System Design And Verification Challenges: Are They On- Or Off-Chip?

What’s next for system integration?

popularity

What are the next natural items for mobile devices to be integrated? From 2002 to today, previously separate items (like GPS, cameras and keyboards) have been integrated into the phone. They caused a frenzy of integration within systems on chips. Now we have the Internet of Things (IoT) adding a trillion devices to the picture. Which ones are to be integrated, if any? What does all this mean for system design and verification?

Last month I had mused about the number of “things” that will be added to the IoT. Last week, I found a box with my old Compaq/HP iPAQ in my garage. (Talk about being an early adopter! Does skipping the Newton make me more of an early majority guy or am I allowed to have been too young when Steve Jobs thought that one up? But I digress.) As Figure 1 below shows, I also found the associated external keyboard, the external camera I had bought for the iPAQ, as well as the external GPS system. At the time, in 2002, I kept my audio separate in an iPod and I also had two different phones, one for the US and one for Asia.


Figure 1: iPAQ from Frank’s garage (from 2002) and Frank’s cell phone (from 2017)

As the picture nicely illustrates, 15 years of system integration has eliminated several external components, cabling and of course three of the four power adapters. It’s all in one phone with wireless earbuds at this point.

So where do we go from here? What’s next to be sucked into the phone? Well, that’s just the thing… It is no longer obvious to me. Back in 2002 I remember the vivid discussions with my friend Roger who has been driving me to the airport for the last 20 years. He always commented on the number of devices I carried, why I needed them and we discussed how many devices we would end up with in the future. I remember arguing that the phone and iPAQ should be integrated first—and I ended up with a Palm Treo a bit later. Turns out the camera and GPS were not too much behind. They made it into the combined PDA phones. Due to storage concerns, I always thought I would have a separate device for music. And sure enough, I still carry my 160 GB iPod sometimes, as it holds the majority of my music collection. However, I listen to music now 90% from the phone using an LTE data connection. The iPod comes out for long flights, as I dislike downloading what I want to listen to in advance. I simply don’t know what musical mood strikes me when passing Newfoundland and may not have downloaded Sibelius, allegedly great while flying over vast tracks of ice and snow.

So Roger’s and my discussions on the way to the airport are no longer about the number of my devices, but more about the capabilities of the smart phone, the new features it enables, and all the apps that were in completely different devices in the past. This amazing progress in complexity in just the last 15 years and has made proper system design and verification nightmarishly difficult.

This is why we need scenario-driven verification (as defined in the Accellera Portable Stimulus group) and implementation using tools like Cadence’s Perspec System Verifier. And this is also why the number of cycles to verify a system and just even assess the architecture options goes way beyond the capabilities of simulation and requires emulation and prototyping to get the job done, Palladium Z1 Enterprise Emulation Platform and Protium S1 FPGA Based Prototyping Platform here we come, and they can be combined with tools like the Interconnect Workbench (IWB), like Renesas did recently, see also a recent blog here.

Today, what was verified as a chip back in 2002, with its singular environment using divide-and-conquer using proper interfaces, now has all become part of the chip verification itself. And looking forward to the next 15 years, while verification may not necessarily include more external components into the smart phone like we did since 2002, continued growth in applications and functionality will require even smarter verification.

As the IoT grows verification itself will have to include smarter verification techniques and modeling to simulate  all of the components to which the devices are interfacing (like connecting to health devices, home automation, cars, etc.). With on-chip verification, off-chip verification will become even more critical.

Talking about system integration on and across chips, to close this blog, here is one from the history books. Just before my iPAQ days, I published a paper with Alberto Sangiovanni Vincentelli called “Virtual Component Co-Design-Applying Function Architecture Co-Design to Automotive Applications” at the IEEE International Vehicle Electronics Conference 2001. In it, we described an example modeled in Cadence’s Virtual Component Co-Design product announced back in 2000 (it came from the Alberto-inspired Felix project from 1997 for which I was the product manager). The example modeled a GSM phone and Power Window mapped into a car network architecture of ECUs. Our focus was the tooling necessary to understand the architecture effects. Figure 2 shows some of the example slides from that time. When presenting this at the SAE Convergence Conference in Detroit in 2000, I remembered the wrinkled foreheads of some automotive executives as we suggested to directly implement the phone in the car. Today that seems so … 2010ish.

Who is laughing now!


Figure 2: Power Window and GSM Functionality mapped into a car network – 2000 SAE Convergence