why does a relay attenuator sounds much worst than an alps black ?

I have just tested a relay attenuator board, and I'm quite diasppointed at the moment. Something I do wrong probably, but I do not know what. If you have any experience with R-2R relay attenuator network, please let me know. If you could point anything wrong on my proto, please let me know.

This is the schematic :


  • schematic prints.pdf
    51.4 KB · Views: 385
a picture of the top :


  • dsc_2033resize.jpg
    74.5 KB · Views: 1,337
Contact quality at very low signal levels is extremely important.. You did use relays designed to reliably switch very low voltages, right?

Choosing a conventional power relay can result in very high distortion due to contact oxidization. The oxidation actually forms diodes. Some (many) relays will not maintain a reliable connection with less than a few mA flowing through the contacts.

Long ago I used some very nice looking relays in a pre-amplifier to select sources, invert phase and mute. None of them worked until I treated all of the contacts with a cleaner and then cramolin. They were brand new and passed dc at a few mA without issues - no high impedance low current audio got through until I treated the contacts. They were the wrong type of relay for the application, contacts being rated at a couple of amps at 24Vdc. Suitable relays will indicate that they are designed to handle very low signal levels and often have a relatively low maximum current and voltage rating due to the contact treatments used. Most reed relays are suitable if selected for the application, although form C types are less reliable - use a form A and form B to make spdt switches if you go this route.

The other potential suspect would be the resistors used. Avoid metal oxide, carbon film for linearity issues due to their construction. Carbon types in general may be too noisy for best performance. Really good resistors like Vishay and the older Holco H series metal films are warranted in this application.
and this is about the bottom :

the resistors are proven 1W Beyschalgs, the relays are Takamisawa NA5W-K. The traces on the PCB are mostly the resistor legs, instead of copper trace.

The board is inserted before a line amplifier which is followed by a 2A3 PSE and AN speakers. The line amp starts with a 100k alps black, turned to full volume. The wires to and from the board are very short, 20-30mm only. The relays are run from 3 alkaline battery BCD encoders. 4.6V measured. It works perfectly on the bench, gives the logaritmic curve very accurately.

The sonic difference is interesting. It looks like to higher the high frequencies a bit. The LP noise is more significant also. Several different disk is tested and on some tracks I did not know which one is better. For sure it was different, much different. Sometime it had a bit more trasnsparent space, but after listeinig it longer I did found it to be true, rather it had too much high frequencies. Brasses decided it, very significantly. The network sounds good, but the alps has much more life, the instrument has more weight.

I was thinking about, maybe I have used wrong relays. I would be very very interested in what relay should I try.

Thanks a lot,



  • dsc_2035resize.jpg
    79.4 KB · Views: 1,197
Hi kevinkr, I have very good experience with these Beyschlag resistors, this should not be a problem.
The relays are miniature signal relays, max. 50mOhm is specified contact resistance. It is sealed.

I'm now thinking about to make a "burn in" circuit. The relays are brand new, during the measurements I switched them no more than 30 time each.


I looked up the NA5W-K and despite the bifurcated gold over silver contacts I see that the relay is designed to switch higher currents its high reliability rating not withstanding. I'd change that to a relay designed expressly to switch low voltages at very low currents. Usually the maximum contact voltage rating will be <30V and the current 50mA or less. Alternately you might find reeds better. There are relays in your existing form factor that will pop right in and have the required specs.

I'm no fan of beyschlag resistors, nothing special imo, and not what I would use in this application.

It may well also be that the attenuator is telling you an "inconvenient truth" (love that line :) ) about something else in your system.

Another thought is that there are a lot of contacts in series with the signal in that binary attenuator.
it is interesting. I have choosen this against Omron G6K ( http://www.omroncomponents.eu/home/products/Relays/PCBSignalRelays/upto1A/G6K/ ) because of the low contact resistance, 50mOhm against 100mOhm (and because I can't source through hole G6K around, RS, Farnell, ELFA no one had on stock).

Against the contact resistance, what you say is also logical for me. Which relay you know as the best for audio signal ?




2005-11-15 7:24 pm
Look for relays that say "telecom" or "dry circuit" in the description. Then look at the specs carefully... when you are looking at the right style of relay... the data sheet should list the MINIMUM current the relay is capable of switching.

The best relay would probably be a mercury wetted reed. I don't know if they are still available though... with mercury and the envoronment issues to consider.



2006-07-06 1:23 am
Hi Giordano,

The relays look good to me. I would be happy to recommend them for general audio use. There are auguably better relays but I believe your problem lies elsewhere...

I think the problem may be due to the higher resistance of the relay attenuator interacting with the input capacitance of your amplifier forming a low pass filter that is audible.

You mention the connection to your amp is very short, but coax cable has considerable capacitance even for short lengths.

You could try measuring the input capacitance and lowering the resistor values. As little as 50pF of capacitance could be an audible problem with your high resistance values.

