[wakibaki]

Quote:

You'll get better results if running the buffers in high bias WB mode(with a jumper instead of resistor).
Otherwise,I'm pretty sure that offset will be lower that 1 mV even without the servo(I've got same topology amp,just with BUF634).

The option to run in NB is something I added at the suggestion of Avro_Arrow. Evidently another member felt they had better results that way. Obviously the link can be made with a 0R resistor.
 
sek@ reports 50mV offset from his build of THE WIRE.
 
The purpose of the servo is not only to control any internally generated offsets, but (in conjunction with the relay) to ensure that no offsets in the source are passed on (amplified) to the phones.
 
Thanks for the input though, let us know if anything else occurs to you...
 
w
 
I bookmarked your blog.

 

[Avro_Arrow]

I asked Wakibaki to add a place for a jumper or resistor.
It gives you the option of running NB, WB or something in between.
It just gives the builder/listener an option to run it the way they want.
 
Perhaps it will turn out that WB works best in all situations and we
can get rid of the jumper...remember this is a prototype...
 
Whether it is in Wide Band or Narrow Band, the slew rate does not change
and band width is well above what is needed for audio.
 
Quote:

You'll get better results if running the buffers in high bias WB mode(with a jumper instead of resistor).
Otherwise,I'm pretty sure that offset will be lower that 1 mV even without the servo(I've got same topology amp,just with BUF634).

 

 

[Avro_Arrow]

So I sat on it, revised it a little, but I'm pretty sure this is the design I'm going to use.
Wakibaki's board is highly integrated, I'm doing a more simple, modular design.
 

 
Other than cleaning up the fonts a little, I think it's ready for the board house.
What do you think..?

 

[sek@]

Hi,
 
Quote:

sek@ reports 50mV offset from his build of THE WIRE.

 
Just to clarify, as I'm afraid this has been mistaken for a normal or expected value.
 
Please keep in mind that the BAL-BAL employs an OPA1632, which has an output common-mode adjustment pin referenced to ground using a capacitor. Thus, at turn-on (and turn-off) the conditions haven't settled yet and the output level is approaching it's final value with a certain time constant. The way this happens depends on which supply settles first and wether there is a high level input signal at this time (as this would then make for a transient). It's also worth mentioning that the time constant is rather short, so the condition settles so quickly that it's probable not even considered an offset voltage at all (but just a turn-on thump, if anything).
 
The stable output offset after settling is usually not dominantly a differential-mode voltage (appearing across the voice coil), at least the differential-mode offset is way lower than 50mV. It's a fully differential circuit, after all. 
 
I made another measurement of the SE-SE (incorporating LME49990 and BUF634) just for the sake of clarifying this issue in this very thread. It showed 0.5mV DC output voltage right after turn on and over the course of a couple of minutes, so it's probably safe to say that my SE-SE simply has half a millivolt output offset.
 
Cheers,
Sebastian.
 
PS: I'll make a more detailed measurement of my BAL-BAL as soon as I have the chance.

 

[sek@]

Quote:

Other than cleaning up the fonts a little, I think it's ready for the board house.
What do you think..?

 
I like the improvements you made.
 
I recommend a design rule check (DRC function in Eagle) before sending to a board house. From over here it looks as if your clearance around pads and wires is rather small. This way you might have to rely on the solder mask resist for easy soldering. Some board houses even supply Eagle design rules as a .drc file to be loaded into the project.
 
Particularly the distance/clearance around the SMD pads looks inconveniently small to me.

The electrolytic caps (C13 through C16) look like they can accomodate up to 220uF (assuming 8mm devices and 25V types). If you want larger capacitance you could go for 10mm devices.
 
Finally, seeing that all connectors except the output pad have the same, larger diameter, I recommend switching to a larger pad type at the output, too. More precisely I'd recommend a larger drill diameter for all the connectors.

 

[Avro_Arrow]

Thanks for the review sek@
 
The clearances were better than it shows in the image I posted...however,
I took your advice and increased the clearance on the ground plane to 16 mils.
Pads were all made the same size and drills were also increased.
I also nudged a few of the components around for better clearance and
added the option for SMD caps in the servo. C13 to C16 were going to be Tantalums.
The power supply is co-located and has 1500uF on the output.
 
Here is the new version:
 

 
 

 

[Avro_Arrow]

Here is the power supply board...just a standard 317/337 dual rail.
 
