Simple Attenuators - Design And Testing

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

  1. matttornado

    matttornado Well-Known Member

    Joined:
    Mar 14, 2012
    Messages:
    907
    Likes Received:
    394
    I'm going to mount R1 resistors on a seperate aluminum plate first, prior to mounting to the case for extra heat sinking.
     
    JohnH likes this.
  2. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    I tried running the attenuator as a loadbox again yesterday, in parallel with a speaker, with amp set to half the speaker ohms. The attenuator was set to max attenuation It worked fine, nice and clear for a -3db reduction in speaker volume. I think thats a good trick!
     
  3. Gene Ballzz

    Gene Ballzz Well-Known Member

    Joined:
    Feb 19, 2013
    Messages:
    1,786
    Likes Received:
    1,183
    Location:
    Las Vegas, NV
    So, can I assume that while giving a -3db reduction, it ends up having all of the load and/or attenuation being reactive, as opposed to simply using the attenuator in between the amp and the speaker on a -3db, non reactive setting? If this is indeed the case, it is definitely a VERY "good trick!"

    Thanks Aagain @JohnH
    Gene
     
  4. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    Yes above about 150hz the amp thinks its driving a normal reactive speaker load of half the ohms and reacts to it normally, and out of the speaker comes the correct normal tone, at -3db reduction.

    What is lost however, and only in theory, is that the bass peak at the speaker is reduced by about 3db max, at about 100hz (depends on the speaker). But I don't think this really makes much noticeable difference - it only affects the fundamental tone of the very lowest guitar notes and not any of their harmonics. In any case, a tweak of a bass or resonance knob would fix it if wanted.

    Interestingly, the -3.5db resistive reduction, if built, gives the right tonal shape including the bass peak, but the amp is not quite getting the full reactiveness to respond to dynamically.

    But I couldn't tell any difference anyway.


    Try it and see what you think!
     
  5. rockgod212

    rockgod212 Member

    Joined:
    Mar 20, 2010
    Messages:
    215
    Likes Received:
    24
    saw this thread, so I built a KF style air brake a while back. I use with my 50 watt 1987 NMV clone, love it. 0823171512-00.jpg
     
  6. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    Thanks for posting that. It looks to be a nice build! When I was working out mine, I learned a lot by studying the Airbrake design.

    Airbrakes are nice and simple. A good thing about them is that they don't get dull at higher attenuation, unlike most Lpad designs. Its because they reverse the Lpad, putting a resistor in series with the speaker instead of across it which would add too much damping. There's a sweet spot in its settings where the what the amp sees in terms of ohms, and what the speaker sees are similar to as if they were directly connected. At this point, the basic tone is in balance and close to that of a directly connected amp/speaker.

    In my version, related to the Airbrake idea, this sweet spot is found at -7db in Stage 1. Instead of varrying the values to move away from this point, I added further stages to add attenuation while keeping this relationship between input and output ohms intact. This maintains consistent tone at each setting. Coils are then added, again in balance, to show the amp a correct reactive load without messing with what the speaker sees.
     
  7. TonyK

    TonyK Member

    Joined:
    Mar 15, 2019
    Messages:
    49
    Likes Received:
    20
    LOL, if I were busy enough playing guitar loud and proud, I probably wouldn't have put the time into building the attenuator either! But in my case, I am now able to play the guitar more frequently and with amp tones I can actually enjoy! Good luck!
     
    Nik Henville likes this.
  8. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    :This post is to assuage my OCD tendencies by exploring whether or not there are any further tweaks to the design values of components that could fine-tune its performance, and to explore its theoretical response in different scenarios to see how consistently it performs.

    The core of the design is the first stage, particularly the two inductor coils L1 and L2, with L1 bypassed by a resistor R9, and how they interact with the resistor stages down-stream.

    So this is mostly about the values of L1, L2 and R9.

    Short answer: If you have one of these built to the current designs, its all good, no benefit in changing because any benefits from tweaks discussed below are typically only a fraction of a db.

    But, we have to pick component values, so might as well pick the optimum, and sometimes the optimum is not available, or might depend on the use of the unit.

    Reminder - here is the current design:

    The values for 16 ohm and 8 ohm builds are in principal, a factor x2 different.
    But above, the values of L1, l2 and R9 are 0.33mH, 0.5mH and 27 ohm for the 8 Ohm build, and 0.7mH, 1.1mH and 56 Ohm for the 16 Ohm build, so a bit more than x2. Also, @Gene Ballzz is having good results with 0.8mH, 1.2mH and 56 Ohm in his 16 ohm units.

