Beware the hidden costs of working with RF ICs

Today's wireless SoCs offer the promise of reduced cost radio telemetry, but Barry Gillibrand warns that for most applications the hidden design costs can dominate.

The explosion of RF ICs onto the world market is mesmerising to say the least.

If Bluetooth or Zigbee or any of the other proprietary parts available is your "bag" then this article is not for you.

The following is aimed at those engineers working in the less "glam" wireless markets mainly concerned with ISM (industrial, scientific and medical) applications.

These cover a multitude of generally non-mass-market products for which nobody has yet written a general protocol.

In fact it could be argued that this would be positively a nuisance given the plethora of different products in this sector of the wireless industry.

So you want to make your next gismo wireless - doesn't everyone?.

What are the design options available and likely costs?.

Which option will get you to market ahead of the competition?.

The apparent "obvious" option, which appears to offer the lowest cost, is the massively promoted RF IC.

After all, the cost of the chip at around $3-5 looks like offering a very cheap radio.

Beware there are usually several external parts required and these can increase the cost considerably not to mention the necessary micro.

Don't fall into the trap of assuming that you can use the one you already have in your application to also run the radio - it will very rarely work as planned.

The marketing blurb from one of the early RF IC offerings made the bold statement that "your grandmother could design a radio with our chips".

Well assuming she had a PhD in RF engineering this was probably accurate.

And at $1.50 a crack it looked like a real cheap deal (RX only) - only three external components, how could you go wrong.

Our company set about making a "working" receiver based on this part and finally after several months of iterations came up with a half-decent AM RX; with a manufactured cost of about GBP 4.50/$9.00, a far cry from an easy design costing $1.50.

The addition of a SAW filter helped to push up the cost somewhat.

So these things are seldom what they seem at first glance.

To be fair to the RF IC guys they have improved their act significantly since the early days and now offer good quality radio chips with considerable help and support available as well as reference designs.

So is the radio design task so easy that you grandma can do it now - unfortunately not.

The apparent deluge of help available has lured many engineers into the belief that this is now straightforward and there is no doubt that very good radios can now be produced using many of the parts available - this is not the issue we are discussing here.

Given that your company has the manpower design resources, expensive RF test equipment etc, and most importantly plenty of that rare commodity - time - then a satisfactory radio can be achieved.

If your application is for considerable volumes because you operate in the consumer or automotive volume then you probably will need to go down the RF IC route.

However, if like thousands of companies you operate in more modest volume areas then a modular solution is going to be certainly quicker and - in the long run - more cost effective.

Our experience over the past 15 years suggests that the usual timescale to produce an approved radio, hardware only, using an RF IC is somewhere between 9 and 12 months.

Once the hardware works, not usually before, the arduous task of writing secure communications software has to be tackled.

You will get very limited support from your RF IC vendor for this part of the project.

They are in the business of selling you chips, not software, so unless you just happen to be using the same micros in the datasheets you're on your own.

If you have written wireless software before you'll be aware of the challenges.

With a normal "wired" design it's usually pretty straightforward: but take away the cables and then the issues start, especially if your project uses two-way comms.

The most common protocol used is RS232, which for a wired system rarely presents a problem; but there are significant difficulties in the wireless version.

For example, RF noise appears as data to a UART and is very difficult to reject using UART interfaces.

Our own experience in writing software and that of the many customers we've seen struggle over the years, shows that design timescales of between 3 and 6 months should be allowed.

The time will depend on the complexity.

The most common misconception with RF IC designs is that they are going to result in the lowest cost design.

Certainly if the piece part cost is the only cost considered then this is always true.

You must of course look further than just the IC cost itself, and you will find that there are usually several other external components some of which like inductors can add significantly to your overall cost.

Final assembly testing is a major consideration and one often left to the last minute.

Do you have the gear to do the tests? Does your subcontractor?.

The micro is another minefield area.

There is usually one lurking about somewhere in your device.

However, radio software adds considerable resource demands on a microcontroller, which will undoubtedly have an effect on the specific things that make your customers buy your product in the first place.

A dedicated micro for the radio must be taken into account.

The typical cost for a TXCVR based on RF IC design in say 10,000 units is going to cost about GBP 8.00 or Eur 11.50 tested.

Your actual amortised cost of course depends on how you account for your design time, equipment costs, approvals etc If time to market is a crucial issue - and it usually is - then there is the issue of lost revenues/profits to be taken into account.

From the comments above it can be seen that a typical wireless design is going to use up between 12 and 18 months.

