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

[pieca]

Hi,

I've got some issues with my build and stepped pots…
I'm using 23 steppes rotary switches  with resistor calculated with harpo's exell file
R37, 39, 41, 43, 45 are bridged
R30 is 56200 ohms
According to the exell file, my switches are going from 105K (-5 Db) to 30K1 (+5.5 Db) center position  (0db) is 56K2
all my resistors are 1% and were measured before soldering

After measures, this is my results :

with all pots centred, I'm not flat…
pots centered :

Bypass :

for sub, 40Hz and 160 Hs pots, i've got something like -2 Db @105k to +2 Db @ 30K1

For 650 Hz : -1.5 Db to +2.5 DB

For 2500K : -2.5 Db to +3.5 Db

For the Air band (all frequencies): from +2.5 Db (105k min position) to +7 Db (30K1 max position)

is everything normal or did i missed something?
I'd prefer not un solder 276 resistors and solder 276 new resistors again…

Thanks for your help

EDIT : same behavior with both channels

 

[frazzman]

Hi all,
Does anyone have any test sweeps of their completed nite eq builds ? My unit just doesn't "sound" right. When I throw it on the mix bus it sounds a bit "flat", to my ears it sounds a bit like I'm loosing some low end punch, then when I bypass or remove the insert the mix immediately sounds "fuller" again. From what I can tell, all bands on both channels are boosting and cutting correctly.

My testing is with the eq engaged, all pots set to 0 (center indent) , air band off.

I've done some testing using RMAA. This is what the plot of my right channel looks like. Left looks pretty much the same too. The curve is not overly flat, there's a roll off to 40hz, which may account for some loss in the lows, does this look right???

 

[frazzman]

Just a second thought ... I just realised I did substitute a number of resistors in the circuit (particularly the precision values around the potentiometer sections on the PCB) that were hard to come by - I.e. I used 20k instead of 20k5, 1.2k instead of 1k27... So if I were to change all the resistors around the potentiometer sections to the precise resistor values would this bring me closer to a flat response ? This is the only thing I can attribute to the circuit not "sounding right"

 

[frazzman]

Hey all,
Still fighting with this build a little bit, appreciate if someone has any input they can offer.

After more testing, it has become apparent I have about a -2db cut @ 40hz on my right channel with EQ engaged and all pots set to 0 (center ident).

Any ideas what could be causing this only on 1 channel? I've checked for shorts and correct component designations. I also replaced the SUB and 40hz pots as there was a fair bit of cramming to fit all these pots on the front panel but it didn't have any impact. I am getting cutting and boosting action on sub and 40hz bands but for some reason when set to 0 I am seeing this dip.

This is my main issue...

The other issue is that with the EQ engaged and all pots set to 0, the frequency curve is not the same as when bypassed, It is almost totally flat in bypass. As I noted earlier, I substitued resistors around the pots section, i.e. 20k instead of 20k5 etc, so I am guessing this is the contributing factor to that issue? I am going to substitute these resistors for exact values.

This still won't fix my 40hz @ -2db cut issue, can anyone offer any advice? I've attached plots in both bypass and with eq engaged and all pots set 0. Thanks a lot

 

[frazzman]

And again in bypass... no issues there

 

[Harpo]

..with EQ engaged and all pots set to 0 (center ident).

These pots with neg.log taper for usual come with +/-20% tolerances. You want the half rotation mechanical center about 10% of its stated value, so this half rotation value + series resistor = value of summing amps feedback resistor. I have an excel file posted earlier in this thread to fix this tolerance related behaviour with hopefully least efford. (and this is a different subject from frequency bands not being exactly 2 octaves apart for the remaining bumps. Next to other improvements this would be a non issue with derived bands, but the OP insisted on keeping it close to the original, at least it was his thread).

 

[frazzman]

..with EQ engaged and all pots set to 0 (center ident).

These pots with neg.log taper for usual come with +/-20% tolerances. You want the half rotation mechanical center about 10% of its stated value, so this half rotation value + series resistor = value of summing amps feedback resistor. I have an excel file posted earlier in this thread to fix this tolerance related behaviour with hopefully least efford. (and this is a different subject from frequency bands not being exactly 2 octaves apart for the remaining bumps. Next to other improvements this would be a non issue with derived bands, but the OP insisted on keeping it close to the original, at least it was his thread).

