DW3220 in combination with LNA (QM14068) and PA (QM14070) controlled by ESP32

Wireless Connectivity Ultra-Wideband
1

Dear Qorvo team,

We are currently designing a UWB solution for maximum range controlled by an ESP32. For this we are integrating the recommended LNA and PA chips (QM14068 and QM14070). We also ordered a reference PCB with the same layout and antenna, but without the LNA and PA. During our testing the version without LNA and PA could reliably range up to 20 meters. Now, our version with the integrated LNA and PA shows very low signal strength and possible range (only about 3 meters with -90dbm strength even when the antennas are very close to each other). You will find our DW3220 schematic for the LNA/PA board below.
This is the driver we are using: GitHub - Makerfabs/Makerfabs-ESP32-UWB-DW3000 · GitHub
These are the settings we enable:

dwt_configuretxrf(&txconfig_options);
dwt_setrxantennadelay(RX_ANT_DLY);
dwt_settxantennadelay(TX_ANT_DLY);
dwt_setrxaftertxdelay(POLL_TX_TO_RESP_RX_DLY_UUS);
dwt_setrxtimeout(RESP_RX_TIMEOUT_UUS);
dwt_configciadiag(DW_CIA_DIAG_LOG_ALL);
dwt_setfinegraintxseq(0);
dwt_setlnapamode(DWT_LNA_ENABLE | DWT_PA_ENABLE);

Please let us know in case you see any issues. Any help would be highly appreciated!

DW3220
DW32202965×1258 255 KB

2

In what jurisdiction or regulatory framework will this be operated?

Those seem to be the best LNA/PA chips out there for this use case.

20 meters is not inspiring.

With our CUWB system, which has a QM14068 LNA but no PA, we get about 60 m ch 5, 40 m ch 9 with the LNA in bypass. With the LNA on, we get 100 m ch 5 and 70 m ch 9. This is at -41.3 dBm/MHz regulatory limits on transmit. Our range testing is actually somewhat better than that, but those are the numbers we advertise.

As a general rule of thumb, the LNA adds about 1.8 times range. A PA going up to -31.3 dBm/MHz will double the range, maybe slightly more. But there are losses in adding both in circuit, so the combination might triple your range.

If your base case is 20 m, triple is only 60 m, which is something that can usually be achieved on ch 5 without any LNA or PA.

Something to consider is that we’ve not found the ceramic UWB antennas to be very good. We have tried the Taiyo Yuden on your schematic and others. We find a well designed PCB planar antenna to be much better and that is what we use. The downside is that you have to design it and that takes time and effort.

Something else to note is that I would expect the LNA to be first after the antenna on the theory you want to get the weak signal amplified off the noise floor as soon as possible and not suffer the insertion loss of the PA in bypass or the other losses in the wiring. Indeed, this is what Qorvo shows in an example block diagram:

image
image330×118 2.86 KB

Note the QM14068 LNA is a 1.8 V device, the QM14070 PA is 3.6 V device with 1.8 V controls. Make sure you are using proper voltages for power and controls.

Also, I am curious why you used a DW3220 which has two active RF ports but you use only one, when a DW3210 would suffice with only one active RF port. Are you somehow mistakenly selecting the wrong RF port?

Yeah, something is very broken, which usually means it will be easy to find.

Steps to help debug this:

  1. Verify all the supply voltages are correct on all the parts.
  2. Verify the EXTRXE goes high during receive and low otherwise.
  3. Verify the EXTPA goes high during transmit and low otherwise.
  4. In receive mode with no signal coming in, use an RF sniffer probe and spectrum analyzer to see if the LNA is self oscillating. Bad PCB layout can do this quite easily and it will severely cut range.
  5. If you can isolate the PA/LNA section, perhaps subject it to a vector network analyzer to check performance.
  6. In the code, instead of relying on EXTRXE and EXTPA automatic controls, use the GPIO pins as software controlled and make sure they operate as you intend.

It is possible that PCB layout could be so bad that your results are what you see just from that.

We are not familiar with that code, so I can’t offer any help there.

We do offer UWB engineering services if you want a professional assessment of your situation. We perform this service for a number of clients developing their own UWB devices. Designing properly working UWB PCBs is non trivial even for experienced RF persons. It is very easy to get things badly wrong at 6+ GHz.

Mike Ciholas, President, Ciholas, Inc
3700 Bell Road, Newburgh, IN 47630 USA
mikec@ciholas.com

2 Likes