Modifying a two pole EQ circuit

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

This site may earn a commission from merchant affiliate links, including eBay, Amazon, and others.

bobkatz

Well-known member
Joined
Jun 25, 2006
Messages
138
Location
Orlando
I wonder if someone can help. I have a subwoofer which has an extreme low bass equalizer, nominally around 25 Hz with both boost and cut. Strangely, while the center detent is at the physical center of the pot, counterclockwise it goes down to -12 dB and clockwise it boosts up to +3 dB. I'm running the pot near the full attenuation and it is very sensitive at that point on the attenuator. I'd like to move the active control position to a place where a tiny move has little effect. So I need a cut of around -10 dB to happen nearer to the center of the rotation. In other words, I need attenuation and I would never need a boost, so the part of the circuit in the CW direction needn't have much effect.

It's a two-pole circuit that's a bit over my head, so I'd appreciate some help. What do you think if I remove R136 and R129? If I have that right, that will give me less gain and so force me to turn those pots more CW to get my same desired level. But I'm not sure of the effect on the crossover frequency, I'm whistling in the wind here.
 

Attachments

  • bass eq.png
    bass eq.png
    63.8 KB · Views: 295
Hi Bob,, hows the system clock?

That circuit looks like a variant on the classic Baxandall tone control (one pole EQ).

C18 sets the LF transition back to unity gain inverting, while C113 sets the HF transition back to unity gain.

In the active bandpass the ratio of impedances to the right of the node common to both pots. divided by the impedance to the left of that node sets the boost/cut.

If you only want only cut and no boost disconnect the wiper of the left half of the pot (pin 2).  If you just want to reduce amount of boost or shift everything toward cut, increase the value of R131.

DO NOT mess with R136 and R129.. these are your unity gain input and feedback resistors for signals above the EQ bandpass...  To visualize what is going on, at higher frequency C113 shorts out the boost/cut network making it unity gain, at very low frequency C18 is open circuit ignoring the boost/cut network.

JR


 
Thanks, John. Happy New Year y'all. I have little experience analyzing ciruits like this, which are bread and butter for you guys.

So you think if I open the wiper on pin 2 (which disables that section of the pot) then all I get is cut?  Would this move the range so that I can move the pot more clockwise to get the same attenuation as before?

How's my System clock?  Not too much jitter today...
 
The easiest way to analyze the boost/cut is like a simple inverting opamp.

The total parallel resistance of network to left of - input (18k in series with 68k in parallel with half of 50k pot). So effective input R varies from 18k at max boost, to approx 47k at max cut. Feedback side varies from 50k at full boost, to 0 ohm at full cut.

To scale whole thing toward cut, we want to make input side higher resistance. Removing the 68k (was in parallel with pot) will shift nominal gain toward cut. Making it about -5dB at the mid point. 

Removing the 68k and lifting the wiper makes it around -9 dB at center.  If that isn't close enough, you can push it to -10dB at midpoint, by increasing 18k to 29k (use 30k), lifting the wiper, and removing 68k in parallel. Note: pots are typically 20% tolerance so this is all +/- a smidge..

You can load the R values into a spread sheet if you want to look at pot law and tweak. Most Baxandall circuits use linear taper pots or some variant with 50% total resistance at mid rotation.

JR



 
> a variant on the classic Baxandall

It's not THE Baxandall; caps in wrong places. And it is a midband bump, not end-band shelves. The limit-cases are similar.

Check my transcription from schematic to SPICE. If correct, the plots show the 0-10 knob settings.

4qq1ae.gif
 
> I'm running the pot near the full attenuation

Then either the amp/speaker is way boomy, or your room has huge bass gain. More than the designer ever expected someone would need/want to compensate.

Assume the EQ is not wrong. Assume your room has excess boom, and fix it. I believe you know the methods and got the room as good as it gets. Anyway excess boom is not bad: it reduces amp power and speaker excursion.

