Skip to the next post for my findings if you don't feel like reading my long story of the day giving background on why I have decided to switch from shunt attenuator to ladder.
.
I have been using direct stepped attenuators which plug onto the amps inputs as volume controls for 10 years and have been recommending a 4k fixed series/ switched shunt style for amps with input impedances up to 100k. Over 100k amps can consider a 20k attenuator to ease the load on the source. The 4k attenuators have the minimum interaction with any cabling and amp circuitry that follows them and are still well tolerated by most active outputs of the source equipment that will be feeding them barring the occasional high output impedance tube circuit. My digital source for the last 6 years has been a modified DEQ2496 running an AKM4395 dac chip direct out with no filters or active buffers. I have tried many different active opamp outputs but have found that just running the dac direct from the analog ouputs through a good cap or transformer to block the dc provides the most transparent sound as is discussed in many different threads about the AKM and also the CS dac chips.
.
http://www.diyaudio.com/forums/digital-line-level/137976-experience-diy-dac.html#post1735158
.
http://www.diyaudio.com/forums/digi...ringer-dcx2496-digital-x-over.html#post185399
.
http://www.diyaudio.com/forums/digi...6-best-solution-output-stage.html#post1828425
.
http://www.diyaudio.com/forums/digi...sonic-noise-spectrum-graphic.html#post1730822
.
It wasn't until I started playing with the transformers that I noticed that the cap out system still sounded better at louder attenuator settings but the trannies sounded better at quieter settings. At -2db the attenuators provide a 21k load but this drops rapidly towards 4k at the middle and quieter settings. Was the single leg cap output stuggling with the now 4k impedance whereas the transformers, which use both output legs, had more power to drive it? This, and the observation that the lowish direct output level allows for a gain matched system using only small amounts of digital attenuation in the 0 to -12db range to trim full on reference listening levels with no analog volume control, which sounds even better due to the elimination of the series resistor which is always in the signal with the shunt type, has prompted me to consider changing to a ladder type attenuator. I have resisted a ladder until now due to the doubling of the cost per step but now see it as the ideal way to provide the optimum and constant load for the new crop of direct out dacs and still offer a straight through setting for digital volume control. Time for a redesign. But what is the ideal trade off between series (output) impedance and load? Skip to the next post for my procedure and findings.
.
Here are a couple recent posts that kick started my thought process on this.
.
Yes, I run a Behringer DEQ or DCX2496 direct from the dac, out through a coupling cap as does DrVega. Panomaniac prefers transformers but it gets expensive with 6 in a DCX. My attenuators are fixed series 4k with a switched shunt. This allows an attenuator to be built with only 1 switch deck. A series attenuator uses 1 deck but has 12-24 solder joints and resistors in the path compared to only 2 for ladder or shunt . Ladder types seem to have technical advantages until you think it through. A 10k ladder is an easier load for the source at quieter levels although I have been reading that measured performance of the dac chip can actually improve when loaded with 3k. (edit: now found to be untrue) The sonic challenge for a passive volume control is the high output impedance which is, in effect a restriction in current that would be useful to drive the signal through the R, C, and L of the following cables and input stages of the amp. E.E. type theorists that don't believe in listening insist that these values are so low in any reasonable cable length so as to be ignored, but thousands of reports of listening tests by people drawn to the cheap price and purity of passives show mixed results when replacing an active volume control due to the increased interactions with a passive. The two keys to getting superior sonics along with the cheaper costs of passive are in choosing the lowest possible output impedance that your source will still drive, and minimizing the the cable length that follows the attenuator. Now, back to ladder vs. shunt. A 10k ladder (or series, or pot) has 10k input at all settings but the output impedance will change from 0 at full volume to 6.6k at -12db to 8k at -18db and continue to rise at quieter levels. The shunt style has a 4k output at all levels and the input is 20k at -2db dropping through 4.6k by -18db. But the 4k actually has a lower output impedance than the 10k series (pot) or ladder from -5db and quieter where most controls end up getting used, staying at 4k while the others are rising.
