Simple Attenuators - Design And Testing

Discussion in 'The Workbench' started by JohnH, Oct 21, 2017.

  1. RickyLee

    RickyLee Well-Known Member VIP Member

    Joined:
    May 22, 2010
    Messages:
    10,242
    Likes Received:
    3,655
    Location:
    SoCal U.S.A.
    This one you have shown here is a bit closer to one I built many years ago. But this one I built is larger and has two rheostats/controls on it with its array of wire wound resistors.

    I had settings/formulas worked out for decibel level adjusts as to where the two controls were ti be set. But all the info is on an old computer. Funny as I seen this thread as I have been thinking about digging that attenuator box out and using it in conjunction with these Eminence FDM speakers I now have . . . .
     
    Dwayne Eash likes this.
  2. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    I looked up abouf those FDM's. A very clever idea based on dialing the distance between magnet and voice coil, to give up to -9db reduction. It would be interesting to know how the results compare to your switched attenuator, at least for the sake of science.

    I was using my rebuilt box to do a bit of practising at whisper volumes before work this morning, while family was sleeping. It sounded fine for what it was doing, and it helped me start my day. Loud testing will be next weekend.
     
    Stephen H and Dwayne Eash like this.
  3. RickyLee

    RickyLee Well-Known Member VIP Member

    Joined:
    May 22, 2010
    Messages:
    10,242
    Likes Received:
    3,655
    Location:
    SoCal U.S.A.
    The last time I had my home made attenuator box out, I noticed the two controls had been changed. I am now curious to tinker with it, but I am wondering if it really matters for harms sake to the amp, what the ohms readings should be on the IN and OUT of this attenuator box.

    If I have a close ohms reading at the IN (will use 8 ohms here, this is where the amplifiers out will be plugged in) with a speaker plugged into OUT ( 8 ohm speaker), then it should be safe to use?

    I do remember there being well over 100W of wire wound resistors in a series/parallel/series type configuration inside this thing.
     
  4. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    I guess if the amp sees 8ohms then it's happy. Without knowing the design, can't tell how much total power it can take since Im guessing some resistors take more than others. Maybe watch for how hot they get.
     
  5. tmingle

    tmingle Well-Known Member

    Joined:
    Jan 24, 2008
    Messages:
    451
    Likes Received:
    468
    Location:
    Mifflintown, Pa.
    I know its been a while. I think the low volume thing has more to do with how your ears perceive the volume(Fletcher-Munson?) Ive mic'ed my amp & recorderdit at 4 different volumes. I then normalized the the volumes in Reaper to compare. It's closer than I thought it would be. Unfortunately, I can't find the clips right now to post.
     
  6. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    I've been trying some thermal tests on the new box, to see what happens. I haven't brought the IR thermometer home yet though for a proper measurement though.

    I'm running a 9V dc power pack into the amp input, with no added load. So the attenuator has a resistance of 9 Ohms on its own. I'm getting 8.5V at the jack, so power = 8.5^2/9 = 8W.

    It has definitely warmed up, but has taken about a half hour to get to a steady state. Its not hot though, not nearly as hot as the front plate on some amps can get. I guessing maybe a 20 degree C rise. but Ill test again properly.

    That power, as a constant dc, would equate to an average dissipation from an amp over several minutes and a max dissipation of considerably more than. Physics says that doubling the power input should less than double the temperature rise, so I'm expecting that this box could soak up a flat out 40W amp, in practice, without adding much more than say 50C rise. TBC.
     
    Last edited: Apr 16, 2018
  7. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    I made a short sound clip, to find out how significant any tonal change is, at -9db.

    Its a low gain setting on the VM, bridge humbucker, played through a looper, with about 5 body and 7 detail. Volume is at 6. Recording is miced with a Rode M1. First without the attenuator, then with it and the recording level normalised later to match volume.

    Actually I cant hear a difference in tone, but the room volume is very much less at this level of reduction:

    https://vocaroo.com/i/s05Qix2IkSdw

    Its not a great recorded tone for this amp, I usually like it thicker. The idea wasn't about absolute tone, but to try to reveal differences particularly in highs, due to the attenuation.
     
  8. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    Who likes numbers? Here's some more, based on better temperature measurements using an Infrared thermometer from the engineering school where I teach.

