Passive attenuator impedance question


Hi,

I would like to build passive attenuators between my DAC and Amp (to reduce potential degradation in the DAC's digital volume control).

The DAC's output impedance is 100 Ohms and the Amp's input impedance is 100K. Using a 9K/1K resistor network I can achieve the desired 20dB attenuation. If my understanding is right, this attenuator will present itself as a 10K load to the DAC and an output impedance of 1K to the Amp. Both DAC-->Attenuator and Attenuator-->Amp will have a nice 100x impedance ratio. I read in another thread that > 10x is preferable.

Do you think these resistor values (9K/1K) are good, or should I go lower or higher? Obviously I am looking for optimal sonics. I realize the exact values will be system dependent; I'm just looking for some direction.

Thanks,
gmudunuri
Your analysis is well done! And accurate to a very close approximation.

The DAC will see a load impedance of 9K + (1K in parallel with 100K), which is extremely close to 10K.

The amp will see a source impedance (the impedance seen "looking back" from its input) of 1K in parallel with (9K + 100 ohms), or about 0.9K, which in terms of impedance compatibility is slightly better than the 1K you assumed.

The one other consideration is that if the DAC utilizes a coupling capacitor at its output, which is particularly likely if its output stage is tube-based, its output impedance may be significantly greater at deep bass frequencies than the specified 100 ohms. But even if that were the case, the nominal ratio of 100x pretty much assures there won't be an issue.

Good luck. Regards,
-- Al
GoldPoint 47 Step 10 kOhm Attenuator

Above link to a very nice attenuator, I believe you can buy them in a chassis with all the connectors etc as well.

Best of Luck

Peter
I don't think you'll benefit in any way by using a passive volume control. I know that when you lower the volume on your dac, you loose some resolution. I've done quite a bit of testing, along with a few other audiophiles that were curious, as well. Using a very detailed system, we couldn't hear any difference whatsoever between low and high settings. I think you're far more likely to lower SQ than raise it. To be fair, we only used Wadia products for our testing. However, I wouldn't expect different results if we tried other brands.

If this is happening Zd542 the dacs volume control is "bit stripping" in the digital domain, this will not happen with a passive volume control in the analog domain.

Cheers George
"07-06-15: Georgelofi

If this is happening Zd542 the dacs volume control is "bit stripping" in the digital domain, this will not happen with a passive volume control in the analog domain."

I know.

"I know that when you lower the volume on your dac, you loose some resolution. I've done quite a bit of testing, along with a few other audiophiles that were curious, as well. Using a very detailed system, we couldn't hear any difference whatsoever between low and high settings. I think you're far more likely to lower SQ than raise it."
Regarding "bit stripping," I would not extrapolate experience with Wadia models to models from other manufacturers, and perhaps even between different Wadia models.

The manual for the Wadia 121 states as follows:
Does the Wadia Volume Control compromise resolution?

The Wadia 121 Decoding Computer uses the latest generation of Wadia’s proprietary digital volume control. The volume level can be varied in the digital domain by means of mathematical manipulation of the signal, eliminating the distortion and noise that are inevitable with even the best analog volume controls. While conventional thinking indicates that reducing the volume digitally can sacrifice low level resolution, Wadia has created an innovative solution. Wadia’s patented digital filtering algorithm produces a 32-bit output. This high-resolution signal is then used in the computations that in turn reduce the volume level. This new signal is fed directly to the DAC chips. Through this innovative method, the Wadia 121 Decoding Computer maintains high resolution even at the lowest volume control settings.
While the manual for the 321 has no such statement but says:
A Volume Setting of 88% on the Wadia 321 will provide optimum performance when it is connected to a Preamplifier (Integrated Amplifier or Receiver).
Best regards,
-- Al
Mr G- Something that may be of interest to you: (http://www.partsconnexion.com/controls_attenuator_khozmo_series.html)
The 121 and 321, I believe are from Wadia's current offerings. I've never used either one, so I really can't comment on how they sound. But I have had many Wadia products over the years, and still have an 861SE and a 302. The owners manuals say to keep the volume above 67 for best sound quality. They also have dip switches (861's are internal, and the 302's are external) that allow you to alter the overall gain so that you can keep the volume setting as high as possible.

Just to make sure we are talking about the same thing, when I did the test, the idea was to see if we could hear a reduction in SQ at volume settings below the recommended 67. Throughout the day of testing, no one, including myself, could hear any difference whatsoever, regardless of low low we set the volume.

