[Dprglist] Broadcom AFBR-S50LV85D vs. ST Micro 53LCX

[Paul Bouchier]

Hi John. I was too lazy to describe the specifics of Broadcom's
anti-aliasing algorithm, but I should have known someone would call me on
it - and who better than you :-)

The short-range or long-range setting, which results in a max unambiguious
range of 6m or 12m respectively, is done by reducing the transmitter
modulation frequency for long range, so TWO_PI is further away. The
downside of that is fewer samples/sec.

If you want to go beyond 12m, they have a dual-frequency mode where they
alternately modulate with one frequency then another that is not
harmonically related. I don't totally understand the phase detection, but
somehow that results in there being only one of the multiple ambiguous
ranges at each modulation frequency that satisfies both frequencies when
considered together. The tradeoff again seems to be fewer samples/sec. For
details, read the app note.

So, not like DDR link training. And your email came through fine - see
below.

Paul

On Thu, Mar 24, 2022 at 4:04 PM <dprglist-request at lists.dprg.org> wrote:

>
> ---------- Forwarded message ----------
> From: John Swindle <swindle at compuserve.com>
> To: "dprglist at lists.dprg.org" <dprglist at lists.dprg.org>
> Cc:
> Bcc:
> Date: Thu, 24 Mar 2022 01:38:30 +0000 (UTC)
> Subject: Re: [Dprglist] Broadcom AFBR-S50LV85D vs. ST Micro 53LCX
> Paul,
>
> Regarding Range Ambiguity: During physical layer training on high-speed
> interfaces such as DDR5 and PCIe where flight times exceed one or two ticks
> of the clock, it is possible to lock onto an alias of the training sequence
> and be off by one (Help me, Off-by-One Kenobi. You're our only hope.) or
> off by two. One trick is to use training sequences that have gaps or
> half-frequency sequences. If the receiver locks onto the wrong clock, the
> half-frequency sequence will fail predictably making it easy to immediately
> adjust the lock.
>
> I suspect the rangefinder is doing the same thing, by simply failing to
> emit some of the pulses. If the receiver reports a result for a pulse that
> wasn't emitted, the receiver is off by a cycle. Easy to detect and easy to
> process. A single detector could handle multiple ranges ongoingly. Same
> thing was done in the 765 Floppy Disk Controller with intentionally
> suppressed clocks.
>
> Pseudorandom sequences prevent aliasing but are harder to correct. Most
> high-speed training is still done with fixed, non-random patterns, though
> DDR5 has two LFSRs.
>
> I suspect that my emails to DPRG list are not posted. Would someone let me
> know they saw this?
>
> Later,
> John Swindle
>
>
> -----Original Message-----
> From: Paul Bouchier via DPRGlist <dprglist at lists.dprg.org>
> To: DPRG <dprglist at lists.dprg.org>
> Sent: Wed, Mar 23, 2022 2:57 pm
> Subject: [Dprglist] Broadcom AFBR-S50LV85D vs. ST Micro 53LCX
>
> At last night's RBNV, Doug P introduced us to the Broadcom AFBR-S50LV85D
> TOF range sensor. It looked interesting, with a max range of 30m, and I
> wondered how it compared to the ST Micro 53LCX, which I'd analyzed
> previously and reviewed in a DPRG list post on Dec 15 2021. That review is
> now on my site at
> https://sites.google.com/site/paulbouchier/home/analyses-blog.
>
> I found an excellent application note here:
>
> https://www.google.com/url?sa=t&rct=j&q=&esrc=s&source=web&cd=&ved=2ahUKEwiv2qCiudz2AhWGl2oFHR7yAq0QFnoECAcQAQ&url=https%3A%2F%2Fwww.avnet.com%2Fwps%2Fwcm%2Fconnect%2Fonesite%2F217f3b0e-788e-44c0-a63a-9595debc8370%2FApplication%2BNote%2BAFBR-S50%2BToF%2BBasics.pdf%3FMOD%3DAJPERES%26CVID%3DnuRdfj6%26CVID%3DnuRdfj6%26CVID%3DnuRdfj6%26CVID%3DnuRdfj6&usg=AOvVaw1D41jdve_frPH2eVGMTVF2
>
> Like the ST 53LCX family of sensors, Broadcom offers a family of sensors
> under the AFBR-S50LV, and the 85D version is optimized for long range. Like
> the 53LCX family, they use "Inferred Time-of-Flight" (iTOF) for distance
> measurement, which as John Swindle observes, isn't direct time-of-flight
> measurement, but is phase measurement of the return of the modulated
> transmit signal cross-correlated with the received signal.
>
> Someone last night observed that the Broadcom sensors have a linear (8 x
> 4) array of sensors and wondered if this is a linear array sensor like the
> old Sharp sensors, which used parallax to measure distance as indicated by
> which sensor in a linear array received the reflected light. Not really.
> The linear array enables the sensor to detect nearby and distant objects,
> whose reflections would fall in different places, so it accommodates
> parallax, and it  looks at several factors including amplitude to choose
> which pixels to use for distance reporting, but it doesn't use parallax for
> distance measurement.
>
> The Broadcom sensor family varies the diameter of the transmit spot in
> different models. The 85D is long range because it uses a very narrow beam
> that consequently produces a stronger reflection from distant objects than
> the ST senstors and has better range in bright sunlight. By contrast, the
> ST sensors don't seem to have such narrow beams - they illuminate the whole
> of the sensed region. This lets the ST sensors offer selectable areas of
> the array for sensing, or to sense multiple areas, producing a multi-zone
> sensing. It's not clear to me whether the Broadcom 85I version can also
> produce multiple sense zones.
>
> Unlike the ST sensors, the Broadcom ones don't appear to offer
> multi-target capability (i.e. multiple returns at different objects
> reported as different objects within the same zone.
>
> The broadcom application note brought up an interesting facet of
> phase-difference distance measurement which didn't appear in the ST
> documentation: Range Ambiguity. There is a range, dependent on the
> modulation frequency, beyond which a reflection will appear to be much
> closer than it is. This is because the phase angle of the return exceeds
> TWO_PI, so it thinks the object is a phase closer. This distance is 6m or
> 12m depending on the short or long range setting of the sensor. Maybe the
> ST sensors, being shorter range, don't suffer from this problem because
> they can't detect anything further away - I don't know. Broadcom has some
> techniques to mitigate this.
>
> In summary, they're similar kinds of sensors, both sophisticated, and with
> versions optimized around various applications. Which is better is
> application-dependent, and ya pays ya money an' ya takes ya choice. 😁
>
> I would have thought that for obstacle sensing for small, slow robots, a
> short-range multi-zone wide-angle sensor would be better, whereas for an
> automobile, a single-zone long-range sensor would be better. But there
> might be other considerations - feel free to chime in with your thoughts.
>
> I hope this was useful.
>
> Regards
>
> Paul
>
>
> _______________________________________________
> DPRGlist mailing list
> DPRGlist at lists.dprg.org
> http://lists.dprg.org/listinfo.cgi/dprglist-dprg.org
> _______________________________________________
> DPRGlist mailing list
> DPRGlist at lists.dprg.org
> http://lists.dprg.org/listinfo.cgi/dprglist-dprg.org
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.dprg.org/pipermail/dprglist-dprg.org/attachments/20220324/2559a206/attachment.html>