So put a boom-cut in the chain. Your preferred setting is around -3dB around 50Hz-90Hz. This can be a simple C-R high-pass. I assume the previous stage can drive 2K load. 1uFd in 2K is -3dB near 72Hz. There's some interaction with the EQ impedances, I like the look of 2uFd 2K, but you may try 1.5uFd or 1uF and see.

The high-pass approximately negates the room gain. The 1uFd-2ufd cap approximates the ~~80Hz range where (apparently) room gain becomes significant. The 6dB/8ve slope is a rough approximation to room gain. Trim this first while judging the 50Hz-200Hz balance and flatness. The EQ peak/dip then fine-tunes the 25Hz area (if you seriously care).

69p3yv.gif
 
PRR said:
> a variant on the classic Baxandall

It's not THE Baxandall; caps in wrong places. And it is a midband bump, not end-band shelves. The limit-cases are similar.

Check my transcription from schematic to SPICE. If correct, the plots show the 0-10 knob settings.

4qq1ae.gif

Dear PRR: You're incredible! Thanks for the translation to spice. That looks very much like the curve I'm getting as of now. It is a peak/dip equalizer, not a shelf.

BK
 
PRR said:
> I'm running the pot near the full attenuation

Then either the amp/speaker is way boomy, or your room has huge bass gain. More than the designer ever expected someone would need/want to compensate.

You'd be surprised what you have to compensate for in real world rooms. Most real world small rooms exhibit extremel irregular response in the bass region. I cannot state what the anechoic response of the subwoofer is, but in this well-treated and trapped room, the raw response at the listening position, the positions of the two subwoofers and the trapping have been optimized considerably. The subwoofer has been located purposely at a position where it compensates for a known dip near 50 Hz. It's a lot harder to deal with dips in a room than peaks so it was a good idea to position this subwoofer to compensate for the weakest link (dip) in this room. Consequently, in this same position, there is a considerable rise below 40 Hz due to corner gain, which is also a well-known phenomenon, but the built-in equalizer in the sub is adequately compensating for that, except that the pot is currently too sensitive to movement at that position of the rotation---not an unusual compromise! The net result after eq and trapping in this room is extraordinarily even, tight, low distortion, high headroom and, shall we say, more accurate than in most professional listening rooms---visitors are always astounded at how even and musical the bass response is.

Nevertheless, this room is never done! I'm always looking for ways to improve it, one of them being this tweak to the subwoofer equalizer and the others being additional trapping and perhaps the custom equalizer you suggest. My next step will probably be a custom digital bass manager, actually, bypassing all the analog circuitry including the one we have just looked at.

The high-pass approximately negates the room gain. The 1uFd-2ufd cap approximates the ~~80Hz range where (apparently) room gain becomes significant. The 6dB/8ve slope is a rough approximation to room gain. Trim this first while judging the 50Hz-200Hz balance and flatness. The EQ peak/dip then fine-tunes the 25Hz area (if you seriously care).

I do seriously care. I have to because I produce masters for mass replication. It is very important in mastering to judge the power and impulse of the bass drum, which extends down well below 40 Hz. I have to make accurate judgments and often have to decide on whether or not even a 13 Hz high pass makes a difference on the mastered product.

BK
 
PRR said:
So put a boom-cut in the chain. Your preferred setting is around -3dB around 50Hz-90Hz. This can be a simple C-R high-pass. I assume the previous stage can drive 2K load. 1uFd in 2K is -3dB near 72Hz. There's some interaction with the EQ impedances, I like the look of 2uFd 2K, but you may try 1.5uFd or 1uF and see.

If we're worried about load, why not a smaller cap and a larger R?

How about 0.5 uF with a 4k variable so the shelf it can be tuned?

BK
 
Baxandall's name is associated with his very widely copied treble/bass shelving tone control. In my defense I used the qualifier "variant".

The peaking variant is generated by applying both the capacitor used by Baxandall to turn off his bass section above bandpass, and the cap used by Baxandall to tell his treble section when to start EQ'ing, to the same single EQ stage.  Note: Don't be tempted to wrap several of these around a single opamp to make a multiband EQ. If not adequately spaced apart they will interact in use for unreliable results. I've seen some pretty ugly EQs attempted this way.  

