Baxandall Volume Control

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

ej_whyte

Well-known member
Joined
Nov 12, 2010
Messages
263
Location
Cambridge, UK
I recently picked Self's Small Signal and have been plowing through it, really good stuff :) I am quite interested in the Baxandall active vol circuit that is described in chapter 9, but have couple of questions.

What resistors are used in calculating the cut-off of the 1st blocking cap? (C1 on schemo). I can't work out whether it is just that left-hand end of RV1 or if R3 should be added in series, not sure how the cap in between the pot and R3 would affect.

Also, why is it that the last cap, C4, has its polarity reversed from what you normally see for a coupling cap? Whilst on the subject actually, I would also like to clear something up: polarised electrolytics introduce distortion when the voltage goes opposite-to-polarity, and therefore add distortion to every negative half of a cycle?

Finally, is it at all feasible to substitute the inverting op-amp stage with a tube stage? I guess it would be advisable to put the C4 blocking cap before the global feedback to RV1.

Cheers
 
This?
http://books.google.com/books?id=PvKPEFu2PVkC&pg=PA237&lpg=PA219&vq=chapter+9

The resistors R3 and of course RV1, but there is a scaling of RV1 "influence" because of the -6,9x gain in the U1A inverting stage, so it finally gets to the 1k-something value as described.

Perhaps you could get by by using 100uF caps there.

Use "tubes" where they are to be used, IMHO. 1K-something impedance driving is a bit hard to swallow for glassware?
 
ej_whyte said:
What resistors are used in calculating the cut-off of the 1st blocking cap? (C1 on schemo). I can't work out whether it is just that left-hand end of RV1 or if R3 should be added in series, not sure how the cap in between the pot and R3 would affect.
A schemo would help.
You have to take into account the minimum value of the input impedance, which occurs when gain is maximum. The input impedance then is the resistor that goes to the pot. Typically, with a 10k pot and 10dB gain, the input res wil be abour 4k. That's the value you have to use in the calculation of the -3dB LF response.
Also, why is it that the last cap, C4, has its polarity reversed from what you normally see for a coupling cap?
Can't answer that without a schemo.
I would advise you to use NP caps; they are much more affordable and available than they were before.
Whilst on the subject actually, I would also like to clear something up: polarised electrolytics introduce distortion when the voltage goes opposite-to-polarity, and therefore add distortion to every negative half of a cycle?
No. Distortion appears both ways, although the presence of even harmonics indicate that the process is asymmetrical.
El caps don't like AC across them, so you have to use much larger caps than the -3dB calculation would indicate. A good rule of thumb is use 10 times the value calcualted for the lowest frequency. In your case, assuming 4kohms, the -3dB calc indicates 2uF, so you would use at least 22uF. Considering the cap will age, why not have another safety margin by making it 47 or even 100uF. The price is just about the same, and it doesn't take more space on a PCB.
Finally, is it at all feasible to substitute the inverting op-amp stage with a tube stage?
Certainly the original Baxendall circuit was tube-based. You have to scale the values according to the tube stage's out impedance, though.
 
Oops just realised I forgot to attach the schemo, its the one TV posted but here it is again  ::)

When the pot is all the way to the left C1 & C2 are in series so I guess the cutoff is just the combination of the resulting ~20uF and R3, I'm just not to sure on when the pot is all the way to the right.

I was already leaning towards NP electros or poly where possible, but thanks for the extra info.

How does the input at max gain come to 1.27kΩ? And why is C4 backwards compared to most blocking caps?

Cheers
 

Attachments

  • Screen shot 2011-11-22 at 12.13.30.png
    Screen shot 2011-11-22 at 12.13.30.png
    31.3 KB · Views: 371
OK, this is not the typical baxendall circuit.
In the extreme position, the input impedance there is given by the FB resistor (the pot) divided the amount of gain.
gain is 4.7k/680 (the gain of the inverting stage), so the input resistance is 10kx680/4700= ca 1.2k.
So you have to calculate the value for -3dB at 20Hz, which gives 6.6uF. That would suggest using a minimum value of 68uF; they have chosen an additional safety factor of 3-4.
The orientation of these el caps is indifferent. they could be in any direction.
 
