Reply
 
Thread Tools Display Modes
  #1   Report Post  
Mikkel C. Simonsen
 
Posts: n/a
Default Attenuator experiments

There were some attenuator discussions here some time ago.

I built a test attenuator using relays. I use 8 relays, so I get a range
of 48dB. The attenuation from 0 to -31dB is in 1dB steps, and after that
the steps are -33, -34, -36, -39, -42, -48dB and off - 39 steps from
full on to off.

I just used E12 resistors, so the actual steps are not that accurate...

My conclusion so far is that 39 steps are enough, but I think 12 steps
are too little and 24 steps is less than I would like also. But as the
amp I use (2 6SN7 gain stages into a 2A3) is much too sensitive, so I
find that I actually need more steps in the low end. I guess I should
just add a 20dB attenuation on the input of the attenuator - or change
the input stage on the amp to lower the gain... Or perhaps a 7-relay
attenuator with a switchable 20- or 30dB pre-attenuator would be a good
idea?

I have read many places that make before break switches are essential,
and that break before make switches cause clicks and other noise. The
relay switch I use is both I guess, as the relays don't have the exact
same switch time. But I get no noises at all - I can adjust the volume
setting with no problems and without muting. Are the clicks only a
problem on direct coupled SS amps?

If anybody's interested the schematic can be seen he
http://stiftsbogtrykkeriet.dk/~mcs/8bitatt.gif

And the mono prototype can be seen here (behind the big, grey monster):
http://stiftsbogtrykkeriet.dk/~mcs/00000005.jpg

Best regards,

Mikkel C. Simonsen
  #2   Report Post  
Mikkel C. Simonsen
 
Posts: n/a
Default

Kirk Patton wrote:

Hi Mikkel,

I really liked your elegant approach to a series (pot-emulating) design . .


Thanks.

I have built, tested, and revised a number of attenuator approaches in the
past year or two, all have been part of an active line-stage. They have
been located in between an active input buffer and an active output stage.
Here are my thoughts . . .

I built a test attenuator using relays. I use 8 relays, so I get a range
of 48dB. The attenuation from 0 to -31dB is in 1dB steps, and after that
the steps are -33, -34, -36, -39, -42, -48dB and off - 39 steps from
full on to off.


Of the basic question is whether or not the attenuator should support the
same step throughout its entire range, I think that in practical purposes
this can be determined by whether or not balance control (by the same
attenuator) is desireable.


I may add a balance control. Since I have two attenuators that can be
controlled individually, adding a balance control would be easy.

If this is the case, then the attenuator steps must of course be
consistant, otherwise any introduced offset between the two attenuators
for balance adjustment will not track correctly across the range.


My plan is to aim for 1dB steps accross the useable area.

If not, then I find that a taper that gives a finer adjustment in the
"money range" than in the extremes is more user-friendly. I personally
like a minimum of fuss when actually using the system, and having to
switch volume control ranges really annoys me.


Any range switching would be done automatically. Right now I just use a
+ and - button. The attenuator is of course linear, so the taper is done
using a lookup table in software. That table could also contain extra
bits for switching ranges. And using a special taper with 0.5dB steps in
some areas and 2dB steps in others could also be done.

I recently built a mechanical 47-position attenuator with the following
curve: -1,-4,-7,-10,-12,-14,-16,-18,-20, -21 . . . (-1dB steps
to -50),-52,-54,-56,-59,-62,-66,-70,-74. The associated active linestage
has about 15dB fixed gain. I didn't include a 0dB setting or an "off"
setting; the former is for impedance reasons and the latter is because when
I don't want to hear it at all, I prefer to push "stop" or lift up the
needle rather than turn the volume all the way down.


With the design I've chosen the "off" step comes for free. That step can
be used for muting (which I often need during the day when the phone
rings) - the attenuator can very easily be returned to the previous
position when the mute button is pressed again. I don't have to turn a
knob after all ;-)

But thanks for the values above. I have been wondering what attenuation
levels are actually used.

I find this gives me a single knob-twist that works well from low
background "dinner-level" to rip-roaring loud, and gives me fine
adjustment in the range where I do my most focused listening. I'm not
using any balance control. My amplifer is moderate-to-lower input
sensitivity (rated 2.5v for 500w output),


500W output? I get by just fine with 3W...

and speakers are highish sensitivity (I think rated 92dB/2.83v/1M).

I just used E12 resistors, so the actual steps are not that accurate...

My conclusion so far is that 39 steps are enough, but I think 12 steps
are too little and 24 steps is less than I would like also. But as the
amp I use (2 6SN7 gain stages into a 2A3) is much too sensitive, so I
find that I actually need more steps in the low end. I guess I should
just add a 20dB attenuation on the input of the attenuator - or change
the input stage on the amp to lower the gain... Or perhaps a 7-relay
attenuator with a switchable 20- or 30dB pre-attenuator would be a good
idea?


I think that this is mainly an impedance/gain structure question. If the
attenuator is in the traditional position in an active-linestage (at the
preamp input) then using a fixed attenuator pad at the power amp input will
probably help S/N at the cost of preamp headroom (which with a sensitive amp
shouldn't be an issue). If, however, this is an all-passsive setup, then
putting in a fixed series resistance before the attenuator and lowering the
attenuator's overall impedance will make the range more usable, and lower
the attenuator's output impedance (good).


Right now my "linestage" is a small box with a rotary input selector, so
it's quite passive...

Given that you have competency in designing the control electronics and are
thus freeing yourself of the limitations of a fixed number of switch
positions, you could simply add more bits (relays) to your existing design
and have as much range and resolution as you need, and simply set the
practical operating range in the control system.


I did some calculations. It seems I could quite a good range with 7
relays and a fixed 25dB attenuator switched by the 8th relay. 0 to -50dB
in 1dB steps and 67dB max. attenuation should be possible.

This would add only minimal cost (unless this is a production design . . .)


The cost isn't that much of a problem. The relays I currently use cost
about $1 each, and the control chips cost even less. But I guess keeping
the number of relays as low as possible would be a good idea.
Production? I don't think so, but you never know... :-)

I have used Hamlin SIP reed relays (available from Digi-Key) for both
attenuators and source-switching with excellent results -- they are
really mechanically quiet as well.


How much do they cost? Reed relays are normally expensive, and only
available in single-pole versions. The relays I use now (made by
Zettler) aren't quiet, but the noise isn't much of a problem (to mee
anyway).

Anyway, hope this gives a few ideas.


It does, thanks for your reply.

Best regards,

Mikkel C. Simonsen
Reply
Thread Tools
Display Modes

Posting Rules

Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT +1. The time now is 02:42 AM.

Powered by: vBulletin
Copyright ©2000 - 2024, Jelsoft Enterprises Ltd.
Copyright ©2004-2024 AudioBanter.com.
The comments are property of their posters.
 

About Us

"It's about Audio and hi-fi"