    I ran dc power into the attenuator with measured voltage 8.55 V and measured resistance 9.2 Ohm. So power in was 7.9W

    I ran for 30 minutes, at which time it had slowly reached a steady state with 35 C max on the top plate (about 33 C average) and average about 25C at the sides. Ambient in the room was 16.5 C.

    That's as much dc as I can put through with what I have to hand. But based on these numbers, I did some maths with convection and radiation and I reckon that with an average of about 25W into it, it would get to about 55C. The hotter these things get, the better the mechanisms of passive cooling work. Given that this is an average value over at least 10 to 15 minutes, and it could easily absorb several times that for a shorter time, then that should be more than adequate for a 40 to 50W amp, given that any practical use has varying amplitude levels and gaps in the playing.

    So I think this little unit is brick-solid electrically, thermally and mechanically, with no switches or moving parts, plus it sounds good. For a few $ I have a 5W Vintage Modern when I need it, running with its awesome 2x12 G12c speakers.

    But Im sure the 1x10's in your 5w combos sound just fine! :)
     
    Last edited: Apr 18, 2018
  9. iron broadsword

    iron broadsword Well-Known Member

    Joined:
    Oct 27, 2011
    Messages:
    2,983
    Likes Received:
    1,475
    Great unit! I built a similar one from this pdf: http://www.hearditontheweb.com/images/pdf/attenuator.pdf

    Two stages, -6db each. Used it for a while with my 900 and then got a weber mass so I could go even further. Now this guy is setup on my mesa which I use for recording. It's handy as hell.. no tone loss and 12db is enough to make a difference in pushing the output tubes past cold lifeless into hot and tasty. The box I used is pretty similar to yours and I also drilled it full of holes for heat dissipation. People talk a lot of shit about these types of attenuators but if all you do is 12db you'll never notice a loss in tone.
     
    JohnH and _ripper_ like this.
  10. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    That's a good arrangement, and it made me do some more tests yesterday with more attenuation, going from -9db to -12db. There's a very important thing to know about tone and resistive attenuators:

    Assuming the load on the amp is correct, then the resistance seen by the speaker has a huge effect on tone, due to the speakers inductance. The two stage one above has this as about 4 Ohms.

    We can use this in the design of the resistor network to control the tendency to lose high treble. If the speaker sees a higher output impedance, the effect is a bit more high end. I took my basic -9db L-pad and added another 4 ohms in series with the speaker. This took it down to -12db, and also lifted the high end, compensating for slight loses Id heard before. I did some modelling and it shows this too. The speaker is seeing about 6.5 Ohms. However, if you do -12db with a simple two resistor L-pad, then the speaker sees about 2 Ohms and it starts to sound dull. Within a range, you can design the resistor network to pick what output impedance you want for a given attenuation, and use this to balance the tone.
     
    Stephen H and Nik Henville like this.
  11. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    I'm still working of this, exploring options and design paarmeters:

    Resistive attenuator designs

    As described above, I’ve been building and testing quite a few purely resistive attenuators, for use on my Marshall Vintage Modern 2266c. This is a 50W amp with a pair of 12” Greenback speakers for a nominal 8 Ohm speaker load. Its partly because I want one for use, and partly to learn by testing, and wishing to start simple.

    The attenuators are very basic, just a single attenuation value with no switching. Typically, there are three resistors in a ‘star’ configuration. This allows me to pick values to target attenuation db, plus both input and output impedance. I maintain the nominal input impedance as seen by the amp at close to 8 ohms, including the speakers. I’ve most recently been targeting around -12db basic attenuation, bringing a 40W amp down to 2.5W for home use.

    Here are three resistor layouts that all do -12db with around 8 Ohms seen by the amp.

    [​IMG]

    Type A on the left, is a simple L-pad and it sounds noticeably dull. Type B in the middle sounds quite good, its clearer with much nearer unattenuated tones. Type C gets brighter again, see analysis below. The main difference between them is the output impedance as seen by the speaker. Based on nominal resistance values, speaker in Type A sees 2 Ohms, B sees 6.6 Ohms and C sees about 26 Ohms. Type C is an extreme example, I tested it and it is a bit bright and harsh, but A and B are quite feasible.

    I’ve built Type B into a box and have been using it all volume settings. Comparison of recordings attenuated and not attenuated, using a looped clip then miced and normalised for equal volume show negligible audible tone difference (though a small response difference is measurable).

