Portable two-band Parametric EQ (ESP/Urei 545)

[JohnRoberts]

It was a real prototype on PCB and metal enclosure, the fpot scraping a lot but with my circuit shown in post #74. Some crazy 1:290 apparently and the OPA1642 (3mV voltage offset per datasheet). With the 1:25 I'm planning now it might be no problem, I hope. Or I might try something else,
Isn't the current noise of the NE5532 too high for this? Considering the high resistor values everywhere. Never tried it. The AD712 produced my favorite sounding EQ curves of what I've tried, but voltage noise specs ended up scaring me off.

I typically used bifet opamps (like TL07x) in parametric SVF designs.

JR

 

[abbey road d enfer]

I typically used bifet opamps (like TL07x) in parametric SVF designs.

JR

The problem with TLO7X is noise voltage. I deemed them not suitable for my 5 and 9-band parametrics. With typical 5k resistors at the input, the noise du to noise urrent is less than that due to noise voltage.
However, today there are FET opamps with noise voltage comparable to 5532.
OPA1652 and even better OPA1656, both only in SOIC package.

 

[JohnRoberts]

The problem with TLO7X is noise voltage. I deemed them not suitable for my 5 and 9-band parametrics. With typical 5k resistors at the input, the noise du to noise urrent is less than that due to noise voltage.
However, today there are FET opamps with noise voltage comparable to 5532.
OPA1652 and even better OPA1656, both only in SOIC package.

I typically used >5k Ohm resistors around SVF to keep capacitor values small.

Do what works for you.... FWIW I haven't designed any parametric EQs this century.

JR

 

[abbey road d enfer]

I typically used >5k Ohm resistors around SVF to keep capacitor values small.

I had to use relatively large caps (about 150nF) for the low range

Do what works for you.... FWIW I haven't designed any parametric EQs this century.

Actually I did a DC-controlled one; turned out the VCA's noise was dominant to a point I gave up.

 

[JohnRoberts]

I had to use relatively large caps (about 150nF) for the low range

I don't recall all but I used 8200pF for the low frequency range in my 1979 kit parametric EQ

Actually I did a DC-controlled one; turned out the VCA's noise was dominant to a point I gave up.

Have you ever considered DPOTs?

There are multiple ways that the SVF parametric filter topology can negatively impact noise floors. I have seen approaches that incorporated an elevated BP section gain to generate a narrow bandwidth (I used a variant on that in my old kit design). Likewise there are variable Q/bandwidth adjustment approaches that result in elevated noise gain in the HP and BP filter outputs. I have seen too many different parametric approaches to list (or even remember).

A possibility with DPOTs or VCAs is to vary Q/bandwidth by shifting the integrator poles further apart to get a broader bandwidth (unsure how narrow you can get by moving them closer together). Sliding integrator poles apart or together should have no obvious negative noise gain consequences.

When considering modern technology pure digital domain filters should be excellent.

JR

 

[abbey road d enfer]

I don't recall all but I used 8200pF for the low frequency range in my 1979 kit parametric EQ

That would have implied rather high value resistors, hence the need for FET opamps.

Have you ever considered DPOTs?

Yes I have. Actually the project was a dynamic notch filter and I was not comfortable with the idea of having the logic needed to drive DPOT's co-existing with the analog path.

When considering modern technology pure digital domain filters should be excellent.

I agree, but the customer wanted a full analog signal path (vinyl mastering).

 

[JohnRoberts]

That would have implied rather high value resistors, hence the need for FET opamps.

Back in those days I was using lots of 100K resistors

Yes I have. Actually the project was a dynamic notch filter and I was not comfortable with the idea of having the logic needed to drive DPOT's co-existing with the analog path.

A couple decades ago, I actually designed a digitally controlled analog (using DPOTs) automatic mixer for friend's "analog only" mixer company. While I got respectable sound quality (to my ears) the company fell on hard times because "analog only" customer base got seriously eroded by digital consoles with tons of features at impossible to ignore prices. That project got abandoned before reaching production prototype stage.

My breadboard layout had both DPOTs and VCAs. I populated the DPOTs first and they worked so well, I forgot about the VCAs.

I agree, but the customer wanted a full analog signal path (vinyl mastering).

That seems like a pretty small market but the customer is always right as long as they have the money to support such requirements.

