ja4gii-AFPSN

For an all-pass filter type PSN transmitter/receiver, an I/Q signal (satisfying a 90-degree phase relationship) of the required bandwidth is required. This filter combines elements that shift 90 degrees in several divisions within the required band, so that the 90-degree phase can be maintained overall. Increasing the number of divisions narrows the 90-degree maintenance interval and compresses the overall 90-degree phase error. In other words, if the interval is wide, the period between the pole elements that move away from 90 degrees becomes longer and the 90-degree error becomes larger. Conversely, if the interval is narrower, the 90-degree error becomes smaller and the reverse side suppression characteristics improve. There are various documents on how many stages to design, and JA3GSE has released the 'Hamstool software', so you can use these to design what you need.

There are also convenient tools on the Internet, such as the 'J-Tek AllPass Filter Designer' tool on the 'GJ3RAX' site, and 'Apf.exe' is uploaded at the bottom of the following URL. Download this and run it to design various AFPSNs.
　　　　　　　　　　　　　　　　　　　　　　　　　https://www.gj3rax.com/apf.htm
　　　　　　　　　　　　　　　　

Enter the required bands F1 to F2, select the required number of stages, and click "Design" to get the answer for C and R.

It has been more than 10 years since I started making PSN machines using all-pass filters, but after trying various things, I have found that there are no many types of all-pass filters to use in PSN transmitters and receivers. In the analog world, the limit is 60 to 70 dB, and even if you theoretically design it to be 80 to 90 dB, even if you can achieve one point (instantaneous), it will immediately deviate due to environmental factors, and the stable point will be fixed at around 60 dB. I have heard that many stages have been made, or that it has been designed to be 90 dB, but if you think about it, SSB radio waves are in the 3 KHz band, so in the case of PSN, the high frequencies are limited by an audio-lopass filter, but even if you use a filter with a lower order, if you set Fc = 3.5 KHz, the 4 KHz component will hardly be reproduced in the radio wave propagation, and the higher the frequency, the smaller the energy. Of course, Fc should be lowered and managed so that it does not exceed 3 KHz. Furthermore, since it is the opposite side component of this, when you consider how much high frequencies need to be kept in phase as an all-pass filter, 4 KHz or more is not necessary at all. (Over-spec) As for the lower end, a good guideline would be 1/2 of the first formant I generate. In my case, it is 90Hz/2 = 45Hz, and anything below this is unnecessary. Providing more stages than necessary only increases the number of fluctuating factors unnecessarily, and is a "a hundred times more harmful than a thousand miles." In that case, in my experience, I think that 6 stages is the most ideal AFPSN for use in amateur radio.

In this section, I will introduce some all-pass filters I have designed for use in hybrid SSB generators, which process only the opposite side of the low frequencies, and process the opposite side of the mid-high frequencies with a filter, in combination with a PSN that can transmit low frequencies sufficiently, even if it is not a filter with a good shape factor, such as a ceramic filter (potato filter), when creating an SSB modulator. I think that with these, you will be able to match with most filters.

2nd stage　AFPSN

2-stage 50dB Circuit diagram file	50dB Characteristics file
2-stage 60dB Circuit diagram file	60dB Characteristics file

3rd stage AFPSN

3 stages 50dB Circuit diagram file	50dB Characteristics file
3 stages 60dB Circuit diagram file	60dB Characteristics file
3 stages 70dB Circuit diagram file	70dB Characteristics file
3 stages 80dB Circuit diagram file	80dB Characteristics file

6th stage　AFPSN
The ultimate -94dB, which can be used to match even potato filters

6 stages 94dB Circuit diagram file

94dB Characteristics file

Most commonly used 6-stage PSN 20Hz to 4KHz (70dB)
Using this 6-stage PSN as an example, we will use LTSpice to examine the amplitude balance of the U and L columns.


The characteristic on the right is when the gain resistance error = 0. From this point on, the gain resistance value is changed and the amplitude difference is extracted from the characteristic, entered into the blue cell of the Excel table, and the converted value of the opposite side suppression ratio is calculated and examined.	The equivalent value is 124dB, so there is no need to worry about it.
The amplitude difference characteristic when the gain resistors on the U row side are 9.9K and 10.1K, making the gain of all stages 1% higher, and the gain on the L row side all stages lower, is a suppression ratio of just under 20dB. However, in reality, this arrangement is never implemented. This is the worst case scenario of using 1% error resistors as is, but it is not realistic. Try setting the gain difference to a small value.	There are many combinations, but here is the gain difference characteristic when the U-row 1/3/5 stages = large gain, 2/4/6 stages = small gain, and the L-row 2/4/6 stages = large gain, 1/3/5 stages = small gain are alternately changed by 1%. This is also low in reality, but the suppression ratio is 45 dB.
When the gain resistors are alternately changed by 0.1%, a further ten-fold increase in the precision of the resistor values results in an improvement of 20 dB, and the suppression ratio becomes 65 dB.	In reality, the resistors are randomly placed, so the suppression will be better than 65 dB. In conclusion, even if you use a gain resistor (10KΩ) with a tolerance of 1%, use one that is at least 10KΩ resistors that are matched to within 10Ω using a multimeter. This does not require matching all the parts, but only matching pairs of two, so they can be easily extracted from a parts box. It is better to use a resistor that is matched with an ohmmeter than a mechanism that adjusts the amplitude with a semi-fixed VR.

振幅補正回路

If an in-band amplitude difference occurs between the I/Q signal outputs, it will cause the opposite side to deteriorate. To prevent this, an amplitude correction circuit is inserted into the pole of each stage. Amplitude correction circuits are not inserted into stages where the pole frequency is outside the phase band. (As a fixed resistor.) For adjustment methods, please refer to the Phase Meter/Amplitude Meter section.

90.1,89.9