    So what puzzled me is that I'm getting this good result by having quite a high output impedance from the attenuator, so far as I understand, 6.6 Ohms is much greater than most amps output Z.

    Amp with reactive load: response

    Aitkens page below shows how the electrical response of an amp varies with frequency when driving a reactive load.

    http://www.aikenamps.com/index.php/parallel-attenuator-loads

    This could be a real speaker, or a load box designed to mimic it. His page is focusing on how this changes if a further resistive load is in parallel, but Im just looking at the full response curves in green. For amps with or without feedback loops, the response varies in a similar fashion to that expected from the speakers reactive impedance. For an amp with a feedback loop (such as mine), there is a lift in response of around 3db from 500hz up to 5khz due to speaker inductance, and a boost in bass due to resonance, as shown here:

    [​IMG]

    Clearly the real amp and reactive speaker load are interacting intimately to make this response.

    With a resistive attenuators at -12db, amp and reactive speaker load are almost completely separated. The amp is seeing almost a totally resistive load and the speaker sees just a resistive output Z. About 15/16ths of the amp power is going into resistors in this case.

    But Spice analysis of this situation shows how the reactive impedance of the speaker could interact with a pure resistive output Z to get very close to unattenuated frequency response, at an attenuated output level.

    I used the reactive load box design from Aitken, scaled to an 8 Ohm version, to represent a real speaker, and tested the calculated frequency response, in terms of signal seen by the load, within the range of output impedances from 2 to 26 Ohms, as for designs A to C above:

    [​IMG]


    Note: in the Spice program I use, I can’t vary more than 2 components at once, so the plots below show frequency response with varying amp output impedance Z but without controlling overall attenuation level – but it’s the relative shapes vs output Z that are important.

    [​IMG]

    The plots show the signal seen by the speaker, with output Z from the attenuator varying from 2 to 26 Ohms.

    The arrangement Type B that I built, with an output impedance of 6.6 Ohms, is very close to the red curve on that graph, but at -12db. It shows just under a 3db rise in treble from 500Hz to 5kHz, ie very close to the full amp/reactive load curve from Aitken. I reckon that helps to explain why it is sounding pretty good!

    As the output Z increases further, about 3.5 more db of high and low relative boost is available.

    This seems like a helpful idea, ie controlling Z to make simple attenuators sound more like an unattenuated amp at least in terms of basic frequency response. So what’s missing? I can’t see any undue risks, since resistor attenuators can be built very robustly and are fine for loading a valve amp. At moderate distortion levels, I think the match is reasonably valid, but at high power-amp distortion, there could be tonal differences. The tubes will still be saturating as they drive into the resistive load, and this will come through in the tone, but Id guess that some of the more subtle interactions of tube/OT/real speaker will be suppressed by the resistances between them.

    -12db Attenuator build

    This is the current -12db build, based on the Type B design. The 5 ohm resistor is 2x10 in parallel, and those and the grounded resistor each dissipate around 12W based on a 40W amp output. They are oversized at 100W rating. The output series resistor is a 10W wire-wound, but it only dissipates 1.25W, so its all overbuilt.

    [​IMG]

    [​IMG]
     
    Last edited: Apr 28, 2018
    Nik Henville likes this.
  12. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    Not sure who's watching but I'm pretty excited. I just took this to a new level (lower), with a switch to go from -12db to -18db. And it sounds clear and full and I was just cranking the VM, maxed out in HDR mode, at about 0.8 of a Watt!

    I've realised that the trick to getting good tone with high attenuation is to keep the speaker seeing a reasonably high output resistance, and a value of about 7 or 8 ohms is about the sweet spot for an 8 ohm speaker. Higher output resistance is brighter, lower is less bright. It's all to do with letting the natural reactance and resonance of the speaker develop, without too much damping (see discussion above).

    Here is this weeks schematic:

    [​IMG]

    I ganged two resistors for R1/R2, since this pair takes most of the power. Also, there is a 4Ohm and a 3.9Ohm shown at R3 and R5. There's nothing critical about that slight difference, its just that the big Chinese power resistors come in slightly non-standard values.

    Assuming 50W total amp power, then the red resistors dissipate about 15W each,the green ones handle about 1.5W each and the blue dissipates less than 0.5W. I used 100w, 10W and 5W types, so it's well over-speced. The switch is a mini-toggle switch, DPDT with the two sides ganged in parallel to double current capacity. But it never sees the full amp power, so it should have no issues.

