• State TI Thinks Resolved
  • Locked Locked
  • Replies 17 replies
  • Answers 5 answers
  • Subscribers 31 subscribers
  • Views 1589 views
  • Users 0 members are here
Support feedback
Options
Options
Related

This thread has been locked.

If you have a related question, please click the "Ask a related question" button in the top right corner. The newly created question will be automatically linked to this question.

PGA460: Noisy waveform on EDD

Yash Darad
Prodigy 120 points
Part Number: PGA460-Q1
Other Parts Discussed in Thread: PGA460, LM1117, LM317, TPS5410

Hi! 

Our setup is as following:

A 12 Volt Car Battery going directly to PGA460 and Epcos TDK Transformer. Transducer is a 40KHZ Osenon Closed Top rated for 160Vpp. Our hardware is made up of 2 PCBs, the base board contains a microcontroller and associated power supply and 2nd PCB contains the PGA, Transformer and Transducer. The two boards are connected using a cable of length 10 cm on TTL UART interface at 3.3 Volts.

Attached is our EEPROM settings files. 

Fullscreen GRID_USER_MEMSPACE-2019-04-24_030318.txt Download 
;GRID_USER_MEMSPACE
00 (USER_DATA1),00
01 (USER_DATA2),00
02 (USER_DATA3),00
03 (USER_DATA4),00
04 (USER_DATA5),00
05 (USER_DATA6),00
06 (USER_DATA7),00
07 (USER_DATA8),00
08 (USER_DATA9),00
09 (USER_DATA10),00
0A (USER_DATA11),00
0B (USER_DATA12),00
0C (USER_DATA13),00
0D (USER_DATA14),00
0E (USER_DATA15),00
0F (USER_DATA16),00
10 (USER_DATA17),00
11 (USER_DATA18),00
12 (USER_DATA19),00
13 (USER_DATA20),00
14 (TVGAIN0),8D
15 (TVGAIN1),EE
16 (TVGAIN2),EF
17 (TVGAIN3),11
18 (TVGAIN4),48
19 (TVGAIN5),28
1A (TVGAIN6),C0
1B (INIT_GAIN),C0
1C (FREQUENCY),32
1D (DEADTIME),00
1E (PULSE_P1),68
1F (PULSE_P2),16
20 (CURR_LIM_P1),47
21 (CURR_LIM_P2),D6
22 (REC_LENGTH),BB
23 (FREQ_DIAG),00
24 (SAT_FDIAG_TH),22
25 (FVOLT_DEC),69
26 (DECPL_TEMP),0F
27 (DSP_SCALE),00
28 (TEMP_TRIM),00
29 (P1_GAIN_CTRL),C1
2A (P2_GAIN_CTRL),09
2B (EE_CRC),0C
40 (EE_CNTRL),00
41 (BPF_A2_MSB),87
42 (BPF_A2_LSB),07
43 (BPF_A3_MSB),F3
44 (BPF_A3_LSB),72
45 (BPF_B1_MSB),06
46 (BPF_B1_LSB),47
47 (LPF_A2_MSB),7C
48 (LPF_A2_LSB),D3
49 (LPF_B1_MSB),01
4A (LPF_B1_LSB),97
4B (TEST_MUX),00
4C (DEV_STAT0),80
4D (DEV_STAT1),00
5F (P1_THR_0),AC
60 (P1_THR_1),CC
61 (P1_THR_2),CC
62 (P1_THR_3),AC
63 (P1_THR_4),CC
64 (P1_THR_5),CC
65 (P1_THR_6),FB
66 (P1_THR_7),5A
67 (P1_THR_8),D3
68 (P1_THR_9),18
69 (P1_THR_10),C6
6A (P1_THR_11),3A
6B (P1_THR_12),3A
6C (P1_THR_13),3A
6D (P1_THR_14),3A
6E (P1_THR_15),07
6F (P2_THR_0),AC
70 (P2_THR_1),CC
71 (P2_THR_2),CC
72 (P2_THR_3),CC
73 (P2_THR_4),CC
74 (P2_THR_5),CC
75 (P2_THR_6),FB
76 (P2_THR_7),DE
77 (P2_THR_8),F3
78 (P2_THR_9),9C
79 (P2_THR_10),E5
7A (P2_THR_11),2F
7B (P2_THR_12),2F
7C (P2_THR_13),2F
7D (P2_THR_14),2F
7E (P2_THR_15),07
7F (THR_CRC),C3
EOF

1.) We receive approximately 1 CRC error (in received data) in 100 scans. Could this be connected to the noisy waveform that we receive? 

2.) Received waveforms are as follows:

Scan 1: This is an EDD of an empty Burst & Listen scan on P2. The detection at approx 1.5 Meters is the ground below. The system was positioned in an open area with no object around it for at least 15 meters on facing side. There are no sources of noise like motors, machines or other ultrasonic sensors (to the best of our knowledge) 

Scan 2: Is an EDD LISTEN Only scan with the XDCR+ connected to GND in the same exact position. 

The noise seen in scan 1 makes it difficult for us to set thresholds that will give us reliable detection and eliminate False Triggers. 