Aah ok thanks, so the impedance at any given position can be calculated by the input side of the pot, in series with feedback side divided by the gain? Example with the pot in the middle:

5k + (5kx680/4700) = 5.72k

Correct?

I just did the 10kx680/4700 though and it comes out as 1.446k. Not a majorly significant difference, however there must be a reason why Self has given it as 1.27k?

How would a typical Bax circuit differ from the example? Is there a paper or any other source on it?

Thanks for the help
 
While I suspect Peter Baxandall designed more than one the circuit in his career, the classic tone control he is now identified with has the useful characteristic that a gain pot spans across the input and feedback sides of a typical inverting gain stage.

This offers noise gain (amount of gain applied to active devices)  that is only as high as it needs to be for a given gain command. This contrasts with the typical passive circuits of the day that were routinely followed by high fixed gain stages so their noise was always present.

JR

 
 
ej_whyte said:
Aah ok thanks, so the impedance at any given position can be calculated by the input side of the pot, in series with feedback side divided by the gain? Example with the pot in the middle:

5k + (5kx680/4700) = 5.72k

Correct?
I just did the 10kx680/4700 though and it comes out as 1.446k. Not a majorly significant difference, however there must be a reason why Self has given it as 1.27k?
My explanation was somewhat rudimentary. In the max position, the pot is submitted to (1+G) times the input voltage, so the impedance is 10k/(1+G), which comes to 1.263k. Then the 22k is in parallels with the cursor. At max gain, it is directly in parallels with the input, which would account for 1.19k. 
How would a typical Bax circuit differ from the example? Is there a paper or any other source on it?
See attached picture. It is more a concept than a real application; there are many derivatives of this concept.
 

Attachments

  • bax VC.jpg
    bax VC.jpg
    40.1 KB · Views: 28
> polarised electrolytics introduce distortion when the voltage goes opposite-to-polarity

The voltage across a sufficiently-large coupling capacitor does not change over the audio cycle.

If it did, we'd lose signal.

Obviously it does, down in the bass. Using Self's 220u/22K, loss is significant at 0.03Hz. At 1,000X times higher, loss (and therefore voltage-change) across capacitor is super-teeny.

> why is C4 backwards compared to most blocking caps?

Do not know which way is "most".

At first glance, the DC voltage here is zero. Polarized electros DO work fine this way. Reverse voltage breakdown takes several volts.

There is always some stray small DC. The NE5532 has NPN inputs. All pins of "U1:B" (note backward sequencing) will be at negative 4mV (nominal) due to bias drop in R3 and follower action. To my eye, this forces current in R1 which will swing U1:A output positive some dozen+ mV. This is certainly no big deal; and I am not sure of my reasoning.

But why think?? Build it, leave the caps out. Measure voltage polarity across the cap pads. Put the cap in with the polarity indicated.

BTW, I think Self's description of this clever volume control deserves re-reading. While Self is not 100% infallable, I'd trust his books more than whatever some John, Abbey or Paul spills on the interwebs.

> substitute the inverting op-amp stage with a tube stage?

With better understanding, you see that this has been designed specifically for 5532 and similar strong opamps. Your screen-clip does not show the stage to the left, but in the older book it is a 5532 to buffer-up to a decent Line Input impedance.

BTW: the point of the plan is to get a nearly linear-in-db response from a linear pot.

Somehow we've muddled along with "audio taper" pots.

Yes, sure, of course, all "opamp" tricks can be done with tubes. Perhaps simpler. Jimmy a Cathodyne for gain-of-5 to the plate. Wrap a linear pot across it. Use something beefy to drive the lowest impedance, probably a cathode follower. You certainly want to go 10X higher in every resistance because even 13K is a non-trivial load for a small tube. You might want to go 100X higher, 130K, but that takes you to a 1Meg pot and probable resistance noise higher than the tube's already significant hiss. And this whole 2-triode trick has maximum gain of 5. Whereas a audio-taper pot and two triodes should give a dirty-gain of 2,000 or be clean at gain of 30-100 and do something useful also.
 
