[USRP-users] ADC + DAC Compare of N200 and B200
marcus.mueller at ettus.com
Sat Oct 14 10:41:16 EDT 2017
it gets even a bit more complicated: While the samples coming out of the
ADS62P4 are pretty raw (though that thing can FIR filter things itself,
but IIRC we don't do anything fancy there), the AD9361 does have a
multistage DSP chain – so, the bitwidth coming out of the AD936x might
not be very representative for the signal that was sampled – though UHD
only picks very specific parameterizations for that chain. So, this
might really be a case of "it's a bit too complicated to reduce it to a
number of bits". For example, the ADS62P4 is (as far as I can tell) a
pipeline ADC, whereas the AD9361 is a ΣΔ ADC.
That doesn't imply the information of a bit of output of both ADCs is
inherently different, but remember that one of the classical reasons to
go for ΣΔ is that you can shape the noise to be at frequencies out of
the band you're interested in very easily – the integrator already kind
of does that for you. From the top of my head, if you build a pipeline
ADC, you basically build something that would lend itself very nicely to
be looked at in a "that's just a (slightly strange) FIR where analog
goes in and digital comes out" and thus, to filtering "on the fly", but
since the pipeline ADC doesn't actually oversample, there's nowhere to
shape your noise "away" to, so that's no better than doing a FIR (with
more degrees of freedom) afterwards.
On 13.10.2017 13:29, Kevin McGuire via USRP-users wrote:
> Derek or anyone,
> Do you know if it's a 12-bits complex on the b200 (6-bit I + 6-bit Q)
> or a 24-bit complex? I always thought it was a 24-bit complex but from
> what you said and what I see in the schematics it is plausible things
> may not be as they seem.
> This is ultimately my question but I have a weirdness of going around
> the world to ask it, haha. The exact precision used of the ADC before
> the FPGA for comparison is what I am after - but was also after any
> details too.
> Your information is most helpful and id assisting me in getting closer
> to understanding. I sincerely appreciate your help/wisdom and I am
> most thankful for people like you (and others) who voluntarily take
> their time to answer questions. I would know nothing if not for the
> help of people.
> I also feel like a lazy dummy because I just realized I could have
> looked at the schematic and seen how the, what appears to be LVDS
> enabling pins, are configured/connected.
> On Oct 13, 2017 5:02 AM, "Derek Kozel" <derek.kozel at ettus.com
> <mailto:derek.kozel at ettus.com>> wrote:
> Hello Kevin,
> No, the N200 and B200 do not change the electrical transport modes
> based on the frequency or bandwidth requested by the application,
> there is no need. The 12 and 14 digital bits are available at all
> frequencies. The actual effective number of bits out to the host
> depends on the ADC performance and the decimation in the FPGA DSP.
> I have not looked in depth at the design decisions that went into
> selecting the interfaces but I'm confident that the hardware
> designers verified the signal integrity.
> You are correct about the use of the two output DAC. For most
> daughterboards they are used for I and Q. For the others such as
> the LFRX and LFTX the two channels can be used either as a complex
> pair or two real valued channels.
> On Fri, Oct 13, 2017 at 5:21 AM, Kevin McGuire via USRP-users
> <usrp-users at lists.ettus.com <mailto:usrp-users at lists.ettus.com>>
> This is connected to me investigating power level but the
> question is specifically about the number of bits per RX and
> TX channel between the N200 and B200. However, I fell into the
> rabbit hole that Alice went into and I seem to be stuck for
> the moment in determining what I am missing.
> I looked at the datasheet for the B200 and the AD9361. The
> AD9361 is advertised as using a 12-bit ADC and DAC. However, I
> see that it can operate in a 6-bit mode using the inputs as
> differentials - i suspect that is useful at ever higher
> frequencies. I counted 24 data pins.
> Now, the N200 uses this ADS62P42 which is a dual channel ADC.
> When looking at the datasheet it has 14 output ports per
> channel and two channels. I think this chip can also run in
> LVDS (differential) or CMOS (single-ended) modes. I am
> guessing once again LVDS gives only 7 effective bits per
> channel and CMOS gives 14 effective bits per channel.
> It seems that the AD9361 has two data ports where each port
> uses 12 of the 24 pins. Each of the data port pins can be an
> input or output. And, from what I can read it seems like these
> data ports can be combined in various ways, such as:
> 12-bit RX channel + 12-bit TX channel [FDD]
> I = 6-bit and Q = 6-bit (single data rate)
> 24-bit RX channel + 24-bit TX channel [TDD]
> I = 12-bit and Q = 12-bit (single data rate)
> 12-bit RX1 + 12-bit RX2 + 12-bit TX1 + 12-bit TX2 [FDD]
> I = 6-bit and Q = 6-bit (double data rate)
> 12-bit RX1 + 12-bit RX2 + 12-bit TX1 + 12-bit TX2 [TDD]
> We can have only two RX or TX active at the same time.
> (double data rate)
> Is this correct? And, which modes are actually used?
> The N200 I believe I have a handle on because it appears more
> straight forward. I can see how the RX provides two
> independent DAC outputs and each are 14-bit. If I have a
> handle on it then one should be the in-phase and the other the
> quadrature or two separate in-phase depending on the daughter
> Then, I also wonder, does the FPGA/firmware switch between
> CMOS and LVDS on either the B200 or the N200 in order to
> improve quality at higher frequencies?
> USRP-users mailing list
> USRP-users at lists.ettus.com <mailto:USRP-users at lists.ettus.com>
> USRP-users mailing list
> USRP-users at lists.ettus.com
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the USRP-users