Top:
 

 
Bottom:
 

 

[qusp]

Quote:

The option to run in NB is something I added at the suggestion of Avro_Arrow. Evidently another member felt they had better results that way. Obviously the link can be made with a 0R resistor.
 
sek@ reports 50mV offset from his build of THE WIRE.
 
The purpose of the servo is not only to control any internally generated offsets, but (in conjunction with the relay) to ensure that no offsets in the source are passed on (amplified) to the phones.
 
Thanks for the input though, let us know if anything else occurs to you...
 
w
 
I bookmarked your blog.

there is no 50mv offset except referenced to ground, which goes nowhere near the headphones, as mentioned above its simply a start up behavior and only on the BAL-BAL;  common mode offset is still for all intents and purposes nil and there is no offset to ground after the VCOM pin wakes up
 
I see no reason for WB mode in audio, just makes the layout more critical for no benefit IMO but some people seem to hear an improvement so best to leave the option I guess.
 
no comment on shaky legs there =D
 
havent had much time for diy lately, got catching up to do at work, keen to get the LPUHP done once thats taken care of, massively parallel LME49600 FTW!
 
edited one word above to make sense
 

 

[sek@]

Quote:

there is no 50mv offset except referenced to ground, which goes nowhere near the headphones, as mentioned above its simply a start up behavior and only on the BAL-BAL;  common mode offset is still for all intents and purposes nil

 
I measured one channel of a BAL-BAL (featuring OPA1632 and LME49610) again using an oscilloscope and +/-12V supplies. I took readings between the non-inverting and the inverting outputs (i.e. the normal headphone output) and from both non-inverting output to ground and inverting output to ground, respectively.
 
The (differential-mode) output offset voltage turned out to be exactly +1mV, the (common-mode) offset voltages to ground were -1.2mV and -0.8mV, respectively.
 
 

I see no reason for WB mode in audio, just makes the layout more critical for no benefit IMO but some people seem to hear an improvement so best to leave the option I guess.

 
I experienced instability with the wrong type of output load (cheap coaxial measurement lead, almost clean sinusoidal oscillation at 13 MHz).
 
I wish THE WIRE had a resistor for tuning the BW pin current. But if the head amp I'm putting it into isn't going to be stable under all loads, I might as well just cut the copper trace and convert the amp to low BW mode.
 
 
Quote:

I also nudged a few of the components around for better clearance and
added the option for SMD caps in the servo. C13 to C16 were going to be Tantalums.
The power supply is co-located and has 1500uF on the output.

Just squeezing the SMD caps onto the round, drilled pads is totally cool. Having actual component pads on the top layer in a single sided job might not be, though, at least not to the board house.
 
Cheers,
Sebastian.
 
Edit: typos corrected and changes in quote incorporated.

 

[Avro_Arrow]

I dropped the SMD back to the bottom layer.
At the board house, all the boards start out double sided...If I don't
put anything in the top layer, they just etch it all off. It's easier
on the home etcher if I keep it all single sided though.
 
Here is the revision and a close up of the servo area to better
see the clearances.
 

 

 

[G.Trenchev]

According to the datasheet,WB mode increases small-signal response and PSRR...and if you're after maximum performance...It's just 15mA anyway

Make sure you put tantalum caps close to the buffers.

 

[wakibaki]

Buffers arrived today, the footprints fit fine. I'm expecting to get some 4-pole slide switches tomorrow which will simplify the fast switch-off of the relay.
 
w

 

[Avro_Arrow]

I think I might build a "Lightspeed Attenuator" for use with this project.
 
Anyone have an opinion on the design?
The threads I have read either love it or think
the idea of using LDR for audio is daft.

 

[wakibaki]

I seem to recall that LDR attenuators have quite high distortion and are problematic to calibrate in terms of gain step and channel-channel matching.
 
I thought about building a self-calibrating one with a PIC controller with A/D, but I was put off by the distortion data, which I think was developed by Nelson Pass.
 
Given that part of the rationale for the design is ultralow distortion, I think it'd be a curious pairing.
 
Curious is not an overriding reason not to do it though, I'd be happy if people only thought I was weird 

.
 
If you want some code to run a PGA volume control using a PIC...
 
w
 
 

 

[Avro_Arrow]

Thats what I like about the modular approach...if it sux, I can toss it.
 
Here is the device the design is based on:
http://www.silonex.com/audiohm/index.html
 
This is the exact device:
http://www.silonex.com/datasheets/specs/images/pdf/104539.pdf
 
I'm sure people think I am weird too...