Thread Tools
Nov 23, 2011, 05:55 PM
Registered User
Thread OP

Smallest FPV Concepts

Recently I've been getting back into RC helicopters and planes, and when I was ordering a "mini wireless spy cam" for my Syma S031 I was amazed by the tiny size and low cost. I looked around the Internet to see just how small these things got and eventually found these threads:
(I started a new thread since I read resurrecting old threads is bad - don't know why.) For some reason that thread didn't get very far, but I see two major points:
  1. smaller cameras exist but require 100+ volts
  2. the Tx can be lightened - a lot.
As for #1, looks like the smallest CMOS video camera is the Awaiba NanEye, at 1.5 cubic millimeters. Resolution is 250 x 250; interface is bit serial LVDS; 44 frames/second; 10 bits/pixel; 1.8V; 4mW. Power requirements look reasnable to me. As for actually getting one, they sell a "Product Evaluation Kit" that includes the camera and USB interface If you could convince Awaiba to sell you the evaluation kit, that would probably be the easiest way to actually get the cam, but I'm guessing it would cost a LOT, given that it was designed for medical endoscopes. Of course that's not an obstacle for everyone, but it is for me.
As for #2, the obnoxiously overpriced nano-cam at is definitely an option, but as Veteran mentioned it wouldn't be too hard to build one. It looks to me like the $1500 video camera uses the same lens and OmniVision sensor as the $100 cams at As for the Tx, I don't see why the AWI2400/AWI5800 shouldn't work, the package is only [25mm]^2, but I think I have a better idea. The OV6930 [1.8mm]^2 sensor is definitely small but (unfortunately) only has analogue NTSC output so it needs its own Tx. (The sensor cannot be unsoldered from the DAC because they are on the same chip!) Instead one could connect a larger digital OmniVision sensor (ex. OV2665, @ [4.8mm]^2) to the Cypress MCU/RF chip and use it as the RC Rx and video Tx. It'll be like a DT receiver with an extra video cam attached. Hopefully the supporting circuitry for the image sensor is only a few surface mount parts but I might be too optimistic in this regard.
As mentioned in the original thread, the video receiver / radio transmitter (I'll call it the "control transceiver") can simply be a matching chip receiving video and transmitting radio commands. The control transceiver can interface with a laptop to show video and accept commands through a USB joystick, it can be used with a modified RC Tx with a video output for TV or computer monitor, or my favorite idea is to make a custom control transceiver with video goggles.
The OV2665 is the smallest digital CMOS chip I could find, but the features list is a bit overkill for this application. Running @ SVGA 600 x 800 - 30 FPS - 16Bits color, that's 28.8MB of raw data per second!? Unless the transciever has that much bandwidth, some compression will be necessary. Unfortunately the chip is unlikely to be powerful enough to do any serious compression, so bandwidth could be saved by:
  1. transmitting a scaled down image (200x200 should still be OK, and the scaling can be done on the sensor chip)
  2. transmitting a dichromic or gray-scale image or using less bits per pixel
  3. transmitting the difference (delta) between adjacent pixels instead of the entire value
  4. convolving the image with an edge detecting kernel to isolate the most important data (as does the human eye)
Of course the lower the visual quality, the less fun and the more serious the project becomes. The good news is that digital transmission doesn't have those annoying artifacts. Also error-correction codes could be used to dispose corrupt pixels. I think black pixels would be far less annoying than that analogue noise you see on typical FPS systems.
They say that nobody is interested in a purely theoretical thread, but I'm going to need some help and motivation before I even consider this project. The following questions are on my mind:
  1. Is all the effort really necessary? Has someone else already taken a stab at this question?
  2. What MCU/RF chip is best suited for this application? Is it easily available? What is needed to program it?
  3. Can the CMOS sensor be connected to the chip directly? Does there need to be some other interface? Is the clock and voltage on the chip appropriate for driving the sensor?
  4. How much bandwidth is available? What is the transmission range? How much does the chip weigh?
  5. What's the best compression algorithm that can run on the chip? Does it use too much electrical power? What visual data should be removed or retained?
  6. What plane is best suited to test the system? Are plans or instructions available?
  7. Any other suggestions, recommendations?
My top skill is software design but I'm OK at using FPGAs to design hardware, building small airplanes, image processing stuff (Fourier transform, high-pass and low-pass filters, convolution, etc...). I also have assistance with micro/precision soldering, hardware testing, and to some degree hardware design and parts sourcing. If I do get around to this, then I'll publish the plans and everything. Maybe even sell a kit one day. For now budget is about $150-250. Any sort of help is welcome.

