A fan of hardware, image processing, clear diary space and most definitely tea, not coffee: we recently sat down with Sondrel’s new Global Program Manager, Ed Loverseed, to talk about his experience managing project programs in the fast-paced world of SoC design and development. In addition to his beverage-consuming habits, we spoke about his enthusiasm for broadcasting technologies, whether he considered AI and neural networks the next big trend to drive industry growth, and his thoughts on future innovations in the hardware space.
Sondrel: You have spent your entire career to date in electronics product and systems development. Why Sondrel at this point in your career?
Ed: Yes, this is true, I have always been involved in electronics product and systems development in one way or another. And Sondrel is the latest step on my journey in this field. My previous position prior to joining Sondrel involved heading up a software development team – and I own to having had a desire to get back to hardware. I like the immediacy of customer driven projects that working in hardware brings, rather than working on product development to a marketing roadmap. Sondrel is clearly a leader in the hardware space – hence one of my reasons for joining.
The other compelling reason is about humans, not systems or sectors. Sondrel is a company with a high concentration of very bright people. This was very attractive to me. You can achieve a lot with a team of talented individuals.
S: Your previous experience has also seen you involved with companies specialising in graphics, imaging and video. Is this a technology area that you have always been interested in? Is this one of the reasons you were attracted to Sondrel as a company?
E: Yes, it is a technology space that has always been of interest to me. There’s something very satisfying about working on a technology where the fruits of your labour are very visible – i.e. in the video broadcasting space. To be able to point to something on the TV, having worked on solutions where you were part of making it real; for example, in the transition from standard definition to high definition.
The worlds of broadcasting and IT are very much merging – you see that in microcosm in broadcast and video related SoC design as well, where previously both at the equipment layer and the chip layer there was a lot of specialist dedicated technology for video and broadcast. Now it’s much more a case of using more common technologies for data processing with a little bit of specialism for video. You see that in all layers – in the software, in the SoCs and in product development.
S: How challenging do you think project execution is, in the current climate?
E: For a new design, the challenge is principally about matching. What I mean by that is having to deal with a lot of unknowns and risks, in terms of quality and suitability of incoming material and IP, and then how long it’s going to take you to do something that has never been done before. Even if you’ve previously developed solutions in a specific technology space, there’s no guarantee of plain sailing if you are laying out a new design. So inevitably there are a lot of uncertainties, and you need to manage those in an environment where client budgets and deadlines are fixed.
The fact Sondrel work on multiple projects of differing sizes has the potential, in this regard, to make my job as a program manager slightly easier when it comes to dealing with uncertainties, because it gives me flexibility. And what I mean by that is the flexibility to ramp up the engineering effort for example, should it be necessary due to an unknown that has been flushed out in the design process. Flexibility of resources is a very valuable tool for me in managing successful outcomes for Sondrel clients.
It’s not interesting and fashionable anymore to refer to Moore’s Law but there is inevitably still an aspect of it lingering when you are talking of challenges. Because as the years go by the technology gets more complex and the feature sizes shrink, everything about design gets harder and more complicated. But teams learn, and the tools that we have to get the job done are better. This does mean that you need to be continually invested in understanding the performance changes and the new tools – you need good people who are learning all the time. And you have to balance the good things that are part of the traditional project management, in learning from your experiences, with the knowledge that the rules are continuously changing on you – so don’t take too much false reassurance from the checklist that got you through the last time. This year’s challenges are never the same as last years, and that’s one of the things that makes hardware design an exciting environment to work in.
S: Have you seen a change in attitudes within the industry regarding the importance of professional project management? Do clients in your opinion pay enough attention to this, or regard it as an unnecessary overhead?
E: The successful companies do see the importance of it! For big-ticket projects, and companies who regularly develop these SoC designs, they know that organisation matters. SoC development is not just about deploying technical experts. It needs to be in scale with what you’re taking on, depending on the package you are delivering. The principles of project management are the same, but you don’t need to apply them in the same way for a 100-man two-year project as compared to a six month three-person project. You need to find ways of scaling the techniques to the size of the problem and the level of risk and complexity that you have. So, in the case of Sondrel, where we deal with different size projects, the smaller projects get the benefit of the learning and the consistency of approach that we always take for sizeable projects without the burden of supporting a dedicated resource for project management.
Good team leads can be the most effective and efficient way to run smaller projects. If one person has in their head the schedule and the technical complexity trade-offs, that’s a good way to run the project as they’re in a good place to be empowered to make good decisions. At the point where it becomes too big for that, it’s better to separate the function of project management. Especially when the work gets difficult – and there are always times in a project where it will get difficult. If you have the detailed technical leads responsible for the project schedule, the order of events and the risk management, they become overloaded just at the point where you need to rely on them. It tends to be their instinct to dive into the nitty gritty of the detail to solve problems, and by necessity they will take their eye off the ball and this can impact on the overall project.
