RFX2401 2.4Ghz front end for Nordic products

[Buriedcode]

Hi,

For a while now I've been working on and off with the NRF24Z1.  Hoping to make my own wireless headphones, but with some some extra 'bells and whistles' like selecting source input.  My crude prototype using small boards for the RF chip, and ADC both mounting on stripboard (ugh) is actually working well.  However, the range is terrible.  The IC uses bidirectional comms, albeit asymmetrical, and the RF part seems quite similar to that of the NRF24L01+.   

The maximum programmable output power is 0dBm, with a whip antenna on the transmitter (thats pumping out audio, probably ~2dBi gain) and a PCB chip antenna on the receiver (taking in audio, probably ~0dBi gain) I get about 3-4m.  It drops out regularly if moved about a lot, and considering this is going in headphones, it will probably move about a fair bit.  I understand this is to be expected because there is no antenna diversity, and in a room there will be a lot of multi-path fading.  I suspect holding the prototype, with my hand quite close to the PCB ceramic antenna probably doesn't help.

Here's the antenna on the ARX board:
https://www.sparkfun.com/products/144

My hope is I can get about 10m max range before it starts to crackle, so I was thinking first, use something other than the PCB antenna perhaps for the next prototype, a meander antenna, or inverted F for the receiver.  This will be mounting inside the ear-cup of the headphone so quite close to ones head, and should also be somewhat omnidirectional.   And secondly, use an external PA/LNA front end to boost the transmitter power.  I would probably have to have a PA at both ends, because the receiver sends data back for acknowledge and re-sends.

So, the point.  Has anyone used this with NRF24Z1 ? http://www.rfaxis.com/downloads/RFX2401C.pdf

Its matched 50ohm in, 50ohm out, so.. in theory I could use the stock matching network from the datasheet for the NRF24Z1, and just a cap on the output to match the antenna.  In order to control the switch, I assume that the VDD_PA from the NRF24Z1 is active when transmitting.  As there is no 'TX_EN' or 'RX_EN' pins on this device, this is probably the only way to tell if its transmitting. 

Also, does anyone have any recommendations to improve range other than 'boost the tx output' ?  I'm not a complete novice with RF, but certainly no expert, and as the headphones are portable, I won't be trying to pump out 100mW.  I only need a solid 10m range, which would probably be 30m LOS.

[radar_macgyver]

I chased a similar range issue with an nRF24L01 for a while. It turned out to be interference with my wifi router. After moving the channels as far apart as possible, the trouble went away. I think a quick test to see what other devices are emitting in the 2.4 GHz band may be a good idea.

[Buriedcode]

MacGuyver, thanks for the reply!  Almost disappointed I didn't see a solution involving a paper clip and a cat to made a rocket propelled grenade...

Well then, that'll be my next test.  The channel list (38 channels, selected consecutively) I generated for it was random, but its handy that I avoid my wifi, or my neighbours.  I'll have to use a wifi scanner to see the bands being swallowed up - its getting rather crowded around here. Also, I tested it right next to a fairly powerful router (used as a bridge) which I'm guessing really did swamp it.

Do you know if chip antennas tend to be somewhat directional?  Polarization I don't think will be an issue - don't plan on turning my head down to one side often - but in testing it really did kill it, even with reflections everywhere.

[Buriedcode]

Well, I changed the channel table to only use the upper half of the 2.4Ghz band (2440 - 2485MHz) with no perceptible change in performance.  I'm fairly confident of the whip antennas specs as I took it straight off an 802.11b/g router.  But I still suspect the chip antenna as being pretty bad. 

If I place my hand about 20cm in front of it, line of sight with the transmitter, the signal gets lost.  I know ones body absorbs 2.45GHz a fair bit, and proximity of hands, plastic, etc.. can pull these chip antennas off frequency, so next up is desoldering that, replacing it with a few cm of RG178 and an SMA to try another whip.  Hardly ideal for headphones, but would at least determine if its low gain, poor antenna matching, or interference.

I'm still looking for a single chip front-end solution.  It seems the chip mentioned in the first post is obsolete, and TI's CC2590 seems to be specifically matched for their devices - I can't see TI giving me advice on how to use that for a rivals device.  In that case I might as well go down the TI route of 'pure path', which is sad because I have 4 of these NRF24Z1's, along with plenty of passives for them.

