<<< Link to Introduction [doc (1)] 

Doc (12) of (13)

 

(12) Building a stepped attenuator.        

(Is this the ultimate form of volume control?)

Stepped attenuators using cheapo switches

If I remember rightly, CD players came onto the domestic market in 1983. At that time I owned a very expensive turntable, arm, moving coil stylus, and battery powered pre-amplifier; along with an amplifier and speakers. It was very quickly apparent to me that the way forward lay with CD’s, and a few years later, I sold the turntable etc and kept the pre-amplifier which had been supplied with various plug-in PCBs, including one for CD’s. Around 1986 or so I read an article in a hi-fi magazine by a chap called Martin Colloms, where he introduced the idea of modifying a traditional turntable pre-amplifier to act in a passive manner, and thus improve the sound quality from a CD player. Initially, I modified an old pre-amplifier, which I had lying around, and was astonished with the improvement over the grossly expensive pre-amp using its plug-in CD board.

Thereafter, I experimented with various types of potentiometer, in new and sexy looking cases, and finally ended up mounting two carbon track potentiometers on the top rear of the CD case, hard wired to the CD outputs and to inter-connecting cables; and hence plugging into the amplifier with only one air-break connection (plus of course, the carbon tracks and wipers).

I had compared various potentiometers, including wire-wound, silver wire-wound, carbon track, and the more recently available conductive plastic types. In line with my recent findings, the conductive plastic type was better for wear qualities than the carbon track type, but both of these seemed relatively noisy, in electrical terms. The wire wound type really was a non-starter all these years ago, due to cost and due to the step sizes on the switches being too large to allow an accurate left / right balance.

In early 2010, I noticed that an English company Hi-fi Collective, sold kits of parts for stepped attenuators, using variously priced switches and traditional resistors. I was about to buy one of their kits, to compliment my recently built first TP, but suddenly had the thought that SMT components with their small size, could be used to minimise noise created by using traditional resistors, assuming that their leads and connections act as a lot of little aerials (quote from V-bro discussing another matter on the Forum!).

I like to listen to all types of music but have classical music as my favourite, which is arguably, the most difficult to mimic in the domestic environment. For many years, I have been trying to achieve the ultimate domestic reproduction, admittedly, at some cost. I had been using a SACD player and large moving coil three way speakers, with a relatively powerful MOSFET amplifier. Hence any improvement over that set-up would have been graciously received! Along came the TP!

At this time V-bro posted on the Forum his almost delirious findings, following his construction and use of ladder type attenuators using traditional resistors, and I was duly impressed.

Also, a clever chap in the far-east, Korea I think, posted a photograph of his stepped attenuators on the Forum, showing SMT resistors soldered across switch terminals. That was it! I ordered up some of the cheaper 24 way switches, they have blue plastic on part of the body, for use as switched attenuators.

My first stepped attenuator

This was my first attempt, as shown above: I used logarithmic values which V-bro kindly posted on the Forum. Inadvertently, I produced series type attenuators; one for each channel. Also these switches were two-pole, and I used the easiest set of terminals, to solder on the resistors.

The results were incredible, and I have not used potentiometers again. The clarity and lack of noise were matched by extremely accurate left / right balance from no sound, up to deafening volumes. All types of potentiometer, which I had previously tried, were guilty of varying their left / right outputs in volume and in frequency! Although the music survives these tragic effects, these effects are very odd, and do affect my enjoyment. The precise balance in volume and in frequency control of the stepped SMT attenuators, were both a delight to behold!

On these switches I used 1206 SMT resistors, and found that the easiest way, to solder the resistors to the terminals, was to hold them with needle nosed tweezers, with the tweezers held horizontally across the middle of the resistors. That allowed the tweezers to slip down between the switch terminals, and hold the resistor firmly against the terminals during the first touch of the soldering iron to locate one side of the resistor. I then soldered the other side and went on to the next resistor, eventually, soldering all the joints again to produce super virtually “welded” sets of joints with decent amounts of solder. If the builder takes his time and adds some patience into the equation, then this soldering exercise is relatively easy to carry out, very pleasant and very rewarding.

