Right L-pad Value?

[Evgen_K]

Hello gentlemen!
I'm going to connect a modern DAC with about 0.9V RMS out to two different vintage ss amps from 60-70s. The out voltage of the DAC will be adjusted with swithcable L-attenuator.
The question is: what is the right resistor value for the signal path in case one amp has 20k ohms input impedance and the other one has 100k ohms?

Also I wonder is the shunt resistor quality makes any sence? I'm going to use a high quality tantalum in signal path, but is it a good idea to use the cheapest carbons as shunts?

[Herald1360]

Leaving aside the logic of spending about ten times as much on tantalum vs standard metal film resistors, why foul things up royally with noisy old carbon resistors anywhere in the signal path?

As to values, as long as the output (source) resistance of the L-pad is significantly lower than the input resistance of the amplifier you should be fine so far as getting the attenuation you might expect. You might need to worry about the output (source) resistance of the DAC though.

[Radio Wrangler]

There is a minimum amount of noise voltage below which any resistor cannot go. It is set by the resistance value, the temperature, and the Maxwell-Boltzman constant. The mechanism comes out of quantum mechanics.

Most resistors make more noise than this limit, the excess noise is powered by residual amounts of DC in most circuits. When you calculate the noise levels, it requires very little DC.

Carbon-composition resistors are about the worst there is. Carbon film resistors are better, but still not good. It is highly ironic that fashion in hifi circles sees these parts as desirable.

Metal film resistors and wirewound resistors come closest to the theoretical noise limitation.

Tantalum nitride and Ruthenium oxide are commonly used materials for thick-film resistors, and they have a lot of excess noise... getting into the levels of carbon resistors. It's also a big joke that the hifi fashionistas pay lots of money for tantalum resistors.

Metal film ones are provably lower noise and are provably highly linear. Their only disadvantage for audio is that they don't cost impressive amounts of money.

The weighting of which resistor, series or shunt, contributes the greater amount of noise depends on the ratio of their values.

For an attenuator designed to present an acceptable load to the DAC, you will probably be using resistor values where the input impedances of those two amplifiers will make a small difference to the attenuation.

If you are using this setup for trying to compare two amplifiers by ear, then be aware that the human impression of sound is highly affected by small changes in level. Human perception of sound is also strongly influenced by what was heard before, so you really need to swap systems many times and for the listeners to not know at the time which they are hearing.

I would make two different attenuators and adjust each to get the same levels out of the two amplifiers.

David

[mhennessy]

Originally Posted by Evgen_K

Also I wonder is the shunt resistor quality makes any sence? I'm going to use a high quality tantalum in signal path, but is it a good idea to use the cheapest carbons as shunts?

The shunt resistor is also in the signal path!

To keep thermal noise to a minimum, choose a pair of resistors that add up to something in the 2k region. The output stage of the DAC should be able to drive that with no increase in distortion, but if in doubt, feel free to check to see which op-amp they used, and how much load its NFB network imposes.

You don't say how much attenuation you need, but for example, if the input sensitivity is in the region of 200mV, then try 1k5 and 510 ohms. Or 3k and 1k...

Having done that, the input impedance of the following amplifier is not a significant factor any more. Meaning the same network will work with any amplifier that requires that sort of signal level.

As has been said, bog-standard metal film resistors are absolutely perfect. Fashion is a cruel master in the DIY audio world - try to avoid the rabbit hole!

But is it actually necessary?

Few amplifiers will actually mind being fed with slightly too much signal.

The main practical downside will be that the volume control will be a bit more "aggressive" in use, perhaps resulting in difficulties at the lower volume settings, and potential channel imbalance at low volumes because the matching of the two ganged tracks might not be as good down there. And of course, you'd need to adjust the volume as you switch between different source equipment - which you usually have to with modern equipment anyway IME. Naturally, you won't be able to turn the volume control as far before the power amplifiers start to clip, but the clipping will take place at the same power output as it did before, so you're not losing anything.

But if those "operational issues" aren't a problem in practice for you, then I wouldn't worry about it. Certainly, I wouldn't be afraid to try it first of all before deciding whether to build an attenuator for operational convenience. Honestly, it would be a very poorly designed amplifier indeed that was overloaded by 900mV at a line input.

[Evgen_K]

Thank you for the advice.
So I give up the idea of using a tantalum and will use metal ones instead both as series and shunt resistors.
The only question left is the right value of serial resistor in the L-pad:
I have 150 ohm DAC out and 100k ohm amp in, 150 ohm DAC out and 20k ohm amp in in case I use two different attenuators or one switchable with two different shunt values as they have slighlty different input sensitivity (170 mV for the first and 125 mV for the second one).

[lesmw0sec]

Given the high DAC output, the question arises as to what the signal requirement is at the amplifier input? Unless it is down at the low millivolt level, just about any resistor would do, as the thermal noise of the resistor would be well below the signal. As Mr. Hennessy points out, keeping the values low would be an advantage.

[mhennessy]

For the series resistor, try 2k.

The required shunt for 170mV is 510R

The required shunt for 125mV is 360R

If you wanted to have a switchable attenuator, build the first, and arrange a switch to place 1k2 in parallel with the 510R resistor.

Another alternative is to split the 510R resistor into two (360R and 150R) and arrange for the switch to short out the 150R resistor.

Don't arrange it to switch between the two (510R and 360R) resistors, as there will be a large jump in level whenever you operate the switch (unless you use a make-before-break type, but those are getting harder to find). Besides, the suggested approaches are better when the switch contacts age and become intermittent.

Those calculations included the output impedance of the DAC, but ignored the input impedance of the amplifiers. If you've studied Thévenin, you'll know that the output impedance of these networks will be 412 and 308 ohms respectively, and the loss caused by the loading of 10k or 100k will be tiny (about 0.3dB worst-case).

I still question the need of this, and I certainly question the need to have two separate networks with this level of precision. The difference between 170mV and 125mV is only 2.7dB.

Either way, here's some possible solutions...

[Evgen_K]

[QUOTE=mhennessy;1315853]

Originally Posted by Evgen_K

The main practical downside will be that the volume control will be a bit more "aggressive" in use, perhaps resulting in difficulties at the lower volume settings, and potential channel imbalance at low volumes because the matching of the two ganged tracks might not be as good down there.

That is the main reason for me to attenuate the signal.

That's a very clear answer, I appreciate your help.
All in all, now I think it would be better to make a simple L-pad with no switch. Simplicity is better.


A bit offtopic, however. There are some ready to use stepped attenuators on market (goldpoint for example) with resistance up to 250 kohms. What are such values needed for in case you mentioned that the smaller value is better?

Speaking about make-before-break switches. There are some made in Kazakhstan and available in Russia called PGK. ПГК-8П4Н-К13Ш model for example (these complicated Russian model names always make me crazy) with quite reasonable price.

[mhennessy]

Higher impedances might be helpful for valve circuitry.