Any help in eliminating the noise source would be much appreciated.

Please let me know if it is possible to have our PCB design reviewed by a TI engineer. 

Thanks & Regards,

Yash Darad

  • 0
    Prodigy 120 points

    Part Number: PGA460-Q1

    Hi Akeem, 

    My team has been working on the PGA460 and is coming close to a POC of our application.

    We are using it in the Transmit and Receive mode and are facing issues with ghost objects even after setting thresholds according to the empty scan. To be precise, our empty scan values average around 40 and the corresponding threshold has been set to 100. The ghost objects are detected around 15% of the time. That is to say, they are not detected every scan. 

    We have been looking into fine tuning the PGA and to this end want to understand how the BPF, LPF & Threshold Comparator Deglitch settings affect the reflected signal.

    For our 40 KHz transducer, By reducing the BPF from 8 to 4, LPF from 2 to 1, we saw that the over all amplitude of reflected signals decreased. This then meant that the reflected signal did not exceed the threshold as often as it did earlier and this reduced the sensitivity of the system. 

    Currently, the Threshold deglitch time is set to 0. Will changing this value allow us to eliminate ghost detentions?

    Please advice.


    Thanks & Regards,

    Yash Darad

  • 0
    TI__Mastermind 27950 points
    Yash,

    The only setting that can generate a false-positive is the time decouple feature. When the time decoupling switch activates internally during active listening, the internal switch noise unintentionally couples to the receiver path, and can report out as a false-positive. Because the time decouple switch time is fixed based on a register setting (bitfield DECPL_T), the false-positive will always trigger at the the same time. Changing the DECPL_T value will also change the time at which the false-positive occurs.

    To eliminate this false positive, set DECPL_T to the maximum value (b'1111) to force it to activate at 65ms, which is the maximum allowable record length time for the preset configurations. This also means you will not be able to utilize the decouple feature.

    The threshold deglitch timer does not generate false-positives because this is value is only tied to a comparator's digital logic, and does not toggle an actual switch internally.
  • 0 in reply to Akeem Whitehead
    Prodigy 120 points

    Hi Akeem,

    The false positive that we are receiving is not from the decoupling pulse. Based on the FAQ, we have set the Decoupling to Time decoupling at 49 ms which is the length of our Scan. Thus it never shows up on our plot.

    Today, for testing, we ran the BOOSTXL EVM in LISTEN ONLY to see if we could capture the kind of waves that we see on our hardware. Please see the attached image.

    We believe that the above kind of signals are responsible for the false positives mentioned earlier. The above image is of 622  plots in which a marginal percentage are above the normal value.

    What can we do to eliminate the abnormal signals.

    Thanks & Regards,

    Yash

  • 0 in reply to Yash Darad
    TI__Mastermind 27950 points
    Yash,

    Thanks for confirming that the time decouple setting is already beyond the record length time.

    If the device were responsible for any false-positives, I would expect a reoccurring or consistent false-positive signature. Based on your plot, the false-positives appear at random. To prove the device is not responsible for the false-positives, you can ground the XDCR+ terminal, and rerun the listen only test. This will prevent any external acoustic noise from becoming amplified at the front end, and only internal device noise will be reported. If you do not see any false-positives under this test condition, then their may be actual in-band acoustic noise in your test environment.

    Another quick check is to run the normal test case in a different area/outside.
  • 0 in reply to Akeem Whitehead
    Prodigy 120 points

    Hi Akeem,

    Following your suggestion, we connected a ground to the Postive terminal of the Transducer (Pin 1 on J12/J14) and run the scan in Listen Only mode with the previous settings.

    The attached plot is for 1536 consecutive Listen Only scans. To add to this, during the period of the scan, the EVK was being powered by the laptop and the laptop was not being charged.

    Attached below is a screenshot of our settings.

    How can we proceed to eliminate the noise that we are seeing on these scans.

    Thanks & Regards,

    Yash Darad

  • 0 in reply to Yash Darad
    TI__Mastermind 27950 points

    Yash,

    Can you try using a non-switching power supply to the device? A battery would be most ideal. The EVM's USB-to-VPWR boost-circuit may be introducing noise at the receiver depending on the delay between each burst/listen command.

  • 0 in reply to Akeem Whitehead
    Prodigy 120 points

    Hi Akeem,

    All the previous tests that we conducted on our custom hardware were done with Battery Supply and LM317 regulating supply for PGA460 and LM1117 for the microcontroller. 

    On our LISTEN ONLY EDD scan with the xdcr+ pin grounded and the above mentioned hardware, we got the following result.

    We got the same kind of EDD on multiple scans on multiple sets of hardware. When xdcr+ was not kept on ground, we got a much cleaner signal, however there was still random noise in the signal. Like the one that was posted in my first post.

    Through trial and error, we arrived at power supply configuration in which an EDD with xdcr+ pin on ground gave us a much cleaner signal. That configuration required that the PGA460 be powered by LM317 and the microcontroller be supplied by a TPS5410.  With this configuration we receive the cleanest EDD yet with the noise floor close to zero (as seen below)

    All tests were done with a Lead Acid Battery as the primary power source.

