[Syzygies]

Ok, I did my homework for y'all on the plane. The part of Mark Levinson's circuit that is obvious from the photo looks like

In Mark Levinson No. 32 Reference Pre-Amplifier they write:

Quote:

An array of precision resistors provides attenuation in 0.1dB steps down to -57.0dB, where step size is increased to 1.0dB. In all, our technology provides over 65,000 steps, allowing us to make the No32's "stepped attenuator" act and sound like a continuously variable control.

This all hangs together when one tries to design something based on this circuit snippet. By selectively tapping into a main path of resistors at various places, one can get a varying voltage divider effect, but in doing so one provides an alternate route for the current. Two problems are apparent:

1. Simply keying binary numbers into the switches doesn't work.

2. The impedance varies.

One can minimize the impedance variation by carefully choosing resistor values, and perhaps by using DPDT switches to balance things, connecting every horizontal resistor to one of two buses, using one for audio out? But this only partially compensates, unless one heavily restricts the set of allowed combinations, which apparently they do:

Go back to the quote. 0.1dB steps down to -57.0dB, etc., is a lot less than 2^16. They must be using a lookup table ROM to translate each dB step to a 16 bit word of switch positions, having carefully optimized for all of the above considerations.

In short, it's not my circuit, and we can't use their circuit to choose 16 volume levels via four toggle switches.

 

[Syzygies]

So I've seen mention of controlling volume by instead varying the resistance in the op amp feedback loops. It sure seems like it would be the easy way to go, for stepped attenuation. What am I missing or not understanding?

 

[z2trillion]

Sorry to revive a dead thread, but I didn't want to start a new thread on this.

Quote:

Originally Posted by Syzygies Ok, I did my homework for y'all on the plane. The part of Mark Levinson's circuit that is obvious from the photo looks like In Mark Levinson No. 32 Reference Pre-Amplifier they write: This all hangs together when one tries to design something based on this circuit snippet. By selectively tapping into a main path of resistors at various places, one can get a varying voltage divider effect, but in doing so one provides an alternate route for the current. Two problems are apparent: 1. Simply keying binary numbers into the switches doesn't work. 2. The impedance varies. One can minimize the impedance variation by carefully choosing resistor values, and perhaps by using DPDT switches to balance things, connecting every horizontal resistor to one of two buses, using one for audio out? But this only partially compensates, unless one heavily restricts the set of allowed combinations, which apparently they do: Go back to the quote. 0.1dB steps down to -57.0dB, etc., is a lot less than 2^16. They must be using a lookup table ROM to translate each dB step to a 16 bit word of switch positions, having carefully optimized for all of the above considerations. In short, it's not my circuit, and we can't use their circuit to choose 16 volume levels via four toggle switches.

Has anyone looked into using a R-2R ladder as a volume control? It seems to me that is what Levinson is using. I'm thinking of building one and using it as a volume control along an encoder and with 40 some relays i bought recently. It seems to me like this would work muhc better than the binary attenuation thingie mentioned previously. Of course, I know very little about EE, hence this post.

 

[Syzygies]

Quote:

Originally Posted by z2trillion Has anyone looked into using a R-2R ladder as a volume control? It seems to me that is what Levinson is using. I'm thinking of building one and using it as a volume control along an encoder and with 40 some relays i bought recently.

Yes, it does look like an R-2R ladder, exactly what is used for digital-to-analog conversion, e.g. playing back a CD. These give linear attenuation curves, while our ears want logarithmic attenuation curves. They take 2^16 linear steps, and choose a sampling of values equally spaced in dB reduction, to achieve their logarithmic curve. Their board is big, expensive, and an engineering triumph (read: a huge time sink to duplicate).

 

[Frenchman]

I'm not too sure it would be that big a hassle. Figuring out the values should be the biggest pain. To do the switching, all you'd need is a simple microcontroller. Take an input from a switch, and when it was pressed, set the selection inputs of a multiplexer so that the multiplexers high input is routed through to the correct analog switch. Conversely, another switch could be monitored so that shifting in either direction is possible.

If you're only looking for 16 possible steps, then you'd need 2 inputs for buttons and 4 outputs to control the multiplexer. All of this could be done with cheap 8 pin PIC microcontroller. Realizing that, the whole circuit would consist of the microcontroller, the multiplexer, 8 analog switches and the passives. Not exactly a small circuit but with creative layout, I think you could cram this all into a portable enclosure. The task would be even easier with a PCB.

Their board is certainly extremely impressive, but I think you might see 2^16 and in that number, forget you're dealing with 1/256 of that amount. I think this could be a fun and rewarding project. It certainly seems like it would be a first here. There have been similar things done, but none of them at such a small scale. Moreover, the state of miniaturization today means that now might be the first time that it actually CAN be done.

 

[z2trillion]

Do relays really do anything to the signal that passes through. If I end up trying to build this, I'm thinking of using something like 20 relays per channel, so any advice would be much appreciated.

 

[Syzygies]

Are they silicon or mechanical relays? Every mechanical contact hurts the sound.

 

[z2trillion]

They're mechanical. Could someone explain why mechanical switches are so much worse than transistors? I've read the wna explanation, but in this case, since the relays are sealed and gold cobalt" plated, most of the things mentioned don't seem to apply.

 

[Frenchman]

I can't say for sure any of this is right, but I might as well give this a try.

