Telephone / 2 Band parametric eq effect

[Chris S]

I'm wanting to build a lo-Fi effect. Specifically I'd like to build an effect that removes everything below 400hz and everything that is above 4000hz. Looking to build something as small / simple as possible as it's a very occasional effect. (Rather than a full blown 3 band eq). Any suggestions where to start looking? Thanks!

[Rob Strand]

This type of filter is very economical but it's not so easy to design

https://orbit.dtu.dk/files/3813126/Gaunholt.pdf

If you look at some cabinet simulators and bass amps you will see this circuit used.   The cab sims will have roughly a 4kHz cut-off but the low-end is 100Hz or lower so you would need to tweak the high-pass filter caps.  It actually requires a bit more than just tweaking the high-pass filters caps.

[tonyharker]

Telephone response is from 300Hz to 3KHz, so you need to be within those frequencies if you want a telephone effect.  Filter designs for this should be easy to find.

[iainpunk]

how about an inverting opamp with and bridge tee filter in the feedback loop and using the original ground connection as its input... that gives a pass band with first order roll off, maybe try this on a breadboard, and fiddle with the caps and resistor to change the frequency response. the last diode parallel to the output volume control adds some asymmetric clipping, for lower-fi



cheers, Iain

[anotherjim]

I have to say "Modem transformer" because they are made to suit the aforementioned comms voice bandwidth. Also, I'm very audio transformer minded at the moment.
Would probably need some RC lo-pass and hi-pass filtering to help though since the roll-off of the transformer alone at the extremes will likely be too gentle to be very obvious. And a boost/buffer in front to make up the transformer insertion loss.
Basically, just keep hitting those 300Hz and 3Khz roll offs as much as you can.

[iainpunk]

you might also want to chose a transformer with a low voltage rating and then add some DC offset current to saturate the core earlier, and asymmetrically.

cheers

edit: use a MCP6042 opamp, its extremely slow, and slew rate distortion adds to the lo-fi aesthetic.

[antonis]

[Chris S]

Thanks all. Some great options.

[Rob Strand]

Our diversion in this thread on telephone ear pieces sparked some memories,
https://www.diystompboxes.com/smfforum/index.php?topic=86385.msg1209862#msg1209862

A lot of "telephone" emulation focuses on the 300Hz to 3kHz bandwidth.  If you listen to such a filter
it's not really that telephone sound.

Some time back there was a similar thread on a megaphone emulation.    The megaphone emuation
has a number of acoustic effects not to mention the horn.

In that thread you can see a telephone system response in PRR's post,

https://www.diystompboxes.com/smfforum/index.php?topic=125885.msg1201540#msg1201540

The key feature to observe, which is missing in a lot of telephone emulations is the large peak
at 1.5kHz.   IIRC this largely comes from the telephone mouthpiece.   It would be worth while
adding a peak around 1.5kHz.     

A simple circuit is as follows, it's another filter to pass the signal through.



A starting point might be R2=8.2k, R1=1.2k, C=33n.  You will need to attenuate the signal to get a level match
with and without this filter as there is quite a bit of boost.

[ElectricDruid]

If we assume that C1=C2, you can simplify that a fair bit:

Freq = 1 / ( 2 x Pi x C x sqrt(R1 x R2))

Q = 0.5 x sqrt(R2 / R1)

[iainpunk]

Our diversion in this thread on telephone ear pieces sparked some memories,
https://www.diystompboxes.com/smfforum/index.php?topic=86385.msg1209862#msg1209862

A lot of "telephone" emulation focuses on the 300Hz to 3kHz bandwidth.  If you listen to such a filter
it's not really that telephone sound.

Some time back there was a similar thread on a megaphone emulation.    The megaphone emuation
has a number of acoustic effects not to mention the horn.

In that thread you can see a telephone system response in PRR's post,

https://www.diystompboxes.com/smfforum/index.php?topic=125885.msg1201540#msg1201540

The key feature to observe, which is missing in a lot of telephone emulations is the large peak
at 1.5kHz.   IIRC this largely comes from the telephone mouthpiece.   It would be worth while
adding a peak around 1.5kHz.     

A simple circuit is as follows, it's another filter to pass the signal through.



