3D "AIR" EQ - "Night EQ" PCB's Complete!

[druu]

I've managed to build up a unit using 50k Lin rotaries and used Harpo's spreadsheet to calculate the values for each position. I've also changed the 56k in the input to 5k6 and added trimmers adjusted to 560R for each band/channel instead of the 5k6 from the original circuit. Centre position on the rotary is 5k6 for 0dB, using 11k3 parallel to reduce the switching noise as calculated on Harpo's spreadsheet. Noise is considerably lower with this setup.

Everything works great, frequencies are accurate enough but I seem to only be getting +/-3dB boost and cut. It seems as though each step is only 0.25dB and not 0.5dB as calculated.

Can anyone advise if there is a resistor value on the mainboard I can alter to get the desired +/-5.5dB?

 

[Harpo]

I've managed to build up a unit using 50k Lin rotaries

switched resistors substituting a rev.log taper pot, not a lin taper pot.

I've also changed the 56k in the input to 5k6 and added trimmers adjusted to 560R for each band/channel instead of the 5k6 from the original circuit. Centre position on the rotary is 5k6 for 0dB, using 11k3 parallel to reduce the switching noise as calculated on Harpo's spreadsheet. Noise is considerably lower with this setup.

5K6 probably refering to R30=R48. C13 previously 22pF increased to 220pF as well ? Keeping the C1 parts value, at least get rid of the 10K R5.

Everything works great, frequencies are accurate enough but I seem to only be getting +/-3dB boost and cut. It seems as though each step is only 0.25dB and not 0.5dB as calculated.

Might be jumper setting for balanced/unbalanced connection not set properly for corresponding following piece of gear with balanced/unbalanced cable in between that causes the missing 6dB. Same loss when switched to bypass ?

 

[druu]

Thanks for the reply Harpo.

Yes, sorry I meant 50k rev log. Values calculated using your v2 spreadsheet. I selected 5.5 max boost and -5.5 max cut but using MBB 23 step rotaries, so require the parallel R, calculated for -6 dB max cut with a value of 11k3 closest E96. This allows for 5k6 centre position value on the rotaries.

R30 has been substituted for 5k6 and R48 not used as I'm using separate relay bypass. Is R48 still required? I've not changed C13 as you mentioned, nor removed 10k R5 but will try this as per your advice. Would this cause the odd boost/cut limitation I'm seeing?

I don't think I have an issue with the bal/unbalanced section. With all band gain controls at centre position, high shelf gain at full CCW and shelf selector OFF, I have a reasonably flat  response and if I feed a -12 signal  from the DAW and back out I still get -12 on the master out with no loss.

In your spreadsheet you mention for the parallel R value, that the dB value must not be higher than regular max.cut. What is the regular max.cut you are referring to? Is it the +/-5.5 values chosen in the section above or the total for the circuit (i.e. 20dB?).

Appreciate your help with this stuff as always..

 

[druu]

I've replaced the 22pF for a 220pF and also removed the 10k at R5. No difference in the gain max boost/cut unfortunately..

 

[druu]

So I've realized that I'm using series string rotaries, calculated based on +/-5.5dB with 23 steps. Also using the parallel R which was calculated as being 11k3. With this I have the following:

closest E96 value to match E15
75000
27000
15000
9100
6200
4300
3300
2400
2000
1800
1500
1000
1100
820
750
680
620
430
510
360
430
390
3900

158590 total R but also includes parallel R if 11k3.

If I measure this as a 'pot', it is 10.5k CCW and 2.8k CW with 5k6 centre position. The spreadsheet has these figures:

"required R in
(to substitute Rseries+RV)"
10548
9958
9401
8875
8379
7910
7468
7050
6656
6283
5932
5600
5287
4991
4712
4448
4199
3964
3743
3533
3336
3149
2973

So R30 I'm using 5k6 and also worth mentioning I'm using relay bypass. With R30 at 5k6, series string resistors at 0R and steps calculated per the spreadsheet with parallel R, I get the same signal when bypassed and when IN.