    Could you help us understand why switching to the TPS5410 on the microcontroller side allowed us to come up with a clean signal? 

    Thanks & Regards,

    Yash Darad

  • 0
    TI__Genius 11015 points
    Hello,

    We have received your question & will respond by tomorrow.
  • 0 in reply to Bharat Aravamudhan
    Prodigy 120 points

    Hi Bharat, any update on this would be really appreciated! 

    Thanks,

    Yash Darad

  • 0 in reply to Yash Darad
    TI__Mastermind 27950 points
    Hi Yash,

    Considering there are no false triggers in the listen only mode while the XDCR+ is shorted to GND, the noise is either acoustical, or electrically coupling onto the transducer due to radiated antenna effects.

    Before we assume the 10cm UART cables to be the culprit, try increasing the delay between each burst+listen cycle to ensure there are no remnant echoes from the previous burst and/or cover the front of the transducer with absorbent material (i.e. soft sponge) to block any potential acoustic energy during the burst+listen data collection.

    If this does not rule out the false positives, you may need to improve the shielding of the system but shielding the 10cm UART cables, and the PCBs. If you can place the entire system in a shielded (grounded) enclosure with a hole to position the transducer through, this will be your best case test setup.
  • 0 in reply to Akeem Whitehead
    Prodigy 120 points

    HI Akeem,

    The following scans were taken in the spot where we usually test our system.

    The first scan (seen below) is a burst & listen scan with foam taped over the transducer.

    The second scan (seen below) is a burst & listen scan without the foam.

    During both the scans, there was no object (other than the foam in the 1st scan) blocking the field of view of the transducer. Both scans were run with same pulse, current & DSP settings.

    Furthermore, we have observed that the scans that we run during the evening (after sunset)  register more noise. As an example, 

    the below scan was taken at 18:22, before sunset. Same location as the previous scans, same settings (burst + listen with an empty field)

    The next scan was taken 30 minutes after sunset. Same location, same settings.

    Not sure what to make of these observations, hoping you can shed some light on the same.

    Thanks & Regards,

    Yash

  • 0 in reply to Yash Darad
    Prodigy 120 points

    Hi Akeem,

    All the previous scans that we shared were done in our office parking lot, which is an open empty area, except for a few trees in the distance. 

    Yesterday we decided to get scans at a different location. We maintained same settings and hardware configuration as the previous scans.

    The following scans were done after sunset.

    SCAN 1

    SCAN 2

    Scan 1 and Scan 2 were done at a different location whereas the scan attached below (Scan 3) was done in our empty parking lot. 

    SCAN 3

    To the best of our knowledge, there aren't any ultrasonic aggressors in the area. Most of the offices are closed at the time we are running these tests.  

    We also ran scans with shielding in place for the entire system and wires, but got similar noisy results. 

    Please Advice.

    Warm Regards,

    Yash Darad

  • 0 in reply to Yash Darad
    TI__Mastermind 27950 points
    Hi Yash,
    Can you add a large low pass filter close to the VPWR pin?
    If VPWR is not stable, then the internal analog regulator (AVDD) which the front end references will not be stable, and exaggerate the amplified noise.
    You may want to use a spectrum analyzer at times when performance is good and when performance is poor for comparison of the environmental noise differences (i.e. before and after sunset).
  • 0 in reply to Akeem Whitehead
    Prodigy 120 points

    Hi Akeem,

    We added a low pass filter close to the VPWR pin of the PGA. That didn't help much as you can see in the below waveform.

    Unfortunately, we don't have access to a spectrum analyzer. Other than a spectrum analyzer, is there anything that we can do to?

    Awaiting your response,

    Yash Darad

  • 0 in reply to Yash Darad
    TI__Mastermind 27950 points
    Hi Yash,

    Considering the false positives only appear in certain test areas at certain times of the day, I am still convinced that the environment contains some form of ultraosnic aggression. If not the environment, your system may contain a sub-module that creates noise in-band to the ultrasonic transducer's frequency.

    Do you have a different transducer (at a different center frequency) you can test for comparison?

    Other than shielding the transducer, shielding the transducer module, low-pass filtering the PGA460 supply, and isolating the PGA460 ground, there are no other preventative measures I can recommend.

    This is a very interesting error case. I will let you know if I think of other debug tests.
  • 0 in reply to Akeem Whitehead
    Prodigy 120 points

    Hi Akeem,

    While testing our system in the evening, we saw Bats flying near by. We believe them to be the cause of the noisy EDD plots. 

    Can you confirm that this is a plausible explanation? And if so then what can we do to filter out that kind of noise?

    Regards,

    Yash Darad

  • 0 in reply to Yash Darad
    TI__Mastermind 27950 points
    Hi Yash,

    Assuming the bats are the cause of the ambient noise, there is no means of filtering out their signals if in-band to the transducer frequency.

    Based on a quick web search, I found that bat cries range between 20-80 kHz over a 50 kHz window for 1 ms. You may need to consider using a higher frequency transducer to prove the bats are at fault or not.