    So do these variations (20%) make a significant difference for good or bad?

    Benchmark reference: A 4x12 cab with G12M's

    To make any quantitative comparisons, we need a benchmark for reference, plus some criteria. I reckon the best target for response would be something like a classic 4x12 cab with Greenbacks in it.

    Mike Lind on the TGP posted some measured load charts for various load boxes, plus actual cabs:

    https://www.thegearpage.net/board/index.php?threads/attenuators-and-load-boxes.1947804/

    So I traced his measured chart for a greenback cab (a 1960AX).

    To bring it into a form I can use, I needed to represent this real impedance curve in the form of an equivalent electrical circuit. The best described is by Aiken, also referenced in a build thread for load boxes on TGP:

    https://www.thegearpage.net/board/index.php?threads/aikens-reactive-dummy-load.1072793/

    Direct link to Aikens page is at the start of that.

    So I've been adjusting Aikens design to match Mike Linds measurements, like this:


    The schematic shows Aikens values for a 16 Ohm cab, with my amended ones in yellow above. This could form the basis of a reactive load-box to be built, as it was intended. But here, we don't need to actually build it. I'm just using the circuit in my analysis for virtual testing. Just as well, because to get that very sharp resonance peak, the required high-current capacitor is large and very expensive.

    In my chart above, red is the calculated impedance vs frequency from the equivalent circuit, and green is a close trace from Mike L's chart. I reckon I got them very close indeed.

    Criteria and testing

    Back to the attenuator analysis. So using the response above as a target, the ideal is for the amp to see an impedance curve as close as possible to the equivalent virtual Greenback cab
    , and the shape of the signal at the speaker also to match the benchmark, when normalised for volume.

    Ive done a bunch of runs, based on 8 Ohm units, exploring L1, L2 and R9

    Here is a plot of impedance, this time with the attenuator set at -14db plus the benchmark - scaled down by 2 for an 8 Ohm rig:


    The values used were L1 =0.37mH, L2 = 0.55mH and R9 = 22 Ohms (equivalent 0.75, 1.1 and 44 in a 16 ohm build)

    Results

    In exploring many options, I've found a couple of consistent results:

    1. R9 is very slightly better at 22 ohms than at 27 ohms. So we might as well use that (nearest would be 47 in a 16 Ohm build). So that can be the new default value although the difference it makes is less than 0.2db.

    2. L1 and L2 keep consistently coming back to being in the ratio L2/L1 = 1.5.This seems to be more important than their actual values, within a close range.

    So I tested the following L1/L2/R9
    0.33/0.5/22
    0.37/0.55/22
    0.4/0.6/22
    and also, for comparison with previous:
    0.33/0.5/27

    I was looking for consistency in the (somewhat arbitrary) frequency range 200hz to 7kHz, which is where these components have most influence, and being above the bass peak, which I am not controlling so closely. I was looking at signal level at the speaker and also at the amp output, based on a few different assumptions about the effective amp output impedance, the speakers and the amount of attenuation.

    I tested at -7, -14, -21 and -28 db. I explored 16, 8 and 4 ohms cabs (with 8 ohm attenuator, the 8 and the 4 correspond to 16 and 8 respectively for a 16 ohm unit). With 8 ohm cab I took the amp effective output impedance at 20 ohms (per my VM) and also at x3 and x1/3 of that.

    Conclusion


    Out of those design options across all those ranges, all the sets of values are very close to the benchmark, within 1.5db (and usually less) inside the frequency range. So they all work fine, and within about 0.3 db of each other.

    The current L1 and L2 values of 0.33/0.5 were very slightly better than the others at -7 db, and also when a 16 Ohm cab is attached (using the dedicated output - diagram red box). Other than that, the higher value pairs 0.37/0.55 or 0.4/0.6 are fractionally better - but nothing much to chose between them.

    I think it would not be possible to hear a significant difference between any of these options. The design seems to have hit a sweet spot, but it is a fairly flat one in this range.

    In general, and also based on Genes experience, a great set of base values would be the 0.4 and 0.6mH for the 8 Ohm build, and 0.8 and 1.2mH for the 16 Ohm build. These are common values in most available product ranges. But if you pick a different value, try to maintain the 1.5 factor L2/L1.

    R9, is slightly better at 22 Ohms (47 for 16 Ohm build)

    But there's no need nor significant benefit in changing any built values.

    Some plots

    This plot shows full volume in red, signal at amp in green and signal at output in blue(normalised for volume), all at the -14db setting:


    Top left is with the base assumptions, with very close matches throughout.

