+- 90 phase shifter

As well as the other suggestions, analog configurations with all-pass filters can give you 90 degree phase shift over a range of frequencies, practical up to maybe the entire audio range (lots of op-amps and resistors and capacitors).

Or, assuming audio, maybe consider a DSP with a Hilbert filter algorithm.

somewhere in my archives is a passive ladder network [similar to the types in the paper Bill Sloman referenced] that gives around 90 phase shift over an extremely wide audio range for use in ?? probably circa

low?

Freq of interest would help immensely. We use coax phase matching. Just roll up some turns according to what you need and there you are..

Just remember what one famous person said.

"No matter where you go, there you are"

Jamie

If its analog audio, you can do it with two parallel paths of allpass filte rs that approximate a 90 degree phase difference over some frequency range. The closer you want to get to DC , the more filter sections you will need. To design this you may need to use an optimizer of some sort. Same approac h can be used in digital, but with another restriction that you can't get 9

Bob

The Hilbert transformer used in DSP is basically the 90 degree phase shift "device", just in software. So DSP is really no different than analog.

Further, my understanding is they used the phase shift between difference networks rather than one filter to produce the 90 degree shift.

As I stated in my other post. this technique is very old, certainly before the 1960s. I got the networks from the IRE journal.

A Hilbert transform filter is anti-symmetric around time 0, and therefore m ust be shifted to the right to make it causal during implementation. Theref ore you need a matching delay in the "direct path". So both methods (fir Hi lbert transform filters and parallel allpass filters) have a structure that is looking at the difference between two parallel branches. It's just that in the FIR case one of the branches is a pure delay.

Bob

The phase sequence filter also only really works for audio, where you don't readily hear slowish phase variations with frequency. You wouldn't want to try using it for instruments except in very special cases.

Cheers

Phil Hobbs

Are you arguing that the 90 degree phase shift networks only produce a 90 degree phase shift when the frequency involved is constant?

Most of the instrument applications where I might have used it involved applications where the frequency involved was essentially constant, but subject to long term - which is to say very slow - variation.

IIRR these networks were invented for single-side band radio where the frequencies involved came from the human vocal tract and changed some twenty times a second. Linear predictive speech coding

seems to sample every 22.5msec - 44.4Hz.

degree phase shift when the frequency involved is constant?

No, that isn't what I said. The phase sequence filter produces two components that are pretty nearly 90 degrees out of phase with each other over as wide a range as you like, if you use enough RC sections. Accordingly you can make SSB mixers over a wide frequency range.

However, that's not the same as a Hilbert transformer, because the phase sequence filter doesn't look like a Hilbert transform plus a delay. That means that it's really only suitable for applications where you want SSB, but don't care a great deal about group delay. Audio is one such application. Digital data, for example, isn't.

Cheers

Phil Hobbs

Hey, We sell a phase sequence filter. (3Hz to 3kHz.) I just tried the step response.. here it is on 4 different time scales. (0.1 ms -100 ms)

If that doesn't work, I can post some other way.

George H.

response.. here it is on 4 different time scales. (0.1 ms -100 ms)

Let me try again...

Geroge H.

response.. here it is on 4 different time scales. (0.1 ms -100 ms)

Cool, I'd like to see that. If you put it in your public folder, and right click on it in the directory listing in the browser, i.e.

Cheers

Phil

response.. here it is on 4 different time scales. (0.1 ms -100 ms)

Wow, those step responses really _are_ gnarly.

Cheers

Phil Hobbs

step response.. here it is on 4 different time scales. (0.1 ms -100 ms)

Grin.. yeah to see all 'ringies' I need a 'scope with a logarithmic time base.

analog.

shift.

I have a circuit for the audio range that does that well with only 4 opamps. I can post it again.

?-)

Now that we've resolved that you making a synchrodyne, with carrier at

-- | James E.Thompson | mens | | Analog Innovations | et | | Analog/Mixed-Signal ASIC's and Discrete Systems | manus | | San Tan Valley, AZ 85142 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at

Howzabout... make a 20F oscillator, use flip/flops to divide it down to a 5F quadrature pair/quartet of signals, and upmix your X with one of those then downmix the 5F + X with a different phase?

If the X signal is narrowband, and you filter the result well, it is a sine wave at X with phase according to the chosen downmix signal.

I am preparing an article about hybrid implementation of wide-band "Hilbert transformer circuits". I refer into it to about 15 authors in the years

ed Pole & Zero of the cells ) were collected in the S.Bedrosian tables, result ing of the solutions of elliptic integrals. My new solution is based on the use of Multiplying C-MOS DAC's. It is fully programmable via an SPI bus. You can select a frequency band in the ratio

Any interest ? Eric Fletcher Consultant for Data Acquisition Systems designs. snipped-for-privacy@gmail.com

Considered a digital implementation?