Hi Harpo,
Thank you for your reply. I know I have two separate issues, one being the -2db 40hz~ cut and the other being that the freq curve isn't flat with pots centred. The 40hz~ cut doesn't seem to be related to the tolerance of the pot(s) as replacing it with another pot nets exactly the same result.

I found your excel file which will help perfectly for offsetting tolerance around center ident variations, but what about the 2db cut on my right channel, this is independent of the pot. Thanks again

 

[Harpo]

double check real world parts values of R18, R20 and R43.

 

[frazzman]

double check real world parts values of R18, R20 and R43.

Thanks Harpo, already checked but I just pulled the resistors from the board and rechecked, they were all within a 5-10% tolerance, this time I even soldered on series some extra resistors to get the values near dead on with the design, but no luck.
Is there anywhere you can think of that a short could exist and result in this issue? Also, could it possibly be an issue around the sub band ? I didn't think so as the frequency plot indicates it's around 40hz but I was wondering if the sub and 40hz band are related in that sense? Thanks again

Edit: maybe a slack workaround but would changing the value of the 5k62 resistor help get me to 0db ?

 

[Harpo]

Is there anywhere you can think of that a short could exist and result in this issue?

a short between opamps output and non-inverting input (unity gain follower) would be missing the LPF part and drop in level would be higher. Same with shunt arm open circuit for the HPF part from overheating caps (you'd have to fry both caps by too long soldering. Only one dead cap would shift frequency response one octave up, not two), so this would be a more unlikely reason.

Also, could it possibly be an issue around the sub band ?

easy enough for you to double check the 10Hz band with the pcb right before your eyes...

maybe a slack workaround but would changing the value of the 5k62 resistor help get me to 0db ?

decreasing R43 will get you closer, but the peak in the center of the dip will be boosted as well.

 

[frazzman]

Is there anywhere you can think of that a short could exist and result in this issue?

a short between opamps output and non-inverting input (unity gain follower) would be missing the LPF part and drop in level would be higher. Same with shunt arm open circuit for the HPF part from overheating caps (you'd have to fry both caps by too long soldering. Only one dead cap would shift frequency response one octave up, not two), so this would be a more unlikely reason.

Also, could it possibly be an issue around the sub band ?

easy enough for you to double check the 10Hz band with the pcb right before your eyes...

maybe a slack workaround but would changing the value of the 5k62 resistor help get me to 0db ?

decreasing R43 will get you closer, but the peak in the center of the dip will be boosted as well.

Thanks Harpo, I'll need to look at this closer, I've been working on the right channel far too long so I decided to leave it for a while and start fixing up the left channel based upon your spreadsheet (wish I had have seen it first time round...). I wanted to adjust the resistor values on each band to try and get 0db center ident for each freq...

Based upon your spreadsheet, I ended up with the following values:

SUB = R pot total: 445,000, R pot center: 56,300, R series: 10,470
40Hz = R pot total: 472,000, R pot center: 54,500, R series 12,270
160Hz = R pot total: 483,000, R pot center: 56,600, R series: 10,170
640Hz = R pot total: 451,000, R pot center: 54,100, R series: 12,670
2.56 Khz = R pot total: 481,000, R pot center: 60,700, R series: 6,070

R fb = 66,770

Since the R series values are so 'irregular' I had to put a few resistors in series to get to the desired values. I replaced R30 & R48 with a 68k resistor (closest to 66770ohms I had) - is there any impact not using an exact value for R30 & R48? The only audiable effect I can hear is that the output is slightly hotter than before

Attached is my frequency response after these resistor value changes, its certainly better but far from flat... I am starting to think that trying to get the center ident to track perfectly at 0db is not a worthwhile exercise. I wanted to use this EQ in a mastering chain but I don't think i'm going to get its response flat enough. The low end roll off seems really unusual too.

How does the plot for the left channel look? Am I being pedantic or have I missed something / misinterpreted your instructions?

Back to working on Right channel later today.... sorry for all the long winded questions

 

[Harpo]

How does the plot for the left channel look?

bumpy road. As previously said, center frequencies between these bandwidth limited gain stages have to be exactly two octaves apart, IE 10,40,160,640Hz and this being a different subject from gain setting. FI your '640 Hz' band looks too high frequency, resulting in a dip below and a peak above this frequency band. With schematic parts values this will result in too low 626.6Hz. Using higher tolerance resistor parts (IE 1K2 instead of 1K27 from schematic) will result in too high 663.1Hz on paper. For the wanted 640Hz and assumed perfect corresponding caps, resistor values needed are 1K243 for R14 and 124R3 for R12 for this band.
Tweaking the series resistor will not fix this behaviour.

