Stepped attenuator - getting the impedance right

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
Hi everyone,

First post!

Right, I'm looking into building a simple passive pre with a stepped attenuator and have done loads and loads of reading.

I've read that the impedance of my attenuator wants to be at least 10 times the output impedance of my source, and that it also wants to be at least a tenth of the input impedance of my power amp. I also read somewhere that 20 times is a good figure to aim for, so I figured that bigger the difference in impedance across each interconnect the better.

The preamp will for the vast majority of the time be fed by the soundcard on my pc. The line-out jack on my soundcard is rated at 32 Ohm. I presume this is the source impedance then?

My power amp has an input impedance of 15 kOhm. So the maximum I can achieve both pc-preamp and then preamp-power amp is about 21 times each way, and this works out as an attenuator impedance of approx 700 Ohm:

32 x 21 = 672
15000/21 = 714

Why then, are most attenuators that are sold as kits rated at 10 kOhm, 20 kOhm, 50k, etc? Also, no-one seems to talk about anything below 10k. I'm nowhere near 10k.

Another problem I have is that because I (think I) need a relatively low total impedance I'm ending up with low resistance values for the quiet-end resistors, like 0.5 and 0.75 Ohm and such. The majority of these high-end metal film resistors I want to buy start at 10 Ohms, nothing lower. I dont want to start mixing and matching brands/specs, I'm far too novice for any of that.

Am I missing something? I must be, surely. Please feel free to pick holes and basically smash to pieces anything and everything that I've said which is complete rubbish. I'm new to all this, but keen to learn.

Thanks,
Leon
 
Hi everyone,

First post!


Why then, are most attenuators that are sold as kits rated at 10 kOhm, 20 kOhm, 50k, etc? Also, no-one seems to talk about anything below 10k. I'm nowhere near 10k.


Thanks,
Leon

Your poweramp is the problem, 15k input is stupid!!! A tube preamp would even have trouble driving that.
The industrie standard is 47k-50K with most tube poweamps at 100k to 500k, I've even seen some at 1Mohm. ie. Rogue.
Talk to the designer of your poweramp to see if the input load resistor can be changed from 15k to 50k or more. Otherwise you'll need a buffer after the passive.

Cheers George
 
No, what you have written is exactly what a beginner should have picked up from sensible reading of the forum.

Your maths is good.

The problem is the impedances of your source and load are not cooperating. This is often the case. Ideally what you would do is build the attenuator with the values you have calculated. If your source had been 1k and your amp 100k, you'd have been home free. These aren't unlikely numbers.

What you can do is compromise.

Pick a kit higher than the output impedance and lower than the input impedance (10k).

Or you can make the effort to build with your calculated values.

The reason that kits are available in 10, 20, 50k impedances is because they're easy to do.

The OTHER reason is that the relative impedances of source and sink in audio are of comparatively small significance just so long as you follow the smaller>bigger>bigger rule.

You can jigger around with this 'till the cows come home, you can put in buffer amplifiers, but this will just be more components which other people will be telling you to take out. My opinion:- go with the 10k.



w
 
#2 on what wakibaki says- just go with 10Kohms. That's the most common value and will work fine. There's nothing unusual about a 15Kohm input impedance on a power amp. Yes, it's lower than you might like, but the nature of solid state power amp circuitry tends to give values far less than the higher numbers associated with ye olde tube amps.

I'm donning my fireproof underwear as I say this, but the 10X and/or 20X rules guarantee nothing, and you can often violate them by huge amounts. If the input impedance of the amp changes significantly with frequency (not a common problem) then you want to drive it from a much lower impedance, lest the response be affected. If the input impedance remains ruler flat with frequency (typical) then you can drive it with near anything. The attenuator steps may not be the same as unloaded, but that's of little consequence. Using a 10Kohm attenuator with a 15Kohm power amp will probably work and sound just fine.

Any solid state preamp should be able to drive a 10Kohm attenuator with no problem. Not so with tube preamps, so watch the capabilities of the particular circuit. Again, if the response isn't affected by loading, the attenuator value becomes less important. If the preamp is affected by loading, be careful if you build a low impedance attenuator. FWIW, I think I once built an attenuator of about 7.5Kohms, but it would be rare to see one below 10Kohms or above 100Kohms.
 
Thanks for the replies guys.

So am I right in taking from this the following points:
1. The 10x rule is a rule of thumb, and not necessarily set in stone.
2. My problems are arising because of a low input impedance on my power amp.
3. I could still go ahead as I am, and build a custom 700 Ohm att if I wanted to.

Hmmmm, what to do. I know people have suggested I just take a 10 kOhm kit, but I have absolutely no trouble calculating resistor values and designing a custom 700 Ohmer.

Would you say a 10k is a bit of a gamble with that 15k input on the power amp? If there's any gamble whatsoever I think I'll stick with my original plan and build a 700 Ohm. If not, I'll go with a 10k.

