Summary: September 2019 This thread started with a few resistors to make a simple attenuator for valve amps, and then developed into multistage resistive and reactive designs. The latest designs work much better than I’d expected and have been built and tested by others. April 2020: After this post, skip to page 27 for the latest updates. http://www.marshallforum.com/thread...design-and-testing.98285/page-27#post-1877676 There are tweaks to the Design M presented below and new front end designs M2 and M3. July 2020 Latest diagram for M2 design: http://www.marshallforum.com/thread...design-and-testing.98285/page-56#post-2000746 Background Passive attenuators are wired between the amp output and the speakers. Their function is to absorb most of the output power of the amp, feeding a smaller amount to the speaker itself. This allows the amp output stage to run at higher power, letting the glorious tone of a good valve output stage develop, but without excessive volume. The attenuator must present a load to the amp that is similar to a speaker and also maintain the tone as volume is reduced. It needs a consistent tonal and dynamic response from low attenuation, down to sub-bedroom level. This is where the simplest designs can be inferior, and the best commercial designs get expensive. With feedback and testing by others I think we have a design that achieves this. For about $100, anyone with workshop skills and the ability to follow a circuit schematic can build this. An important point: Anyone who builds it does so at their own risk, and takes responsibility for working out their own wiring for their own private, non-commercial use, and completing it safely. Attenuator M This design has been built by myself and others since January 2019. It needs to be matched to the output tap of the amp, eg 8 Ohm or 16 Ohms. Component values for both are given, which differ by a factor of two: Attenuator M 190110 by JohnH posted Jan 10, 2019 at 10:52 PM (EDIT 23/9/2019: a minor tweak for a very slight improvement - see post #528 on p27: L1=0.4mH, L2=0.6mH, R9=22 Ohm. or 0.8mH/1.2mH/47 Ohm for a 16 Ohm unit) There are a number of attenuation stages, engaged or bypassed by switches. Stage 1 is the key reactive stage and includes two inductor coils. This stage on its own, reduces power by a factor of 5, or -7db, reducing a 50W amp to 10W. The inductor coils are configured so that the impedance presented to the amp is similar to that of a real speaker (values based on various Celestions), particularly how impedance rises with frequency. After Stage 1, several more stages are provided. These can be mixed and matched, but the design shown is based around additional -3.5db, -7db and -14db stages. By combining these switches in combination, and with Stage 1, reductions of up to -31.5db can be achieved in small, equal steps of -3.5db, at which point a 50W amp, at full power, is playing quietly at about 35mW. At the output, one or two speakers can be used. The most even attenuation steps and consistent tone is if an 8 Ohm attenuator is used with an 8 Ohm speaker (or two x 16 Ohm), or 16 with 16. But both versions are safe to use with either 8 or 16 Ohm speakers. In the red box, some additional parts are engaged to a third output socket, which will make a slight tweak to the tone for use when a 16 Ohm speaker is used with an 8 Ohm attenuator. These add a couple of db to correct the high treble, whereas if a 16 Ohm speaker is used with 8 Ohm attenuator with the standard output, there is a bit more mids (not much difference though – its optional). Another optional aspect is the bypass switching. As shown, full bypass can be achieved, and also the -3.5 stage can be run on its own, as a small resistive reduction. This gives the widest range of volume settings. But for many users (including myself), full bypass and -3.5db may not be needed, and switching can be simplified if the -7db stage is always engaged and all further switched steps are below that. Simplified version This is the same design, omitting the bypass switching and the 3rd output. Minimum attenuation is -7db. Its an easier build and it should still meet most needs. But if you need -3db, a good work-around is to switch it to max attenuation and run it in parallel with the speaker as a dummy load, using a lower Ohm amp tap. Attenuator M-lite by JohnH posted Aug 24, 2019 at 3:43 PM Component references are the same. Component values and power ratings The table below shows the maximum expected power being dissipated by each resistor. The component ratings need to have a good margin above this. I suggest a factor of at least 3 for case-mounted aluminium resistors, bolted (using thermal grease) to