Last edited by 2ambitious; Nov 23, 2011 at 06:00 PM. Reason: Links didn't go on auto
Sign up now
to remove ads between posts
Nov 24, 2011, 01:29 AM
Registered User
Indeed, too ambitious
Nov 25, 2011, 10:28 PM
Registered User
Thread OP
Is that a dare!? Actually I've been looking around and I think option two is looking really promising. If nobody objects I'll keep posting my progress and hopefully someone will get on board, as far as I see there is no reason not to make a submicro FPV So far I have this:

Radio Choice
Reading about receivers here is the order of technology in increasing mass:
microwave tuned memory wire => radio sensor chip (theory) => 6 pin PIC w/ IR diode (26 mg) => PIC with RF circuitry (60mg) => RF/MCU (115mg and very expensive for what appears to be a microchip, but I digress.) => macro receivers

While interesting, this doesn't change the fact that we need a video transmitter so I looked at the Cypress website. Turns out the CYRF69103 is the only Transceiver/MCU they make, but there is also a wafer version, which to the best of my understanding is just the silicon chip w/o the casing. That probably means you have to solder through a high-power microscope with gold solder. probably not worth the effort. Either way it has a digital bandwidth of 1MBps using Gaussian smoothed frequency keyed modulation. I'll come back to that in a second. The next step was to do an Internet search for RF/CMU. frPICs look too big. The STM32W and si4012 look small at 6x6 mm and 8x4 mm respectively, but they advertise a fraction of the Cypress's bandwidth. Amtel has a few, eg. ATA5423, but still compare to the Cypress. Some of the Texas Instruments chips look comparable.
The next question was how much video can the CYRF69103 actually transmit. These are the main stats:
Transmission power: 4dBm (=Class 2 Bluetooth), ~10m range
Transmission bandwidth: 1 MBps = 0.125 MBytes/s

I also considered how usable the microprocessor is. It appears to have enough I/O ports for the camera data bus and control bus. I really hope the programming tools aren't too low level / expensive because I have no intention of reading the 73-page-long datasheet!

Video Transmission Format
1MBps turns out to be quite a lot when you're using a proper video compressor. I took a sample video [VGA, 360 frames / 12 seconds = 30 fps], loaded it into VideoDub, disabled audio, set the
compression codec to Xvid MPEG-4, set the target bitrate to 1000 kBps, and saved. The result: the file shrank to 0.39% of the original size! Of course I doubt the poor little MCU could do 1
millionth as much compression with its 256 bytes of memory (which leads me to think... meta-heuristically evolved nano FPGA!) so I'm going to write an image-processing program using 256 bytes
of data memory to see which data is best removed. I've calculated the following example formats are viable (@998,400MBps)
QQVGA 26FPS 2bit/pixel
QQVGA 13FPS 4bit/pixel
half-QVGA 13FPS 2bit/pixel
half-QVGA 26FPS 1bit/pixel
QVGA 13FPS 1bit/pixel

Of course by modern standards, the video quality will be pathetic but nevertheless usable. I'll be doing tests to see how tolerable the quality is. So far I've found a few videos online that hint at the quality. I'm presuming it will be worse than analogue but much more stable and free of interference. Maybe CRC codes will help with QC, maybe not.

Technical Note on Color Space
Got a bit too complex so I moved it to my website (will post link when available).

