Re: [Discuss-gnuradio] Daughterboard progress

[Matt Ettus]

Quoting cfk <address@hidden>:

> 1. I have tried to find pin 1 (or pin A1) on the Molex web site to no avail.
> I did find in the Molex reference for their spice model the designators
> A1..A32 on one side and B1..B32 on the other. Do you *really* want to change
> the pin numbering? I will certainly go with whatever you decide. I suspect
> the PADS database will allow me to avoid mirror imaging the connector.

Unfortunately, I din't know about the Molex numbering scheme, so all of my
schematics are in teh 1..64 format.  You can number it how you like in your
schematics.

> 2. It may be that we might want to use the AUX_DAC_A line instead of a
> MAX518 I2C DAC for the AGC. Is this what you might want on this
> daughterboard?

It was intended that the AUX_DAC lines be used for this kind of purpose. 
Unfortunately the AGC tuning range on the 4702 is up to 4 volts and the AD9862
will only output up to 3.3 on the DAC outputs.  So there are a number of
options here, the most straightforward being to use the MAX518 anyway. 
Otherwise you could use an opamp to adjust the output range of the AUX_DACs, or
maybe even a resistor and diode[s] to the +5V rail to extend the range.

> 3. I wonder if you have a source for T1 (or T2, I assume they are
> identical). Having a source for this part would allow me to get the
> footprint right. Without it, I currently have an 8pin DIP that we can solder
> the leads into. That usually works for transformers, but the "right"
> footprint is always better. I know that we could use many different
> transformers. I would just feel more comfortable with at least one source.

T1 and T2 in the posted schematics is the ADT1-1WT (high freq SMT version of the
T1-1T) from Minicircuits, but it isn't critical, and many applications will not
use a transformer at all.  This one probably will, though, since the IF freq is
36 MHz and you don't care about DC.

> 4. I assume that in the case of the 4702, that we are going to go through a
> transformer with its Ifout1 and Ifout2 on one side, and then to VINP_A and
> VINN_A on the other side?

Yes

> Perhaps with or without impedance resistors?

The resistors are there to terminate the input at 50 ohms, which is important
when using the coax connectors.

> I think the 4702 is 300ohm output, so we will need to go from 300ohms to
> 50ohms to match the AD9862 input. I'll take a swag at that and send you the
> schematic and layout.

The 9862 has a 200 ohm input impedance when you don't use the terminating
resistors.  A 1:1 xformer (like the ADT1-1WT) would not provide a perfect match
to the 300 ohms of the 4702, but might be close enough.  Also, is it 300 ohms
differential or 300 ohms to ground each (600 ohms differential)?

On the other hand, the resistors provide a nice constant impedance at the
expense of some additional loss, so matching to 50 ohms wouldn't be so
horrible.

> 5. Do you want me to link the schematic and layout files from my web site
> http://home.pacbell.net/cfk similarly to the way you have done with the Rx &
> Tx daughterboard schematics. I can update them from my end and include the
> Orcad data base files for the schematic and the layout in addition to the
> .pdf files.

That would be great.  Unfortunately our wiki does not allow you to directly
upload files.

> 6. I dont know what the output level of the 4702 is yet. Do you know if it
> is more or less then 100mv P-P? If greater then 100mv, no problem. If less
> then 100mv, we may lose some dynamic range without a buffer.

I don't know.  I do know that the 9862 input range is either 2V p-p for best
noise performance or 1V for lowest distortion.  The PGA can provide a 0 to 20
dB gain, so that would equate to 200mV or 100mV inputs, respectively.

A couple of other things...

Since the 4702 is rather large and there is no directly comparable wide-range TX
system, you might consider just making this a full-sized RX-only board (115mm).

Also, I'm not sure if these are the vertical or horizontal mount versions, but
it might be possible to fit 2 of them on one board.  Since the output is real,
not I/Q, you could put one into each channel.  This would allow a fully
populated deluxe USRP system (2 RX boards) to have 4 input channels.  For those
only needing one input, you could just not populate the second 4702 and
associated circuitry.

Matt