.
2 feet cables seem short but in my experience, you will still hear them robbing you slightly. If aesthetics and convenience are secondary to getting the best sonics, I would recommend mounting the volume controls to the boards in the empty spot between the connectors and the heat sink. Or at least hardwiring from a housing front panel to the amp to keep the wires to 6 inches.
.
I can run my AK4395 and CS43122 dacs with transformer or cap coupled direct out and change back and forth just by moving the interconnects. This has allowed me to discover that with both legs of the dac contributing through the trannies in a balanced to single ended conversion, they have enough power to drive my 4k direct attenuators much better. One leg out through the caps sounds slightly closed in by comparison into the 4k load the attenuators provide at quieter settings. At louder settings, say -2db, the attenuators input impedance is 20.4k. At this load the beauty of the caps can sound better and are much cheaper than the transformers. The AKM chip happens to have the perfect level to connect straight to the Sure 2X100 with no volume control. This arrangement reveals that the Dayton foil caps do sound better than the best transformers but only if the load is eased to 10-20k at which point, digital volume control in the range of 0 to -12db sounds better than having any value of added series resistance of any style of analog volume control which would favor the availability of a straight through setting with a ladder. Which is not to say that you can use a control with 20k in series with the signal in order to ease the load. I also have 20k series/ switched shunt direct attenuators. I changed the input terminating resistors and gain setting resistors R14,34,11,32,30,16 in an amp to 150k in order to compare 4k versus 20k attenuators and still maintain a suitable 7:1 ratio. I found that the 4k series resistor allowed more of the fine detail through. These are direct attenuators that plug right onto the amp's input connectors. The difference would be much greater if there were any length of interconnect involved, favoring the 4k value further. So where does this leave us? Since the direct out dacs are much cheaper to run with one leg through caps and have a lowish level that needs to be run nearly wide open through a terminating load no lower than 10k, I have decided start work on a new ladder type attenuator. Ladders cost twice as much per step but do have a sonic advantage over shunt types, of reduced series (output) impedance in the wide open ranges from 0 to -12db that I now need. I have never tried the current crop of series attenuators which can duplicate a ladder with one deck, due to the massive number of solder joints they introduce into the circuit but I guess I should also explore these as I have learned over the years, and more so with everything new I experiment with, you never know what something will sound like based on theory, until you listen to it. It is possible they could sound ok at wide open settings which would place most of the solder on the ground side. So I need to find out what the minimum value that sounds good with the direct out AKM is ( the CS chips have slightly more level and power so should be less fussy) and pick between ladder and series. Whatever I build next will use the excellent and affordable Susumu surface mount resistors as they sound better than any through hole resistors in this critical application.
.
Single ended ladder attenuators costs twice as much per step than shunt due to needing 2 switch decks versus 1 which adds up when considering 6-8 channels in an active cross system. The difference in input and output impedence is only signifcantly better for the ladder from 0 to -12db or so. Any levels quieter than this, which is where most controls end up being used at -20 to -30db, and the impedances are nearly the same for any given step so the shunt can sound just as good at half the price. The price/ performance of the shunt style really shines when using a balanced signal where the shunt attenuator can still be realized with 1 deck while the balanced ladder attenuator needs 3 to 4. It is only now, with the lowish gain Tripath amps and direct out dacs, that I have the need for the 0 to -12db range which favors the ladder or series attenuator.
.
I have been using direct stepped attenuators which plug onto the amps inputs as volume controls for 10 years and have been recommending a 4k fixed series/ switched shunt style for amps with input impedances up to 100k. Over 100k amps can consider a 20k attenuator to ease the load on the source. The 4k attenuators have the minimum interaction with any cabling and amp circuitry that follows them and are still well tolerated by most active outputs of the source equipment that will be feeding them barring the occasional high output impedance tube circuit. My digital source for the last 6 years has been a modified DEQ2496 running an AKM4395 dac chip direct out with no filters or active buffers. I have tried many different active opamp outputs but have found that just running the dac direct from the analog ouputs through a good cap or transformer to block the dc provides the most transparent sound as is discussed in many different threads about the AKM and also the CS dac chips.