    There's a couple of extra 'Easter Eggs' with this design:

    1. Provided the amp is used with its 8 Ohm tap, you can safely plug any Ohms of speaker into it. There is enough resistor separation from amp to speaker that the amp always sees close to 8 Ohms, within about 1/3 of an Ohm.

    2. My toggle switch is actually an on-off-on, so in the centre position, it is disconnected. This leaves the speaker seeing about 15 Ohms, a little brighter tone but also perfect for running a 16 Ohm cab at an attenuation of about -15db.

    Here it is as built today:

    [​IMG]

    This is a pretty handy and compact box as it is now, just right for home, bringing a 50W amp down to about 3W or 0.8W (-12db or -18db). With a bigger enclosure, a useful development could be to add a -6db option, and maybe a full bypass switch. A neat version with all of that would be a two stage attenuator, with one stage calculated for -6db and the other for -12db, each with an on/off switch. Then four settings would be possible, 0db, -6db, -12db and -18db.
     
    Last edited: May 6, 2018
    Nik Henville and _ripper_ like this.
  13. gearhead

    gearhead Member

    Joined:
    Mar 24, 2013
    Messages:
    43
    Likes Received:
    17
    Location:
    Dallas, TX
    John, the pic you posted with the large power resistors looks like my dummy load for testing amplifier output. I have (2) 100 watt, 8 ohm resistors mounted to an aluminum heatsink with an 80mm fan on it. and have a cable with a 1/4" mono plug and alligator clips so I can connect one or both resistors for 4, 8 or 16 ohm load. I am only using it to look at sine wave clipping, voltage and form at power. Now it's a purely resistive load. Speakers are a reactive load, meaning that the resistance changes at certain frequencies (usually rated minimum ohms @ the resonance frequency of the speaker), due to the capacitance of the voice coil winding (it's an R-C circuit). An 8 ohm speaker is not 8 ohm ALL of the time. I've never tried to measure the impedance at different frequencies, that may be an interesting experiment one day.
    Not sure I would want to run one of my expensive tube amps at close to full volume on my dummy load. But since you're connecting speakers with yours, then they give it some capacitance. Should be fine for what you are doing.
     
  14. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    A bit more:

    I tested the approximate output resistance of the VM this week. As I have recently come to expect, its quite high and I was getting 16 to 20 Ohms with the 8 Ohm tap. It's is consistent with a valve output stage being close to a constant current source, until it clips. I did this test by loading the amp with the attenuator box and no speaker, which using the switch, gives about 9 or 8.5Ohms. A smallish signal was fed in, not wanting to push the amp in this situation. With the lower ohms load, the output voltage drops almost in proportion, and back figuring the implied voltage divider of amp output impedance and load, lead to these values.

    This highish measured output Z is further evidence that if you want to make a simple passive/resistive attenuator sound half decent, you have to keep the output impedance up high, so the speaker sees something more like the original amp output impedance and is not damped by low value resistors. Doing this, gets the design to be capable of matching the full tone, at least for signals in the clean range.

    This is this weeks schematic:

    [​IMG]

    R7 was added, taking the attenuation values down another 3 db to -21db max, and brightening the tone with a 13 to 14 Ohm output resistance.

    Inside:

    [​IMG]

    I tested again, a looped signal, the VM fully cranked in HDR, volume at 8 and close miced. Jeez that's loud!, even with industrial ear protection. Then attenuated -21db, and its just enjoyably loud. But the frequency response and audible tones are very similar. The traces do vary a little (probably due to the speakers being driven or not) but there's no overall trend of losses of highs or lows and once recordings are matched for volume, they sound the same to me. I did another run with the presence cranked up, and that was a much more significant change in tone.
     