I am not a circuit historian but I thought prior to Peter Baxandall's ubiquitous tone control, most EQs were passive stages followed by make up gain (similar to the tone circuits built into electric guitars). I don't suggest that all active feedback EQ stages are "Baxandalls", but Bob's EQ sure looks like a minor variation to Baxandall's classic tone control to me.  

=======
If PRR is graciously willing to similate this in spice I will defer to his computer.

FWIW, Placing a cap in series with the input of a typical Baxandall tone control can interact with the changing impedance as the EQ boost/cut is changed. I have actually used this interaction beneficially with bass tone controls to provide a LF pole below the shelf region that shifts higher in frequency when more bass boost is commanded. This can somewhat mitigate the build up of infrasonic energy from shelved boost, while still delivering a wide response when set flat or cut.  If Bob lifts the pot wiper on his input side, his input impedance will not vary, so whatever pole he adds will be stable.
-----

I've seen Bob used as an endorsee in bass trap ads, so I'll trust him to sort out his room. An old friend of mine has a business selling listening room treatment panels, but this is a pretty small industry so I assume Bob already knows the typical suspects.

JR

PPS: One well known mixer company (that I didn't work for) used a Baxandall tone circuit immediately following an insert jack in a small mixer without buffering. A measurable frequency response error was caused by this input load that changes over frequency and with boost/cut, if the output impedance of the product patched into the insert was above a few hundred ohms. I discovered this quite by accident when switching my bench rig between 600 ohm and low Z source impedance while measuring the competition's specs. IMO flaky design but their typical customers never knew to blame them when the sound quality changed after patching in some other company's gear.   
 
JohnRoberts said:
FWIW, Placing a cap in series with the input of a typical Baxandall tone control can interact with the changing impedance as the EQ boost/cut is changed.

PRR noted that there is some interaction. I'm going to start with the mod of the peak/dip section, either lift the wiper or remove the 68K as John recommended. Having an additional bass rolloff is attractive to me and since it's passive it's not difficult to wire in and play with.
 
> You'd be surprised...

Not really.

I may have been speaking to the general case rather than your specific case.

You've optimized to raise dips, a usually-wise path. Since bass response in non-huge rooms is all peaky/dippy, you pushed up the peaks. What you have overall is a rise, greater than the speaker designer expected you to find. i.e., "your system is way-boomy", by intent.

Without a measured response, or "simple cause", the first-approximation is a general roll-off.

I said the knob would then trim 25Hz, but actually this filter is so broad that you get half your dB-trim an octave up from 25Hz. -12dB at 25Hz implies -6dB at 50Hz. Bass dip/peak is generally narrower than an octave. Any speaker that approximates 25Hz in less than 50 cubic feet is probably using a third-octave vent bandwidth. Room bass resonance Q, unless absorption is ample and very effective, is far higher than 1. I think this knob is for lower aspirations than yours. And I'm not inclined to re-design it for sharper performance; I think that needs more parts. I'd be thinking about 1/6 octave EQ. But setting and verification is difficult.

Digital bass smells bad to me on general old-curmudgeon grounds. Get a BIG cone, a BIG room, and let the bass fall out naturally. But in realistic somewhat-profitable situations, I can see how digital could be a much better solution. Not only because you can do very fine specific kinks, but the system can THUMP itself and deconvoljivate its own impulse, instead of a week of running sweeps and soldering coil-taps.

> why not a smaller cap and a larger R?

Then go 0.05u and ~~40K, AND a buffer. Which is extra wiring and slight added sonic corruption.

As John says, the bandpass trim has a complicated input impedance, and possibly quite low. 5.11K at very high frequency, 475K at very low freq, and very likely changing right IN the frequency band of interest.

The ~~2K impedance looks low relative to the bandpass components, and SPICE showed <1dB deviation at the interface between low-cut and bandpass. The deviation itself is insignificant; as John also says, the frequency can shift, which complicates tuning. And for you to buy a few half-uFd of good cap is cheaper than my free labor to do a detailed comparison, how small a cap (how high an impedance) we could get away with.