Keeping the client informed and involved is also important. For example, if a piece of IP doesn’t come in on time, or something isn’t functioning correctly, you’ve got to be careful not to make a false assumption about what the client wants you to do. For example, it may be more important for them to avoid delay, and so they would rather accept a compromise in the design. In all these types of circumstance, managing and trying to do the technical stuff is too much.
S: When you are tackling a complex SoC – where do you think the most prevalent ‘pain points’ are and what do you suggest are the key ways to keep your project on track?
E: The first pain point is that it’s hard to get a good clean start off the blocks, particularly if you’re doing a new design. It’s different with an iteration, of course, where much of the design will be the same. With a new design, there are so many choices to make - the technology, IP, eco-system partners – and these are all interdependent. It always takes longer than you hoped for those things to resolve themselves to a level of stability where an execution team can take it on and move the design forward.
The second is a client clearly understanding that the choices they make and their attitude to risk regarding the maturity of what is being brought in makes a big difference to how quickly you can move through the stages of architectural design, verification, then physical design and prototyping. If you’re starting from a clean sheet of paper, you have an enormous possibility space and an enormous temptation to do a lot of new things – just because you can. There is a reason for wanting to push many of those things because you are investing a lot of money in a new SoC so you want to get the best possible return on investment. But new things bring a lot of risk. They need to be proven or pipe-cleaned, and if you are the first to do so then indirectly you can be paying for that in dollars and in months. It’s a balance. Because you will have hard constraints – usually in the form of timelines and budget. You need to moderate your enthusiasm for new technology to make sure you can achieve your aim, but not necessarily by using the ‘newest’ of everything.
What are the key ways to keep client projects on track? There are of course many – but I would pull out two key things. The first is to manage your program to flush problems out early – bring them out in the open at the earliest opportunity to give you time to decide how to manage them. The other is to engage the project team with the client’s priorities at all levels – that way everyone is consciously working towards these goals instead of pedalling very hard in the wrong direction. Sometimes you can be set off in the right direction, but inevitably focus and priorities shift as the project progresses, and you need to start rebalancing the emphasis of what you are doing. In a large team this can easily happen. Formal and informal channels need to be used to give the whole team visibility of the project status – for example meeting minutes to be circulated, and on the informal side, encouraging the team to mix across all disciplines to avoid team members having a narrow view of the project only relating to their ’bit’.
S: Where do you imagine chip design will take us in the next 10 years? What do you think will become ‘the norm’ out of the myriad new products and solutions coming to market, particularly involving AI?
E: A guy that I used to work with did his PhD in Neural Networks at Queens, Belfast, and came out of his three years doing that with the conclusion that using neural nets was the second-best way of solving every problem! He went on to say that it was a fascinating technology, and you can do a lot with it, but you always come out of the end of it thinking you could solve the problem quicker and in a more straightforward way. But that was 20 years ago. I think the thing that has changed is not the fundamentals of neural networks, what’s changed is that hardware has now scaled to the point where you can apply that in spades, rather than with ‘a dropper’ – and at that point you can say that is the probably the best way of solving problems. Now you can afford to put enough AI processing resource into a problem for you to be able to do a very good job of it, rather than an average one. Yes, I think it’s going to become much more of a component within products, and the single biggest change within the architecture and processing of products. In broadcast technology, for example, it is beginning to be used in video processing algorithms for compression and filtering where previously it would have been neglected because an implementation using AI was too big and too expensive; to not only be competitive, but to be realised.
AI is definitely going to be the single biggest driver of architecture and processing development, and will become much more of a component within products.
At the other end, going from being esoteric to part of the toolkit is doing things with multi-chip modules and stacked dies and so on, which was relatively uncommon. The approach of doing SoCs with memories embedded in the package and taking advantage of the consequent performance gains is becoming much more accessible. More components off the shelf for this are becoming available, and present another way of driving performance.
In ten years’ time, I think it will be more common than not to see AI functionality within an SoC. What is not clear yet, is what will be the norm in the way this is incorporated. Will it be a standard IP product available for vendors, or will it become an integral part of other IP, as part of CPUs, DSP cores etc? On how this is incorporated, I don’t think there’s any clear consensus at the moment.
S: What quality have you most admired in a former boss?
E: I admire openness, consistency and clarity, especially in tough times. Organisations shrink and grow and will always go through difficult times – I’ve seen that handled in different ways. A former boss and CEO had those qualities, where in tough times he stood up and explained the reasons behind decisions that had wrong awry and the pain points that would occur.
S: What makes you smile in the morning – apart from the coffee machine being in full working order?
E: I’m a tea drinker so the coffee machine holds no appeal to me! The thing that makes me smile in the morning is seeing some open space in my calendar – it's useful to be able to see you can have some moments of calm when you know you can step back, refocus and gain perspective.
Ed joins a team with a longstanding history of managing and delivering large projects successfully. As a world-leading IC design services consultancy, Sondrel are capable of managing and delivering a range of different projects across the design spectrum.
Whether you're looking to develop and build an advanced SoC, create a secure, low power IoT ASIC, or solve your time-to-market pressures with specialist engineering support, Sondrel's extensive knowledge & experience will get your silicon to market on time and on budget.