 

[frazzman]

How does the plot for the left channel look?

bumpy road. As previously said, center frequencies between these bandwidth limited gain stages have to be exactly two octaves apart, IE 10,40,160,640Hz and this being a different subject from gain setting. FI your '640 Hz' band looks too high frequency, resulting in a dip below and a peak above this frequency band. With schematic parts values this will result in too low 626.6Hz. Using higher tolerance resistor parts (IE 1K2 instead of 1K27 from schematic) will result in too high 663.1Hz on paper. For the wanted 640Hz and assumed perfect corresponding caps, resistor values needed are 1K243 for R14 and 124R3 for R12 for this band.
Tweaking the series resistor will not fix this behaviour.

Thank you Harpo, I managed to fix the troublesome 40hz-ish dip, looks like there was a dry joint on one of legs of a cap on the 40hz band, not sure but I reheated all joints in this section and the fault is gone.

So I ordered precision value resistors (0.1%) tolerance, inline with the values in the schematic - so I take it that once I replace all higher tolerance resistors currently fitted, and then change series resistors inline with the center point measurements for each pot - I should be well on my way to a flat curve ?  Have I missed anything else ? I think with this project it's a very easy build but all the work is in refining it

 

[Harpo]

How does the plot for the left channel look?

bumpy road. As previously said, center frequencies between these bandwidth limited gain stages have to be exactly two octaves apart, IE 10,40,160,640Hz and this being a different subject from gain setting. FI your '640 Hz' band looks too high frequency, resulting in a dip below and a peak above this frequency band. With schematic parts values this will result in too low 626.6Hz. Using higher tolerance resistor parts (IE 1K2 instead of 1K27 from schematic) will result in too high 663.1Hz on paper. For the wanted 640Hz and assumed perfect corresponding caps, resistor values needed are 1K243 for R14 and 124R3 for R12 for this band.
Tweaking the series resistor will not fix this behaviour.

Thank you Harpo, I managed to fix the troublesome 40hz-ish dip, looks like there was a dry joint on one of legs of a cap on the 40hz band, not sure but I reheated all joints in this section and the fault is gone.

So I ordered precision value resistors (0.1%) tolerance, inline with the values in the schematic - so I take it that once I replace all higher tolerance resistors currently fitted, and then change series resistors inline with the center point measurements for each pot - I should be well on my way to a flat curve ?  Have I missed anything else ? I think with this project it's a very easy build but all the work is in refining it

in the values needed ? Just run your numbers for the correct values needed when assuming other parts perfect (schematic values are for higher tolerance parts, IE 1K27 would better be 1K241 for your E192 0.1% resistor). This circuit is highly tolerance sensible, so not an easy build.

 

[frazzman]

How does the plot for the left channel look?

bumpy road. As previously said, center frequencies between these bandwidth limited gain stages have to be exactly two octaves apart, IE 10,40,160,640Hz and this being a different subject from gain setting. FI your '640 Hz' band looks too high frequency, resulting in a dip below and a peak above this frequency band. With schematic parts values this will result in too low 626.6Hz. Using higher tolerance resistor parts (IE 1K2 instead of 1K27 from schematic) will result in too high 663.1Hz on paper. For the wanted 640Hz and assumed perfect corresponding caps, resistor values needed are 1K243 for R14 and 124R3 for R12 for this band.
Tweaking the series resistor will not fix this behaviour.

Thank you Harpo, I managed to fix the troublesome 40hz-ish dip, looks like there was a dry joint on one of legs of a cap on the 40hz band, not sure but I reheated all joints in this section and the fault is gone.

So I ordered precision value resistors (0.1%) tolerance, inline with the values in the schematic - so I take it that once I replace all higher tolerance resistors currently fitted, and then change series resistors inline with the center point measurements for each pot - I should be well on my way to a flat curve ?  Have I missed anything else ? I think with this project it's a very easy build but all the work is in refining it

in the values needed ? Just run your numbers for the correct values needed when assuming other parts perfect (schematic values are for higher tolerance parts, IE 1K27 would better be 1K241 for your E192 0.1% resistor). This circuit is highly tolerance sensible, so not an easy build.

