Tales of a system design that worked

Screen Shot 2018-01-29 at 10.58.33 PM.png

Early in my career at National Semiconductor (now merged with Texas Instrument), we were spitting out 7+ different audio chips each quarter in a fiscal year. Some of them were brand new design or some were just a little tweak such as addition of a small feature i.e low power or shutdown mode. Typically, these chips came in with different packages from microSMD to DIP. With handful of application engineers and test engineers always needed more time, the challenge was how to characterize these chips on time with better efficiency. My application manager at that time, Jeff Bridges, came up with a solution: simply to build automation test system using Audio Precision as the brain of the system (now a standard tool to test audio performances of almost any audio system and chips) with GPIB (General Purpose Interface Bus) equipped power supplies, data acquisition units, function generators, and a digital oscilloscope. (simple block diagram block shown below). Not just build one, but build 10 of these systems to support different regions in the world that are doing manual audio testing for our products. Screen Shot 2018-01-29 at 10.43.54 PM

I was chosen as the lead hardware and software engineer. As a newly graduated EE, I was eager to take in the big responsibility, but this turned out to be no joke at all. As most people in the group weren’t software engineer, I was left alone to figure out how the whole system supposed to interface with hardware and software. With time clock started ticking, myself and my technician at the time, started the daunting job. Luckily, technician, Orville has tons of experience building cabling and test racks, while he started on purchasing equipments and cabling, I went to white board and started brainstorming, how the boards and daughter card should looks like? Schematics were drawn in Protel (now Altium). I started working on the layout of the motherboard and daughter cards. The idea was that we will have one giant (12”x8”) motherboard with relays to route different audio test paths with interface pins which connect to test rack. After that I focused on the software aspect, the audio precision used at that time supported built in visual basic programming; this became the logical choice to automate other instruments as well. It turned out that VB is a great scripting language to control GPIB capable instruments. The only thing was that libraries have to be written from scratch for each instrument. For example, in order to set voltage and current for a power supply or capture a scope shot with a required relays configuration, raw SCPI commands from these instruments has to be properly captured and then wrap around on user friendly functions. The whole project needed several files linked together in the main program. The good portion of the year went by and finally we were able to characterize numerous products including LM4930.

The LM4930 is an integrated audio subsystem that supports voice and digital audio functions. The LM4930 includes a high quality 16 bits I2S input stereo DAC, a voice band codec, a stereo headphone amplifier and a high-power mono speaker amplifier. I believe it is still in production as of today by Ti despite the fact, they have closed the audio group.

As a first big project, I really enjoyed the whole system experience from idea to actually building the test system (a picture of this test system shown above). Another happy side effect came out when datasheet electrical table entirely can be produced with one command. Various performance curves now be generated much faster than on the bench. Soon it became part of a regular project development cycle where product and other application engineers used this test system on a regular basis.

One of the main lesson that I learned was that during development of layout and design, it is best to engage relevant people early on where they can chime in their expertise before you send the board to fab. For example, I did the routing of some of amplifier control lines without 50 ohm impedance matching. Even though, SPI interface can be run at lower frequency, but at 24MHz, it needs proper termination and controlled trace impedances. This ended up being requiring spin out of a new version of the daughter board.


This entry was posted in Automation. Bookmark the permalink.

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s