Yeh this is why I picked this circuit out, useful pot tracking for stereo channels and less noise when set to lower gain. The pot tracking gets screwed at the bottom of travel but the ~40dB range before this happens given in Self's example fits well for the application I will be using it for.

Why is C4 so large? When combined with R4 it gives a cut-off of 0.03Hz, wouldn't a 0.3Hz cut-off from a 22u cap suffice? I guess the lower the better really.

Cheers
 
ej_whyte said:
Why is C4 so large? When combined with R4 it gives a cut-off of 0.03Hz, wouldn't a 0.3Hz cut-off from a 22u cap suffice? I guess the lower the better really.
The distortion mechanism in electrolytics depends primarily on the voltage developed across the cap. Making the cap >> than what is strictly needed for proper -3dB LF corner ensures this voltage is extremely low compared to the signal, so the distortion ratio is very small.
 
I have been keeping the electrolytics to ~0.2Hz with a poly at 2Hz in the first filter in the system, following the rule of electros having a cutoff 10x lower. I've tweaked the design slightly to change the cut-off freqs and adjusted the gain to trimmable between 20-26dB, any glaring mistakes?

screenshot20111127at230.png


Slightly OT, but the GND pin on a typical DOA pinout, is that meant to be signal or power ground?

Cheers
 
ej_whyte said:
I have been keeping the electrolytics to ~0.2Hz with a poly at 2Hz in the first filter in the system, following the rule of electros having a cutoff 10x lower. I've tweaked the design slightly to change the cut-off freqs and adjusted the gain to trimmable between 20-26dB, any glaring mistakes?
What's your target load impedance? I would increase the output cap to 100uF for 10k load.
Slightly OT, but the GND pin on a typical DOA pinout, is that meant to be signal or power ground?
In most cases, it's a power ground, used to bias current mirrors and current sources. But there may be compensation caps connected there. Do you want to know for the PCB software? In doubt make it a signal.
 
Heres a basic overview of what it will be feeding. Straight-from-the-data-sheet THAT 1646/1206 combo for a switchable insert, summed with another similar channel in to an inverting opamp.

Cheers
 

Attachments

  • Overview.jpg
    Overview.jpg
    150.3 KB · Views: 109
ej_whyte said:
Heres a basic overview of what it will be feeding. Straight-from-the-data-sheet THAT 1646/1206 combo for a switchable insert, summed with another similar channel in to an inverting opamp.
The load impedance for the Bax volume ctrl stage can be as low as 1kohms. Then the -3dB point would be 0.7Hz, which is 4x your design rule. You may want a larger cap (1000uF) or increase the 1k res (R102).
BTW there's a mistake with the res going from output to non-inverting input of IC24b. The res shoud be between output an dinverting input and the non-inverting input should go to gnd.
Also I see you have a DC blocking cap at the output of the Bax VC, but none at the output of IC23. I know the offset is supposed to be much smaller than the input stage, but are you confident with that? You may have a click when switching the insert in.
 
OK cool thanks, I think i'll increase R102 to 3.9k to bring the cutoff down and that should be nice and easy for the 1206 (I need to look in to what load these can take). Good point on the cap for IC23 and the resistor mistake, I haven't gone through those parts in detail yet and just quickly fleshed them out for that overview.

R31 is not essential if im just running short pcb tracks not long cable runs right?

Cheers
 
I'm trying to use this Baxandall active volume control in a stereo line preamp, using a stereo 10k lin pot and I unfortunately get a mismatch between both chanels (up to 0,5db sometimes)

  I must say in my version R2 is 12k instead of 4k7
index.php


  I'd like to find a way to trim or correct this mismatch between left and right chanels..
  We can find a similar gain control in the SSL4000 line preamps (around op amp T3) :
http://www.gyraf.dk/schematics/SSL_82E01_Channel_Amplifier.GIF
  Is the 5k pot labelled "pre" used as  a trim? or maybe to set unity gain if the gain pot has a center detent?
 
> there should be a end limit resistor

End-gain is limited by R1 R2 ratio.

Gillis does not say *where* the mismatch is happening. This topology does not really reduce the need for dual-pot tracking though mid-turn. While tracking may be "better" with Lin instead of Audio taper, it may not be good enough.
 
Back
Top