Cheers, Ralph
Actually given his specific complaints it does not seem like cable capacitance is the issue here, he complained in an early post of emphasized (bright) treble if I interpreted what he was saying correctly.

The minimum contact current for these relays is 100uA based on the spec, below that and performance is likely to by iffy. My experience with very similar relays in audio work has not been all that positive.
I've found relays designed to "dry" switch very low level signals generally work better. Mercury whetted relays work great, but I have no idea whether or not they can be easily obtained these days due to ROHS.

The resistance values used in most attenuators associated with tube gear are high enough to assure in most cases that the signal current flowing through the contacts will be <100uA at any signal level at any point in the attenuator.

The first thing I would do though would be to try some different resistors and see how that affects the sound. I would also take a look at what the source impedance of the attenuator is at commonly used attenuations as compared to the existing volume control, and how if at all that might interact with the input of the amplifier circuit.
The glass capsule around the reed generally guarantees relatively low capacitance to the coil assembly which is the primary concern. The coil voltage is dc and it is not hard to provide a relatively clean supply to the coils - in any case the mutual inductance between the coil and contacts is so small that even any noise modulating the supply is not going to couple effectively into the switch path. These are frequently used in very low level circuitry in telco, ATE, and audio test gear without issues.

You can use rf reed relays and ground the capsule shields if you are concerned, but I believe this is overkill.
I'm more and more getting to the point that there is an impedance matching problem. In theory it should work, the network is designed for 100k termination, and the alps provides that.

In the other hand, yesterday at my friend's system it was sensitive to noise (mainly from mains) pick up, but we were able to place it with minimum noise. Today I spent 2 hour now on how to place it behind my preamp, but I have so much noise pick up that I can not evaluate it. Unfortunately just behind/under the line amp I have the mains distribution. I know it is not lucky, but I had 0 problem with it so far.

I'm afraid that this network must be right in the tube's legs with it's termination good noise floor.

Tomorrow I make a 2 resistor devider to see if that picks up similar noise or not. I less and less believe that it will be a right choice for tube based chain.


In the other hand, kevinkr you are right, I tried to explain we had more treble for sure with the network, not less. The difference is so high that it is not questionnable. Do not misunderstand, the relay netwrork is not bad, but it is like a very good interconnect against a not so good interconnect.

These relays are classified 'signal relay' also, but I have some Omron G6K now also.

I try to solder to gether the closed contact legs tomorrow. If that works, so it improves much, I will change to the G6K and I'll let you know.



its really interesting again that all answers so far just deal with ominous component ´qualities´ and no answer dealt with the circuit schem aka the component values itself! Very High-Enderish indeed and most sadly probabely useless :shutup:
I don´t think, that the chosen resistor types and relays are responsible for the ´bad´ sound. The choosen resistor values are imo the problem. They are way too high! This leads to a too high output resistance value and hence to reduced bandwidth.

I´d suggest to redesign the resistor values so that You have a lowish output resistance. Anything above 2kohms is rather problematic.
The values which You chose after Jos´ attenuator are very high. He works with an output impedance of 22kohms, reducing to ~2kOhms only for the lowest 32 values (those You probabely never need anyway!). This might work with high input impedances within a tube amp, but gives problems, as soon as You have to run the signal through a cable (i.e capacitances). Too the change in output resistance leads to a nonlinear stepping when You don´t run the attenuator into the specified 100kOhms input impedance of the following device (most devices work with 47kOhms).
A general problem of these logarithmic attenuators is that to achieve a high level of step linearity the resistor values have to be very tightly tolerated.
Besides a constant hum (I still couldn´t figure out where it came from..probabely interferences of the µController) this was the main reason I dropped Jos´s device and made one myself. The R-2R-network as well as a new logarithmic attenuator work imo both better because of lower and constant -througout the whole attenuation spectrum- impedance values.


ps. This attenuator is not an R-2R-network! That would work solely with 2 equal valued resistor-values...as its name implies.
Calvin said:
The choosen resistor values are imo the problem. They are way too high! This leads to a too high output resistance value and hence to reduced bandwidth.

And in your opinion this reduced bandwidth is responsible for bright sound and elevated LP noise?

As much as i hate high impedance volume controls, this does not seem to be the issue here.


Your setup is not very clean. The Alps pot is fully open but it still imposes its sonic contribution upon the setup. You should desolder it completely.

You are in a much better position to find out the reason for the poor sound that any of us. Eliminate the effect of the relays by hard-wiring the attenuator to a particular position and compare again with the Alps.

It may turn out that the Alps is responsible for a more mellow, less detailed sound which you actually prefer. Nothing strange here.

My experiences with relays have generally been bad, even in low resistance attenuators. High resistance just exacerbates the audibility. Just one set of bad contacts is enough to ruin the sound, let alone several.

A few years ago someone here published a subjective appraisal of several relays by a German magazine. Not surprisingly the perceived quality of sound varied a lot.

