DC blocking caps

GroupDIY Audio Forum

Help Support GroupDIY Audio Forum:

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

sr1200

Well-known member
Joined
Dec 6, 2010
Messages
2,094
Location
Long Island, NY USA
Im working on an experiment to rack some console modules and im trying to implement a balancing circuit at the output of the unit. At the output I'd like to figure out what (if any) value I should use as a blocking cap before I hit the balancing circuit (THAT 1646).

I've seen folks use 47uf, 100uf and even some things down around 22uf.  Is there any advantage/disadvantage to these values?
And electrolytic vs tantalum (or is that getting into audiophoolery)?
 
The choice depend of bandwidth/phase behavior you want.

The formulae is 1/(2*pi*r*c) where r is input impedance of next stage c capacitance in f.

Sometimes you'll see very low values (sometimes in the 1 or below hertz area): this is not really for bandwidth but phase related as hp filter generate phase modification up to two decade over the fc (eg: with 1hz fc up to 10hz (first decade) there is lot of phase shift and still some minor modifications until 100hz (second decade)-usually acceptable).

About choice of technology: the real downside of electrolitycs is that they dry with age so be ready to change them in 5 to 10 years depending of your application , build and range of product you choose (+105°c are usually better).

About their sound... yes they have a kind of sound but every component have one. Do you need to stay away from them? Just take a look at ANY schematic of say R.Neve (or SSL, or AMS/Neve, or...) products from the 80's till now and see/ear by yourself if it don't make 'good' sound, and how many are in the signal path...

Aside from the fact you have to be prepared to change them in some years... go with them.
 
Thank you for the reply. 
Given the THAT 1646 has a 5k input impedance, we get  1/(2*pi*5000*"X") = what?  Not quite sure what the formula is telling you.

If i plug in 47uf into the formula i get .000000677255077.  not really shure what this means.
 
sr1200 said:
Thank you for the reply. 
Given the THAT 1646 has a 5k input impedance, we get  1/(2*pi*5000*"X") = what?  Not quite sure what the formula is telling you.

If i plug in 47uf into the formula i get .000000677255077.  not really shure what this means.
Perhaps a mistake with scientific notation, your answer seems to be off by several decimal places.

JR
 
sr1200 said:
Thank you for the reply. 
Given the THAT 1646 has a 5k input impedance, we get  1/(2*pi*5000*"X") = what?  Not quite sure what the formula is telling you.

If i plug in 47uf into the formula i get .000000677255077.  not really shure what this means.

It is telling you the frequency at which the response is 3dB down. The answer for 47uF is 0.67Hz

Cheers

Ian
 
Sr1200,
sorry to not have told what the results are for...

As Ian answered this is the -3db point of FC of your hp filter (determine the low end bandwith).

Given you are making an RC filter this is a 1 pole (6db/octave) kind of filter. Try to find a plot of phase/versus frequency to have an idea of the artefact being created.

When you use the formula C is in Farad, 47µf is equal to 0.000047f.  ;)
 
Ahhhhh and thats why my decimal was off too.  Thanks so much everyone.  I read about 22 pages on this last night, and your answers (all of you) made more sense than all of that.  ;D
 
sr1200 said:
Im working on an experiment to rack some console modules and im trying to implement a balancing circuit at the output of the unit. At the output I'd like to figure out what (if any) value I should use as a blocking cap before I hit the balancing circuit (THAT 1646).

I've seen folks use 47uf, 100uf and even some things down around 22uf.  Is there any advantage/disadvantage to these values?

A DC block is arguably needed in front of the 1646 if there is "significant" DC offset coming from previous stages. The real question is, "how much is significant?"  I suppose that single-digit millivolts is probably not significant.

And thinking forward: THAT1646 creates a differential output from its single-ended (w.r.t. ground) input. Since it is unity gain, the input offset won't be multiplied, so at worst you will see a single-digit millivolt offset in the two outputs. And that offset will be completely canceled by the input differential amplifier in the receiving device.
 