Camera Sensor Choice
I went back to the OmniVision website and found that I had totally neglected to consider a truly nanoscopic camera, the OVM7690. This is a "wafer level camera" meaning that the lens and assembly are made together with the image sensor chip. Here are the stats:
Package: 2.517 x 2.967 x 2.465mm
Interface: 20 pin ball-grid-array (power, ground, 8 data, XVCLK, PWDN, VSYNC, HREF, PCLK, SCL, SDA, 3 unknown)
Image: VGA 30 fps 8-bit color
Power supply: 2.6v - 3.0v 100mW
I/O pin v: 1.7v - 3.0v
Clock: 6 - 27 MHz
DigiKey: part#884-1019-1-ND; $12.34 per unit

I noticed a few things: the full datasheet is completely missing from the Internet, the DigiKey price is suspiciously arbitrary, and assuming it has the same density as silicon, it should weigh 0.0184084696 ml * 2.3290 g/ml = 42.873 mg. That's not that bad! The clock rate and voltage superficially appear to work with the Cypress chip, although I need to look into that further.

There are also other manufacturers of these nanocams, like Aptina, but I haven't found any smaller cameras yet!

Next steps
I figure the only practical way to connect the camera to the MCU is with a PCB (even then it's gonna be a nightmare to solder under a microscope). PCBs are heavy but lucky they make them as thin as 0.2mm.
The next step is to check that the video formats I thought of are usable, and if they can be programmed easily enough. Then to get the datasheet for the OVM7690, first asking DigiKey and then the manufacturer. If things still look good at that point I'll try working out a circuit to connect the two devices. The final step is to confirm the design and to solder it together at a nearby lab. For a test plane I was thinking something along the lines of the 2.8 gram Plantraco Micro Butterfly.

PS is there some way to tag this thread with key terms so people can find it easier? I'm just worried the chance of someone coming across it is small, I dunno
Nov 26, 2011, 02:07 AM
coriolan's Avatar
Interesting project, might find useful parts here:
Dec 03, 2011, 07:09 PM
Registered User
Thread OP
Thanks, but after all I decided to go with Plantraco (Based on Micro Butterfly). I am Canadian after all Here is the parts breakdown:
$3.49.......Motor-4mm X 11mm, 13 Ohm...................0.65g
$3.89.......32mm Direct Drive Propeller......................0.055g
(pricy but w/e):
$5.39.......NanoAct Magnetic Actuator Kit................0.07g
$7.99.......20mAh battery......................................0.8g
N/A..........reciever chip........................................0.23g
$20.76............................................ ...................1.805g

2.6g-1.805g = 795 milligrams for the camera + airframe + rudder, which I'll build myself. I also though about the PicooZ but it's too heavy at 10g. The smallest FPV I know of is 3g! This one should be ~2.6 with some careful work.
I've ordered all the parts and I'm just waiting for them to come. Also I've done some tests and looks like the video should be at least recognizable with delta encoding and run-length encoding. Control transceiver is custom build with the same Cypress chip connected to a Xilinx FPGA board with video DAC and joystick ADC.
Also I attached a visual comparison of useful chips and parts:
Dec 04, 2011, 03:27 PM
Registered User
sketchaholic's Avatar
Here's a small FPV setup:

Originally Posted by FPV Hobby webiste
Total 1 gram weight SUB-NANO Camera + Tx combo set.

Camera :
0.6 gram weight.
PAL 720*576 pixels resolution , NTSC 720*480 pixels resolution.
3.3 - 5 volts (1S LiPo works well)

Transmitter :
50 mw outputpower (Theoretical range is more than 3000 meters with 11 dBi patch antenna)
0.4 gram weight with light antenna
Only Video transmission , no Audio.
One channel (2.370 gHz channel)
3.3 - 5 Volts (1S LiPo works well)

As you know , WIFI uses 2.401 - 2.473 gHz band. For minimum interference with WIFI , our Tx works on 2.370 gHz frequency. (Special product for FPVhobby)
Dec 04, 2011, 05:33 PM
Registered User
Thread OP
Yeah I saw that website but only the 3g model, that's impressive and not that expensive! But I'm aiming to go 0.25-7 grams including receiver and motor MOSFET! And I've seen videos of those tiny cameras and they're noisy, I'm hoping I can eliminate that by using a digital feed (and saving weight on a DAC).
Dec 06, 2011, 06:28 AM
Registered User
boingk's Avatar
Its all very interesting... but my big question is why?

You can do everything I can think of at 5 or even 10 grams total weight that you can do at 2.6g total - including better tolerance of any air that isn't totally calm.

Just trying to think of this as a project with a practical use - what/where/how would you use it?

If it was me I'd throw the 1g Camera/Tx unit on a 12~18"/1oz plane and be done with it.