a heavy metal chassis or heatsink , and a factor of 5 or more for air-cooled resistors. These values fit with the spec in the schematic diagram above. Wire for hookup and also the winding of air-cored inductors should be 18 gage for 50W attenuators, and this is also OK for a 100W one, if built to the 16 Ohm values. For switches, use at least 5A rating (at 125V ac) for a 50W 8 Ohm build. Get the best jacks you can find. Attenuator M Power 190331 by JohnH posted Mar 31, 2019 at 9:43 PM Cooling With amps > 30W at high power, the unit will heat up as it dissipates power. A good size die-cast aluminium case is best. Once components are positioned, then a number of additional large vent holes should be drilled, in the top and in the base, with feet to raise up the base. This will help to promote good convective flow of air out through the top, replaced by cool air at the base. The best colours for cooling are dark. Build My current build is in a case 170 x 120 x 55mm of thick aluminium: AttenuatorM Inside 190217 by JohnH posted Feb 17, 2019 at 1:20 PM AttenuatorM Outside 190217 by JohnH posted Feb 17, 2019 at 1:19 PM A 'watch-it': don't mount the air-cored inductors using normal steel bolts, since this can significant change their inductance. Use nylon, or stainless steel bolts, or another method such as zip-ties. (or, to experiment, put an M3 or M4 steel bolt through inductor L2, which in theory should increase treble slightly!) Performance In the schematic above, there is a graph showing a calculated frequency response at each attenuation level from 0 to -31.5 db. These use a spreadsheet to calculate the signal at each stage of the circuit, as a series of voltage dividers, using complex number theory to assess magnitudes and phase angles. The speaker was represented, for analysis, by an equivalent load model, by Aiken: http://www.aikenamps.com/index.php/designing-a-reactive-speaker-load-emulator ...adjusted to match the measured performance of a G12M 4x12 cab. The plots are based on small signals, with the amp output impedance assumed to be 20 Ohms, for an 8 Ohm tap, based on measurements of my VM2266C. These calcs were used to adjust the values in the design. Here are some sound clips Attenuator M: Max attenuation to non-attenuated: https://vocaroo.com/i/s1QgVDnl1XQi Attenuator M: Normalised: https://vocaroo.com/i/s1cnuucLz8yK It’s a simple looped riff, played twice at each attenuation setting from -31db up to full unattenuated in 3.5db steps. The second file based on the same recording, with each stage normalised for volume so you can hear any differences in the tone. The VM2266c was on LDR mode, body at 6, detail at 9, master vol at 6, tones and presence at 6, using my LP bridge pickup, miced off a speaker. Attenuator M Frequency Plots 190302 by JohnH posted Mar 2, 2019 at 12:38 PM The plots are taken from the sound sample posted above. The lower set of data are the basic plots, from full volume down to -31db (db scale is arbitrary, but relative db's are right). The upper plots are intended to show the differences between responses. I took the -7db recording as the base case, so this is shown as a flat line. The others are the various other settings, with the -7db trace subtracted. The ideal for these traces is therefore also a flat line. And for all the traces below -7db down to -31db, this is what is happening, there is virtually no further tonal change at all as you attenuate down as far as you want. It measures as consistent. The -3.5db and full-volume traces show some variarion relative to -7db. The peaks are consistent though. I think we are seeing extra resonances and distortion generated in the speaker itself at this high volume, and no attenuator can capture those. The -3.5 trace (resistive) shows a very slight treble fall-off, hard to hear in practice. How it works The tone of a guitar speaker in a valve amp depends on low damping due to a high effective amp output impedance. This allows the natural speaker inductance and resonance to develop a rise in treble, and a bass resonance. So the output impedance of the attenuator needs to be consistent and representative of a real amp, which is quite high. Most simple attenuators do not get this right, and it is rarely discussed. Based on this, a good resistive attenuator can be designed, as shown through the first few pages of the thread, which follow. At high volume, the amp reacts to changing impedance of the speaker. This is what the inductor coils do in the attenuator. This design matches impedance of a real speaker from low mids up to high treble. It doesn’t show the amp the bass resonance however, which is not necessary since it is developed at the speaker.