Take a monthly estimate of your expected gross margin per product or system and work out what the total profits figure would have been if you had already had the product on the market.

You will probably get a nasty shock.

The final hurdle to leap once you have your finished radio design is regulatory approval.

In Europe this is achieved by either a self declaration route - risky if you get caught out because you didn't do the tests properly, but low cost - or by submitting the product to an approved test house who will carry out the necessary EMC/radio tests.

This will cost you about GBP 2000 if you get it right first time: most don't so you'll have retest costs.

Failing the tests can be costly, time consuming and very frustrating.

You will probably already have announced the imminent availability of the new product to the sales team if not the customers.

Soon disappointment sets in.

Given all the above it is worth repeating that if you are in a highly competitive market like consumer goods or automotive, then the RF IC is going to be your only choice to get the cost right.

If you're in a less cutthroat market then a module could well be your best bet all things considered.

But which module should you use: dumb or intelligent?.

That's not meant as a slight on the supplier's abilities.

Some devices offer built-in/embedded software whereas others concentrate on just the RF and leave you to do your own thing with the software.

At the low cost end of the market there are AM devices - usually simple AM TX to RX one-way links.

These still need software to run well but the task is not so tough and in any case the performance isn't expected to be that hot.

Low end FM radios also exist: these are a little up the food chain from AM and perform reasonably well over ranges up to a couple of hundred meters or so if the software is well thought out.

The real challenge comes with two-way, transceiver (TXCVR) applications.

These are becoming the norm in wireless applications since the RF IC explosion because it is now feasible to design a low cost TXCVR at about the same level of a decent FM TX-RX pair.

The software can be a nightmare as there are so many things to consider like packet sizes, transmission speed, range, retries, encoding, decoding and many more.

Expect many sleepless nights as well as missed deadlines.

LPRS has sold tens of thousands of "dumb" modules in the past, mostly TX-RX but when faced with TXCVR applications, almost without exception, some form of software solution had to be provided to the customer to get the system working satisfactorily.

Very few users ever successfully wrote their own software.

This was the reason we developed Easy-Radio software to offer a complete wireless link solution.

Several intelligent modules are now on the market offering "out of the box" or "plug and play" solutions.

After nearly 3 years in this area we have discovered that there is no such thing as a "one size fits all" - there is always a degree of customisation required.

The flexibility of the ER module series has allowed many of our customers to achieve the customisation they needed.

Over the period we have seen many applications and Easy-Radio has evolved to include many of them in the latest firmware version released in January 2005.

This revision offers several unique features/benefits - never seen before in the embedded wireless module market.

The ER400 and 900 radios allow user control over frequency of operation with, for example, ten channels between 433 and 434MHz for Europe.

This means interference from other systems can be avoided.

Power output can be selected between 1 and 10mW.

If the range needed is short then reducing the output power can conserve battery power.

As well as RS232 baud, over-air datarates can be now be modified by the user.

The faster the datarate is set the shorter the range, so if data speed isn't your main consideration then it can be lowered with the bonus of improved range.

All the programmable features are achieved using a Windows based program that is very simple to use.

Version 2.0 incorporates the following additional features.

Repeater mode allows data to be relayed to increase range.

Encryption allows the user to configure devices to transmit secure data.

DCS (digital channel separation) can prevent unwanted channels being downloaded.

Faster ER commands allow the user to use a slimmer configuration protocol, which improves setup speeds.

User configurable UART rate allows virtually any RS232 baud rate to be set via the PC software as well as the previous presets.

Adjustable over-air datarate allows users to trade speed for range.

Reconfigurable channel frequencies allow systems to listen on one frequency, transmit on another.

User defined maximum packet size allows users to speed up transmission by defining when to send data.

Simultaneous upload/download means that a message can be uploaded or received over air while the module delivers its message received, to speed up data flow.

And the software also allows the module to be used as a raw data transceiver, using your own radio software but still being able to modify frequency, power output and datarate.

When considering the design options for your next wireless project you will need to address the following points.

Calculate the likely time to market and consequent loss of revenue, market position, competition etc If considering RF ICs, work out the total cost of associated circuit components, purchase or hire of RF equipment, testing, approvals costs etc Compare this with the range of dumb/intelligent radio modules available.

Look closely at the amortised design cost over the lifetime of the product.

As a rough guide, any application which uses fewer than 20,000 units per year is going to work out more cost effective using a module, and an intelligent version will be more cost effective still.

To check your own application numbers visit the LPRS website, click on the "Design cost calculator" and follow the on-screen instructions.

Back to top