.
http://www.diyaudio.com/forums/digital-line-level/137976-experience-diy-dac.html#post1735158
.
http://www.diyaudio.com/forums/digi...ringer-dcx2496-digital-x-over.html#post185399
.
http://www.diyaudio.com/forums/digi...6-best-solution-output-stage.html#post1828425
.
http://www.diyaudio.com/forums/digi...sonic-noise-spectrum-graphic.html#post1730822
.
It wasn't until I started playing with the transformers that I noticed that the cap out system still sounded better at louder attenuator settings but the trannies sounded better at quieter settings. At -2db the attenuators provide a 21k load but this drops rapidly towards 4k at the middle and quieter settings. Was the single leg cap output stuggling with the now 4k impedance whereas the transformers, which use both output legs, had more power to drive it? This, and the observation that the lowish direct output level allows for a gain matched system using only small amounts of digital attenuation in the 0 to -12db range to trim full on reference listening levels with no analog volume control, which sounds even better due to the elimination of the series resistor which is always in the signal with the shunt type, has prompted me to consider changing to a ladder type attenuator. I have resisted a ladder until now due to the doubling of the cost per step but now see it as the ideal way to provide the optimum and constant load for the new crop of direct out dacs and still offer a straight through setting for digital volume control. Time for a redesign. But what is the ideal trade off between series (output) impedance and load? Skip to the next post for my procedure and findings.
.
Here are a couple recent posts that kick started my thought process on this.
.
Yes, I run a Behringer DEQ or DCX2496 direct from the dac, out through a coupling cap as does DrVega. Panomaniac prefers transformers but it gets expensive with 6 in a DCX. My attenuators are fixed series 4k with a switched shunt. This allows an attenuator to be built with only 1 switch deck. A series attenuator uses 1 deck but has 12-24 solder joints and resistors in the path compared to only 2 for ladder or shunt . Ladder types seem to have technical advantages until you think it through. A 10k ladder is an easier load for the source at quieter levels although I have been reading that measured performance of the dac chip can actually improve when loaded with 3k. (edit: now found to be untrue) The sonic challenge for a passive volume control is the high output impedance which is, in effect a restriction in current that would be useful to drive the signal through the R, C, and L of the following cables and input stages of the amp. E.E. type theorists that don't believe in listening insist that these values are so low in any reasonable cable length so as to be ignored, but thousands of reports of listening tests by people drawn to the cheap price and purity of passives show mixed results when replacing an active volume control due to the increased interactions with a passive. The two keys to getting superior sonics along with the cheaper costs of passive are in choosing the lowest possible output impedance that your source will still drive, and minimizing the the cable length that follows the attenuator. Now, back to ladder vs. shunt. A 10k ladder (or series, or pot) has 10k input at all settings but the output impedance will change from 0 at full volume to 6.6k at -12db to 8k at -18db and continue to rise at quieter levels. The shunt style has a 4k output at all levels and the input is 20k at -2db dropping through 4.6k by -18db. But the 4k actually has a lower output impedance than the 10k series (pot) or ladder from -5db and quieter where most controls end up getting used, staying at 4k while the others are rising.
.
2 feet cables seem short but in my experience, you will still hear them robbing you slightly. If aesthetics and convenience are secondary to getting the best sonics, I would recommend mounting the volume controls to the boards in the empty spot between the connectors and the heat sink. Or at least hardwiring from a housing front panel to the amp to keep the wires to 6 inches.
.
