[AndrewFischer]

Quote:

Originally Posted by Syzygies I've only seen single throw DIP switches; one would lose the top row of resistors in the circuit.

Greyhill 76 and 78 series DIP switches come come in SPDT. They also come DPDT for stereo

http://embrace.grayhill.com/embrace/IMAGES/PDF/B-13.pdf
http://embrace.grayhill.com/embrace/IMAGES/PDF/B-14.pdf

Switch is rated for 20,000 cycles and derates to 10,000 cycles under load.
I think this is typical for a dip switch. They aren't designed to be switched all the time. Might need to put the switches in sockes.

 

[Syzygies]

It's probably a good thing I didn't give values for a practical design like 25K, forcing any of us to check the values several times before building.

To hopefully belabor the obvious, so people can check me, my method is this: Fix R0 to be the desired input impedance. Pick R1 so the pair R0, R1 gives 1 attenuation step, here 2.5 dB. Pick R2 so the circuit here to the right has the same impedance R0 with this switch up or down.

Move left one stage. Pick R3 so the pair R0, R3 gives 2 attenuation steps, here 5 dB. Pick R4 so the circuit here to the right has the same impedance R0 with this switch up or down.

This recursive step requires some thought: Since the circuit to the right has impedance R0 in all switch positons, we can simplify the current stage so it looks like the rightmost stage. The output voltage is then the input voltage to the next stage, so the attenuations add.

The remaining stages go the same: Move left one stage. Pick R5 so the pair R0, R5 gives 4 attenuation steps, here 10 dB. Pick R6 so the circuit here to the right has the same impedance R0 with this switch up or down. Move left one stage. Pick R7 so the pair R0, R7 gives 8 attenuation steps, here 20 dB. Pick R8 so the circuit here to the right has the same impedance R0 with this switch up or down.

Quote:

Originally Posted by rickcr42 .I can see the potential for the accidental full volume out scenario

Yup, this did cross my mind, too. The nicest slide switches might be better suited for this than toggles.

On the other hand, in my personal experience with commercial products I've seen digital controls go to to full on, far more often than mechanical controls. (I have tape to this day over the infrared eye on my Adcom preamp, for example.) If I made momentary up/down controls for a digital attenuator, my fear would be that a circuit malfunction would start it racing up on its own.

Quote:

Originally Posted by dsavitsk Last, I think the natural evolution of this is to use transistors as the switches allowing for a very very inexpensive digitally controlled rotary stepped attenuator.

Quote:

Originally Posted by rickcr42 and why not cmos analog switches ? Just a prepackaged FET switch that uses little power and you can even get "clickless" versions plus they are a plug and play solution.

This is a very appealing idea to me. Are you saying that an FET switch can be cleaner than the best gold or silver contact mechanical switch? We're going through 4 to 6 of them at once, if they add distortion it defeats any advantages to a stepped attenuator.

 

[Syzygies]

Quote:

Originally Posted by rickcr42 and why not cmos analog switches ?

I keep coming back to the following control idea: Basic mono linear pot sticking out of the enclosure, hooked up to the attenuator board. On board, a basic linear A to D chip converts to digital, to feed the cmos switches controlling a six stage (logarithmic) binary attenuator circuit like I've proposed. Voila, a small, inexpensive high grade logarithmic stepped attenuator with a familiar interface.

 

[DCameronMauch]

I love this design. Especially the constant load impedance. For those having problems figuring out the resistor values, I generated some equations:

V=10^(-dB/20)
R1=(1-V)(R0/V)
R2=(R0/R1)(R0+R1)

So for an attenuation of 2.5dB:

V=10^(-2.5/20)=10^-0.125=.7499
R1=(1-.7499)(10000/.7499)=.2501*13335=3335
R2=(10000/3335)(10000+3335)=2.999*13335=39985

You can see these computed values are very close to those in the schematic. To continue down the chain, R0 stays 10k, but R1 and R2 become R3 and R4 in the computations. And so on.

 

[DCameronMauch]

I'm thinking of using an optical encoder type rotary control connected to a quadrature decoder/counter to cmos switches. I'm at work right now, so I don't have the exact part numbers off hand. But I will post them later. This creates a very simple solution.

optical encoder: Bourns EN series rotary optical encoder
decoder/counter: Aligent Tech HCTL-2000 ???

 

[Syzygies]

Quote:

Originally Posted by DCameronMauch You can see these computed values are very close to those in the schematic.

I agree with your formula values; thanks for checking!

I chose Dale Vishay values out of a Mouser catalog, as close as I could get, playing both ways on a spreadsheet to see what happens. I worked the next stage using the actual values chosen so far, trying to bring the errors back to zero, etc...

Of course, best to write a program to try all combinations. I'd put the most emphasis on keeping the clusters of impedances tightly bunched, as the circuit itself depends on this. Rounding up or down at random from formula values probably works fine, but there's an aesthetic appeal to making the optimal choice.

 

[Syzygies]

Quote:

Originally Posted by DCameronMauch V=10^(-dB/20) R1=(1-V)(R0/V) R2=(R0/R1)(R0+R1)

Here's the way of writing the resistor value ratios I find clearest:

...where R=10^(dB/20) for a single attenuation step of dB. Then multiply all values by the desired impedance.

Here's a first draft for a home-etched one-sided layout:

Here's a tighter layout:

Although ideally this should be no cause for concern, I find it unsettling that this circuit acts as a current divider, with very little of the current making its way past the output at low volumes. In a basic voltage divider attenuator, all of the current is raging past the output. Here, the output is sticking its toe in an idle brook.