    Nik Henville likes this.
  15. 1199RS

    1199RS New Member

    Joined:
    Jul 29, 2015
    Messages:
    17
    Likes Received:
    5
    Location:
    Dundee Scotland uk
    Hi john have been following this thread as it is very interesting reading. I have been using a hotplate for a while now and when pushing master volume i get two tubes red-plating so i made a resistive load box from 4 68ohm 100w resistors all wired in parallel. Then into a di box and into the front end of a jubilee set clean. This has eliminated the red plating but there is some ground loop buzz as i am not using an isolating transformer. Your little schematic looks like a far better and easier option. I have been playing with some values as i think i need to drop some more db as it is for a 100w plexi. I thought the 2nd voltage divider could change to 12 ohms to drop more voltage. My circuit calc app has output watts as 0.7 with it. Just curious as to how that will affect what the speaker sees and tonal implications.
    Cant physically try it as i am currently stuck offshore due to fog
    Thanks for the read
     
    JohnH likes this.
  16. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    Thanks for your interest. I don't know about the red-plating, generally that wouldn't be fixed by load boxes or attenuators, which encourage you to drive the amp harder.

    But with this resistive attenuator scheme, Ive got spreadsheets happening to work them out to any level of db reduction. I'm on a roll, so I'm going to build another one, starting again. This will be 3 stage. A fixed -7db stage, then two switchable stages of around -7db and -13 db. That will give four settings in combination of -7db, -14db, -20db and -27db. based on a 50W amp, this will give power levels of 10W (rehearsal), 2.5W (loud home use), 0.5W (quiet home use) and 0.1W (late night). Im not building in a bypass, 0db switch. There's no point since if this is needed, you just don't use the attenuator. There will be 2x output jacks, so a pair of 16Ohm speakers can be used.

    In all of them, the design needs to keep the ohms seen by the amp reasonably close to the spec. I'm allowing around 7.5 to 9 Ohms for the 8 ohm tap. Also, the output resistance needs to be kept up highish. Based on some tests and calcs, I'm targeting 18 to 20 Ohms output resistance in all settings for this new build, into an 8 Ohm speaker, to let the natural treble and bass resonance from the speaker develop.

    Based on the design above that you saw, dropping the 18ohm to 12 will reduce power, at the expense of output resistance also dropping so it may sound a bit duller. Better to recalculate all value to suit. If you are interested to see what might be needed, give me some power or db values that you'd like to achieve with your amp and Ill run some numbers. Also what ohms cab would you use?

    Note that my build is quite compact, and with the case-mounted power resistors and ventilation, Its fine for a 40-50W amp. but I think that for 100W, although the components may still be ok on paper, more serious cooling is needed which could mean a more spacious box with better ventilation or heatsinks.
     
    Last edited: Jun 1, 2018
  17. 1199RS

    1199RS New Member

    Joined:
    Jul 29, 2015
    Messages:
    17
    Likes Received:
    5
    Location:
    Dundee Scotland uk
    Hi John thanks for speedy reply. I usually practice around 85db at night and during the day up to 100 is ok for the neighbors My amp is putting out around 120 or so db at full tilt but usually have master around 6 or 7. Using an 8ohm cab but have a standard 1960 as well.

    I have some designs for an aluminium heat sink box to get rid of excess heat. Your box looks around 5" or so in length going by the resistors what temperature does it get up to with sustained playing?

    The design you are working on now sounds just what I'm after, I thought I would parallel more resistors in the first stage to build in a factor of safety.

    Cheers, kev.
     
  18. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    The 100W resistors are 60mm body length, and the current box is 120x90 mm. The way I play, cleanish with breakup at about 6 volume, its hard to get it to heat up. But I did run a loud distorted thrash through it at max for a while and it heated up then, but not to a point where you couldnt touch it for an extended time, maybe 50C temp rise. But while the step from 50 to 100W is a small step in volume, Id expect its a big factor on heat, going from 'quite hot' to 'OUCH #@!!'.
     
  19. 1199RS

    1199RS New Member

    Joined:
    Jul 29, 2015
    Messages:
    17
    Likes Received:
    5
    Location:
    Dundee Scotland uk
    Ha ha ye I imagine so as my load box gets to around 70 deg c after an half an hour or so of playing. I have not used thermal paste to seat the resistors yet as was waiting to see how it functioned first.
    I have the first two stages cascaded on mine so only cleans I get are when I roll back guitar volume.
    I think I will add cooling fins on the box this time much like the hot plate to help dissipate some of the heat.
     
  20. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    4,693
    Likes Received:
    2,777
    Location:
    Wilton NSW
    Im waiting on a new batch of power-resistors from China, which take a while. But given the interest, Ill post the new schematic soon.
     
    1199RS likes this.

Share This Page