 

However, the terminals immediately to each side of the bolts, which hold this type of switch together, are set a larger distance apart, compared to all the other terminal spacings. If I had been thinking sensibly and properly approached this difficulty, I would have ordered resistors in a, suitable jacket size for this larger gap. Far too sensible for me, I “battered” on (“breenged in” Scottish vernacular!) and created a solder bridge! The solder bridge has not broken in the last year or so. In any case, I am about to replace these attenuators using my favourite switches, noted below.

On these early attempts, I used 1% resistors and matched each pair of resistors using a digital multi-meter. On all my other attempts, six using the cheapo switches, I did not bother with this procedure, and was caught out only once when the highest value resistor was 5% out, and this was very apparent during listening tests! I have now returned to the matching process, for completeness; some would say, insane perfectionism! Actually to be fair in this matter, it is very pleasant being able to rotate the volume controls with such precise results. I also think that it is advisable to spend time selecting the resistors, to allow proper left / right hand balancing of the hi-fi system, from very low to very high music volumes, as described below.

Whilst posting on the Forum about my first success, I realised from a post by V-bro that these were series attenuators, not the ladder type. However, I thought that I would move on to the ladder type next, using both poles of the two pole cheapish switch, perhaps bending the terminals if found necessary. In the meantime, I had one of my electronic disasters! I managed to destroy a TP while testing, and the attenuators slipped into the background, while I repaired the TP or tried to!

During that trouble-shooting exercise, I was forced to alter the value of some of the resistors on the attenuators, and found these cheaper switches to be of poor quality, and would not recommend them for use over an extended period of time. The wafers are very thin, the contacts float in molten plastic if re-soldering is attempted, the contact is a pathetically small piece of spring metal, and they are bigger than the Elma switches which I now use, phew! The Elma switches have a solid action, are precise and feel vastly better in every way.

Having said all that, my TP1 has used these cheapo switches for 18 months or so, in daily use, without showing a “hitch”. Never the less, they are now very sloppy in action; it feels like you are stirring porridge with a cardboard pole, not butchering a haggis!

Attenuators using Elma switches

Stepped attenuators using Elma switches

Theoretically, I presume, that the smaller 0805 resistors used on these switches should have some sort of improvement over the larger 1206 type, as the ring of resistors and solder is considerably smaller and narrower. Back to V-bro’s little aerials!

Unfortunately, the geometry of the Elma switches, does not easily suit the use of SMT components as ladder type attenuators, due to the distance being quite large between the two poles. No doubt some “hero” or other will manage to get this to work, in some way or another. If you do, please post on the Forum! Actually, once I have finished my current hair-brained project, which will take some time to complete, I intend to try and build a balanced ladder attenuator using four pole twenty four position Elma switches, with SMT resistors.

Accordingly, I have decided to stick with the Elma switches as series type attenuators in the meantime. In fact, I suppose that I would have great difficulty in noticing the difference in the sound quality compared to the ladder type of attenuator, due to the additional resistors in the signal path using the series type. Indeed, the performance of these attenuators is way beyond any potentiometer in any case; and that ultimate side by side comparison (series versus ladder attenuators) will have to wait in the meantime. Dream on HB, and look away sound card users!

With careful judgement at the top, and the bottom of the log curve, the number of resistors in the signal path can be cut down considerably. Or these extra spaces could be used to lessen the difference in volume between each switch position.

I used two mono 24 way switches, with a shortened version of 10K log, using half the log length, thus doubling the sensitivity of the attenuator, which gave very small changes in the perceived volume between switch positions. This allows for reasonably precise setting of the stereo balance.