I can run my AK4395 and CS43122 dacs with transformer or cap coupled direct out and change back and forth just by moving the interconnects. This has allowed me to discover that with both legs of the dac contributing through the trannies in a balanced to single ended conversion, they have enough power to drive my 4k direct attenuators much better. One leg out through the caps sounds slightly closed in by comparison into the 4k load the attenuators provide at quieter settings. At louder settings, say -2db, the attenuators input impedance is 20.4k. At this load the beauty of the caps can sound better and are much cheaper than the transformers. The AKM chip happens to have the perfect level to connect straight to the Sure 2X100 with no volume control. This arrangement reveals that the Dayton foil caps do sound better than the best transformers but only if the load is eased to 10-20k at which point, digital volume control in the range of 0 to -12db sounds better than having any value of added series resistance of any style of analog volume control which would favor the availability of a straight through setting with a ladder. Which is not to say that you can use a control with 20k in series with the signal in order to ease the load. I also have 20k series/ switched shunt direct attenuators. I changed the input terminating resistors and gain setting resistors R14,34,11,32,30,16 in an amp to 150k in order to compare 4k versus 20k attenuators and still maintain a suitable 7:1 ratio. I found that the 4k series resistor allowed more of the fine detail through. These are direct attenuators that plug right onto the amp's input connectors. The difference would be much greater if there were any length of interconnect involved, favoring the 4k value further. So where does this leave us? Since the direct out dacs are much cheaper to run with one leg through caps and have a lowish level that needs to be run nearly wide open through a terminating load no lower than 10k, I have decided start work on a new ladder type attenuator. Ladders cost twice as much per step but do have a sonic advantage over shunt types, of reduced series (output) impedance in the wide open ranges from 0 to -12db that I now need. I have never tried the current crop of series attenuators which can duplicate a ladder with one deck, due to the massive number of solder joints they introduce into the circuit but I guess I should also explore these as I have learned over the years, and more so with everything new I experiment with, you never know what something will sound like based on theory, until you listen to it. It is possible they could sound ok at wide open settings which would place most of the solder on the ground side. So I need to find out what the minimum value that sounds good with the direct out AKM is ( the CS chips have slightly more level and power so should be less fussy) and pick between ladder and series. Whatever I build next will use the excellent and affordable Susumu surface mount resistors as they sound better than any through hole resistors in this critical application.
.
Single ended ladder attenuators costs twice as much per step than shunt due to needing 2 switch decks versus 1 which adds up when considering 6-8 channels in an active cross system. The difference in input and output impedence is only signifcantly better for the ladder from 0 to -12db or so. Any levels quieter than this, which is where most controls end up being used at -20 to -30db, and the impedances are nearly the same for any given step so the shunt can sound just as good at half the price. The price/ performance of the shunt style really shines when using a balanced signal where the shunt attenuator can still be realized with 1 deck while the balanced ladder attenuator needs 3 to 4. It is only now, with the lowish gain Tripath amps and direct out dacs, that I have the need for the 0 to -12db range which favors the ladder or series attenuator.
Which impedance for direct out dacs?
In order to find the best sounding balance of load versus output impedance, I built a pair of ladder direct attenuators where each step had the same -6db attenuation, but different input impedances from 5k to 30k in 6 steps. It turned out to be a really fun project as I was able to get everything built without ordering any new parts. I originally would have liked to listen at -2db but a -6db level made finding the resistors easy as this is half and half, meaning that both legs of the L pad use the same resitance value. No formulas to bother with, all I had to do was grab a pair of resistors at half the desired input impedance for each setting of 5, 10, 15, 20, 25, and 30k. I was able to cut a couple traces on the boards I have been using to change them from shunt to ladder. The cheap but excellent Lorlin switches came apart with no fuss to rob an extra wiper, creating a two pole switch of sorts. Just what I needed. I was so stoked when that came together easily without having to order new parts. I started listening with my modified Sure 2X100 amp set up with input gain setting resistors of 150k:150k (virtual 150k? load) in order to minimize the interaction here (they sound better than 20k: 20k for some reason too, even with the terminating resistor preceding them), and a 20k input terminating resistor. I broke out some 8 foot speaker cables so the amp with the attenators on it could sit right in front of me on the coffee table. The attenuators allowed me to switch on the fly with out moving from the listening position which made everything go pretty quickly. I let them break in for about 1/2 hour on each setting and then got right to listening. The difference was a little less than I expected but still worth the trouble in order to dial in the values of a new attenuator design. The AKM chip sounded best between 15k and 20k even into the 20k amp. This is the beauty of not having any cable following your passive volume control. Minimal interactions. I find that the most revealing aspect of music that can really sort out miniscule differences is in the decay of the reverb and recorded ambience and this was best at 15k. There was also an improvement in the body and weight of the images that interestingly faded toward the same type of increasing opacity toward both extremes of the attenuator. The CS43122 chip was also best at around 15k and maybe shading more toward the 10k side than 20k. I was also surprised that the transformers liked the 15k value the same as the caps.