If you were to look at it as a transmission line problem, then you could simplify the system as a transmission line of a certain impedence with a short length of line with a different impedence followed by the first type of line. First, you're going to get some reflections on the boundries, and then as the switch operates more times, then the switch will start to scratch particles off the surfaces. This creates a bunch of little capacitors, which introduces a complex value to the impedence and distortion. Transistors and other semiconductors keep the characteristic impedences closely matched.

Anyway, that's just a guess from a current EE undergrad so it could be just a bunch of 2AM BS.

 

[blip]

Very interesting thread. I like the idea that we, as DIY, can break free of traditional interface confines if it better suits engineering. Hope that something related to this gets off the ground.

My one thought, in response to some concerns raised earlier about the risk of accidentally setting full-volume while intentionally or unintentionally changing the switches. It seems to me like the ideal way to combat this would be to have a sort of "clutch" that would take the switches out of the signal chain while they are being adjusted. But I can't noodle how this could be done while maintaining the existing volume level.

Another option would be to have a safety switch which, when closed, would prevent the volume from changing. Again, I'm out of my depth here so I don't have a good way to do this... but I thought that the thought might spark someone elses thought process on the interface.

 

[rickcr42]

Quote:

They're mechanical. Could someone explain why mechanical switches are so much worse than transistors? I've read the wna explanation, but in this case, since the relays are sealed and gold cobalt" plated, most of the things mentioned don't seem to apply.

the only reason you see transistor/cmos/fet switches is cost and not superior audio contact.
They are NOT superior to relays but a good audio quality relay is far more expensive than an electronic switch,takes up more room and will in the end require more power from the supply to drive multiple relays than multiple cmos switches-again costing more.

But consider the audio path itself and then decide what you would want an audio signal to go through :

1-Another active stage which is what an electronic switch really is even though it has no gain,and ALL active devices add something to the sound plus if you know what the mechanism is that actually does the switching (an fet) there is always the potential for power supply bleed through from the control pin

2-mechanical switches with precious metal contact areas and NO power supply interaction and No additional active stages in the signal path.Signal purity as high as can be expected short of hardwiring each stage (every single wire,jack,plug,switch is a mechanical contact and a simple jack is more of an audio butcher than a simple switch)

I know which way I would go if ultimate sound quality and cost no object was my goal but if I were to build a 48 channel full feature recording desk or an complex AV switcher I would most likely opt for cheap and low current or the cost and heat buildup would make for a very complex and very expensive project.

BTW-the biggest audio butcher in my system is the TVs electronic volume control.In level matched direct comparison there is no doubt what it is doing and it is not good

 

[Publius]

Quote:

Originally Posted by rickcr42 1-Another active stage which is what an electronic switch really is even though it has no gain,and ALL active devices add something to the sound plus if you know what the mechanism is that actually does the switching (an fet) there is always the potential for power supply bleed through from the control pin

Have you actually looked at some FET specs? This for instance can take a +-10V change in gate voltage and only deviate in resistance by +- 2 mohm (out of 5). Utterly insignificant for an attenuator IMHO.

 

[rickcr42]

you can look at all the specs you want but ANY active stage will add to the sound of the signal going through it.It can be no other way no matter what the data sheets say.

as usual and for those who have not read this disclaimer before : this is DIY and anything any member wants to build and by any method is cool.But since this is a discussion forum with opinions and questions asked there will be answers from all sides of an issue-as there should be.

but to design a piece of audio gear solely on the specs is a recipe for poor sound sice the very nature of "audio' means to listen.

Electronic switching while it has come a long way is still a basic fet used as a switch just like they were twenty years ago.
The draw of the "solid state switching element' is ease of assembly,low cost for large scale construction and is easily integrated into control logic but it is sonically inferior to a heavy gold contact mechanical relay.this is not "on paper' but by actual listening.If the system the switch is integrated into does not have the resolution to show (hear) a discernable difference then trying for 'ultimate" anything is moot anyway.
Then it comes down to cost and convenience as with most of our "progress" in audio which is mostly a step back in sonics but easy for the layman to implement because being an "all in one' monolithic solution,all the work is done for you and put into a cute little box that is added to other little "boxes" until you have a final design.

Yes the mechanical the relay has a finite life span,an amount of cycles before it degrades the signal and eventually wear out but for me personally I will take ten to twenty years of superior sound over 100 years of "pretty good".

Just my opinion and as always YMMV

 

[Publius]

I never argued designing by the numbers - I'm just annoyed at statements like
Quote:

there is always the potential for power supply bleed through from the control pin

which imply any sort of engineering consensus for what you believe, when in fact no consensus exists. That's all I'm saying. You can only justify using switches because they sound better.

There is a bleed through current (Igs); for the MOSFET in question, it's 100nA. Apparantly it gets smaller for small-signal MOSFETs. Based on a rather primitive analysis, it looks like it depends on the source resistance driving the signal, the resistance of the switch, and the load resistance. Given source=10 ohms, load=10k, switch=5mohm and leakage current of 100nA, the output differs from the input by something like 1uV. Note that is a very worst case scenario.

 

[rickcr42]

did you actually read the word "potential" ?

there is a potential for every damn electronic design to blow up but that does not mean it will !

Ease up dude and grab some damn coffee or something.If DIY annoys you so much maybe you need a new hobby