One thing I may try, with is ambitious, but a few years ago I did start work on wireless headphones using two NRF24L01+'s at each end, running off different frequency hopping tables, but synced by a microcontroller.  Should give a 'peak' over-the-air data  rate of 4Mbps, perhaps enough for stereo audio, although the algorithm for it was horrifically complicated :/

[radar_macgyver]

Sorry to hear that didn't help. If you suspect the antenna, you could try using a small whip made of a quarter-wave paperclip, or better yet a resistor lead (paper clips don't solder well). The nRF24L01 modules I use couple their meander antenna to the balun with a chip capacitor, this is a convenient point at which to break the circuit to try a different antenna.

Signal dropouts (even those not due to orientation) are more likely due to polarization mismatch. Applications where one station's orientation might change usually use circular polarization or diversity receivers for this reason.



Here's a good video showing how circular polarization can help avoid multipath interference (scrub to about the 10 minute mark).

[stmdude]

In that case I might as well go down the TI route of 'pure path', which is sad because I have 4 of these NRF24Z1's, along with plenty of passives for them.

One thing I may try, with is ambitious, but a few years ago I did start work on wireless headphones using two NRF24L01+'s at each end, running off different frequency hopping tables, but synced by a microcontroller.  Should give a 'peak' over-the-air data  rate of 4Mbps, perhaps enough for stereo audio, although the algorithm for it was horrifically complicated :/

If you enjoyed the NRF24L01+'s, and you have passives for the NRF24Z1 (which is NRD, btw), might I suggest the nRF51822 or nRF52822 chips?
They're a all-in-one package CPU+Radio. Even has built in LDO and DC/DC (for the RF parts). It should re-use a bunch of the passives as well.

Nordic has a nice-enough SDK for them, but recently, I've started banging on the radio manually (on a L2 level), as BLE or ESB didn't quite match my requirements.

I haven't had range-issues, but right now, my antenna is a piece of wire or a random length (yep, didn't really care about range).  Still, I'm getting about 15 meters (unobstructed) in an office-environment (aka, 2.4GHz is somewhat crowded).

If you care about power, with a little bit of effort, you can get them down to about 3-4uA @ 1.8V when not RX/TXing.

[Buriedcode]

Hi there,  yes I saw that nordic have really gone for the BTLE market, and rightly so.  Whilst bluetooth headphones is another project.. (you can buy them cheap, but not all have apt-x support, so, just buying a module..) for just custom protocols, that 2Mbps is a real limit.  In fact, wireless digital audio is probably the only common application that requires a bandwidth > 2Mbps, and given how integrated SOC's are nowadays, they can cram in some pretty damn good audio compression to give high quality audio over less bandwidth.

I think I'm just flogging a dead horse here haha, that's the trouble with starting projects, buying up parts, then shelving it for a few years - you come back to it, and realize why you stopped.  Still the 'ol NRF24L01's should still be handy. 

[stmdude]

The nRF24L01 is still kind of available, but it's now called nRF24L01+ .  The difference with the + variant is that it can do enhanced shockburst, in addition to the shockburst/gazelle that the non-plus did.

For some projects, I treat the nRF51822 chips as nRF24L01+'s with a built-in CPU.

How come you need 2Mbps though?  48KHz * 2channels * 16bits = 192Kbps..  If you drop the rate of the Nordic chips to 1Mbps, or even 250Kbps, the range increases drastically..

[Buriedcode]

How come you need 2Mbps though?  48KHz * 2channels * 16bits = 192Kbps..  If you drop the rate of the Nordic chips to 1Mbps, or even 250Kbps, the range increases drastically..

Well 16 bit x 48Ksps x 2 = 1536kbps. I believe you're working out bytes per second   Along with the overhead of preamble, address CRC etc.. 2Mbps is just not enough for two channel audio.  Especially when using bidirectional comms for acknowledge, and re-send request, not even time to resend any lost data.  I would even say that 4Mbps isn't even enough for a robust link, which is why I think I haven't seen many products using TI's purepath stuff.