How does the deviation from ideal linearity work for high quality resistors? Are they more accurate or less accurate with tiny currents?

 

[Syzygies]

I'm gearing up to try to build a Pimeta amp, 12 AAA cells battery, a circuit that trickle charges the cells in two banks from an iPod power supply when the amp is off, and a binary stepped attenuator, all to fit in a Hammond 1455L1201 case, 120 x 95 x 27.5 mm interior. I won't cut the case till I see that the attenuator works, though a Blue Velvet just misses fitting into this case.

I'm convinced that I'll never want more than four bits of attenuation, and I'd believe that some problems will start to emerge with more than four bits, because of the large resistor values that would be needed. (I'm hoping problems don't emerge with four bits!) I'll play with amp gain for my source and headphones so a 37 dB range gives me the settings I'll actually use.

It would be appealing to have a slide-in PCB above the sideways Pimeta board (tall caps in back), to support a row of horizontally mounted right-angle slide switches such as the Alcoswitch MSS2250RG, and leaving room to fit affordable but oversized exotic resistors such as the Holco H4 resistors from Michael Percy Audio. This just fits; I don't like the jumpers, but I'm home etching here:

 

[Syzygies]

Quote:

Originally Posted by rickcr42 heh,been done man .Called "Volume" and "Trim"

Quote:

Originally Posted by Syzygies Thanks. Do you have a link to the best resistor layout for this?

Has anyone seen a constant impedance way to wire "Volume" and "Trim" stepped attenuators? I'm trying to avoid the current division in my binary attenuator; everything I can think of here has milder forms of the same issues. Perhaps that's good news, if these are accepted as something that works. I'd still love to see the status quo wiring diagram.

 

[AndrewFischer]

Quote:

Originally Posted by Syzygies Has anyone seen a constant impedance way to wire "Volume" and "Trim" stepped attenuators? I'm trying to avoid the current division in my binary attenuator; everything I can think of here has milder forms of the same issues. Perhaps that's good news, if these are accepted as something that works. I'd still love to see the status quo wiring diagram.

The HP attenuator I posted a picture of is a constant 600 ohms. However that only holds true if the output is a 600 ohm load. There is a schematic in the Manual. If only I knew where that was...

 

[Syzygies]

From the White Noise catalog http://www.wnaudio.com/cat.pdf, p.19:

Quote:

Notice that, contrary to myth, it is the number of switch contacts in the signal path, not the number of resistors, that determines attenuator quality. Switch contacts are exposed to oxygen, salt, grease, and organic pollutants in the air and they are subject to wear. Switch contacts have a far more deleterious effect on sound quality than the relatively benign resistors. Shunt attenuators, with only one switch contact in the signal path at any setting, are therefore much to be preferred over L-pad attenuators, which have two switch contacts in the signal path at all times.

If this is right it pretty much kills my binary attenuator idea, unless it can cleanly be digitally switched, as various people proposed.

 

[AndrewB]

Don't write your idea off that quickly. Try to use DPDT switches with gold contacts. The potential for arcing of pitting the switch contacts is pretty small.

 

[Syzygies]

Quote:

Originally Posted by AndrewB Don't write your idea off that quickly. Try to use DPDT switches with gold contacts.

Thanks for the encouragement. The only way to find out for sure is to proceed with the best parts as if I'm sure it will work. I'm ok with that, whatever the odds.

I'd love to use the Hammond 1455J1201 case for my first amp project, a PIMETA. Leaving room for 3.5mm input and output jacks on the front panel, I can just fit four E-Switch PC-mount DPDT toggles, gold contacts, part number 200AWMDP1T1A1M2RE, $6.64 each from Digi-Key:

This is competing with a Panasonic pot; the ALPS Blue Velvet won't fit. Every other switch option I looked at is bigger and pricier, including DIP switches. My on/off switch and charging jack will move to the side. That's fine, the above is exactly what I need to have peeking out of my pocket.

I'll socket the PIMETA resistors so I can play with gain to get the attenuation of this thing within a useful range for my iPod and Ety 4p's.

 

[DCameronMauch]

Mark Levinson uses these CMOS analog switches in their reference No32 preamp attenuation board.

Edit: You would have to account for their 45ohm on-resistance though.

 

[Syzygies]

Quote:

Originally Posted by DCameronMauch Mark Levinson uses these CMOS analog switches in their reference No32 preamp attenuation board.

Wow! You can actually see their circuit, at least one of the layers of traces, I'll take this on the plane tomorrow together with the datasheet, see if I can figure out what they're doing. What are those empty pads, are they connecting to other trace layers? Great puzzle!

They claim 0.1 dB steps most of the way down, and 2^16 volume levels, so this must also be truly a multiplicative circuit, not additive like a D to A converter. They're using 16 DPDT switches in a very uniform layout, and 64 of the resistors look very similar in role, hmmm. That's exactly my resistor to switch ratio, this all looks binary. I wonder if there's more than one way to do this, or if I've stumbled onto their circuit?

What I'd love (and the others that have chimed in) is a small, affordable attenuator board suitable for our portable projects. I'm trying mechanical first as it is the easiest way to see what works.

[You can tell I'm a newbie because everything I think about only involves R. I'm hoping to get better at visualizing R,L,C so I can play with amps. There are "L" attenuators out there, but they're huge, not suitable for us.]