JR

 

[domingo]

How relevant is the opamp's 'gain bandwidth' for this application? Some opamps are very fast like above 20MHz, I wonder how beneficial the extra bandwidth in real terms compared to a humble ie. 5.5MHz. I understand that faster opamps normally have higher slew rate and output current, but I wonder for directs benefits for this application.

Here I share some short noise tests with 5 opamps: OPA1656, OPA2145, AD8620 and OPA2132. The results seem similar and noise at the lowest fq produced by the 68K quite dramatic.

 

[abbey road d enfer]

How relevant is the opamp's 'gain bandwidth' for this application? Some opamps are very fast like above 20MHz, I wonder how beneficial the extra bandwidth in real terms compared to a humble ie. 5.5MHz. I understand that faster opamps normally have higher slew rate and output current,

Not necessarily regarding output capability; output current is unrelated with BW. OTOH many newer opamps have improved output capability.
It allows reducing the resistor values for optimizing noise.

but I wonder for directs benefits for this application.

In terms of audio perception I don't think there's any advantage using high speed opamps. The integrators noise is dominated by the time- constant, only the summer may benefit from extended BW but since it operates at moderate gain (about 30dB at min BW), even with 5.5MHz it means performance will be kept up to ca. 180kHz.
It must be said that older opamps, such as 748's and LM358 were actually challenged, but not the TLO's and 5532's.

Here I share some short noise tests with 5 opamps: OPA1656, OPA2145, AD8620 and OPA2132. The results seem similar and noise at the lowest fq produced by the 68K quite dramatic.

A good enough reason to keep resistor values low, which is where opamp's output capability becomes a concern.

 

[abbey road d enfer]

A couple decades ago, I actually designed a digitally controlled analog (using DPOTs) automatic mixer for friend's "analog only" mixer company. While I got respectable sound quality (to my ears)

My concern was with the dynamic part of the design.
Setting frequency and BW was not an issue, but the dynamic part required converting the detector voltage, with possible interference, and the possible zipper noise. Being a strictly analog designer, it was too challenging for me.

That seems like a pretty small market but the customer is always right as long as they have the money to support such requirements.

Vinyl mastering engineers are ready to invest in 100k+ custom monitor speakers. They would spend a significant fraction for a dedicated all analog dynamic notch.
Not enough to make a business out of it, but filling a niche.

 

[domingo]

It allows reducing the resistor values for optimizing noise.

I see! Is there an easy way know or estimation to predict how low input resistors can be for a certain opamp's gain bandwidth? Or is it something to try and see if it runs out of steam? I like the idea of having a minimum Q of around 17, equivalent to a chromatic semi-tone. But this requires around 160R at the summing inverting input (Q pot) when using 5.1K general bandwidth resistors. If you can see what I mean.

My breadboard layout had both DPOTs and VCAs. I populated the DPOTs first and they worked so well, I forgot about the VCAs.

The discussion on DPOTs is fascinating. What was your experience with steps using DPOTs @JohnRoberts? Do you remember how many steps did you use and for what range more or less? I wonder if steps are really noticeable because I have an ARM MU in my system already, so very tempting to explore this in the future, once I manage to set the analog filter properly with carbon pots on PCB.

 

[abbey road d enfer]

I see! Is there an easy way know or estimation to predict how low input resistors can be for a certain opamp's gain bandwidth? Or is it something to try and see if it runs out of steam?

Resistor value is not limited by GBW. Limitations come from the drive capability of opamps.
At max boost the level at the outputs of the mixer and the integrators is a few dB below the output, you could dimension the resistors for the lowest load recommended by the manufacturer. For 5532 it's 300R IIRC.. For OPA1656 it's about 150R.

I like the idea of having a minimum Q of around 17, equivalent to a chromatic semi-tone. But this requires around 160R at the summing inverting input (Q pot) when using 5.1K general bandwidth resistors. If you can see what I mean.

I don't think there's any issue with that. The limitation comes from the GBW product. High "Q" * means high gain in the mixer.

BTW I hate using "Q" for this parameter. Q is valid only for transfer functions that go asymptotically to minus infinity, which is not the case for EQ's. Actually Q varies with the amount of boost/cut.
BW is a much more "organic" description of perception. It has been debated a lot, and never came to a common standard. Still, I prefer using BW over Q.

