EMRFD Message Archive 10762

Message Date From Subject
10762 2015-02-06 13:56:03 bob_ledoux Some Thoughts on Phase Delay Networks for DC Receivers
Some builders use fixed components for critical active phase delay networks.  I use trimmer pots to optimize circuit performance and simplify future adjustment.

On page 9.13 of EMRFD, Rick Cambell makes the argument that 40db of opposite sideband reception makes for a good sounding single signal DC receiver.  On page 9.29 is the statement that this performance can be realized by using 1% components.

On page 113 of the Norcal NC2030 transceiver builder's manual Dan Tayloe says, "After the phasing strip has been aligned, the opposite side band will be rejected a minimum of 45 db,with much of the bandpass having 50 to 55 db of opposite sideband rejection."  The Tayloe phasing network is adjusted, by ear, using a pair of trimmers to maximize the unwanted sideband suppression.

I was curious about the impact of component variation on phasing network performance.  On page 9.31 of EMRFD we see four graphs to display the performance variation using 1% to 0.1% tolerance parts.

You can duplicate my simulations using Jim Tonne's QuaNet software that is included on the CD with the 2015 ARRL handbook.

I started by duplicating the phasing network in the NC2030 transceiver.  Two pair of op amps cover a range of 300 to about 800 Hz.  

The QuadNet Network Design menu has a bottom frequency of 300 Hz and a Top frequency of 800 Hz.  The Order is set to 4.   QuadNet treats each op amp as an "order" while this term refers to an op amp pair in EMRFD.  So we have a phasing network consisting of two op amp pairs.

Using zero tolerance calculated resistor and capacitor values the program generates a minimum suppression of -67.57db.

If resistors are replaced with nearest 1% values, assuming again these components have zero tolerance, the suppression falls by 20db to about -47db.

Using the Monte Carlo technique we can see the potential variation assuming a 1% error in the selected components. The performance falls another 10db at the higher frequencies to about -37db.  At low frequencies its falls to about -41db.

These simulations demonstrate how small component variations impact phasing network performance. These are also discussed in EMRFD on page 9.13. Few experimenters have test equipment capable of reading to these tolerances.  I solve this problem by using lower tolerance parts and trimmers to adjust by ear.

But if -40db suppression provides a good sounding receiver, are we vain if we seek the other potential -37db?

10764 2015-02-06 15:14:21 Tayloe, Dan (NSN ... Re: Some Thoughts on Phase Delay Networks for DC Receivers
I think the reason that I found the phasing adjustments so useful was that I was using the dual 1:4 mix part as a detector.  Rick, KK7B recently noted that delay in that part caused as much as 3 degrees of phase errors.

In addition, I have found that the R/C roll off of the detector to be a bit asymmetrical when the detection bandwidth is cranked down.  I think of it as detector "back spin".  The adjustable phasing fixes that as well.

- Dan, N7VE

Sent from my Windows Phone

10765 2015-02-06 16:41:58 kb1gmx Re: Some Thoughts on Phase Delay Networks for DC Receivers
To answer the last question No.  the 37dB  was for a simpler microR2 which is
by all means the simpler minimum performing but most complete of all the 
R2 series radios.  All the rest do minimally 7-10db better.

My microR2 SSB is running at 37 near the limit for the network (order 2) can do.

My two miniR2s using a order 3 all pass are at 44-45(kanga kit 1% parts) and 48
(fixed frequency unit, hand picked parts) dB measured real world numbers.  Its 
hard by ear to tell them apart.  Half of every small is small.  I stopped at that level
as it was better than the signals on the air.

In both cases that alone didn't determine the opposing sideband rejection.
Getting the I/Q  LO at 90 degrees and equal levels went far in getting the attainable.
In the one miniR2 the resistor values were measured on a bridge to very high accuracy
(better than 1%) and the resulting all pass measured at 54db. but getting the levels to it
and I/Q phase correct its also part of the deal.  Any error anywhere nets a reduced result.
Then any distortion would also lower the result.  When you get into the mid 40s you have
manage everything to a fine degree.  Run the numbers its dirt simple to make a 60dB
4th order all pass for SSB (or CW).  Then calculate how mush error in amplitudes 
applied and phase error accumulated up to that point. decreases the system result.

I like QSD/Tayloe but the propagation delay in the flipflops, FST switches, and board 
layout all contribute increasing phase error with increasing frequency.  Its correctable
but consider how it would have to be done. 

You can dial it all in but if your tuning it over a wide range you need to consider what 
changes and how it impacts the result.  

Its not a single point fix.