A starting point might be R2=8.2k, R1=1.2k, C=33n.  You will need to attenuate the signal to get a level match
with and without this filter as there is quite a bit of boost.

that's basically the non-inverting version of what i posted earlier.
the inverting one also features first pass rolloff above and below the resonant peak that doesn't shelve, like the non-inverting does.
an important plus point of the non inverting one is that it doesn't need an added input buffer or preceding low impedance stage, since it has way better input impedance (if done right), but the low impedance might be fun because the resonance is dependent on guitar pickups and tone controls!

the values i posted earlier has resonance at 1073Hz changing the 10k resistor to an 8.2k for a 1.2kHz resonant frequency.

cheers, Iain

[Andon]

Possibly another [more involved] consideration:

http://www.geofex.com/FX_images/lofi.gif

https://www.diystompboxes.com/smfforum/index.php?topic=98773.0

[Rob Strand]

If we assume that C1=C2, you can simplify that a fair bit:

Yeah thanks, I actually did that to check the values.

If we say put a fixed attenuator at the input to balance the level, a blend control from input from the input side to the output side is a way to tweak the peak level.   That would be a way to get an idea of what peak level sounded best.   The caps could be tweaked as well to tune the peak frequency.   Worth trying the next smaller size cap say 27n.

You can see from PRR's pic the level can vary.

[Rob Strand]

that's basically the non-inverting version of what i posted earlier.
the inverting one also features first pass rolloff above and below the resonant peak that doesn't shelve, like the non-inverting does.
an important plus point of the non inverting one is that it doesn't need an added input buffer or preceding low impedance stage, since it has way better input impedance (if done right), but the low impedance might be fun because the resonance is dependent on guitar pickups and tone controls!

the values i posted earlier has resonance at 1073Hz changing the 10k resistor to an 8.2k for a 1.2kHz resonant frequency.

It could end-up sounding similar with less parts.   For a bit more HF roll-off you could and another passive RC filter at the output.  The inverting one probably should have a buffer at the input for consistent results but you might get away without it as well.

[Rob Strand]

Here's  what you can get by shifting the frequency of iainpunk's band-pass circuit then adding some simple passive post filters.



The first circuit is a simple 2-stage filter: high pass 300Hz, low-pass 3kHz.  Following that is a peak booster for the mic response.   I've used a gyrator here only so it was easy to tweak the peak.  It does more or less the same as the Bridge-T 1.5kHz boost I posted.

You can see simplified filter with the mods is a good approximation.  I didn't try to make it 100% equal since the idea is to answer the question that post filters on the band-pass version should be close, which it is.

The whole idea of the 300Hz to 3kHz filters and the height of the 1.5kHz peak comes from common knowledge of telephones. 
It cannot make any claim of achieving the desired effect.    Taking that circuit as a starting point and playing with the values will soon show what needs to be tweaked.

[Rob Strand]

FYI,

For a bit of fun I compared the filter with this track,
https://youtu.be/DupyAkOZLYA?t=81

This track has some extreme filtering.    To get an idea I had to move the highpass to 2kHz and bandpass to 3kHz.
The band pass at 3kHz gives it the right smallness but with the highpass at 300Hz it's nowhere near right,
highpass at 1kHz is still too fat, and high-pass at 2kHz starts to be more on the money.  I have a feeling if
I moved the low-pass up a bit I could get the same effect by moving the high-pass up to 3.0kHz, and removing
the peak filter altogether - [confirmed].

The amount of highpass depends on the source as well.   I was using a male radio announcer source, which had a lot of bass.

The biggest problem I see is the target sound you have in your head at the time will greatly affect where you put the filters!

[anotherjim]

It does depend on what your idea of telephone sound is. The 300Hz-3Khz band only really became imposed by the use of modems and 8kbps digital PCM (which brings compression into the mix). The real original limits were imposed by the carbon granule mouthpieces and simple plate diaphragm speaker and all the transmission line losses and effects. Speech intelligibility was the only aim. Apparently, it suited female voices better hence predominantly female telephonists and operators.

[ElectricDruid]

Apparently, it suited female voices better hence predominantly female telephonists and operators.

I don't believe that was the only reason for even a moment!!

[Rob Strand]

I don't believe that was the only reason for even a moment!!

I'm glad my sister didn't understand frequency response otherwise it would be an excuse to get priority of the phone.

[PRR]

Ah! Here is the plot for 302 vs 500, *complete* subscriber set and line(s).


The 302 was intelligible but the bass cut made all men sound similar. And to the point of the other(?) thread, the 1900 and 2800 cps (we didn't Hz yet) bumps, cascaded for effect, give Q similar to some music instruments.