The +/-3dB is still an issue - is this down to the circuit seeing only a 10k 'pot'?

 

[druu]

Just to give an idea of what I'm getting with the bands:

https://imgur.com/a/AAGZs

Just to recap;
- R30 = 5k6
- 23 step rotary switches calculated using Harpo's spreadsheet for +/-5.5dB max boost/cut, including parallel R value of 11k3 (for -6dB max cut between steps)
- Series string trimmers set to full CCW (i.e. 0R)
- Rotary switch values - CCW @ 10.5k, centre position @ 5k6 and CW @ 2k8
- Relay bypassed used but no difference between feeding a -12dB 1kHz tone whether IN or bypassed so centre position on all bands ('air' gain at full CCW) is at correct unity gain.

This is occurring on both channels as per the screenshot. After reading the whole thread it seems a few have experienced this issue but unfortunately no success stories on how it was overcome (e.g. here..).

Any ideas?

 

[frazzman]

I too have had similar issues to druu, and the other users who reported similar behaviour.

Has anyone successfully built a unit with stepped rotaries using Harpos logic?

 

[dagoose]

Guys.. it's not a standard series attenuator. I think that is where your problems might be.
It's just two resistors in parallel with the one switched into the circuit.

 

[druu]

Guys.. it's not a standard series attenuator. I think that is where your problems might be.
It's just two resistors in parallel with the one switched into the circuit.

Thanks Jeff. Totally get that but my series string rotary steps match the required figures from Harpo’s spreadsheet, all the way from CCW to CW. It’s basicallt the equivalent of the ladder type rotary..

 

[druu]

hmm.. so that is what i'm missing here.  :-[
They are indeed.. totaly missed that!
Thanks! 8)

EDIT: i shorted r39 and now i get a bit more boost (max 2.5 db) but still no 5db as it should be with 31k6.  ???
When i replace the 31k6 with 10k i get something like 5.5 db boost which is good. If i use the sheet with 10k max in the boost it says it's boosting 15db, which is totally different then the 5.5db i see.
With 330k for cut (= -15db according sheet) i get a cut of something like 3.5/4 db but with the 100k= -5db (according top sheet) i get only about 1db of cut.

anyone? I really don't get it..

Jeff, did you have to tweak more values to overcome this issue?

 

[dagoose]

Jeff, did you have to tweak more values to overcome this issue?

To be honest, it was quite some time ago but I think I didn't really tweak a lot of things.
The only thing that came to my mind is that I had an issue with  more boost then cut caused by the 5.62k resistor that had to be bridged. For the rest, these are the values that I used IIRC. This is a copy/paste from a small txt doc I had in my EQ3D map.

step 1 to 2 = 36k5
step 2 to 3 = 28k7
step 3 to 4 = 23k2
step 4 to 5 = 18k2
step 5 to 6 = 14k7
step 6 to 7 = 11k5
step 7 to 8 = 9k09
step 8 to 9 = 7k32
step 9 to 10 = 5k76
step 10 to 11 = 4k64
step 11 to sum = 17k8 (replaces R37,39,41,43,45)

less great than ugly, but its not that hard to do. See the little wiring sketch on the right side. The resistors for each step position at the switches are listed in the green cells, with a resistor in parallel (value in the orange cell) to get it less jumpy between switch positions. Each of these building blocks substitute a 470K rev.log. pot (RV36, RV38, .., RV44) with its 5.62K series resistor (R37, R39, .., R45) for each freq.band gain setting part of the summing amp.

Maybe you want to use a 12 pos. lorlin type switch instead, limited to 11 pos. for a center position at step 6, so you just replace the default value '23' with '11' for 'step positions' in the yellow cell. Maybe you want a different cut/boost range (within limits), so replace the default values '5.5' and '-5.5' with maybe '5' and '-12.487' for 'max.boost' and 'max.cut' in the yellow cells. Each time you update a value in the yellow cells for your needs, the whole chart is recalculated and resistor values picked for the closest E96 value to match your plan.