Cheers - boingk
Dec 07, 2011, 09:08 PM
Registered User
Thread OP
This is more of an electronics project than an "airplane" project, but the fun in it is the challenge! To the best of my knowledge the smallest FPV is the 3g micro delfly so I see a perfect opportunity to set a new world record! I know that a 1 oz. plane can do FPV so where's the fun in building one

Anyway I got the parts two days ago, and I'm attaching a photo and PDF summary of the thing that I'm using to explain the thing to people

(The white square is 1x1mm smallest production camera)
PS. I hope martin and lad dont mind that i borrowed their pics..
Last edited by 2ambitious; Dec 07, 2011 at 09:25 PM.
Dec 12, 2011, 07:30 PM
Registered User
Thread OP


So I've been told it's pricy to solder ball grid array parts, so I looked everywhere for a camera similar to OVM7690 with peripheral pins but no luck - the smallest I could find was the OmniVision module in the #11 1280x720 HD key chain cam, which is still wayyy bigger (I'm guessing 100-250mg vs 20-30mg). COMedia makes that "1g world record" module posted above and other similarly tiny ones but they're all analogue. So it looks like I'm gonna have to stick with the CameraCube, even thought it might cost 50 - 100 bucks to solder (although I'm looking into getting a discount.) Many people have used toaster ovens to do reflow soldering but I was told it's chancy because the solder beads crack easily and if the pressure applied to the device is just a little uneven then some pins will have a higher resistance than others. Either way I might reconsider depending on the final price quote for soldering.
I found an interesting link about someone else who's succeeded in using this cam. So it must be possible! I looked around to see if I could get a toy or demo kit with the cam already soldered to the board but it looks like the product is too new! There is something like that on DigiKey but it looks sketchy. I even looked into the tiny cameras on optical mice but they aren't that tiny after all.
Anyway to make sure I don't get stuck with buggy hardware, I had the idea of soldering the OVM7690 to a strip of flexible circuit and then soldering that to my transceiver. That way I can solder and unsolder as many times as necessary to get it to work. Another idea was to take some weight off by snipping unused pins, using ultra-small SMTs (the yellow rectangle in the pic), and cutting a square in the middle of the (already tiny 0.2mm) PCB. I've seen these tiny PCBs in quartz watches and they weight practically nothing, so the total weight should be pretty darn low! The only thing I'm still considering is filtering motor noise..
So I'll probably post an update once I get the hardware soldered (after new years due to a waiting list) and it's time to code!
Dec 13, 2011, 09:08 PM
Registered User
boingk's Avatar
If the fun is in the build rather than the flying then I totally understand you, its sometimes just plain fun to build something. I myself have around half a dozen projects on the go!

Be sure to keep us up to date with the project, this should be very interesting indeed.

Cheers - boingk
Dec 25, 2011, 10:50 AM
AndyKunz's Avatar
Even smaller camera

Too bad it's only B&W.

Dec 25, 2011, 12:53 PM
Love is the message
gizmointhesky's Avatar
Last edited by gizmointhesky; Mar 03, 2012 at 04:42 AM.
Dec 26, 2011, 09:23 PM
Registered User
Thread OP
Smaller? Hmm this one says 0.9g but the super nano cam above is 0.6g, and my omnivision is 0.06g and thats a worst case scenario!
Dec 27, 2011, 02:05 AM
Love is the message
gizmointhesky's Avatar
Last edited by gizmointhesky; Mar 03, 2012 at 04:40 AM.

Quick Reply

Thread Tools

Similar Threads
Category Thread Thread Starter Forum Replies Last Post
Build Log 1:4 Scale Schweizer SGS 2-33 by Aviation Concepts Lbuff1 Scale Sailplanes 21 Sep 11, 2020 08:59 PM
Discussion Smallest power kit for 4M/8 pound glider ? Smallest tow plane ? gyrokiteguy Scale Sailplanes 4 Jan 16, 2011 08:18 AM
Question Smallest FPV system + plane hazchem88 FPV Aircraft 14 May 09, 2010 09:55 PM
Discussion Could this be the world's smallest FPV heli? tvdude310 FPV Talk 0 Apr 29, 2008 05:12 AM