe_284951]

Hey Odissey1. Thanks for your feedback on the subject. In order to reduce the power supply dependency, I've made some alterations to the use of the analog pins on the PCB. The IDAC is now connected to the dedicated IDAC output pin P0[6] and the reference voltage to the dedicated Opamp Out pin P0[0]. However the dependency remained, although in reduced form. The source of this dependency seems to be situated around the reference voltage output, Ex4 at P0[0]. The (external) output impedance of the opamp is too high in order to ignore the changes in current (+ and - 10µA) induced by the IDAC at Ex3 (with the EC cell connected between Ex3 and Ex4). Subsequently, I've added a low cost unity gain Opamp in hardware (MCP606T) to buffer the reference voltage, Ex4, which solved the problem of the power supply dependency. Seems like the impedance of the routing is slightly a function of the power supply.(However it would be nice to hear that such is a possibility)  (Note, our application is directly fed from a battery, without DC/DC converters. Hence the need for power supply immunity). Thanks again for your support, your remarks triggered the idea which led to this solution.