Following on from a thread created specifically to sort the maths aspect of calculating phase shifts ( http://www.diyaudio.com/forums/showthread.php?t=152926 - many thanks to all who responded) I've started this thread towards understanding the phase shifts going on with all pass filters.
Now from my understanding, an all pass filter will....
1. Apply the same gain to all frequencies (unity gain)
2. Solely change the phase (the amount of phase shift is frequency dependent)
3. Phase shift the 'centre frequency' by 90° (the centre frequency can be determined with the formula 1/2TT x CR .....to see where C & R fit into the picture, please look at fig 2.4 here... http://sound.westhost.com/pcmm.htm#s22
4. The maximum 'usable' real world phase shift, amounts to approx 120° (source = http://www.maxim-ic.com/appnotes.cfm/an_pk/559 ...the text immediately under the third formula down - phase formula)
Ok, as can be seen there is a non-inverting & inverting variants of all pass filters. The phase shift for a given frequency will depend on the circuit used.
All good so far.
What's troubling me is some of the results the calculations are providing.
For example on the previous thread, using a resistance value of 68K & 10nF, yields a centre frequency of about 234Hz...where the phase will be shifted 90 degrees...*but* (& here's where I'm struggling), wrt to the inverting vs non-inverting .....the former gives -90° shift & the latter +90°?!!
For the non-inverting, does that +90° phase shift really mean the phase is lagging by 270°, ultimately meaning a 90° lead? (If so, then how does this marry with the aformention 120° useable maximum phase shift from an all pass?!)
Can anyone please explain real world +ve phase shifts versus -ve phase shifts wrt all pass circuits?
Now from my understanding, an all pass filter will....
1. Apply the same gain to all frequencies (unity gain)
2. Solely change the phase (the amount of phase shift is frequency dependent)
3. Phase shift the 'centre frequency' by 90° (the centre frequency can be determined with the formula 1/2TT x CR .....to see where C & R fit into the picture, please look at fig 2.4 here... http://sound.westhost.com/pcmm.htm#s22
4. The maximum 'usable' real world phase shift, amounts to approx 120° (source = http://www.maxim-ic.com/appnotes.cfm/an_pk/559 ...the text immediately under the third formula down - phase formula)
Ok, as can be seen there is a non-inverting & inverting variants of all pass filters. The phase shift for a given frequency will depend on the circuit used.
All good so far.
What's troubling me is some of the results the calculations are providing.
For example on the previous thread, using a resistance value of 68K & 10nF, yields a centre frequency of about 234Hz...where the phase will be shifted 90 degrees...*but* (& here's where I'm struggling), wrt to the inverting vs non-inverting .....the former gives -90° shift & the latter +90°?!!
For the non-inverting, does that +90° phase shift really mean the phase is lagging by 270°, ultimately meaning a 90° lead? (If so, then how does this marry with the aformention 120° useable maximum phase shift from an all pass?!)
Can anyone please explain real world +ve phase shifts versus -ve phase shifts wrt all pass circuits?
For some reason I can not edit my last post. Its all wrong anyway
Hopefully you can see this one. Green is "inverting".... red is "non-inverting" (this is only true at lower frequencies). You can see on the graph that the are both 90 degrees out of phase with the signal. Green is leading and red is lagging.
Hopefully you can see this one. Green is "inverting".... red is "non-inverting" (this is only true at lower frequencies). You can see on the graph that the are both 90 degrees out of phase with the signal. Green is leading and red is lagging.Thanks lepomis....as it goes, I've no problem grasping what the lags/leads are, but more related to when is a 'lag' not a lag, but a 'lead'!!!
When it comes to pondering phase changes all pass, if the phase from either the inverting/non-inverting all pass circuit is lagging from 1° thru 180° vs the original ...I'm comfortable with that.
But where do the 'phase leads' come from? Are they simply *BIG* lags, but viewed as leads? (eg a 270° lag can be alternatively viewed as a 90° lead )- do you see where I'm coming from?
So, does an all pass purely lag the phase....?
(PS the reason you couldn't edit your post is, because this forum gives you a time 'window' to nip in there & correct typos etc - but then locks the post after a certain period of time...so just don't go posting stuff yu'd rather not have the internet populace viewing forever!)
When it comes to pondering phase changes all pass, if the phase from either the inverting/non-inverting all pass circuit is lagging from 1° thru 180° vs the original ...I'm comfortable with that.
But where do the 'phase leads' come from? Are they simply *BIG* lags, but viewed as leads? (eg a 270° lag can be alternatively viewed as a 90° lead )- do you see where I'm coming from?
So, does an all pass purely lag the phase....?
