455khz phase shift network

[neazoi]

hello I need to build a 0, 90, 180, 270 degrees phase shift network for 455KHz.

I could do it with a dual flip flop dividing a 1.82mhz crystal down to 455khz in 4 phases but the 1.82mhz crystal is obsolete.

Another way I am thinking is to make a 455KHz crystal oscillator and take the 180 degrees phase out of the collector to drive an amplifier.
Then another amplifier will take the 0 degrees from the emitter and drive another amplifier. At the ends of the two amplifiers I will have 0, 90, 180, 270 degrees phase shift.

Is my consideration right or do you propose another schematic?

 

[flatulent]

You will only have 1 and 180 no matter how many stages and takeoff points. How accurate is the phase shift requirement? You could use some rc networks for the phase shift if the accuracy needed is not high.

You could generate the 4x frequency with a PLL and divide by 4 in the feedback loop and then use your Johnson counter method.

 

[betwixt]

I'm guessing this is for a "third method" (Polyphase) SSB generator. The flip-flop method is a good one but I would suggest you use either 3.64MHz or 7.28MHz and divide by either 2 or 4 before the quadrature generator to ensure it is fed with a good 50:50 square wave. These crystals are not expensive, you can have them made in single quantity to any reasonable frequency you like.

Brian.

 

[neazoi]

You will only have 1 and 180 no matter how many stages and takeoff points. How accurate is the phase shift requirement? You could use some rc networks for the phase shift if the accuracy needed is not high.

You could generate the 4x frequency with a PLL and divide by 4 in the feedback loop and then use your Johnson counter method.

It should be quite accurate but I do not know if good accuracy RC can achieve this. Anyway the AUDIO polyphase network for this SSB modulator is made of 1% capacitors tolerance and 0.1% resistors tolerance, so I suppose this is quite accurate...? The 0-180deg input phases of the audio section are taken out of the collector and emitter of a transistor as these should be out of phase 180deg (how accurate is this?)

To be honest I tried the divider method and the first two phases were accurately placed, but the last two seemed to be out of desired phase a bit on the oscilloscope.

Do you believe accurate RC components (NP0 caps and 0.1% resistors) will be able to provide the accuracy of the flip flop divider, as betwixt has described previously??

 

[FvM]

The phase accuracy of a frequency divider will be affected by logic delays, if they are not cancelled by design. With a good design, the phase error will be reduced to the delay skew of the respective logic family, which shouldn't exceed a few ns with fast devices (e.g. 74HC family), but may be still too high with slow ones (as CD4000). If design involves systematic delay differences (ripple clock, different propagation delays for individual phases), even a fast logic family will give poor results.

 

[G0HZU]

Get a commonly available (and cheap) 16.384MHz crystal and divide by 9 to get 1.820MHz. Then use flip flop dividers to get all 4 phases at 455kHz.

The divide by 9 can be done with a 74HC161 device. You could do the whole circuit including the crystal for little more than $1.

You would have to pull the 16.384MHz crystal down by 4kHz but this should be easy to do.

 

[neazoi]

I was also thinking of mixing two transistor crystal oscillators with 455KHz crystal frequency difference to get a 455KHz output signal. then a 455KHz ceramic filter could remove the unwanted mixing products and a passive phase network could do the rest of the phase dividing.
Or I could just use a 455KHz resonator transistor oscillator. The problem in this case is the RF shift network.

By the way the only simple RF phasing network I have found is the one attached. But Is this bridge network a 4-phase divider or two?
Also I have no component values for it

 

[FvM]

The network gives only meaningful results for a +/- 45° phase shift, so it's apparently intended to generate a 90° carrier phase difference.

 

[neazoi]

The network gives only meaningful results for a +/- 45° phase shift, so it's apparently intended to generate a 90° carrier phase difference.

I have found this phase shift oscillator Phase-shift oscillator - Wikipedia, the free encyclopedia topology that uses RC to create different phases of an input signal but I have not found a calculator to easily determine the values.
I could use one of these opamp circuits but I like discrete components design

 

[mvs sarma]

I saw one article deriving all the needed 4 signals using 455KHz resonator.
the sch is attached here. it was published in some QEX mag of 2009. it is a complete SDR fixed at 455 IF frequency.

 

[neazoi]

I guess the 4 phases comming out of the IC2?

 

[mvs sarma]

I guess the 4 phases comming out of the IC2?

YES, and the article is attached as pdf thanks to the author. even if language is Slovenian, the schematics could help
i made a minor correction on power feed to 5532 and the 4066bias section, as it felt original, there appears to be a mistake.

 

[betwixt]

No!

The four phases are coming out of IC1 not IC2. The schematic is an IQ demodulator using IC2 as a mixer. The output of IC2 is actually at audio frequencies.
IC1 pins 4 and 6 are 180 degrees out of phase and so are pins 8 and 10, they are the same signal and it's inverted copy.
The 90 degree phase shift is created by L1, C2 and C3. So what IC1 produces is 0 and 180 degrees on the top part, 90 and 270 degrees on the lower part.

Brian.

 

[neazoi]

No!

The four phases are coming out of IC1 not IC2. The schematic is an IQ demodulator using IC2 as a mixer. The output of IC2 is actually at audio frequencies.
IC1 pins 4 and 6 are 180 degrees out of phase and so are pins 8 and 10, they are the same signal and it's inverted copy.
The 90 degree phase shift is created by L1, C2 and C3. So what IC1 produces is 0 and 180 degrees on the top part, 90 and 270 degrees on the lower part.

Brian.

I am susceptible using this phase shifting using LC configuration. the tolerances must be high and the C must be NP0 for phase shift maintainance. the divide solution seems more stable...?

 

[mvs sarma]

I totally agree and that is stable across a broad range depends only on the device selected, like 74HC(AC)74.

 

[betwixt]

I use a quadrature divider in my PAL color encoder which works in the same way. If you cross-couple two 'D' flip flops you can get exact 90 degree shifts but you have to start with four times the frequency. In other words you inject 1.82MHz and get four 455KHz outputs, each 90 degrees shifted from each other.

In reality, a few degrees of phase change will hardly be noticeable, unlike RF balanced modulators, the carrier will not leak through the system if it is slightly out of balance. It isn't like a tuned circuit where the frequency on the band or reaching the filters might be wrong, only the relative phase of the I and Q changes.

Brian.

 

[BigBoss]

Polyphase filter will correct the phase erros by its nature even if it is driven by imprecise quadrature signals.
To do this, 455x2=910kHz Oscillator must be divided by 2 first.Or 1820kHz is divided by 4.

 

[mvs sarma]

No!

The four phases are coming out of IC1 not IC2. The schematic is an IQ demodulator using IC2 as a mixer. The output of IC2 is actually at audio frequencies.
IC1 pins 4 and 6 are 180 degrees out of phase and so are pins 8 and 10, they are the same signal and it's inverted copy.
The 90 degree phase shift is created by L1, C2 and C3. So what IC1 produces is 0 and 180 degrees on the top part, 90 and 270 degrees on the lower part.

Brian.

Sorry my reply was wrong, and i get corrected , Thanks Brian