Can I run into any trouble with a very low impedance attenuator?

Cheers,
Leon
 
There is another potential snag. The input impedance of the amp might not change much with frequency, but it could change with level instead. If so, this would give distortion.

You said that the output impedance of your source is 32 ohms. Is this so, or is it designed to drive 32 ohms? Either could be true, and people often get the two confused. Specifications will often say "output impedance X", when what they actually mean is "required load impedance >= X".
 
Hi,
the output impedance of the attenuator is not the quoted attenuator impedance. It varies, a lot.

The maximum output impedance is {Rs + attenuator resistance} / 4

A 10k pot and a source of 200r will have an output impedance that varies from 0r0 to 2550r.

It would be nice if that 10k could interface with 20times that maximum i.e. 50K, but 15k will do. It is only ~5.88times, but it will work.
 
Hmmm well thanks for all the input guys, it's extremely helpful, but this is beginning to go over my head a bit.

Is there a way I can determine with absolute certainty what the output impedance of my soundcard is?

If the input impedance of my power amp can vary with level, I take it that if it drops too much then that’s when I could get distortion? Hence the standard practice of having the att impedance an order of magnitude lower than the rated input on the amp?

I’ve made myself a little spreadsheet where I can enter a total impedance and pick the dB attenuation at each of the 24 steps, and it will calculate exact resistor values for all the positions on the switch. Then it pastes those values into another table where I can change the resistor values to the nearest values that are actually available to buy, and it shows me how the total impedance and dB for each step is affected. After messing around with it trying out different combinations I’m finding that a 1.5 kOhm attenuator would be easy to do, all the resistors I’d need are readily available and I can easily get a nice attenuation curve down to -60dB. This appears a good compromise.

I still feel there’s a lot more to this than I’m considering though, and I don’t wanna blow a load of money on resistors and find them unsuitable. Maybe experimenting with a cheap-*** pot (as suggested above) is a good idea for me.

Probably another huge can of worms, but what does a buffer do (basically), what does it entail to build (basically)?

Apologies for my lack of knowledge, I’ll get there eventually.

Cheers,
Leon
 
I think you've got your head around this just fine and most of the design factors have been covered. I'd second the idea of just getting some 1K and 10K pots and have a listen to be sure everything is well behaved. FWIW, I don't like buffers and rarely (never say never) use them. It makes no sense, given the quality of buffer one can build, but I always convince myself that the sound is better without them. Maybe I just like the top end rolled ever so slightly (though the numbers say I can't possible hear it).

Something about output impedances and drive capability. Most/all ICs can be configured to have a very low (<5 ohms) output impedance, but the drive capability alluded to above is much more limited. Worse, even if the IC can drive the low impedance load, the specifications may be dismal in terms of distortion and response. Some later op amps can drive very low impedance loads with no problem, but I've no idea what sound cards use. If it can drive low impedance phones, it may be OK, but performance usually degrades with falling impedance.

You can easily measure output impedance by simply generating a fixed tone at a given amplitude, then loading the output with a resistor, say 500 ohms, and measuring the amplitude drop. Ohms law will take it from there. I'm not going to tell you how, as this is a really valuable exercise to work out if you haven't already. You should also add a resistor to your spreadsheet in series with the attenuator to represent the card output impedance. That way you can see how much voltage you lose and what the maximum output really is.

Conrad
 
Thanks for the encouragement.

I've taken on board everything that's been said, gonna try a couple of cheap pots first to make sure I'm pitching in the right area.

One thing which does bother me is this: I'll be using this passive pre to switch between my pc and tv. Any idea what kind of output impedance I can expect from the cable tv box? It'll be wired to the tv with composite video and standard audio L-R, so i'll just be routing the red and white audio rca's directly into this preamp.

Worried now that the cable tv box could have a pretty high output impedance.

Thoughts anyone?

Cheers,
Leon
 
The lack of a PE connection, due to double insulation, is part of the problem. I am not talking ground loops, but direct injection of a hum current. There is likely to be a capacitor (in the mains filter) between 'ground' of an audio output, and live mains. This could inject several hundred uA of hum - I believe the limit is 1mA. Sony equipment is particularly prone to this - maybe they have good EMC filtering?
 
I'd expect modern TVs to be good, since they were designed to be used in home theater and similar situations. My Sharp work flawlessly with no hum or problems. Still, I know with older sets it's been an issue, though I don't know the exact mechanism. I have found TV sound to be surprisingly good.

One other thing about attenuators. IMO, stepped attenuators are invariably better than pots because you can get the gain matched near-to-perfect. Compared to that one single thing, resistor type/brand, step accuracy and such, are minor issues. Nothing will help your sound stage like getting everything in the signal chain matched. Anything else you improve is just icing on the cake.

CH
 
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.