| << return to blog entries
2010-05-17 13:00:22 (3374 views) Last week was the scariest and most complex since the project started. This post will try to put the fragments back together in a coherent fashion but time is scarce lately, we're working more than 12 hours a day and it's already 2am so don't expect clarity...
PART 1: The panic
While doing some RF testing with the DL2000 in its plastic housing with all the rubber parts including the new keypad, we discovered to our horror that the range, with or without the enclosure, was very short. In particular, when the wearer faced away from the remote, a distance of just 3 meters constituted the maximum distance limit for the signal to be received. This problem alone would basically make our product useless, so this was a moment of panic.
After spending 24 hours on the phone with the radio module manufacturers, the antenna manufacturers and one RF testing lab it seemed this module's antenna was the problem, as it was chosen for size and not range. The manufacturer said sorry but you should have chosen another type of product.
This was very very disturbing because months ago we tested these modules extensively. Did we not try to make the signal pass through the wearer's body when the wearer is facing away? What could we have missed? What???
PART 2: Redesign
We were almost ready to give up on this module and switch to a power amplified one. This would be installed in the remote (no space in the receiver because it's too big). It would greatly increase range but require LOTS and lots of labor. For instance, a PCB redesign, a remote control enclosure redesign and prototyping, more current consumption and non-standard, labor intensive assembly methods. Parts of the module would have to be unmounted to save space before assembly and a wire antenna would have to be hand soldered after removing a special antenna connector that's quite useless since any antenna you can connect to it is much bigger than our entire PCB. After going through the motor thing with Mr Tang the LAST thing we want is non-standard manufacturing methods!!!
Despite these worries we created the new module in our CAD package in order to lay out components from scratch. We created a frankenprototype with this new module connected with wires to our existing board. Outdoor testing was phenomenal, we were able to reliably activate the receiver over 100 meters away, from the opposite side of a pond. Thankfully the RGB LED is bright enough that the receiver is clearly visible at this distance (not through clothing obviously). So,
- big disappointment with our previous module
- great range with the power amplified one
- anticipating many problems with a complex assembly process, OR, we could keep the extra thickness and make a big remote that's easy to assemble
PART 3: Discovery
HOWEVER things changed drastically the following morning. Tried to desolder the chip antenna that was on the old module, and replace it with a wire about 12cm in length (the wave length of the 2.4GHz signal). That immediately boosted the signal in a wonderful way, the signal now passed through one, sometimes two thick concrete chinese walls (as opposed to zero) and this is with the regular circuitry, without amplifier. Tried different lengths, and orientations. Even tried random bits of wire and metal, anything at all worked better than the antenna supplied with the module, even just soldering an LED in place of the antenna! How can random things work so well and the dedicated 2.4GHz antenna so poorly?
Eventually the breakthrough came when we attempted to solder the original antenna vertically, something that would never be done in a production environment as these are chip antennas meant to be soldered horizontally and are quite fragile.
Turned out that introducing an angle (any angle) between the antenna and the board gave a result that was just as good as any wire...
With this unbelievable discovery we got back on the phone with testing centers, the manufacturer... nobody knew why this was happening.
Tried to desolder the antenna on the RECEIVER as well, and introduce an angle. So now both antennas were skewed. Amazing range. We crossed the pond without power amplification. Hurray!!!
But, why???
After talking with the RF guys some more, in desperation, we just went on the net at 4am and looked up "chip antenna ground plane". Turns out chip antennas need at least 3-4mm separation between their longest sides and the ground plane, and this is in the spec sheet of near all chip antennas. We're no RF engineers - but it's crazy that people who do this for a living wouldn't know such basic rule.
Our ground plane is as big and as near the antenna (but not below it) as possible, thanks to the manufacturer's recommendation. So that's why the angle improves transmission - we're moving the antenna AWAY from the ground plane. The manufacturer even reviewed our board design prior to this discovery and their suggestion was to ADD a ground plane and merge them together with vias. Well - the solution was not more ground but less ground.
We are making some new board samples that have much more space between the ground plane and the antenna. Our hope is that we will get as much range as we're getting now by manually tweaking antenna position.
However, the RF modules themselves contain a ground plane, and this is spaced only 2mm from the antenna. If this ground plane is large enough to interfere with resonance then we may not be able to solve this 100%.
Impatiently, since we can't wait until the new PCBs, we tried to scrape off 4mm of copper from the ground plane on our PCBs to see if we could match the effect of antenna desoldering. This took hours and ruined several tools and a few antennas.
In the end this PCB carving removed 3-4 mm of ground plane from near the antenna and resulted in about 80% of the range we get with the rotated antenna. Unfortunately because the receiver too must undergo the same process you must calculate 80% * 80% = 64%. That means that rotating both antennas still yields almost twice the power.
It seems that antenna rotation still improves range by a whole lot even when compared to a properly sized host PCB ground plane, and this is due to the ground plane on the RF module.
This may seem very complicated especially if you are not an engineer - but let's try to make this very simple.
Initially the range sucked because we screwed up the PCB design due to the manufacturer's recommendations. Then, we discovered a way to get not only good but superb range out of these modules, however it involves messing with the modules in a way we cannot be doing in a production environment. So we're now making new boards and trying to see how far we can get, and the good news is that we won't need a total redesign of the remote.
Comments
Post new comment
 |
By dyonisos at 2010-05-17 16:45:26 Reply 
