[mike1127]

I need to replace the potentiometer (volume control) in my Hifiman EF5 and I'm going to upgrade it and move it into a separate box as there is no room in the main box for anything larger than miniscule.
 
One issue is that I prefer the sound in the high gain setting but I use a fairly efficient headphone (the Audeze LCD2) so I need to keep the volume really turned down. I have about 15 degrees of usable range on it, near the very bottom.
 
I'm looking at getting a TKD pot. Hopefully is a quality item because it costs $75 from Percy... so what I am wondering is, do high-quality stereo pots tend to be well balanced between channels at the low end and do they sound good at the low end?
 
The cheap stock pot in the EF5 has a channel imbalance near the bottom of its range which is a problem because it further limits the usable range for me.
 
Mike

 

[dude_500]

My Alps RK17 has pretty severe imbalance in the lower 10-15 degrees, which is where I typically run my beta-22. I don't know about the particular pot you mention, but from what I've read just about all pots suffer from some low-end balance issues that is often noticable, even pots that cost hundreds or thousands.
 
Stepped attenuators/relay pots are really the way to go in my opinion. I put a 128 step relay pot in my blue hawaii, flawless balance at every setting. It's definitely a noticeable user experience improvement over my b22.

 

[tomb]

http://www.tangentsoft.net/audio/atten.html
 
In particular, refer to the spreadsheet and graphs that Tangent has prepared here:
http://www.tangentsoft.net/audio/misc/pot-curves.zip
 
It has response and channel differences for several pots and attenuators, including the RK27 and the DACT CT2.  One of the mysteries of the industry is 1) the lack of audio taper stereo volume pots, and 2) the fact that pots offered by a company can vary widely in performance from model to model.  The RK27 may measure as good as any pot (Tangent says, "excellent"), but then some of the others ALPS pots as Tangent puts it, are "outright junk."

 

[Face]

Check out Goldpoint. http://goldpt.com/

Sent from my DROID3 using Tapatalk 2

 

[mike1127]

A couple of clarifications:
 
With the cheap pot that's in the stock EF5, I don't need the bottom 15 degrees. It's too quiet there anyway. That's where most of the imbalance is. The next 15 degrees above that is useful (ranging from "about the quietest I would want" to "about the loudest I would want").
 
I don't think a stepped attenuator (at least in off-the-shelf form) is going to work because I need such precision at the low end .. like 2 dB per step.. but my understanding is that most stepped attenuators have larger steps near the bottom. Customizing a stepped attentuator is not realistic because I am not that good with a soldering iron. I am actually paying a guy to do this to make sure it is done well. His fee is very high and I can't afford to pay him to assemble a custom stepped attenuator.
 
Mike

 

[mike1127]

Quote:

http://www.tangentsoft.net/audio/atten.html
 
In particular, refer to the spreadsheet and graphs that Tangent has prepared here:
http://www.tangentsoft.net/audio/misc/pot-curves.zip
 
It has response and channel differences for several pots and attenuators, including the RK27 and the DACT CT2.  One of the mysteries of the industry is 1) the lack of audio taper stereo volume pots, and 2) the fact that pots offered by a company can vary widely in performance from model to model.  The RK27 may measure as good as any pot (Tangent says, "excellent"), but then some of the others ALPS pots as Tangent puts it, are "outright junk."

I looked at the graphs (spreadsheet) but what I can't figure out is what the "approx" column is. Do you know? It's the column plotted as a purple line in all the plots.
Mike

 

[tomb]

Quote:

I looked at the graphs (spreadsheet) but what I can't figure out is what the "approx" column is. Do you know? It's the column plotted as a purple line in all the plots.
Mike

You'll have to get Tangent to describe it in more detail, but essentially, the purple, "approx" graph is a plot/values of a curve-fitted equation for the measured data.  If the relationship between the knob travel and pot response was linear, then a simple linear-regression line-fit would suffice to "linearize" the scatter of the data.  The response of a volume pot is much more complex, though, and follows a logarithmic relationship.  So, a curve-fit program is used.  He mentions several programs in the text.
 
One way to look at the purple curves/lines or "approx" column is that these are the values of the pot's performance if it perfectly followed the theoretical response for its design.  You're always going to get data "scatter" when you measure a system in the real world.  (It's one reason pure-objectivists fall short of predicting real-world performance.)  However, if you can "curve-fit" the data to a reasonably approximate curve, you should be able to predict performance over a wide range of conditions - without specifically measuring the data at every instance.  For instance for the RK27, the approximate curve-fit matches very well with the plotted data.  The actual equation that resulted from the curve-fit can give you an excellent approximation of the pot response at any point that was not measured -   5 degrees, 48 degrees, 92.5 degrees, etc., etc.
 
If you've ever computer-programmed, think of the difference between a look-up table to determine values and then cumbersome interpolation routines to estimate in-between values, vs. a single, self-contained equation. 
 
Similarly, you can use a curve-fit to demonstrate deviation of the measured data from "ideal," which is what I think Tangent is trying to show here.