My two cents would be to use just basic 85 degree caps of sufficiently high voltage rating (at least 25V). Blocking caps in the usual dual supply op amp circuits don't get tasked heavily and thus don't get hot, and the series resistance of the cheaper caps helps to keep your circuit stable. I've had to actually add resistors in order to make low ESR caps work properly in my console, and sonically standard electrolytics are fine in the signal path as long as the circuit is designed properly (e.g. draining resistors where they are needed).
 
While this may be a little esoteric, I think the concern about cap coupling the 1646 input is that the small added impedance of the capacitor in series with one input leg imbalances the internal front end differential.  Probably only a subtle difference and I have never used one, but people who have advocate DC coupling and maybe even adding a DC servo, if DC performance is that important.

JR

http://www.proaudiodesignforum.com/forum/php/viewtopic.php?f=7&t=123
 
John Roberts,

you are saying that it could be better to have the preceding stage direct coupled or dc servoed or try to make the 1646 dc servoed/direct coupled?
 
living sounds said:
My two cents would be to use just basic 85 degree caps of sufficiently high voltage rating (at least 25V). Blocking caps in the usual dual supply op amp circuits don't get tasked heavily and thus don't get hot, and the series resistance of the cheaper caps helps to keep your circuit stable. I've had to actually add resistors in order to make low ESR caps work properly in my console, and sonically standard electrolytics are fine in the signal path as long as the circuit is designed properly (e.g. draining resistors where they are needed).
I would disagree on a few points. Even when not dissipating energy, capacitors dry out because of the internal temperature of the unit they are in. 105°C will outlast 85°C type. Electrolytic capacitors made today have much longer life expectancy than those made in the 60's or 70's, though.
Unstability issues with low-ESR capacitors cannot possibly happen in a half-decent designed product, except when used in rail decoupling, but not in signal DC-blocking.
 
KrIVIUM2323 said:
John Roberts,

you are saying that it could be better to have the preceding stage direct coupled or dc servoed or try to make the 1646 dc servoed/direct coupled?
Like Andy said it depends on the DC offset coming from your previous stage.  If DC error is small you should be able to get away with DC coupling. 

If large DC offset you can use an op amp buffer, for small DC error (or if you are a perfectionist) you can add a servo.

JR

1606-1646 datat sheet said:
2. Both devices must be driven from a
low-impedance source, preferably directly from
opamp outputs, to maintain the specified performance.
 
KrIVIUM2323 said:
The choice depend of bandwidth/phase behavior you want.

The formulae is 1/(2*pi*r*c) where r is input impedance of next stage c capacitance in f.
Frequency/phase response is just one side of the coin. The sonic deterioration due to electrolytics in the signal path is proportional to the amount of signal developped across them. That's why "Sometimes you'll see very low values (sometimes in the 1 or below hertz area): this is not really for bandwidth". For a given current circulating through the cap, the higher the capacitance, the lower the voltage.


Sometimes you'll see very low values (sometimes in the 1 or below hertz area): this is not really for bandwidth but phase related as hp filter generate phase modification up to two decade over the fc (eg: with 1hz fc up to 10hz (first decade) there is lot of phase shift and still some minor modifications until 100hz (second decade)-usually acceptable). 
I won't start a debate on the audibility of phase-shift of signals that are below aural perception; unless there is something very wrong in the transmission path (strong asymmetry of response, non-linearity of medium under very high pressure, low-frequency FIR filters), the amplitude perception always dominates whatever phase perception there may be. The actual reason for oversizing elcaps is a) high manufacturing tolerances: your 100uF cap can be as low as 70uF or as high as 150
b) indeed elcaps dry with age; increasing the nominal value increases the normal operation life expectancy
c) in systems with global NFB, the cumulated phase-shift due to coupling caps may cause stability issues; tube amp designers know it well
d) ancient "radio" designers noticed that using caps with just the calculated value did not sound right
All in all, this led to the "10x rule" (design for 2Hz) that was prevalent for several decades; now, with the cost going down and capacitance going up, one can go for the 100x rule (design for 0.2Hz).


