Shlomo Gadot is CEO of Percello, an Israeli startup company designing a low-cost, high performance femtocell 3G W-CDMA baseband processor chip. He describes how he is working with a team of mobile handset chip designers to develop a dedicated 3G femtocell chip, believes femtocells will be dramatically cheaper in the near future, gives some market forecasts, industry predictions and even tells us what’s keeping him awake at night.
First things first. How do you pronounce the company name Percello?
"Per-Sello" (which is short for personal cell) is the way we normally say it. But we also answer to "Per-Chello". It’s Latin for “excitement”.
What exactly does Percello make/sell?
We are developing the baseband processor for 3G W-CDMA and later on a dual mode 3G/LTE femtocell. Currently we are focused on the HSPA+ solution which will support peak data rates up to 21 Mbit/s downlink and 5.76 Mbit/s uplink. Our focus is to produce a single chip which will integrate all the components this type of system need including Security (TPM compliant), Timing (NTP, PTP) and RF interface. This will dramatically reduce the chipcount and not require extra glue logic. The first release of the chip will support 8 concurrent users and soon we will support 16. We’ve included all the mechanisms needed for self configuration. The Percello product has the lowest price and lowest power with highest feature list in the market.
Tell us something of the background to Percello. How did it originate?
The Percello team is based on the core team of Modem-Art. Modem-Art was a start-up company which developed a baseband processor for 3G W-CDMA for the handset market. Modem-Art was acquired by Agere in 2005 for $145M. I was the CEO of Modem-Art and had the opportunity to take the best 25 people out of 70 to start Percello in November 2007. Today, the company now has 30 employees who have all the expertise required to deliver such a product. This team is bringing the philosophy of the handset to the basestation market – aspects such as a fully integrated solution, low cost and low power. We fight for every penny on the solution cost and on any mW in the power cosumption. All these characteristics are really needed for a femtocell because it is a commodity solution rather than a regular base station.
What is your channel to market? Do you sell direct and to whom?
We sell directly to Femtocell vendors, Residential Gateway vendors, Base station vendors and Original Device Manfacturers in China and Taiwan. Each of them has different processes and different complexity for us to sell to. This market has started the convergence of network equipment vendors (e.g. Cisco, Thomson, 2Wire, Netgear) and cellular network vendors (e.g. Nokia-Siemens Networks, NEC, Panasonic, Ericsson)
When will your first chip be available in production and what features/capacity will it have?
We received the first chips from the Fab (factory) in the second week of February. It already has passed all the initial tests we have prepared in advance. We’ll be ready for commercial production before the end of 2009 with femtocell companies including Ubiquisys.
The chips supports 8 simultaneous users, that is 8 voice or packet sessions in parallel, with a peak data rate of 21 Mbit/s downlink and 5.76 Mbit/s uplink in total. The chip includes all security features, authentication, timing features ( NTP,PTP), interfaces to existing RF and self configuration features ( GSM sniffer, UMTS sniffer).
For a fully integrated femtocell, it requires Flash memory, DDR memory, RF chip, Crystal oscillator and an internet connection. We are working with few RF vendors and Rakon amongst others with a view to reducing the BOM below $50 by Q1 2010. This assumes volume production of millions of units.
Our current chip consume less than 1.5W to handle 8 concurrent sessions. Our 16 sessions/users solution will be pin compatible with no extra requirement from the customer.
Any other peculiarities of your chip?
Our cell range is up to 500m. Most existing femtocells are limited to 200m. We’ve had some operators asking us to increase this range, for use in providing outdoor coverage in rural areas (say a village with a few houses). This is perhaps considered more of a picocell application, rather than femtocell.
Isn’t it a bit risky only making femtocell chips? Other specialist vendors have two or three streams of investment in case one isn’t popular.
This chip can also serve the picocell market including small office and home office ( SOHO). It doesn't make sense to support multi-applications because you get an inefficient solution with high power consumption. The multi-application solution slows down the market since it could not reach the price target and power target and above all the integration this type of solution needs.
Can you provide a comparison between the different chipset approaches: Design your own (like RadioFrame), general purpose DSP platform (like TI), specialist femto DSP (like picoChip/Percello).
The TI (Texas Instruments) chip is a DSP solution. This is not appropriate solution to this type of market in the long term which is very sensitive to price and power. Also the TI solution will not include all the peripherals needed for this application – such as an interface to RF, timing solution for NTP or 1588, USIM interface etc. The power consumption is a very important factor since it dictates the price of the total package and whether you need heat sink or maybe air flow.