Firstly, the 10k log curve was calibrated using a normal potentiometer, I set the music to the desired level (both highest and lowest levels), and measured the values of resistance at these chosen points on the test potentiometer. This was easily done, by using a numbered knob, noted the value on the knob relative to the zero marking; then switched off, disconnected and got the old multi-meter out. Effectively, I had now produced a suitable 24 position attenuator which had the usable range selected from a 48 position attenuator!

The easiest way, to solder the 0805 resistors onto the terminals of this switch, was to use Audiophool’s "sticky wand", using a 1.2mm chisel tip on my soldering iron. Again, taking my time and being patient, I developed a successful technique, and I enjoyed the process. I have now built six of these attenuators on the Elma switches, and my soldering technique has improved considerably during this work. Consequently, the soldering of 0805 SMT components onto the TP board, now seems straightforward!

In the above photograph, the resistors appear to be badly placed on the switch terminals. However, there is a problem in soldering the SMT resistors onto the terminals when approaching either of the two nuts and bolts which hold the entire switch together. The terminals near these bolts require to be carefully bent away from the bolts in a gradual manner, in order to get the resistors past the bolts.

Was it worth all this work? The results were phenomenal, and it is difficult to imagine how any other volume control could begin to compete with such quality, except the ladder type using SMT components.

A point to watch, and take care with, is the soldered connections of the input wiring to the attenuator. My tip is to leave a space on the appropriate terminals to allow threading through of these tiny wires prior to soldering. Initially, I soldered all the resistors and then used liquid flux and a slightly hotter than normal iron temperature to clear the appropriate terminals of most of the solder, along with careful use of solder wick (that wonderful stuff!) before threading the tiny cable through the terminals, and soldering.

Another tip is to make sure that the cables connected to the ends of the trail of resistors, do not put physical strain on these end terminals. The low end of the trail where there are two wires connected to the switch terminal (ground) is particularly at risk. I had de-soldered a couple of attenuators and replaced them with others, with different ranges, and the cables had become heavily saturated in solder. Inadvertently, I had created a downwards pull on the terminal. I tested the attenuators on their full range with a multi-meter, installed them in the case, and on playing music, one of the attenuators produced full volume, deafening and frightening! Confusingly, it passed the multi-meter test again? However, when I was fiddling around re-testing every single joint, I noticed that when I pressed more firmly on the last contact, the meter read full insulation, not the 470ohms I expected. I inspected the two joints under high magnification in strong light, and found that one joint did in fact move, apart by a tiny amount, when gently pressed; how irritating.

OK, I had soldered them together perfectly, so I used my standard test: Set the solder station to 400degC, touched one end of the resistor, and watch if it would come off, onto the soldering iron. Yes it did, and the other joint now showed the end cap of the resistor embedded in the soldered joint. Thus, whilst the switch was not turned, the joint measured OK. When the switch was moved, the broken joint opened: the two ground wires were still soldered solidly together, but not to the terminal; hence the only components in the signal path were the resistor trail. Obvious, once understood.

I would heartily recommend the use of Elma switches with SMT components.

I don’t think I have missed and hit the wall!

 

A short break!

Britney in a necklace

Wow, she is a beauty!

Come on, no more slacking, back to electronics.

Selector switch

For the selector switch, I used an Elma 2 pole 6 way switch, cut down to 3 ways using the Elma pin system. For simplicity, I now connect all the input grounds together on the back of the RCA sockets, as per the suggestion of “tuo” on the Forum; a very knowledgeable and helpful chap.

 

Imbalance of sound between the left and right channels whilst listening to music

 

Every TP I have built, and indeed every amplifier I have used, has always had a small imbalance of power output between the left and right channels. This is not unexpected as there are a lot of components used in amplifier circuits, and unless every component is very accurately matched this situation is inevitable. Also, there are plenty of other variables in the hi-fi chain which could be out of balance and also affect the overall system balance from left to right. The source of this imbalance can easily be determined by interchanging each set of electronic devices in the hi-fi chain and comparing them, whilst trying to hear if the listening room is affecting this balance.

