Differences

This shows you the differences between two versions of the page.

--- resources:eval:user-guides:ad-trxboost1-ebz:practical [06 Jul 2016 23:32] – [Maximum input to the FIRs] Robin Getz
+++ resources:eval:user-guides:ad-trxboost1-ebz:practical [15 May 2017 18:55] (current) – [AD9361 Block Diagram] Lars-Peter Clausen
@@ Line 7: / Line 7: @@
 This block diagram is shown to help with the discussion below, to better understand the Tx and Rx signal chain.
-{{:resources:eval:user-guides:ad-fmcomms2-ebz:fmcomms2.svg?800x650|}}
+{{:resources:eval:user-guides:ad-fmcomms2-ebz:ad9361.svg?650x450|}}
 ===== Transmit =====
+You would not be using the TRXBOOST board, unless you wanted maximum output power, so let's understand how to do that without sacrificing dynamic range, or causing distortion.
 ==== Maximum input to the FIRs ====
-The first step on understanding what sort of digital input can be provided to the FIR filters without causing the FIR/Half Bands to saturate/overflow. Normally this gain is added to compensate for losses in the half bands, and analog filters.
+The first step on understanding what sort of digital input can be provided to the FIR filters without causing the FIR/Half Bands to saturate/overflow (and cause distortion). Normally the FIR filter has some small gain to compensate for losses in the half bands, and analog filters.
 Luckily - the [[../ad-fmcomms2-ebz/software/filters|MATLAB Filter Design Wizard for AD9361]] will tell you this.
@@ Line 19: / Line 21: @@
 {{:resources:eval:user-guides:ad-trxboost1-ebz:filterwizard_snip.png}}
-Where that comes from is the compensation of the analog filters. If we just look at the composite response, the overall flatness is within 0.05 dB (very flat) - even when we zoom in (the next picture), we can see it is very flat. However, we can see that in the passband, the analog filters are causing ~ 0.5dB of droop (which the FIR is compensating for).
+This gain is required to compensate the losses in the pass band the analog filters cuase. If we just look at the composite response, the overall flatness is within 0.05 dB (very flat) - even when we zoom in (the next picture), we can see it is very, very flat (still 0.05dB). However, we can see that in the passband, the analog filters are causing ~ 0.5dB of droop (which the FIR is compensating for).
 {{:resources:eval:user-guides:ad-trxboost1-ebz:fvtool_composite.png?420|}}
 {{:resources:eval:user-guides:ad-trxboost1-ebz:fvtool_composite_zoom.png?420|}}
-We can clearly see the FIR compensation if we look at things just by themselves. There is a small amount of peaking (0.5683dB) to make up for the droop. Since this is frequency dependent - with a CW tone at low frequency, we may not notice it, but with wide band signals - this will cause distortion.
+We can clearly see the FIR compensation if we look at the FIR just by itself. There is a small amount of peaking (0.5683dB) to make up for the droop. Since this is frequency dependent - with a CW tone at low frequency, we may not notice it, but with wideband signals - this will cause saturate/overflow and which can be measured as distortion.
 {{:resources:eval:user-guides:ad-trxboost1-ebz:fvtool_fir.png?420|}}
 {{:resources:eval:user-guides:ad-trxboost1-ebz:fvtool_fir_zoom.png?420|}}
-This indicates that the maximum signal that can go into the Tx FIR (if this was the Tx FIR being designed) is either 0.93667 of full scale, or -0.5683 dBFS. This means, that rather than a 12-bit (4096 bits) system, you really have a 3836.6 bit system, or a 11.9056 bit system. Backing off a little on the dynamic range (.1 bits) is pretty common for many sorts of systems, including communications systems.
+This indicates that the maximum signal that can go into the Tx FIR is either 0.93667 of full scale, or -0.5683 dBFS (assuming 0dBFS is full scale). This means, that rather than a 12-bit (4096 bits) system, you really have a 3836.6 bit system, or a 11.9056 bit system (loosing those 261 bits out of 4096 bits doesn't really change the dynamic range much from a practical standpoint). Backing off a little on the dynamic range (.1 bits) is pretty common for many sorts of systems, including communications systems.
-This specific number will vary for your specific analog RF bandwidth, your specific half band settings, and overall RF setup.
+This specific number will vary for your specific analog RF bandwidth, your specific half band settings, and overall RF setup. You should check the [[../ad-fmcomms2-ebz/software/filters|MATLAB Filter Design Wizard for AD9361]] for your actual design.
-==== The Digital to Analog Converter ====
+==== The Digital to Analog Converter (DAC) and modulators ====
 Also keep in mind that in this specific case - scaling back 0.93667 ensures that there isn't any overflow/saturation in the digital sections, and will still cause full scale signals to be provided to the DAC.
+As these Full scale I and Q signals (assume for now ±1), as these baseband signals are modulated with the LO in phase and quadrature, and added together, the magnitude of the resulting signals grows to ±<m>sqrt{2}</m> with a phase of ±π.
+This is why there is an attenuation block in the modulators.
+To take advantage of the dynamic range in the system, you want to drive as close to full scale into the DAC, but not overpower the output stage (causing saturation, and again ultimately distortion). To do this - it's nearly always required to have some attenuation in the output section.
 ===== Receive =====

Wiki

Analog Devices Wiki

Differences

Analog Devices Wiki

Differences

Page Tools