With that said I did make a wireless guitar system using them, and yes I have the NRF24L01+ version.  Guitar has less bandwidth so that's 512kbps which allowed less samples per packet, resends, frequency hopping, lower latency etc.  I should probably dig it out as it wasn't half bad!  There were still frequent errors, but maybe I should have worked a bit more on the blacklisting channels algorithm.  Was one of those half arsed 'oh it sort of works, shelve it' projects.

About the original topic:  I got hold of some PS3 wireless mics, and these use the NRF24Z1's.   A quick test of range (USB receiver recording on the PC, me wandering around with the mic) and it seems to do 10m easily before any hint of crackle (corrupt samples).   The mic's have a single straight PCB monopole roughly 24mm long. The receivers use a meander antenna.  Neither seemed to dislike my hand near them (when testing at range with frequent errors).  Its hardly a definitive test as I'm not sending known data, and recording it, but it at least shows that without addition PA/LNA these things should be good for 10m LOS, or at least everywhere in a small room.

So, that means its either the antennas I'm using (very likely as chip antennas I always thought were fairly bad..), the board layout (perhaps, as the guy I bought them from knows his stuff, but perhaps didn't do any matching) or the power supply.  I tested the power supply on one transmitter - just using 2 AA's, and tantalum caps with no difference in range compared to a (MCP1640 3V to 5V + 3.3V).  Borrowing a scope to check for ripple on that soon.

[AF6LJ]

 
I see I need a larger shunt for my bullshit meter.
That patch antenna, that is even too small for five Gigs.
The patch antennas I have seen for 2.4GHZ were close to 30cm square.

[macboy]

The nRF24L01 is still kind of available, but it's now called nRF24L01+ .  The difference with the + variant is that it can do enhanced shockburst, in addition to the shockburst/gazelle that the non-plus did.

For some projects, I treat the nRF51822 chips as nRF24L01+'s with a built-in CPU.

How come you need 2Mbps though?  48KHz * 2channels * 16bits = 192Kbps..  If you drop the rate of the Nordic chips to 1Mbps, or even 250Kbps, the range increases drastically..

You've confused your bits and Bytes. 48Khz * 2 ch * 16b = 1536 kbps or 192 kBps.

[stmdude]

Well 16 bit x 48Ksps x 2 = 1536kbps. I believe you're working out bytes per second

You've confused your bits and Bytes. 48Khz * 2 ch * 16b = 1536 kbps or 192 kBps.

Yep, I'm an idiot. That's what I get for writing an audio-mixer in software today. Brain is in "smallest piece of information is a byte" mode..

[Buriedcode]

I see I need a larger shunt for my bullshit meter.
That patch antenna, that is even too small for five Gigs.
The patch antennas I have seen for 2.4GHZ were close to 30cm square.

Who said anything about patch antenna's?  The power output of these things is limited to 0dBm, with a quarter-wave monopole.  The receiver has a meander antenna, one for each receiver (two mics = two NRF24Z1'a).  For FR4, guided wavelength is 3/4 of the free air wavelength.. 0.75 * C/(2.45*10^9) = 0.75 * 122.5 = 92mm.  so a lambda/4 antenna should be... ~23mm ?  I am *not* an RF guy, only dabbled in it when adding 433/868/2.4Ghz wireless control to things, often frustrated by how situations where an external antenna with SMA would be ideal, but application wouldn't allow.  So my maths could easily be out as I've only done a few PCB atennaes with no facility to tune/match them accurately. 

[AF6LJ]

Watch the video...
In regards to the OP's issue, I would start by using a higher gain transmitter antenna.

[Buriedcode]

My apologies sir.

Yes a higher gain transmitter would be the first port of call rather than slapping on a PA.  I've found some PlayStation 3 microphones that actually use the NRF24Z1, and wrote some micro code to sniff the SPI bus to both the transmitter and receiver to see the config.  Was also a good chance to see how a professional product has these things set up.  The transmitter - the microphone - uses a single ~23mm monopole on FR4.  The receiver uses meander antenna's (two on the board, symmetrical, one going left, the other going right) and the config says both Tx and Rx are at full power - 0dB.

The range isn't great, but it doesn't have to be for that application, it seems to have no cut-outs or glitches providing the two are in the same room, roughly 10m x 7m.  It performs significantly better than the test boards I have, with one using a chip antenna, the other a 2.0dbi whip on an SMA.  That could be a number of things, most likely the chip antenna, poor matching, and less than ideal PCB layout on the prototype.