 

The easy way to solve this problem, in an overall sense, is to adjust the balance from left to right using one of the attenuators, or so I thought. In practice, the imbalance never coincides with exactly one step of my attenuators. Returning to the use of potentiometers as an infinitely variable type of attenuator was not an option for me, as I find that they are notoriously inaccurate in left to right balance at different sound levels, and at different frequencies; sometimes making a complete mockery of the music in comparison with a stepped attenuator.

 

I could have added a potentiometer, on one channel, after the stepped attenuator, and used it for this balancing procedure. However, in my almost religious determination to minimise the “component count” which the signal has to use or pass through, I did not want to add any unnecessary components to the hi-fi chain, particularly a potentiometer!

 

Continuing, the only reliable way I have found to determine a value for this imbalance, is to connect everything up in its final (listening) position, and measure it by listening tests; recording how many steps, or part of a step, it takes on my stepped attenuators to bring everything into perfect balance from left to right.

 

As each step on my attenuators is on a logarithmic curve to make each step sound louder in a linear manner, I took the liberty of ignoring this theoretical fact, and decided to make a value judgement (guess!) at the appropriate position on one of my attenuators, which would balance the sound.

 

 

My first method of resolving the balance problem was to alter the resistors at each end of my 24 step attenuators. This is fine in theory, but is more than a slight “pain in the tonsils” in practical terms, as the alterations to the attenuators are very difficult in situ, inside a case; the tiny wires are connected to the same terminals as the resistors at each end of the attenuator. Therefore, it is necessary to de-solder the wires, then the resistors, and repeat these actions in reverse order.

 

 

My current solution to this problem: I then investigated the possibility of altering either the modulator feedback resistors, or the input feedback resistors on the TP PCB. The modulator feedback resistors are easier to access on the finished TP than the input feedback resistors, but it is necessary to replace at least twice the number of modulator feedback resistors, than the input feedback resistors. I therefore opted to replace the input feedback resistors. As they control the input gain directly in proportion to the input resistor, it seemed to me that one simple change could solve all matters of imbalance, with respect to my empirical in-situ tests at the listening position. As SMT 0805 are used for these resistors, it makes it a relatively simple and safe job to replace the resistors.

 

The chip and associated 0805 size SMT components

       

On my first attempt using this method, I estimated the overall imbalance to be around 0.9 x one step on my attenuators. I had used 33k resistors for the input feedback resistors, and then ordered up 100 of these tiny guys at 5% accuracy, hoping to find two with a difference of:

 

(0.9 x 1/48 x 33k)  =  0.62k, remember my attenuator selects 24 steps out of a 48 step attenuator. I know a direct proportion of 33k log is never going to be exactly correct; please read on!

 

Thus, I needed a resistor of value =  (33 – 0.62)k = 32.38k

 

On testing all one hundred resistors, I found the highest to be exactly 33k, and the lowest to be 32.6k (a difference of only 0.4k), with 80% measuring 32.8; not the difference I needed. I expected at least one resistor out of the hundred I had bought, would go down to at least minus 2.5%, half the inaccuracy quoted at +/-5%; how irritating!

 

Ie (0.025 x 33k)  =  0.825k, equivalent to minus 2.5%.

 

In the meantime, I soldered these wee guys onto the PCB, and was pleased to see that my rough calculation seemed to provide the expected result, but still not accurate enough to stop the whole matter still being very annoying! Now I will order more 33k at 5% resistors, and try again, since I have at least proved that this seems to be the best route to take, to satisfy my listening conditions. The alternative would be to rake around the web, and try to find the desired resistor in an unusual and rare value; a process with which I can not be bothered!

 

If I was in the business of building and selling superb quality amplifiers such as the TP, which I am not (and never intend to be!), I would use this method to very accurately balance each amplifier’s outputs. Then it would be up to the customer to resolve all his own balance issues within a system; at least he would start with a very accurately balanced amplifier.

