See Williams+Taylor's filter book, 3rd edition. Fig 7-22 is an allpass that gives 90 degrees shift from 300 to 3K Hz, +-0.1 degrees. It uses six opamps and 24 passives.
Now if you are telling us that you design audio electronics and don't have a copy of this book, well, that's the most bizarre thing you're ever posted here.
John
Why do you persist in giving Asinine Allison a forum?
...Jim Thompson
-- | James E.Thompson, P.E. | mens | | Analog Innovations, Inc. | et | | Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus | | Phoenix, Arizona 85048 Skype: Contacts Only | | | Voice:(480)460-2350 Fax: Available upon request | Brass Rat | | E-mail Icon at http://www.analog-innovations.com | 1962 | Only Jerks Need to Satisfy Their Woeful Egos by Feeding Trolls
"John Larkin" "Phil Allison"
** Play fair you lying cut - post the schem.All such filters I have studied or designed use TWO paths that differ by 90 degrees.
..... Phil
And why do you persist in objecting to persons whose views and posting styles differ from your own? I would have thought that someone such as yourself, known to have had experience with autistic children, would have a bit more understanding of and patience with those whose experience/genetics/capabilities/perspective etc differ from your own, for whatever reason.
My hat is off to John for engaging Mr. Allison in a constructive rather than destructive manner.
And what sort of person needs to satisfy their woeful egos by blasting everyone whose views, experience and posting styles differ from their own? Oops, there I go again, following your example rather than Johns :-).
...Glen Walpert
Mooncalf ;-)
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine Sometimes I even put it in the food
Buy the book. It's really worth having.
Yes, they do.
I'm not a signals expert, but my impressions are that...
A true 90 degree phase shifter is non-causal, like an ideal lowpass filter. Its transfer function can be defined, and it can be approximated in real life, but the implementation must introduce time delay to get around the causality (ie, predicting the future) dilemma.
You can make a 90 degree shifter by taking the Fourier transform of a signal, swapping the real and imaginary components, and then reverse transforming. That can be implemented with an FFT, again with time delay to sort of prime the pump full of samples. The bandwidth will be limited on the high end by the Nyquist criterion, and on the low end by the number of samples used in the FFT.
In an FPGA, you can do a FIR implementation of a Hilbert 90 degree phase shifter, with high/low frequency limits similar to the FFT idea, and time delay like any FIR filter. I think the Xilinx software will build optimized Hilbert filters for you, with some clever folding tricks. The time delay is usually worked around by delaying the original signal by an equal amount.
Some signal processing pros here can correct me or fill in details.
John
Do you have a reference? Googling on the concept of non-causal filters only seems to turn up material relating to digital filters. Does this apply only to digital filters, or to analog filter circuits as well?
-- Greg
You can google "Hilbert transform" and "ideal lowpass filter" for some bits on causality. Both analog and digital filters are causal, because they are real.
An ideal lowpass filter has an impulse response that rings before the impulse hits - in fact the impulse response rings over all of time, past and future - so it's impossible to build. An ideal Hilbert
90-degree phase shifter has a positive 1/t tail after the impulse and a negative 1/t tail going backwards in time before the input, also impossible in real life. So any real-life approximation has to add delay; the better the approximation to the shape of the ideal impulse response, the more you have to shift the impulse response to the right to minimize chopping important stuff off the left-hand end.John
Most SSB 90-degree filters are really two filters whose phase difference is 90 degrees. That's much easier to do over a multi-octave bandwidth.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 hobbs at electrooptical dot net http://electrooptical.net
Yep. I do that for image-reject mixers, simply for convenience (and lack of trim).
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine Sometimes I even put it in the food
I left out "differential".
...Jim Thompson
--
| James E.Thompson, P.E. | mens |
| Analog Innovations, Inc. | et |
| Analog/Mixed-Signal ASIC\'s and Discrete Systems | manus |
| Phoenix, Arizona 85048 Skype: Contacts Only | |
| Voice:(480)460-2350 Fax: Available upon request | Brass Rat |
| E-mail Icon at http://www.analog-innovations.com | 1962 |
I love to cook with wine Sometimes I even put it in the food
Not to forget Mike Gingell's polyphase network, re-invented by many since
1969, that makes a good job of it and avoids the need for cascaded active elements.ftp://ftp.vnet.net/pub/users/gingell/polyphas/ ftp://ftp.vnet.net/pub/users/gingell/polyphas/UK69_1.pdf
Chris
I think that cleverly takes out, for free, the delay inherent in trying to create a real version of a non-causal filter. A 90 degree shifter would need a pure delay line alongside it to line things up.
Inside an FPGA, doing a FIR Hilbert phase shifter, I think you can use the same folded data delay line to tap off the FIR coefficients and to delay the original signal. That could be handy for AC power calculations.
John
Needs quite a few taps when you want 3 decades of bandwidth, no?
Cheers
Phil
:
how
ete
0Hz 5 . eh.
s/gingell/polyphas/UK69_1.pdf
This looks like the 'phase sequence filter' from AoE (2nd ed. pg
295).George H.
Yes, it's intimidating. It's probably more sensible to do an IIR simulation of the analog all-pass thing, or a tracking PLL and a clocked delay line if there's just one frequency involved.
We need FPGAs with 1000x the resources of what we can afford today.
John
It might be simpler to FFT, chop off the negative frequencies, and double the positive ones.
Cheers
Phil Hobbs
Apologies for replying to my own post, but I realized I should be more explicit. The above procedure will get you the analytic signal, i.e. the signal plus j times its Hilbert transform. The actual Hilbert transform deletes the DC and multiplies positive frequency components by
-j and negative frequency ones by j (depending on your sign convention). Once you have the analytic signal, it's two real multiplications and an add to get any phase shift you like.
Cheers
Phil Hobbs
-- Dr Philip C D Hobbs Principal ElectroOptical Innovations 55 Orchard Rd Briarcliff Manor NY 10510 845-480-2058 email: hobbs@electrooptical.net http://electrooptical.net
-- snip --
ftp://ftp.vnet.net/pub/users/gingell/polyphas/ftp://ftp.vnet.net/pub/users/gingell/polyphas/UK69_1.pdf
This looks like the 'phase sequence filter' from AoE (2nd ed. pg
295).George H.
Yes, they're the same arrangement. Does AoE give any credit to M. Gingell?
The same polyphase network also appears to be ascribed to HA5WH by the ARRL:
Chris
ce
ve
ell?
RL:
No references in AoE that I ever found. I've used this circuit and would have been happy to have the above ARRL link. I just mucked around with a simulation until it gave a phase ripple I found acceptable, then built and tested it. The first air wire prototype had a nice tubular shape.
The problem with this circuit is there is no phase reference to the input. (Well not quit true. There is certainly a relation between output and input, but the phase twists around several times, going from the lowest the highest frequencies, and you lose track after the first 360.)
George H.
ElectronDepot website is not affiliated with any of the manufacturers or service providers discussed here. All logos and trade names are the property of their respective owners.