I still can't figure out why they wired a dual-pot that way. It appears to be identical to a single pot in the conventional "Baxandall variant". It isn't like they had a boxload of these pots: it seems to be 30% Reverse Audio taper, not a common Dual-Audio taper. There does not seem to be wiper-redundancy: if either wiper lifts, the bass bump/dip still goes wrong, and the >100Hz response is always near-flat so there is no danger of SPLATT deafening the user.
 
> I'm not inclined to re-design it for sharper performance; I think that needs more parts.

I got curious.

First: the "normalized" version is simmed below: symmetric boost/cut, simple single linear pot. The resistor values are generally good for modern chips and should be left alone. The center frequency may be moved by scaling both caps the same amount: 0.05uFd and 0.005uFd is 250Hz, 0.5 nano and 0.05 nano (50pFd) is nominally 25KHz (though at 50pFd, stray capacitance will skew actual results).

Generally, this plan gives 6dB/8ve max slopes. It is possible to induce a narrow pip centered on a broad hill, but that's not a useful thing. Funny shape and very bad impedances. And for less than 12dB action, the slope is less than 6dB/8ve.

Room modes for a not-huge room will be a series like 25 35 50Hz... or even denser, and each mode may peak or dip depending where you are re: the source. So the granularity is sub-octave. The peak/dip ratio depends on absorption, and you may be in much better shape than most rooms. In my experience, existing rooms show peak/dip ratios over 10db, and >20dB nulls happen. If peaks >6dB are rare, and nulls deeper than 6dB are not fully corrected, you still may need 12dB of slope in much less than an octave. You may need 20dB or 30dB per octave. So 6dB/8ve filters are "general shape", not "detailing". And while I won't prove it here, significantly steeper slopes will need "more parts"; at least separate the filter from the boost/cut summer.
 

Attachments

  • katz-EQ-sym.gif
    katz-EQ-sym.gif
    13 KB · Views: 86
PRR said:
I still can't figure out why they wired a dual-pot that way. It appears to be identical to a single pot in the conventional "Baxandall variant". It isn't like they had a boxload of these pots: it seems to be 30% Reverse Audio taper, not a common Dual-Audio taper. There does not seem to be wiper-redundancy: if either wiper lifts, the bass bump/dip still goes wrong, and the >100Hz response is always near-flat so there is no danger of SPLATT deafening the user.
Couldn't it be one of these special-order pots with half the travel at 0 ohms and the full variation spreading on the remaining 150° of rotation? Like those used on balance control on several HiFi amps? The justification would be that the pot's tolerance would not impair the flatness at mid-rotation...
 
I'm very gratified at the  expert advice being thrown at this problem! Is it true that PRR has redesigned the circuit with a single pot and a single amp to perform much better at 25 Hz than the original circuit? I could "whack and slash" PRRs mod pretty easily! Then why (as others have asked) did the original designers do a dual pot solution? PRRs sim of the dual pot solution was right on the money, with 3 dB boost at the cw position and 12 dB cut CCW, and I suspect he used a linear pot for his sim... 

Was the dual-pot solution of the original designer to give it more dip than boost? If I could get the circuit to stay further away from 50 Hz (higher Q) and operate closer to 25 Hz I'd be a happier camper. It's the 25 to 40 Hz range that's being boosted by my room gain.  In the end, we're talking about lots of whacks and slashes on the existing PC board, an interim solution until I do my digital one, but it would be fun to try.


BK
 
Oops. I took both of PRR's spice simulations into photoshop and normalized them to the same scale and it appears that both of them have the same slope and center frequency. PRR's single pot solution has a symmetrical cut and boost. So for me to have less effect at 50 Hz I would have to turn this into a narrower q (two pole?) circuit, right....

BK
 
This approach is always going to be one pole. To get more poles requires more poles... ;)

You can in principle narrow up the bandpass (Q) somewhat by moving the caps that set the HF and LF skirts closer together. At some point they will start diminishing the max boost or cut available as the skirts overlap and interact with each other.