Next I exchanged the 20k amp input termination that I had installed as it sounded better than 100k when running without a volume control. The chips do enjoy some load. I went back up to 100k resistors to further eliminate the interaction with the varying output impedance of the attenuator. This sounds much better! Just what I have been missing. Let the volume control be the entire amp input termination. Surprisingly again, both chips still sounded best into 15k even though the higher ouput impedance of the 25 and 30k settings was now more of a non issue with the increased amp input impedance. So 15k ladder it is. 10k may be a little better at quieter, -20 to -30db levels if that is where your system ends up but the Tripath amps I now find to sound the best can run with a wide range of gains as long as you also scale the supply voltage so I should always be able set my gain stucture to play a little too loud at -0db.
In order to find the best sounding balance of load versus output impedance, I built a pair of ladder direct attenuators where each step had the same -6db attenuation, but different input impedances from 5k to 30k in 6 steps. It turned out to be a really fun project as I was able to get everything built without ordering any new parts. I originally would have liked to listen at -2db but a -6db level made finding the resistors easy as this is half and half, meaning that both legs of the L pad use the same resitance value. No formulas to bother with, all I had to do was grab a pair of resistors at half the desired input impedance for each setting of 5, 10, 15, 20, 25, and 30k. I was able to cut a couple traces on the boards I have been using to change them from shunt to ladder. The cheap but excellent Lorlin switches came apart with no fuss to rob an extra wiper, creating a two pole switch of sorts. Just what I needed. I was so stoked when that came together easily without having to order new parts. I started listening with my modified Sure 2X100 amp set up with input gain setting resistors of 150k:150k (virtual 150k? load) in order to minimize the interaction here (they sound better than 20k: 20k for some reason too, even with the terminating resistor preceding them), and a 20k input terminating resistor. I broke out some 8 foot speaker cables so the amp with the attenators on it could sit right in front of me on the coffee table. The attenuators allowed me to switch on the fly with out moving from the listening position which made everything go pretty quickly. I let them break in for about 1/2 hour on each setting and then got right to listening. The difference was a little less than I expected but still worth the trouble in order to dial in the values of a new attenuator design. The AKM chip sounded best between 15k and 20k even into the 20k amp. This is the beauty of not having any cable following your passive volume control. Minimal interactions. I find that the most revealing aspect of music that can really sort out miniscule differences is in the decay of the reverb and recorded ambience and this was best at 15k. There was also an improvement in the body and weight of the images that interestingly faded toward the same type of increasing opacity toward both extremes of the attenuator. The CS43122 chip was also best at around 15k and maybe shading more toward the 10k side than 20k. I was also surprised that the transformers liked the 15k value the same as the caps.
Next I exchanged the 20k amp input termination that I had installed as it sounded better than 100k when running without a volume control. The chips do enjoy some load. I went back up to 100k resistors to further eliminate the interaction with the varying output impedance of the attenuator. This sounds much better! Just what I have been missing. Let the volume control be the entire amp input termination. Surprisingly again, both chips still sounded best into 15k even though the higher ouput impedance of the 25 and 30k settings was now more of a non issue with the increased amp input impedance. So 15k ladder it is. 10k may be a little better at quieter, -20 to -30db levels if that is where your system ends up but the Tripath amps I now find to sound the best can run with a wide range of gains as long as you also scale the supply voltage so I should always be able set my gain stucture to play a little too loud at -0db.
Attachments
- Status
- Not open for further replies.