Someone should probably run such a test again. It will be particularly interesting to include Mercury wetted and reed relays.

My limited first hand experience shows that the "sound of a relay" varies a lot depending on whether the coil is energised or not.


2002-12-23 11:22 pm
Hi Giordano,

Based on my own experiences, my thoughts align with those of analog_sa's.

With an entirely passive arrangement as you have used here, it seems extremely unlikely to me that all frequencies *other than the extreme HF* are being attenuated by a similar degree.
This would need to be the case as far as I can see, as there is no apparent amplifying circuitry to 'enhance' the HF, and you imply that the balance of sound *other than the HF 'enhancement'* has remained the same.

Whilst passive components like you used here (in any audio circuits which I have had dealings with) can and do 'veil' certain frequencies by a small amount, I find it inconceivable that this would happen to the extent which you describe, and so evenly across the entire spectrum, except for the HF.

Most of the circuit-related comments mentioned here would simply result in a roll-off at the extremes of the audio frequencies, which as analog_sa has pointed out is the converse of what you hear.

Over the years, I experimented with dozens of HQ pots (and the Alps you mention was certainly not sonically benign, although I forget my impressions of it) and even the very costly and finest hand-selected Penny & Giles pot which I have, 'veils' the upper audio frequencies. This gives a distinctly 'duller' or less bright sound with slightly reduced upper frequencies, as a result.

For many years now, I have made stepped attenuators, mostly with 'manual' Shallco switches and resistors which were chosen for their sonic attributes, and every one was a world of difference where 'veiling' is concerned compared with the best of pots, especially in the HF regions. In fact, apart from the usual greater accuracy in attenuation and better channel-balance, the greatest improvements otherwise were certainly always at the HF end, where the 'transparency' improved a lot.

It is my guess that you have grown accustomed to this HF 'veiling' which existed until now, and that HFs now seem more prominent as this 'veiling' has been substantially lessened.

More recently I constructed a remotely controlled attenuator, using the best components known, and it shares the same characteristics of a very open and revealing top end, as my earlier manual attenuators do.

I hope that this helps.

My experiences pretty closely parallel the comments of both analog_sa and bobken as I have indicated in earlier posts.

Some have commented that the network values are excessively high, to those individuals I would comment that this is standard practice in tube gear where presumably tube sources have neither the low source impedance nor drive capability of most op-amp based solid state gear.

Note: Several have commented on high frequency roll-off, if you read the original poster's comments you would realize this is not what he is talking about, but rather the reverse. (I mentioned this in an earlier post along with a lot of other things.)

I also use all hand built step attenuators which in this case switch 2 resistor networks, which means I never have more than two resistors or two contacts in the path. Mine are hand built using Holco H4 resistors (the older version) and elma switches which unfortunately are NOT the last word in reliability. I have also used the dact attenuators in some commercial work I did in the past - these also work well. These incidentally have all been 100K types as I want to retain compatibility with tube tuners, tube tape machines, and some of my phono stages which in the past did not have particularly low source impedances.

The suggestion to hardwire with relays removed is a good one - that will either exonerate the relays (unlikely) - I too have had lots of problems with relays in the past -- or not.

I had some problems with mepco-electra mf resistors some years ago, ultimately it turned out that the end terminations were done incorrectly and the resistors in question were generating a lot of distortion. (I measured 0.15% for several of these as opposed to below the tester residual for same value different brand. Circuit was just a simple 6dB attenuator.)

If you have a good pc sound card and audio tester (audiotester.de) software you can have a look at the distortion spectra as well as frequency response anomalies the volume control is introducing at different attenuation settings. At this point I think this is almost a must - eliminating the guess work as to what is going on here. Use a unity gain buffer ahead of the sound card input so you can have a high input impedance. In a pinch the 5532 (dual) will work - better still the AD8610/8620 or the LM4562 (dual) - any will give you distortion levels far below what you are trying to measure.

I do subscribe to the possibility that the attenuator is simply unmasking things that your volume control pot has hidden until now.
I've never heard a transparent volume control pot, and don't use them if I can avoid them. You should remove it from the circuit for evaluative purposes.
I'm surprised.
The relay is the problem. Sure. We selected a comfortable level and soldered the shorted legs together. All the problems disappeared. Significantly. Some elevated treble remained, but it is not a problem any more. It is not too much, just very lightly more. Interestingly, on some disks the alps black won, on some disk the network. I heard the network to be better most. On orchestral music I heard better transparency and better treble.

This means I do not give up :) I try the Omron G6K relays. If it stays the same, I'll build it in. I hope by building it in, close to the tube it will improve further.

In the other hand, if would have started to solve somehow to remote control a rotary switch, I would be in better position with less efforts :), but I'm not good in mechanics at all.

Remote control is a must for us now, our 11 month old baby sleeps on music sometime and to get up, decrease volume and sit down back with her in my hands is not really comfortable, also we listen to music on a decent level, but we should decrease volume easier on high volume parts. That is why I do it all.