 

[compuryan]

Personally, I'd go for an RK27. The low end imbalance I experienced with it was gone well before the time I reached normal listening levels. It is a high quality product. The more expensive pots/stepped attenuators don't offer a significant improvement in sound quality IMO. I've heard several of them. The extra money you'd spend would be much better invested elsewhere in your system.

 

[mike1127]

Quote:

Originally Posted by tomb /img/forum/go_quote.gif
 
The response of a volume pot is much more complex, though, and follows a logarithmic relationship.

I think there's an additional complexity, which is that a log pot can't stay log across its whole range -- it's okay at the high end, but it would never get to zero at the low end
 
EDIT (what I mean is -- let's say that every 15 degrees to the left drops the output 2dB. After rotating it 360 degrees you would have dropped the output by 48 dB. That's not zero! Suppose it made any physical sense to keep rotating it to the left, and it continued to follow the rule that 15 degrees = 2dB. You could rotate it another 360 degree but not be at zero. You would be at -96 dB. A small number, yes. But zero? Not even close! Rotate it a million times to the left. Not zero! you get my point.  A sort of version of Xeno's Paradox, although not the same paradox.)
 
. So the lower 15 degrees or so show a higher rate of attenuation. By ear, this is true of the cheap ALPS pot in my Hifiman EF5 (which is now noisy and has a pretty severe channel imbalance in the bottom 15 degrees). The useful range, for my amp, is from about 15 degrees to 30 degrees. Fortunately most of the channel imbalance occurs below 15 degrees.
 
I didn't completely read the first link you gave so I didn't realize he explains what the "approx" field is. The data is so smooth that I probably wouldn't have been tempted to curve-fit, myself. (I just realized one interesting thing this lets him do, which is show the fitted curve on the linear plot.) One confusing thing is that he modeled (curve-fit) the extra-steep drop near zero of the ALPS RK27 but not the others. Maybe that's because he was using the ALPS datasheet to get the specs on that drop?

 

[IPodPJ]

You should just go with a DACT CT-2 if you want accurate channel balance. If you must have a pot nothing will likely top the Alps RK50 but they aren't cheap.

 

[mike1127]

Folks, a stock stepped attenuator isn't going to work. I need most of my adjustability at the bottom end. Maybe a custom stepped attenuator, but I don't want to get into that.

 

[mike1127]

Quote:

You should just go with a DACT CT-2 if you want accurate channel balance. If you must have a pot nothing will likely top the Alps RK50 but they aren't cheap.

Yeah RK50 is off the table. It's not THAT big a project.
 
RKD makes a pot that Percy sells for $75. (Parts Conexion has the same thing for $99.) Any experience with that one?

 

[Face]

Quote:

Folks, a stock stepped attenuator isn't going to work. I need most of my adjustability at the bottom end. Maybe a custom stepped attenuator, but I don't want to get into that.

http://goldpt.com/info.html

 

[tomb]

Quote:

I think there's an additional complexity, which is that a log pot can't stay log across its whole range -- it's okay at the high end, but it would never get to zero at the low end
 
EDIT (what I mean is -- let's say that every 15 degrees to the left drops the output 2dB. After rotating it 360 degrees you would have dropped the output by 48 dB. That's not zero! Suppose it made any physical sense to keep rotating it to the left, and it continued to follow the rule that 15 degrees = 2dB. You could rotate it another 360 degree but not be at zero. You would be at -96 dB. A small number, yes. But zero? Not even close! Rotate it a million times to the left. Not zero! you get my point.  A sort of version of Xeno's Paradox, although not the same paradox.)

I'm not into scientific paradoxes - especially ones where you're not supposed to reach zero through infiinity.  The Engineer knows better - and can always get close enough to zero to make it work.

 
You've had several helpful comments, especially about the RK27.  It's about as good as you're going to get on the low end with a stereo volume pot.  Yet, that doesn't seem to be satisfactory.
 
There's another option and something we did quite successfully with the Starving Student.  You can use some input resistors in the signal lines (do not cross connect them to ground) to "push" the pot into its most optimum range for the volume level you want.  A reasonable starting point is to use 1X the impedance of the pot for the resistor values, then try multiples after that in resistor values - 2X, 3X, and so forth.  This doesn't alter the behavior of the pot in its volume attentuation characteristics independent of the load.  The resistors simply pre-attenuate to a more favorable position in the pot's travel for the volume level you want.

 

[mike1127]

Oh, I see several ideas here. The impression I get is that most quality pots will be okay on the low end. I can further improve the situation by using resistors in series as both Tom and Face (through his link) pointed out. I don't want to cut the gain too much, however, because I would like the amp to be useful for inefficient headphones (it was primarily designed by Hifiman for their own inefficient headphones).
 
I would like to be able to use a stepped attenuator, but they have very large steps in the bottom 20 dB or so. That means I need to cut the gain by 20 dB with resistors in order to bring the useful range into the range where there are 2 dB steps in the stepped attenuator. That would mean the amp is pretty much no longer useful with inefficient headphones. So I have to pick my poison.
 
Another option is to set up a switch which switches the attenuation resistors into the circuit or takes them out. I prefer to keep it simpler than that, however.
 
I'm thinking a good quality pot is my best option, with maybe 10 dB of cut gain.
 
Mike