 

[dagoose]

I think this was the sheet that I used for calculations. (rename the jpg to xls)

 

[druu]

Thanks for the reply Jeff.

I'm a little confused as to how you got your values mentioned above but at the same time, you're using a ladder type with less steps so would differ.

I've just tried to simulate with the values I'm using in the circuit (mentioned above a few posts) but also getting spot on values (see attached).. more than confused now.. it must really not like the series string assembly of the rotary, but wondering why this is the case..

 

[Harpo]

I've just tried to simulate with the values I'm using in the circuit...

... missing the 1k54 shunt arm resistor between your R2 and R3 (original R24/R25) for the -22dB pad (and ignoring the -6dB balanced line receiver and the last inverting +6dB stage that should keep it at unity with correct jumper settings).
Your tested response for the range 1Hz...1.33Hz doesn't make much sense for the 640Hz band (and missing all remaining bands that add up to this band as well due to the non-inverting gain stages with their -6dB slopes that stop to fall when unity gain is archieved. Max.cut is about -5.5dB with all remaining bands set for +/-0dB.)

 

[druu]

Sorry Harpo, I had the previous sim zoomed to show the gain steps. I've attached another slightly updated but I didn't realize the missing 1k54. In this updated one I added real life resistance steps at each step point on Rgain and also added the 11k3 parallel R to simulate my series string rotary.

Can you advise on what I should be seeing if I measure Vrms at XLR in vs XLR out? I do seem to have a variance there, if I feed 1.0v input, I get around 0.7v output, a gain drop of 3dB?

 

[druu]

OK! I've managed to replicate the issue in the schem based on your suggestions Harpo. I added all bands (except the Air band or selector) and also added the 1k5 arm shunt, which you were spot on about as I had simulated without it and it was clearly showing unity at +22dB on the sim.

Now that I've replicated, is there any way I can try to remedy this and somehow get it back to being +/-5.5dB as calculated in the spreadsheet?

 

[druu]

Spice schematic..

https://imgur.com/a/nQPH5

 

[Harpo]

.. but I didn't realize the missing 1k54.

.. that is still missing in your sim schematic. These fixed resistors 18.2K and 1.54K form a -22dB voltage divider (the 1.54K better would be a 1.58K with other parts -more unlikely- being perfect) that substitute a 20K trimmer from design stage.

In this updated one I added real life resistance steps at each step point on Rgain and also added the 11k3 parallel R to simulate my series string rotary.

.. showing, the 0.5dB steps should work as calculated.

Can you advise on what I should be seeing if I measure Vrms at XLR in vs XLR out? I do seem to have a variance there, if I feed 1.0v input, I get around 0.7v output, a gain drop of 3dB?

measured at what frequency ? You could do a frequency sweep across the audible band to see the overall outcome with all level settings centered) As already said, your sim is missing the remaining bands that add up to the frequency under test by a more or lesser amount.
Imagine one of the paralleled frequency setting caps missing a connection to the pcb trace will shift the response of this bandwidth limited gain stage up by one octave, causing this 3dB bump (and a 3dB dip 2 octaves below), just one idea of a possible cause.

oops, answering while you posted ..

 

[druu]

Thanks Harpo. Did you see my latest posts above?

To answer your question on frequency, this was a 1kHz tone sent from my DAW.

Also, the simulation above should show this isn't a pcb related issue.

 

[druu]

Ok so I've just tried without the parallel R 11k3 on the rotary switches and when running the simulation without these on any band, shows a -6dB drop in total frequency response but gain steps are returned to the normal 0.5dB for each step.

So it seems the parallel R is causing a total drop of -6dB in each band as a side effect, effectively halving each bands gain range?