Integrating the Trimming & Satisfying the Customer’s Needs, Part 3

In Part 2 (Integrating the Trimming & Satisfying the Customer’s Needs, Part 2), we looked at the method to use fused-links programming to trim a voltage regulator's voltage reference. This trimming resulted in changes to the UVLO (under voltage lockout) and ENABLE thresholds of the power supply IC. Due to time constraints, I verified that only some of the logic sequences worked well. As my available time expired, I decided to fix this bug during the starting phase of the mass production of the unit at my company's power supply facility.

I decided to interface my testing machine with an automatic handler in order to gather statistical information on a high number of samples. The data that I gathered was related to the effect of each trimming sequence on the parameter that I wanted to control (i.e., the internal reference voltage). This resulted in the trimming table below that shows the real results:

Table 1

Trimming table (practical)

Trimming table (practical)

After implementing this fine tuning, all the trimming sequences were effective, but this was true on a population of 100 units per sequence, i.e., on 1,500 devices. Then the mass production started and I observed a 20 percent yield loss due to the trimming procedure. To solve this issue, I switched the testing mode from automatic to manual.

I noticed that the problem disappeared, so I concluded that the problem was in the handler test board when running the automatic testing. My guess was that this was probably due to an increased resistive path between the equipment generating the logical trimming sequence situated in the ATE (Automatic Test Equipment) and the device under test. This increased resistance due to a longer path (test board ATE + automatic handler interfacing board) modified the voltage levels of the pulse sequence utilized to force open the fuse-links as part of the bit-trimming sequence.

Hence the voltage level utilized to supply a bit “1” was modified from the desired pulse value of 16V to ~15.2V. This difference was apparently enough to cause the trouble. See Figure 1:

Figure 1

The voltage drop on the path of the trimming signal caused an ineffective fusing of the fusible MOSFETs, labeled CREF in the schematic.

The voltage drop on the path of the trimming signal caused an ineffective fusing of the fusible MOSFETs, labeled CREF in the schematic.

I solved the problem by soldering a capacitor and a resistor near to the device to be programmed. See Figure 2. This R-C filter was generally helpful: It held the supply voltage up high enough until it was needed for the programming action. It guaranteed the right voltage levels of each trimming sequence.

Figure 2

An RC group soldered near the device under test

An RC group soldered near the device under test

The final result of this action was winning the business: The customer requirements have been satisfied and the client obtained the requested product.

Have you ever experienced a similar situation? As an engineer have you ever modified your hardware in real time or on the fly to perform a task? What strategies do you follow for your trials, by adapting to the particular conditions you have to face?

Related posts:

6 comments on “Integrating the Trimming & Satisfying the Customer’s Needs, Part 3

  1. etnapowers
    November 13, 2013

    I experienced many times as an engineer to have modify the hardware or the software depending on the situation. Many collegues experienced the same issue, it's quite common.

  2. etnapowers
    November 13, 2013

    The main difficulty to face in this kind of situation is to set up the system in order to achieve two objectives at the same time: Starting the mass production of the devices and supplying to the customer the devices that match the customer's requirements.

  3. amrutah
    November 13, 2013

    Paolo,   Thanks for the post.

          As I understand the problem, we are here trying to trim the Reference Voltage to some particular value.  I understand that while trimming there was some resistive path which was causing the Reference to be trimmed to a wrong value.  What I don't understand is having some additional RC shoud actually increase the settling time of the reference, then how does it help to finish the trimming within the same test time?

  4. amrutah
    November 13, 2013


      The application will define a load for the Reference Voltage generator. Adding a RC network has modified the Load as comapared to the actual application, how are we guaranting that the trimming that is done is good for the final product?

  5. etnapowers
    November 14, 2013

    @amrutah: the additional path was not only resistive , the RC filter acted as a low pass filter and moreover the capacitor closer to the trimming bridge was able to supply the electric charge to fuse the MOSFETS of the trimming bridge and achieve an effective trimming procedure.

  6. etnapowers
    November 14, 2013

    @amrutah: good point , the design team to solve this issue decided to integrate and apply the RC load to the Enable Pin network, this did not modified the load of the actual application.

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.