 

[JohnRoberts]

The discussion on DPOTs is fascinating. What was your experience with steps using DPOTs @JohnRoberts? Do you remember how many steps did you use and for what range more or less? I wonder if steps are really noticeable because I have an ARM MU in my system already, so very tempting to explore this in the future, once I manage to set the analog filter properly with carbon pots on PCB.

The computer I did the PCB layout on lost its hard drive years ago, but my recollection is that I used 12 bit DPOTs. I included hooks in the software architecture for later adding zero crossing synchronization and the MCU had enough A/Ds that I could use them for synchronization, but during debugging I didn't perceive a pressing need for it. FWIW the target application was spoken word. I used TV talking heads audio for source material.

My speculation is that I was updating the gain changes so frequently that the gain steps were tiny and not very audible. I expected more critical listening tests later as this SKU got closer to production, but that never happened.

JR

 

[domingo]

my recollection is that I used 12 bit DPOTs

That's 4096 steps pretty amazing! The highest dpots I've seen online are 10-bit and those seem already quite exclusive.

BTW I hate using "Q" for this parameter.

Me too! I only used the term to make clear I was referring to the BW resistors at the potentiometer's section, rather than the big resistors that set the general bandwidth of the filter.

I've noticed in sims that the resolution of the freq pot changes a lot depending on the value of those 'general bandwidth resistors'–when including a low frequency bypass resistor between opamps as it is in the current circuit. When using 5.1K as general resistor value, the range with more resolution tends to concentrate in lower frequencies. When using 22K instead resolution of the pots concentrates in higher frequencies. Below some examples.

This seems convenient to use one setup for lows and the other for highs, since normally when filtering ambient sound the extremes are where most resolution is needed. The only issue is that, to concentrate resolution on the highs, big noisy (22K) general resistors are needed. I wonder if the same concentration of resolution on the highs can be achieved with lower resistors. Maybe I should fiddle more with possibilities, but these two circuits are my complementary candidates to prototype during the next days.

Thanks a lot for your contributions!

 

[abbey road d enfer]

I've noticed in sims that the resolution of the freq pot changes a lot depending on the value of those 'general bandwidth resistors'–

Think about the maths. At the beginning of travel, the resistor from wiper to integrator input is high compared to the pot's resistance, so the resolution is that of the pot. When the pot is at mid position, that's where the resistor loads is the most, resulting in max deflection from half-value. Now if the value of the resistor is way smaller than teh pot's value, the deflection will be higher, resulting in a big leap from mid to top rotation, hence loss of resolution towards the top.

This seems convenient to use one setup for lows and the other for highs, since normally when filtering ambient sound the extremes are where most resolution is needed.

I'm not sure about it.

The only issue is that, to concentrate resolution on the highs, big noisy (22K) general resistors are needed.

Higher resolution in the highs mean "un-slugging" the pot.

I wonder if the same concentration of resolution on the highs can be achieved with lower resistors.

Have you actually checked the audibility of said noise? You may find that it's not unacceptable.
Maybe you could use lower value pots, like 1K?

 

[domingo]

When the pot is at mid position, that's where the resistor loads is the most, resulting in max deflection from half-value.

This is very challenging for me to understand, because I don't know exactly how the fpot affects the integer as a resistor divider to be honest. I see that at the beginning of the travel the resistance between the summing amp and the (first) integer is dominated by the wiper resistor, for the pot's resistance between input and wiper is close to zero. As the fpot turns, its resistance will dominate over the wiper resistor. But this doesn't account for the section of the fpot that goes to ground, which decreases with the travel and affects the integer in a way that I can't grasp. Then there is the added complexity of the high value resistor we added in parallel to the fpots and wiper resistors.

In my ignorance I'm a bit tied to sims and experiments. I found this another scheme that centers resolution to the middle of the travel. My experience is that when working ambient sounds (field recordings, not music), most filtering is done at the end highs and lows. The middle specially human voices tends to remain mostly untouched, so I presume that un-slugging the pots at the extremes of the low/high filters would be ideal. But probably I have to try all many options to find something workable.

Have you actually checked the audibility of said noise?

I never really found that noise was a problem really because it comes up along with a desired signal.