    Lower left and top right show the same, with amp impedance x1/3 or x3. it is very interesting to see how the responses continue to track, illustrating how this design can respond to different amps and also dynamic changes in the amp.

    Lower left is using half the cab impedance, ie a 4 ohm cab in an 8 ohm attenuator - a good results, a little bit brighter response as expected. This plot also shows what to expect from an 8 ohm cab into the 16 ohm attenuator.
     
    Last edited: Sep 23, 2019
  9. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    I've got a new idea for the front end of the attenuator circuit that reduces the number of coils needed, with the same performance achieved, and then leads to an enhancement with the abilty to add a resonant circuit.

    To explain, it best to step through the journey so far. The following four circuits are based on the 8 Ohm design with four stages. They show the front end stage 1, then the other stages all consistent as per the current Attenuator M. Plots are shown of calculated frequency response.

    Red is the signal seen by a speaker directly connected to an amp
    The upper black and green lines are the signal at the amp feeding into the attenuator
    The lower blue and black lines are the signal after the attenuator, using attenuation values of -7db and -14db.

    Ideally, all these plots would be exactly the same shape, with the two attenuated plots being shifted down, and the upper plots of full-volume speaker and amp signal feeding the attenuator all identical. All should show the characteristic bass resonance peak, and the general rise in treble of a natural valve guitar amp driving a real guitar speaker

    Resistive attenuator

    Attenuator Resistive 191005.gif


    This is the simplest, and what is interesting about it is how it achieves a basically correct and very consistent tone at all settings, just with resistors. This is the necessary key to all these designs and is the main point of difference to most comparable commercial designs.

    But, although nominally the right tone is coming out of the speaker, the amp itself is only seeing a flat resistive load. ie the upper curves below the red full-response curve are nearly flat, which means that the amp cant respond dynamically to the differences in impedance.

    It still works though!


    Attenuator M - reactive input stage

    This is the current design since January 2019, now well tested by many builders. It captures the rise in treble impedance and lets the amp see this and respond dynamically to it. The bass resonance is developed at the speaker as in the resistive design, but is not seen at the amp. It continues to work well.

    Attenuator M 191005.gif

    See how the green and black signals at the amp show a treble rise but not a bass resonance peak.

    The basis of this is to split the treble inductance into two coils, one in series and one in parallel. This achieves a balance of impedances that allows further stages of resistive attenuation to be added, without affecting tone.

    The coils are fairly compact and inexpensive, but to do this double arrangement while also capturing a bass resonance in each would be very expensive.


    Attenuator M2 - single treble coil

    So here's the new idea. The front end of Attenuator M is amended so that the twin coil series/parallel arrangement is replaced with a single coil, and a resistor pair. This results in one less coil - simpler, cheaper and more compact. Plus, it appears that it will work just as well (not tested to date though):

    Attenuator M2 191005.gif

    (EDIT 14/10/19: R2B is better at 18 Ohms)

    I don't see any down-sides to this variation. The numbers look very consistent and I believe it will sound virtually the same as the base attenuator M with twin coils.

    Attenuator M3 - with bass resonant circuit

    Having reduced the reactive circuits from two to one, it is now feasible to consider adding a bass-resonant circuit to it, like this:

    Attenuator M3 191005.gif

    (EDIT 14/10/19: R2B is better at 18 Ohms)

    The circuit is based on getting close to the impedance of a a Marshall 1960 with 4x12 G12M's. Now the amp can see a bass resonant peak, and if it is driving hard, it should respond to it.

    It should work fine. What I don't know yet is whether it is really worth doing this. The current circuits don't seem to lack for not having a resonant circuit. So this is just an exploratory proposal. Its not a cheap thing to add because to do the resonant circuit right, it needs capacitors and inductors with very low losses. Ideally, high quality bipolar film caps should be used, and two x 100uF may be needed to get the 200uF value specified. And the resonant coil should be iron core, preferably toroidal for very low dc resistance <0.5 ohm.

    Interesting though I think! It would make the whole unit slightly better for use as a free-standing load box with no speaker, and might just add a few extra tads of authenticity to the bass-note overdrive tone.

    16 Ohm versions

    All the above is applicable at 16 Ohms, with inductors and resistors x2 and for M3, the capacitor value halved. This might make the cost a bit more feasible since one 100uF could be used instead of two to get 200uF.
     
    Last edited: Oct 13, 2019
    6StringStewie and Nik Henville like this.
  10. Nik Henville

    Nik Henville Well-Known Member

    Joined:
    Jul 18, 2018
    Messages:
    958
    Likes Received:
    1,457
    Location:
    Eastern Seaboard, the nook-shotten isle of Albion
    Blimey - now THAT is impressive stuff...