Hi Harpo,

I've followed all up on all your suggestions and have replaced all of low tolerance resistors with high tolerance values with precision values as per schematic - I.e. Replace 1k5 > 1k54, 1k2 > 1k27 etch. The whole board is now populated with precise resistor values.

I have changed all 5k6 resistors on each frequency band inline with the values generated using  your spreadsheet. I put multiple resistors in series to achieve the values generated in your spreadsheet as most required values were not standard resistor values. I was very particular in this process to ensure that values matched the spreadsheet exactly.

Unfortunately after hours completing this, I still can't get the curve to behave. It's a a little flatter nut still not right... Is there something I am missing ?

If I used alpha rotary switches instead of the pots could this alleviate my problem? I'm really surprised that there haven't been more builders mentioning this issue.

Thanks a lot

 

[shot]

I had similar issue with very irregular curve, but mine was built using pots from AML.
I found out that those pots were so off that I had to use very different values of series and summing resistors. I had a calculated that value for some resistors had to be around 14k instead of 5k6 to be correct!
It's all explained in Harpo's excel document how to calculate needed values for those resistors.

What I did is I replaced all problematic resistors with combo of resistor+trimmer so I was able to finetune my unit to be almost flat. Almost flat, since when measuring with Smaart software I was able to zoom in so that I can still see that it's off by 0.2db at max. (I won't mention how ugly now my unit looks with all those trimmers above pcb, hehe!)

I still do have problem with my unit that it cannot cut more than 4db, and boosts by +16db. But I'm fine with that. It's mystery but it'll stay as is...

You should measure your switches. If you used high tolerance resistors, your switch is probably behaving a bit off the specified value. This eq is very picky on resistor values! Measure your switches in center position and insert those values into Harpo's excel spreadsheet. You'll probably have an idea what to do next then.

 

[frazzman]

I had similar issue with very irregular curve, but mine was built using pots from AML.
I found out that those pots were so off that I had to use very different values of series and summing resistors. I had a calculated that value for some resistors had to be around 14k instead of 5k6 to be correct!
It's all explained in Harpo's excel document how to calculate needed values for those resistors.

What I did is I replaced all problematic resistors with combo of resistor+trimmer so I was able to finetune my unit to be almost flat. Almost flat, since when measuring with Smaart software I was able to zoom in so that I can still see that it's off by 0.2db at max. (I won't mention how ugly now my unit looks with all those trimmers above pcb, hehe!)

I still do have problem with my unit that it cannot cut more than 4db, and boosts by +16db. But I'm fine with that. It's mystery but it'll stay as is...

You should measure your switches. If you used high tolerance resistors, your switch is probably behaving a bit off the specified value. This eq is very picky on resistor values! Measure your switches in center position and insert those values into Harpo's excel spreadsheet. You'll probably have an idea what to do next then.

Hi Shot,

Thanks a lot for your reply. Just to clarify I used pots from AML as well. I used Harpos spreadsheet to calculate the series resistors for each band. Given the measurements of my pots I ended up at with series resistors ranging from 6-12k roughly. Even with these fitted, the EQ is still not flat. I also replaced all high tolerance resistors throughout the board with low tolerance precision resistors.

I was thinking to myself last night that it makes the most sense to introduce a trimmer into the 5k6 position of each band. How did you go about doing that ? Like for example of you needed say 12k470 , would you wire a 10k + 470r + 2k trimmer in series ?

I built this eq in no time at all but I've spent about 10 x longer trying to refine the frequency response... I really wanted to get it right as I can tell this is a good sounding eq so it would be a shame not to have it right.

I was thinking about removing all the pots and wiring up rotary switches but that would be a lot of work so hopefully I can introduce trimmers to fine tune... You don't happen to have any pics of how you wired in your trimmers, I'd be interested to see how you arranged it all

Thanks for your help

 

[frazzman]

Just a second thought ... Why couldn't you just replace the series resistor on each band with say a 15k trim pot and use that solely for adjustment rather than adding other resistors in series??

 

[shot]

I actually replaced a lot of resistors for resistor+trimmer combo!
All of the series resistors, summing resistor R30, bypass resistor R48 (and R24 not relevant for this issue - I was hunting my boost/cut asymetry issue).

I went for single turn 5K trimmers just cause I had a lot of them around.  Surely 2K would suit even better but hey - I was in a rush to do it same day. For summing and bypass I went with single 100K trimmer.

