DIYHiFi.org

(I've not read the whole thread, so apologies if this has been stated before...)

Another option to consider is a constant-impedance attenuator instead of a potentiometer. Perhaps using one of the Shallco switches (http://www.shallco.com/frameset.html) or a pre-configured 46 step unit from Marchandelec (http://www.marchandelec.com/att.html).

A major advantage of using a constant-impedance attenuator would be that the impedance and hence the dc offset would be stable, so once set to zero the offset would not change and hence the coupling capacitor could be eliminated. In reality the offset may drift a little with temperature, but it is likely to be a very small.

And when replacing the potentiometer with an attenuator the other instant benefit would be the better channel-to-channel matching confered by the use of close tolerance resistors in the attenuator.

Not cheap though.

John, I have always recommended a coupling cap on the input of the power amp, or on the output of the preamp.
Not both.
The truth is, it is possible to null DC at a volume pot position (picking a normal volume setting), but DC will change according to the volume pot setting.
As the goal was not to use an input buffer of any kind or an input coupling cap between the pot and the AD815, the options are:

1 - Output coupling cap on the pre or input coupling cap on the power amp.
2 - A (big) cap from the gain resistor to ground.
3 - A servo.

I have tried a servo with the AD815, on the bench, but it takes an uncomfortable amount of time for the output DC offset to null, at every rotation of the volume pot.
All in all, the safest option is 1.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

You mean an L-Pad attenuator?

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

That's curious, Carlos... what was the TC of the servo?

I have considered another option for your challenge here... if one were to put the volume control after the 815, then the variable component of the DC offset problem is mitigated. With that topology either your manual 'null circuit' DC injector or an active servo would be effective over the entire volume range. The penalty is [of course there's one] the variable series resistance (a 2k-10k pot comes to mind, but with the current capability of the 815, maybe a few hundred ohms w/b better) between the pre and pwr amp. Considering interconnect cable capacitance, maybe the additional series resistance is not such a bad thing.

Cheers

A simple complementary JFET buffer as in this paper (page5)
http://home.ca.inter.net/~lloyd.maclean ... ECHINF.pdf
will make all that redundant. Only four parts!
I bet it sounds better too.

I know, and I've thought about it. Since the beginning.
The AD815 has lower DC offset if it's input impedance is very low (as the feedback and gain resistors are very low).
So, removing the volume pot from it's inputs, the DC null circuit and all that, to have a low output DC offset the input buffer would have to drive a very low impedance (a low value resistor to ground on the NI input of the AD815).



You'd better not bet.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

Igo51,

A servo does not work well when it is expected that DC offset can change abruptly.
Optimizing the DC servo so that it has the minimal sonic impact (working at a very low frequency) only does things worse in terms of effectiveness.

High output impedance is something that I don't want in my preamp, I would always avoid such 'solution'.
Unless it was with a 100 Ohm dual log pot. Which the AD815 would drive without problems. Heh...
I have a 5 meter shielded interconnect between the pre and the power amp. The power amp is right in the middle of the speakers, with short speaker cables. This is how my system works and sounds best, because I can't have all the system between the speakers.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

HHhhhmmmm, ~50R max Zo (R6 eliminated) seems reasonable enough. Zero net gain parts count. Easy to add to any of the boards that have been done. Allows for user choice of direct or cap coupled output. No active devices - save for the twisty-turny part Boutique part compatible Too easy?

Cheers

I would say too easy indeed. Get me some decent, readily available 100R dual gang log pots, I'll buy some.
You can make an attenuator, though.
Btw this would not dispense the DC null circuit, as you should have a 'normal' input impedance on the pre (far from ideally matched impedances on both inputs of the chip).

Read this post:
viewtopic.php?p=28614#p28614

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

Hi Carlos,
you're speakers must be very close together.
Or are these headphones?
When I refer to short speaker cables I intend 600mm to 1000mm long from amp terminal to driver terminal.
Not quite in the territory of plate amps attached to the back of active speakers, but close enough to almost remove the effect cables have on the amplifier. i.e. low capacitance AND low inductance.

_________________
regards
Andrew T.

Well, let's put this into context...
From the amp to each speaker goes a 2 meter cable. The speakers are separated around 3m from each other.
I say 'short' speaker cables because from my hi-fi rack to the left speaker I would need ~5.5m speaker cable, to the right speaker ~2.5m. I had this setup many years ago, with an integrated amp. It's far from ideal.
By having the power amp central to the speakers I can have equal length of cable, and shorter.
Indeed, something I may do some day, with two amps, one near each speaker, I could have less than 1 meter speaker cables. As there is no dedicated place where to put them there, It's a question of finding a way that the wife doesn't see the amps and doesn't complain.
My main system is in the living room, so everybody has to be happy, not just me.

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

If your definition of 'some' is in the 1,2000 piece range, then have a look at the ALPS RK27 selection online
So, is the new challenge to source this part?

Absolutely... even with the asymmetric input bias currents of the 815. But, now the Zin configuration is more stable and the DC null achieved with the injected bias voltage should not vary.

Cheers

1200??!
I'll take a look at them.



Yes!

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

I have downloaded the datasheet of the Alps RK27 and I can't see anything lower than 5K...

_________________
Carlos Filipe

"Any intelligent fool can make things bigger, more complex, and more violent. It takes a touch of genius - and a lot of courage - to move in the opposite direction." Albert Einstein

I was thinking of a constant-impedance balanced bridged-H. Unfortunately I don't have a link to post to illustrate it. The balanced attenuator requires four set resistors and three stepped resistances (which are switched at the same time), and therefore it can not be made from a pot. The advantage is that the input and output impedance of the attenuator does not change from step to step. So from zero attenuation to maximum attenuation the input and output impedance of the attenuator is constant.

Actually, if you need to provide an input path to ground you would need the double T version as opposed to the H version, and that would mean yet another stepped resistance to add in to the design, and it gets complex and expensive.

They are popular with the RF brigade, scientific community, and pro audio.