 

OK, I am a confirmed perfectionist and lunatic! I just want the ultimate in mimicked sound in my listening room.

 

In my next document, I will delve into the “ins and outs” of power supplies in my usual cavalier manner, and will probably annoy quite a few people who hold cherished semi-religious beliefs in these matters. Tough, “take it on the chin” guys!

 

Haggis Basher.

Appendix 1: HB’s list of resistor values.

I have calibrated these values to work for my set-up, using approximately +/-60Vrail on the TP in simple stereo into 8 Ohm speakers. These are large speakers at nominal power handling of 300W each. The greatest signal input from some electronic music (not playing too loud) at position 2, up to position 23 for very low signal strength classical music. At least these values should give the novice a good starting point, if not the final answer.

My SMD resistor values for a series switched attenuator, 24 positions

 

24 way switch position.

 

Theoretical

 value of

log curve.

 

0805 / 1206 Size Resistors

 

 

 

 

cumulative

difference

Each

Resistor.

1

-

-

-

-

2

445

470

+25

470

3

499

503

+4

33

4

560

559

-1

56

5

634

634

0

75

6

716

716

0

82

7

806

798

-8

82

8

897

898

+1

100

9

1k

998

-2

100

10

1.12k

1.12k

0

120

11

1.27k

1.27k

0

150

12

1.42k

1.42k

0

150

13

1.58k

1.57k

-10

150

14

1.77k

1.77k

0

200

15

2k

1.99k

-10

220

16

2.23k

2.23k

0

240

17

2.49k

2.53k

+40

300

18

2.8k

2.83k

+30

300

19

3.16k

3.13k

-30

300

20

3.525k

3.52k

-5

390

21

3.93k

3.91k

-20

390

22

4.44k

4.42k

-20

510

23

5.06k

5.1k

+40

680

24

10k

10.2k

+200

5.1k

Three steps have been omitted between step 23 and step 24; and all the steps below step 2.

 

 

Appendix 2: Resistance values for a full range series switched attenuator, 24 positions

I have included this table for those of you who wish to carry out your own trials; as I find this kind of experimentation intriguing, and it may be infectious!

 

24 way switch position.

 

Theoretical

 value of

log curve.

 

0805 / 1206 Size Resistors

 

 

 

 

cumulative

difference

Each

Resistor

1

0

0

0

0

2

8.5

10

+1.5

10

3

22

22

0

12

4

40

37

-3

15

5

68

70

+2

33

6

100

103

+3

33

7

145

150

+5

47

8

195

197

+2

47

9

252

253

+1

56

10

320

221

+1

68

11

398

403

+4

82

12

499

503

+4

100

13

634

623

-11

120

14

806

803

-3

180

15

1k

1k

+3

200

16

1.27k

1.27k

0

270

17

1.58k

1.57k

-10

300

18

2k

1.96k

-40

390

19

2.49k

2.47k

-20

510

20

3.16k

3.15k

-10

680

21

3.93k

3.9k

-30

750

22

5.06k

5.1k

+40

1.2k

23

7.15k

7.1k

-50

2k

24

10k

10.1k

+100

3k

 

 

Tributes: I am indebted to my mentors on the 41Hz Forum, for all the kind help, and the general support and encouragement they gave me, during construction of my TPs. Also, I am indebted to my mentors on the 41Hz Forum for their generous help in editing all my documents, for accuracy and formatting, without which it would not have been possible, to produce these documents.

Opinions: The opinions expressed and any conclusions reached in this document, are mine and mine alone, except when I have quoted someone directly; I am an individual with no commercial connections to anyone or any company.

Disclaimer: I will not in any way be held responsible for any damage to any audio equipment or for any damage to any other electrical equipment or for any damage to any other property or for injury to any person or persons. I will not in any way be held responsible for what you do with electric power. Electrical power can be lethal! If you are not professionally qualified to handle mains power, then do not! Get help!

  

<<< Link to document (11)                                 Link to document (13) >>>