About choice of technology: the real downside of electrolitycs is that they dry with age so be ready to change them in 5 to 10 years depending of your application , build and range of product you choose (+105°c are usually better). 
Agreed; currently made elcaps are somewhat better, though. I would think properly used modern elcaps would last about 30-40 years. We have some hindsight now, with products manufactured in the 80's, that show no sign of need of recapping.


About their sound... yes they have a kind of sound but every component have one. Do you need to stay away from them? Just take a look at ANY schematic of say R.Neve (or SSL, or AMS/Neve, or...) products from the 80's till now and see/ear by yourself if it don't make 'good' sound, and how many are in the signal path... 
These products were designed in view of the various concerns I expressed earlier, certainly Rupert Neve did care about audibility of elcaps and understood that it depends as much on the implementation as in the technology.
 
abbey road d enfer said:
  I won't start a debate on the audibility of phase-shift of signals that are below aural perception; unless there is something very wrong in the transmission path (strong asymmetry of response, non-linearity of medium under very high pressure, low-frequency FIR filters), the amplitude perception always dominates whatever phase perception there may be.

Interesting statement. What evidence is there for this?

Cheers

Ian
 
JohnRoberts said:
While this may be a little esoteric, I think the concern about cap coupling the 1646 input is that the small added impedance of the capacitor in series with one input leg imbalances the internal front end differential.  Probably only a subtle difference and I have never used one, but people who have advocate DC coupling and maybe even adding a DC servo, if DC performance is that important.

I suppose one could put a same-value/type cap between the 1646's GND pin and ground to balance that impedance.  Or one could use one of the newish electrolytic caps designed for switching power supplies that have milliohm ESR like the Nichicon SS parts.

But between the DC offset reduction offered by the sense caps and the fact that you're talking to a differential receiver, I don't think there is anything to worry about here. Drive the 1646 input from an op-amp and you're done.
 
Well, now I hope i dont have an entirely worse situation on my hands.  The ouput of the device that it's coming from has around 21k impedance on the output that im using a U pad to drop 6db to offset the THAT's +6db.  Guess I'll plug it in this week and see what happens.
 
sr1200 said:
Well, now I hope i dont have an entirely worse situation on my hands.  The ouput of the device that it's coming from has around 21k impedance on the output that im using a U pad to drop 6db to offset the THAT's +6db.  Guess I'll plug it in this week and see what happens.
I suspect that will not provide a low Z source.  :eek:

Perhaps if you used the 1606 instead. It has two separate 5K inputs (+ and -).  Adding external 5k resistors in series with each 5k input would drop your -6dB.

Putting identical caps in series with each resistor would probably not upset the differential balance.

This seems a little too hard to be the easy way...  8)

JR
 
sr1200 said:
Well, now I hope i dont have an entirely worse situation on my hands.  The ouput of the device that it's coming from has around 21k impedance on the output that im using a U pad to drop 6db to offset the THAT's +6db.  Guess I'll plug it in this week and see what happens.
I had that discussion with Gary Hebert at THAT; he sent me a SPICE model for simulation that showed the issues when the source Z is not zero. CMRR is all over the place. However, if the receiving end has good rejection, it may go unnoticed.
 
JohnRoberts said:
Perhaps if you used the 1606 instead. It has two separate 5K inputs (+ and -). 

And I just carefully looked at the data sheet, and my suggestion that one could put a "balancing" cap on the 1646's GND pin and ground won't work.  On the 1606, yes, the IN- signal connects to a 5k resistor which connects to the internal op-amp. On the 1646, the GND pin connects to that resistor, too, but it also connects to the ground node  between the sense inputs. Subtle difference, yes.
 

Latest posts

Back
Top