Taking the above into consideration, my advice to the OP is to see if you have a problem that needs to be fixed before you start. If you hear a noticeable difference at lower volumes, then it may make sense to try a passive volume control. But if you can't hear a difference, putting the volume control in the signal path, can't do anything positive.
Just a few thoughts as I'm a passive guy ... your impedance's are fine as Al has confirmed but we don't know anything about the V/out of the source and your V/in of the amp



Because you've eliminated the ... Pre which usually takes the responsibility of driving the IC to the amp ... your source is now the driving everything



I think the IC between the Passive's output and the amp's input is much more critical than the Digital vs Analog volume control debate



The IC from your passive attenuator should be very low in capacitance and as short as possible to keep from rolling off the highs ... Al can expound on this and the V/out V/in through the passive



On the digital vs analog volume control debate ... all volume control/attenuators sound best at full volume as they have little or no resistance in the circuit to degrade the sound when set to FULL



If a digital volume controls throws away bits and compromises the sound ... then I say a analogue volume controls have to many resistors or resistance in the circuit at low volume to attenuate the volume and compromises the sound equally ... 6 of one a half dozen of another



I cheat ... after my source I use a Burson class A no neg feedback buffer feeding the volume controls in my Phase Linear D500 to do the driving



The P/Linear's volume controls are a compromise but I also have a modded P/L D500 which I've bypassed the V/Controls on and will be using a Slagle auto former set up that I'm just finishing up



The Slagle A/F doesn't suffer the bit drop out of the digital V/C or the resistance issues of a typical resistor based passive V/control and it does a better job of shifting less as you turn the V/C up or down



Something not mentioned or considered is the shift in impedance as the V/C is turned up and down and again Al can easily explain this better than I
@Dave- "then I say a analogue volume controls have to many resistors or resistance in the circuit at low volume to attenuate the volume and compromises the sound equally...." The stepped attenuator I noted, as most others, only has two very high quality resistors in the path at any level setting.
07-06-15: Davehrab
The IC from your passive attenuator should be very low in capacitance and as short as possible to keep from rolling off the highs ... Al can expound on this
Thanks, Dave. Good point, which is certainly often a major consideration when a resistance-based passive attenuator is being used. In this case, though, given that the cable is being driven from an impedance of less than 1K (about 900 ohms, as I mentioned earlier), I would not expect the capacitance of that cable to result in audibly significant effects under most circumstances (i.e., unless cable length is particularly long and capacitance per unit length is particularly high).

Also, the OP may wish to consider constructing the attenuator such that it can be connected directly to the input jacks of the amp, with no intervening cable. That is how the Rothwells and other such fixed in-line resistive attenuators are typically used. Although eliminating the cable, or at least keeping the cable short and using one having low capacitance, assumes much greater importance with something like the Rothwells, because the resistor values they employ are much higher than the 9K/1K values the OP will be using. Low values such as 9K/1K are suitable for his particular application because of the low output impedance of his DAC.
Something not mentioned or considered is the shift in impedance as the V/C is turned up and down and again Al can easily explain this better than I
In this case, though, note that the resistive attenuator would be fixed, consisting of just a pair of resistors. The digital volume control in the DAC would continue to be used.
... we don't know anything about the V/out of the source and your V/in of the amp.... Al can expound on ... the V/out V/in through the passive
But note the OP's reference to "the desired 20 db attenuation." Presumably he has determined that to be what is necessary.

Best regards,
-- Al

Wow thats a lot of responses! Thank you!

Just to clarify, I do not wish to use a variable (stepped) attenuator. Currently I am using between 20dB to 35dB of digital attenuation in the DAC. All I want is to introduce fixed passive attenuation of 20dB. This will allow me to use 0dB to 15dB digital attenuation. Theoretically this should reduce potential digital degradation (if is any), but the attenuator could introduce its own.

To be honest, I don't detect loss of SQ in my system even when the DAC attenuates 35dB. Maybe my system or ears are not resolving enough :-). Having said that I just wanted to experiment with a fixed attenuator (built right into an XLR male connector to minimize signal degradation). Should be a fairly inexpensive experiment even with high quality resistors.

@Almarg
I am a little confused as to which circuit to use with balanced cables. Some websites suggest soldering the R2 resistor (1K in my case) between the + and - terminals and a R1/2 resistors (9/2 = 4.5K in my case) in series with the + and - wires. Other online sources suggest that using two R1 -- R2 voltage divider networks. One between + and ground and another between - and ground. Not sure which one to go with. Thoughts?
Gmudunuri, I suspect that either of those approaches would work well, but I'd suggest going with the three resistor approach. Not only because it would use one fewer resistor per channel, but because of some probably very minor but perhaps significant theoretical advantages.

One being that the four resistor approach may cause or contribute to signal current flowing in the ground connection between the two components (i.e., in the shield of the balanced cable), which may to some extent counteract the potential advantages of a balanced interface. Another being that with the four resistor approach if the exact values of the two 1K resistors are not well matched the result might be a slight degradation of common mode noise rejection. In the three resistor configuration the exact value of the 1K resistor is non-critical. You would still want to purchase 4.5K's having relatively tight tolerances, though, perhaps 1%.

There are also differences of a factor of 2 in impedance between the two configurations, both looking forward from the DAC (favoring the four resistor configuration), and looking back from the amp (favoring the three resistor configuration). But I don't think either difference would be significant with your particular components.

Good luck. Regards,
-- Al
I just received IRC 0.1% 1/4 Watt resistors. Went with double the resistence: 2K and 9K. Will try to put together the attenuator over the weekend and report my findings.