Before we reinvent the wheel... what do you want? A more friendly sweet spot on the adjustment pot, or a different EQ? If you patch in a parametric can you empirically get a noticeably better result? A fixed EQ to mimic that might be a productive path.

YMMV

JR 
 
> redesigned the circuit with a single pot and a single amp to perform much better at 25 Hz than the original circuit?

No. The dual-pot is wired as two half-pots which function same-as a single pot. Abbey's speculation is interesting but I dunno.

The asymmetry is due to the odd-value parts around the pot(s), the 18.2K 68.1K 100K oddballs. Toss them away, it is symmetric. Basically the 18K limits the boost. You'd get a not-dissimilar result by driving a nail in the faceplate to block the knob pointer from going past "7" on a 0-10 scale. The 68K and 100K must balance the mid-turn upset caused by the 18K.

Since they skewed the natural symmetry, yet you say response is flat at mid-turn, it appears they use non-linear pots. If you like the symmetric case, linear is perfect and a single section seems sufficient.

The increase of dB/degree at the ends is intrinsic, generally useful (except for you), and could only be "fixed" with an S-curve pot. Actually, some +/-20dB graphic EQs do favor S-curve pots; otherwise all the action is at the ends and a very up/down knob-shape actually gives modest effect. But if you need fine-trim, you want to be working near mid-turn, not at the end. Which is another reason to get your general shape out to a different stage, so this one can work near-center.

> You can in principle narrow up

Yes, but I think the curves are the same or gentler either and all ways (with one exception). You can't get a significantly steeper slope: there is some optimum and these relative values are very close to it. For the +/-12dB case (or the +3/-12dB cut-down), cap ratio near 10 is as steep as it gets. Other ratios give broad curves tending to flat-top.

There is an exception if "5.11K" is made 1 ohm and everything is "ideal". At the VERY extreme you get a very broad plateau with a very narrow pip. I can't see any musical use, anyway 1-ohm is not a practical value for real-world chips and 50K pots. As the pot comes off the stop, the pip vanishes. There are ways to get narrow pips without the broad plateau and with practical values, so this is just a simulator freak.

You sure can get steeper slopes in one opamp. I guess the issue is: that opamp can't also be doing the boost/cut summation. Sticking that variable in will interact with the resonance and generally defeat it. It happens that for "small" Q, near unity, the +/- variable is about right for both purposes.

Room-nodes and box vent bumps will need a narrower band, and you want 2 to 10 of them in the 10Hz-100Hz band. That suggests a different topology, with narrow dedicated fix-gain filters and a single add/subtract stage. If you can get your general slope in another stage, and you have some frequency-shift in each filter (a non-trivial thing), each add/subtract can probably be +/-6dB.

> how even and musical the bass response is.

I wish you were closer.

I have a recording of pipe-organ, a good instrument in an unusually massive church. Omni-condensers plausibly flat to 20Hz. I can see honest 32Hz tones in the DAW, and I heard them live. But all my speakers make it to 50Hz (at best) by sacrificing everything lower. So I've never heard all the bass which I know is in the file. And one track starts by dropping a toe on the leftmost pedal and holding it for most of a minute. Even at >50Hz, it's impressive; but it was more impressive live. The 64 and 96Hz overtones should be decoration on the 32Hz foundation, not fake fundamental.

 
PRR said:
I wish you were closer.

I have a recording of pipe-organ, a good instrument in an unusually massive church. Omni-condensers plausibly flat to 20Hz. I can see honest 32Hz tones in the DAW, and I heard them live. But all my speakers make it to 50Hz (at best) by sacrificing everything lower. So I've never heard all the bass which I know is in the file. And one track starts by dropping a toe on the leftmost pedal and holding it for most of a minute. Even at >50Hz, it's impressive; but it was more impressive live. The 64 and 96Hz overtones should be decoration on the 32Hz foundation, not fake fundamental.

I wish you were closer. Pipe organ fundamentals exercise room modes in the toughest way, much tougher than passing tones. If you have a 63 Hz note we still might hear problems in my room. Good thing I don't master many pipe organs.
 
Back
Top