If you want to use trimmer only (like 15K or such) than be sure to use multi-turn trims. Even with my 5K single turns it was a bit pain to get an exact value.

I soldered resistor (or two) to the legs of the trimmer using center wiper and one outer leg - basic rheostat configuration. Than I'd adjust trimmer so that the series value (resistor+trimmer) is exactly as calculated by spreadsheet. And only then I would solder this combo in place of the resistor. Looks messy but it works. After all is in place, I would trim more precisely passing pink noise and matching it on the transfer function analysis in Smaart software.
I went to replace R48 (bypass) since I was fiddling with R24 and I messed up gain in bypass state. Probably you won't have to do it (but it can't hurt - do it if you feel need to reach absolute unity gain in bypass).

Here's a photo of my unit. It's hard to capture third perspective with camera but I hope it can be informative enough.
Note that some trimmers that have legs too close to other wires have shrinkwrap on their legs. Under the shrinkwrap there's a resistor!
And don't forget to clip the third unused leg of the trimmer just so it doesn't move to a contact with components underside!


imagen

 

[Harpo]

in the values needed ? Just run your numbers for the correct values needed when assuming other parts perfect (schematic values are for higher tolerance parts, IE 1K27 would better be 1K241 for your E192 0.1% resistor). This circuit is highly tolerance sensible, so not an easy build.

Hi Harpo,

I've followed all up on all your suggestions and have replaced all of low tolerance resistors with high tolerance values with precision values as per schematic - I.e. Replace 1k5 > 1k54, 1k2 > 1k27 etch. The whole board is now populated with precise resistor values.
I have changed all 5k6 resistors on each frequency band inline with the values generated using  your spreadsheet. I put multiple resistors in series to achieve the values generated in your spreadsheet as most required values were not standard resistor values. I was very particular in this process to ensure that values matched the spreadsheet exactly.

Unfortunately after hours completing this, I still can't get the curve to behave. It's a a little flatter nut still not right... Is there something I am missing ?

Didn't say so. I said 'run your numbers for the correct values needed', done by [-3dB cutoff freqency=1/(2*PI()*R*C)] with R in ohm and C in Farad.
In each bandwidthlimited gain stage you have a 1st.order HPF, set by the value of the paralleled caps in series with the shunt arm resistor. When you assume the caps value -more unlikely- being perfect 1.00uF||1.00uF=2.00uF=0.000002F, the resistor values for the 10,40,160 and 640Hz bands needed are 7958, 1989, 497.4 and 124.3 ohm. Just for examples sake, if all caps values are maybe 1% high, IE 2.02uF instead of 2.00uF, resistor values would need to change accordingly to 7879, 1970, 492.4 and 123.1 ohm. If all caps values are maybe 1% low, IE 1.98uF instead of 2.00uF, resistor values would need to change accordingly to 8038, 2010, 502.4 and 125.6 ohm. The so called '2K5' shelfing band is a 1280Hz HPF for real with a different from other bands voltage gain setting, so shunt arm resistor value needed is 62.17 ohm with a perfect 2.00uF cap.
IE measure the real world caps value of each band and calculate the resistor value needed by Rx=1/(2*PI*frequency*measured capacitance in F) as being the lesser evil to tweaking caps. Using a multiturn trimmer of sufficient value instead of the shunt arm resistor might be an easier workaround if you don't have the needed value at hand.
Same goes for the LPF at the same -3dB cutoff frequency as the HPF, using same formula, set by the two paralleled 100nF caps=0.2uF in parallel to the gain stages feedback resistor, giving ideal value 79.58K, 19.89K, 4.974K and 1.243K if the caps values are spot on. IE measure the real world 0.2uF caps value and calculate the resistor value needed. Again, a multiturn trimmer (in parallel to a higher value fixed feedback resistor to prevent an open loop gain stage from a failing trimmer) might help. As a followup, any ratio change of Rfb/Rshunt will result in a needed tweaking of the from schematic 5.62K series resistor in order to correct the change in voltage gain of the corresponding band stage.
(all this hassle could have been a walk in the park with derived filters that by definition just have this wanted -6dB/oct slope ...)

If I used alpha rotary switches instead of the pots could this alleviate my problem?

You have to get the frequencies correct first. Tweaking a bands gain value and/or an off-center level pot by a series resistor is an easy further step (FI by using a multiturn trimmer of sufficient value to substitute the 5.62Ks from schematic).