(PS the reason you couldn't edit your post is, because this forum gives you a time 'window' to nip in there & correct typos etc - but then locks the post after a certain period of time...so just don't go posting stuff yu'd rather not have the internet populace viewing forever!)
Last edited:
I am pretty ignorant when it come to electronics, especially phase stuff, but with this circuit having only one capacitor.... I don't think you can get a big phase lag. Hopefully someone here can correct me if I am wrong... or at least can give you some more insight.
Yes, I know that - but we're comparing a 'signal voltage' against a 'signal voltage' here (ie input to circuit vs its output)
Bada bing! Doh... I can't believe i missed that *obvious* one!
With that in mind...I'm off to tap some examples into my spreadsheet!
Thanks!
Voltage doesn't get from one place to another without a current inbetween.
Voltage to current , current to voltage... Either transformation or both can
cause amplification.... Usually both.
Both types of phase lags have to be considered based upon the cause.
Cause comes before the effect, looking at it the other way round only
makes you go crazy... Maybe phase leads in repeating cycles, but time
doesn't run backward, ever... Trust me.
On the other hand, outputs are sometimes inputs too....
Maybe you would find a read up on scattering parameters relevant?
http://cp.literature.agilent.com/litweb/pdf/5989-9034EN.pdf
Check out Slide #19. I think thats where the light bulb finally turned
on for me. S parameters finally started to make sense.
In my situation...I am more or less just interested in a single frequency. Here's my guitar string being twanged (I can't bring myself to say well & truly 'plucked'), as viewed on a scope when used with a Guitar sustainer....
http://www.youtube.com/watch?v=u6gm5fY1wM0 ....after a second or two, pretty much a pure sine wve.
Just a bit more info about why I'm so focused on phase right now ...I'm making a guitar sustainer & what I've found is that some strings sustain much less than others. At first, I thought this was just a drive level (power) deficiency ...but having eliminated that (ie chucked loads of power at it - but no better) It transpires this is a phase issue - therefore, I'm toying with the idea of compensating the phase for every (lingering) note played.
The stages involved with a sustainer are as follows...
guitar string vibrates-> into a guitar pickup-> guitar pickup outputs a signal-> into the sustainer circuit-> which drives a sustainer coil-> the string vibrates.
Now obviously there's an awful amount of phase lag/lead possibilities there, but here's my thinking...
within my circuit ...after the guitar signal has been amplified a little, I rectify it into a DC level....obviously when a sustained note is totally in phase (end to end), the DC level will be at its greatest (because the incoming signal will be at it's largest). I'm therefore pondering how to vary a variable resistor in an all pass to move the phase about to get the maximum DC on a long sustained note.
Last edited:
Yes, I know about phase locked loops...but the problem here is that I'm not 'creating' a signal from scracth (which is what phase locked loops do?), ...but also, where would I seek my 'reference' signal from in that longish signal chain!
I guess an obvious point would be the sustainer coil (ie the last part in the overall signal chain), but the sustainer coil (which stimulates the string into resonance via magnetism) is a coil - & that in iteslf will have some phase shift going on internally! Nightmare.
Phase shift and frequency response are interrelated in minimum-phase systems (a pickup is one, the sustainer coil is one - an allpass isn't) and can therefore be compensated by frequency response equalisation. Though it might be quite tricky doing it for the pickup-sustainer combination.
Regards
Charles
Regards
Charles
Hi Charles - your username suggests you know what you're on about...but having read it a few times - I don't understand what you're getting at?
What do you mean....
"Phase shift and frequency response are interrelated in minimum-phase systems"
Also, when you say such a problem can be compensated with frequency equalisation - how/why? (I'm not disagreeing... I just seek enlightenment!).
I'd always taken frequency equalisation to mean 'varying the amplitude of the signal depending on frequency'. If this is what you mean, then I can guarantee you, that this won't work in this particular situation!
How do I know this?
Well, by focusing in on one string/frequency - specifically the top E string (330Hz) - I noticed that the results for this string vibrations with my basic circuit was below par - at first, I simply thought it was a 'drive' problem, so cranked up the drive (ie the amplitude across the drive coil)...whilst this did obviously increase the string vibration intensity, it still wasn't satisfactory. I then created an all pass cct with a variable resistor so I could move the centre frequency - I found that the string vibrated much better at a certain 'point' when rotating the VR.
I conclude therefore that for the top E string at least, the only way to improve the string vibrations significantly, is to alter the phase (& *not* just the amplitude) of the signal across the coil.
Last edited:
A linear and time-invariant system having the inverse causal and stable, as well.
Pretty much every linear, stable circuit that you can build.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.