Picochip – Their solution is still missing the right integration, the right price and power consumption required. Today solution needs heat sink. It doesn’t clear if they yet support 8 simultaneous voice users in parallel of data sessions with data rate of 21Mbs. Their architecture requires extra parts such as 2 DDR memory chips, one for the DSP and one for the main processor – this is unacceptable in the long term.
Some femtocell vendors have designed their own chipsets (I’m thinking of RadioFrame) because they believe they can control costs/features better this way. What are your thoughts here?
In the earlier days of femtocells, when the primary chipset vendor was picoChip, then this was true because picoChip didn't go to the optimal cost structure. But look what happened in the mobile handset industry or other technologies. A system house always has bought their baseband solution from chip house. Examples include Qualcomm, Broadcom. The same in wifi market were everybody buys from Atheros and others.
What do you think of the Femto Forum’s “Femtocell platform API”, which might allow femtocell vendors to swap easily between chipset suppliers?
We have been driving this approach since it will make it easier for software vendors to align with our solution. It is still won’t be easy but much easier.
How can Percello continue to differentiate if the API becomes popular?
On features: Data rate, number of users, cost, power, integration to name a few.
The large chipset vendors haven’t invested in femtocells yet. When do you think they will become interested (if at all)?
Qualcomm has already started working on femtocell chip. Their latest Press Release says they will have solution within 18 months. Broadcom generally waits for the market to develop and then acquires small companies with that technology – they traditionally prefer to acquire than design from scratch.
Regarding your RF Technology choice. You are focussing on only 3G UMTS chipsets today – any plans for WiMAX or CDMA?
We are making 3G UMTS and the evolution (LTE). We will have in the future some other flavours. Right now we are focussed entirely on W-CDMA.
Next generation: What format would be needed for an LTE (e.g. joint GSM/UMTS/LTE or just standalone LTE), and when would this be needed for femtocells?
Dual Mode 3G and LTE is essential for backward support of existing networks. The move to LTE is evolution and not revolution. I’d expect LTE to be mainstream by 2012 or 2013 with femtocells following a year later.
Do you see a need for greater integration onto a single chipset, e.g. more features/reduced chip count?
In the short term there’s no need for more integration.
The long term may need more integration depend on what Residential Gateways will look like. This may happen after dual mode UMTS/LTE products.
There’s a lot of focus on the cost of femtocells – operators are still seeking the elusive $100 device cost. What pricepoint do you think the industry has achieved to date, and what impact will your chipset/pricing have?
I think that we are already close to this target price (of $100) or within this target using the Percello chipset. The integration needed is the RF as a single chip which several companies are already shooting for (e.g. Lime, Sandband and Bitwave). Rakon (who supply crystals) also need to push their price down. This is only a question of market pressure and we will then be below the $50 BOM.
Enterprise Market: Percello have announced capacity of up to 16 channels on a single chip – by far the highest offered to date. Why have you chosen such a large capacity compared to the mainstream suppliers?
I think that this offer will also help in the residential case where the operator will decide to leave the femto open for neighbours. It is good from the technical point of view and you may start consider that as a small picocell. In the enterprise case, 16 users is the optimal case we can cover with almost no penalty on the price.
Domestic Market: Do you have a different chipset for domestic use (say a 4 channel offering), or do you take an approach like Intel by disabling some circuitry/features for markets with simpler requirements?
As I said before, from customer point of view there is not much difference since we have the same package and same pinout. It is more of a business issue and trade off between quantities and management of 2 production line.
What do you believe will be the timeframe for volume take-up of femtocells: Market forecasts vary widely from almost zero (Total Telecom) to 40 million (ABI research) by 2012. What’s your view of where consumer demand is likely to be strongest and what the trigger points for high volume are likely to be.
My view is that 2009 will be less than 500K femtocells, 2010 around 3-5 million and 2011 will be the trigger point of more than 10 million.
Whilst the US market is today primarily focussed on coverage issues, Europe and Japanese operators are seriously looking at femtocells for data services.
Any views on the future, predictions, issues to watch out for?
I think that we will see the 3G solution last longer than anyone expects. The life cycle of each cellular technology generation is more than 10 years. I can see 3G remaining dominant in the market for another 8-10 years. It could even achieve higher data rates like 42Mb/s using MiMO technology.
I can see some type of mergers where the giant companies will acquire companies such as Airvana, Ubiquisys, ip.access and some other large silicon providers acquiring the small silicon houses where these provide more complete solutions.
Finally, as CEO of Percello, what’s keeping you awake at nights!
My biggest concern today is Qualcomm, who I see entering the femtocell chipset market next year. They are likely to be providing some tough competition in 18 months.
[Update: We caught up with Shlomo again in October 2009 - read Shlomo's interview about Percello femtocell chipset progress next]