As a test, I configured the SMA test board with the channels/address/config for receiving from the PlayStation mic, and range was ok, but not as far as the original PCB printed antenna.  It was very sensitive to orientation.  Also the receiver PCB is mounted vertically,  about 10 degree's leaning back, so the meander antennas are seen ( from the transmitter) from the top-side copper, the Z-axis.   So it seems they really did do a rather good job of maximizing range with printed antenna's. But I fear this is the maximum range these devices are capable off given how sensitive they are to bit errors.  I'm sure a less time-critical application that allows for multiple resends would achieve far greater range with the given 0dBm output power, 'coding gain' can do quite a bit for range.

As my transmitter is stationary and powered from a wallwart, power is no issue, so I could attempt to design a PCB, with an RF front end (the chip in the first post) and also have the option of a printed antenna, or just an SMA for external.  External would of course be easier as I do not have the equipment to tune/match antennas so I would be working blind.  But this may not help much, as the receiver must send back data indicated acknowledgements and channel quality.   Adding a PA to the receiver, whilst that would extend range, would eat up precious power  - trying to keep the power consumption of the headphones to < 60mA.  I'm not after '100m' here, but perhaps coverage on one floor, and with multipath fading/reflections, I'm taking a wild guess at +6dB gain required, the 20dB gain seems excessive.

So, I'm thinking... the play station mic design is tried and tested, I'll try to modify that for Rx operation (requires reworking a QFN surrounded by 0402..).  The monopole will be quite near the users head, but at least 10mm from anything, I'm sure the housing will detune the antenna somewhat.  And for the transmitter, use a board with a PA/LNA, and an external antenna.

Lastly, the external antenna.  A whip is fine for omnidirectional comms, but the transmitter will most likely sit, on a desk/surface, close to a wall.  It is doubtful a signal will be needed behind it, or to the extreme sides. A more prudent design, to try and maximize range, would be a semi-directional antenna, sort-of hemispherical pattern.  One you put up against a wall of a room, and it fills the room, without sending half of the power into the wall (reflections..).  I know things can get very hairy with DIY antenna's, but would a patch antenna provide better coverage than say, a 'rubber ducky'?  I'm sure the routers antenna is a good 2.0dBi.. but either I don't have a good match to 50ohm, or its just not as omni as the meander antenna.

So, thoughts?  I have a habit of over complicating matters (thus the idea of just 'more power!) but it owuld be pretty cool to get some use out of these things, and uncompressed wireless audio is just plain cool.

[AF6LJ]

For the base station component in your system you might consider a corner reflector antenna. It will give you some gain and give you a 90 degree radiation pattern which sounds like it will meet your needs.https://www.google.com/?gws_rd=ssl#q=corner+reflector+antenna+2.4ghz] [url]https://www.google.com/?gws_rd=ssl#q=corner+reflector+antenna+2.4ghz[/url]
These are easily constructed, just keep your transmission line as short as possible.
https://en.wikipedia.org/wiki/Corner_reflector_antenna

Sadly; there are a lot of bad examples on the web.
for 2.4GHZ you can use a solid sheet of metal as the reflector, to simplify construction.

[NiHaoMike]

Maybe too late at this stage, but what about use an ESP8266 and just run the audio over IP?

[Buriedcode]

Maybe too late at this stage, but what about use an ESP8266 and just run the audio over IP?

I have been meaning to use the ESP8266 as a wifi radio, a standalone unit to pull MP3's off the network drives, but haven't thought about it for wireless headphones.  For a start, the power consumption is too much for a portable application.  Then there's latency - although given the speed, short UDP packets would have very low latency.  TCP would be more robust but with greater latency.  I'm aiming for about 15ms max. 

I believe the reason the internet is rather silent on digital wireless headphones is because, bluetooth has finally got good enough, and ultimately if it pairs with phones, PC's, laptops etc.. few will go for proprietary.  And the companies that do make proprietary systems (logitech, plantronics sennheiser) are unlikely to post schematics or datasheets for the devices they use.

As I mentioned in a previous post, it has really got to the point where its far cheaper and easier to just buy decent bluetooth headphones, I'm just continuing this because I've come too far to stop :/  Plus the only real advantage to a diy build is features - I can add them, or not include them.