Re: [Fwd: Re: [Discuss-gnuradio] Gain Range error / Tx LO Offset]

[Matt Ettus]

On 01/21/2010 12:40 PM, Eric Schneider wrote:
> On Wed, 2010-01-20 at 18:11 -0800, Matt Ettus wrote:
> > Because 0.2 amplitude doesn't clip.  On a typical 1800, it will start to
> > clip somewhere around 0.5 to 0.6, and a typical 2400 will be much closer
> > to 1.0.  But in any case, you want to back off on the power to get the
> > linearity you need.  How much linearity you need is very application
> > dependent.
>
> Okay, my use of "clip" wasn't precise.  I was just parroting some recent
> posts regarding 0.2 amplitudes and "clipping".  I haven't measured it
> but eyeballing the compression point does seem to be around 0.5 as you
> state.
>
> Linearity does seem to be an issue above 0.1 or so, based on the
> intermod products I'm seeing.


Well, it depends on what you mean by "issue".  The RFX2400 typical 
midband OIP3 is about 37 dBm, which is about all you can expect for a 
small, low voltage, wideband design like this.


> > There are users that are doing nonlinear modulations where clipping is
> > ok, like GMSK, and in that case they need to be able to get the full
> > power output (well over 20dBm for most boards).  If we scale the analog
> > electronics to give good linearity at digital amplitude of 1.0, then you
> > would not be able to run in saturated mode if you wanted to.  So we
> > scaled it so that it is a compromise between the two usage cases.
>
> Okay, that makes perfect sense.  Just to be clear, I am in no way
> challenging the design; just in case I may have come across that way.

Understood.  I just think it is important to make sure you are really 
seeing what you think you are seeing.

> For my application, I would like to have high linearity and keep most of
> the DAC range. (I don't need high output power)  It would seem that
> reducing the DAC/9862 output gain is a good way (the only way?) to do
> that.

There are 14 bit DACs on the USRP1 and 16 bit on the USRP2.  Even if you 
were to back off to an amplitude of 0.1 (which is probably more backoff 
than you need), that leaves you with 11 bits on the USRP1 and 13 on the 
USRP2.


> Thus my questions regarding testing offset/balance via a flowgraph, and
> if the tx settings are stored on the DB EEPROM (as opposed to having to
> reset them after every power cycle).

I am not sure if you are asking about DC offset correction (which shows 
up as carrier feedthrough on the output) or if you are asking about the 
DC bias points of the differential baseband signals.  I'll try to answer 
both questions...


So DC offset and IQ balance are independent of intermod, and reducing
your amplitude won't help those.  The original intention was to store 
those values on the EEPROM, but the reality is that there are too many 
variables, so active correction is a lot better.

Since the RFX (and WBX) boards have independent LOs for RX and TX, it is 
pretty easy to measure DC offset and IQ balance of the TX.  Just make 
sure the LO for RX is set differently than the TX so you can distinguish 
RX errors from TX errors.  Corrections from there should be relatively 
easy for DC offset and IQ amplitude balance.  IQ phase imbalance is a 
little harder.


If you actually mean the DC bias point, then there is a way to handle 
this, see the next answer.

> But the core question is: Am I correct to assume that I should be able
> to lower the DAC gain as long as I make the required offset adjustments
> as well?


By default we use 20mA as the DAC output current.  You can lower this 
all the way down to 2mA which would give you 20dB lower power.  The 
problem is that reducing the differential amplitude (which is what you 
want) also happens to reduce the DC bias point (which you don't want), 
and the inputs of the IQ modulator can be somewhat sensitive to this. 
You can actually reduce the DAC gain by about 3-6dB without too much 
problem.  You should try it in your case to see if it gives you the 
effect you want.

If you wish to reduce it further, you have another option, as long as 
you will _always_ want to operate it at that reduced amplitude.  You can 
replace R103 and R110 (which are zero ohms) with a higher resistance so 
that the bias point remains 400mV.  See page 18 of the AD8349 datasheet 
for more info on how to do this.


> > On the WBX we have closed-loop analog transmit power control, so you
> > could always run at .75 for great linearity and then can control your
> > power with the low-speed DAC over a 25-35 dB range.
>
> I'm really looking forward to playing with that board.  Quite a design
> feat!

Thank you.

Matt