    :hippie::pirate::uk:
     
    JohnH likes this.
  11. Jordan Prysmiki

    Jordan Prysmiki New Member

    Joined:
    Feb 14, 2018
    Messages:
    21
    Likes Received:
    10
    Hello John. I just purchased an amp with a reactive alternator built in. 15watt Frenzel HBX AC15. Friggin awesome amp by the way. The switch goes full power, half, and 1/3 power.

    Wondering if you could help me figure out how to add inductors in to make it reactive?
     
  12. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    looks like a fun amp! Im not seeing the built-attenuator described on the website. Do you have a link?
     
  13. Jordan Prysmiki

    Jordan Prysmiki New Member

    Joined:
    Feb 14, 2018
    Messages:
    21
    Likes Received:
    10
    You know that's funny because I saw they offered three attenuator as a paid upgrade on eBay but don't mention it on their site..

    They unfortunately would not send schematics either but I can maybe snap some pictures if that would help.
     
  14. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    OK, lets have a look, but Ill probably end up suggesting to make a separate attenuator, based on whichever one of the 4/8/16 output options suits you best, then ignore the built in one.
     
  15. tristanc

    tristanc New Member

    Joined:
    Oct 6, 2019
    Messages:
    1
    Likes Received:
    3
    Hi John, just a quick note to say thanks for all your work on this. I've just today finished your version M, before your recent updates (already had the parts). Works really well so far - great being able to blend between pre- and power-amp distortion easily.

    I reduced the resistor wattages as my amps are 18W and 2W... And got the closest inductor I could for L2 (0.47). Other than that, it's stock in a Hammond enclosure.

    IMG_0398.jpg

    IMG_0399.jpg

    IMG_0400.jpg

    Tristan
     
    Last edited: Nov 14, 2019
    Jordan Prysmiki, JCarno and JohnH like this.
  16. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    A nice build! Thanks for sharing.
     
    Jordan Prysmiki and JCarno like this.
  17. Jordan Prysmiki

    Jordan Prysmiki New Member

    Joined:
    Feb 14, 2018
    Messages:
    21
    Likes Received:
    10
    Nice work bro!
     
  18. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    Ive been crunching more numbers on the M2 and M3 versions posted above. This confirmed what I posted, now checked at all attenuation levels and at various speaker ohms and amp output impedances. .

    What I do is to put the whole circuit including a speaker sim into a dedicated spreadsheet that crunches all the resistances and reactances. Then I check it against a Spice model. Once that's working, i run macros to step through all the scenarios of stages on or off, with both 8 and 16 ohm speakers. I can then see plots of all the performance at once and quickly fine-tune the values. R2B got tweaked from 15 to 18, which I noted in the earlier post from last week.

    Anyway, it's looking like M2 is definately a good alternative to version M. It performs at least as well and is a bit simpler and cheaper.

    Version M3 is intriguing. It puts in the bass peak as seen by the amp, but this is already there in all versions at the speaker. But I might not be able to resist getting the big capacitors and iron-core inductor, just to try it, even though the cost of those parts alone more than doubles the cost of the whole build .

    Another way to assess this, requiring ear protection and an empty house, is to do more listening/recording tests with the current build, at full and attenuated volume to try to identify any difference in bass overdrive tone.
     
  19. matttornado

    matttornado Well-Known Member

    Joined:
    Mar 14, 2012
    Messages:
    907
    Likes Received:
    394
    Hi JohnH.

    I'm still gathering some parts.... question for you?
    How come R2 is rated for 50 watts but R9 only 10 watts? They are in series, right next to each other so I'm just wondering if R9 should be at least 50 watts as well?

    Or does the L1 Inductor in parallel with R9 take up some of the power?

    Almost ready to start building!
     
  20. JohnH

    JohnH Well-Known Member

    Joined:
    May 31, 2008
    Messages:
    3,636
    Likes Received:
    1,573
    Location:
    Wilton NSW
    hi @matttornado , that's a good question. Actually neither L1 nor R9 dissipate much power. Most of it goes straight through them without stopping, just like a piece of wire can send a large amount of power through without getting very hot itself. And at say 500 hz, where most of the power is, L1 has an impedance of only about 1 Ohm, so it effectively shunts R9, bypassing it. Its only at high frequency that L1 has a higher impedance, channelling more of the current through R9

    Good luck with the build!
     

Share This Page

  1. This site uses cookies to help personalise content, tailor your experience and to keep you logged in if you register.
    By continuing to use this site, you are consenting to our use of cookies.
    Dismiss Notice