Simple Attenuators - Design And Testing

[Esa Martikainen]

When here in Europe I have searched audio components often I have found them from from Holland and for my attenuator 15mH coil came from "audio.nl" and its link is here. Nowadays shipping is fast but quite costly as well.

https://audio.nl/Aprodview.asp?type=product&Product=1342

Esa

 

[Esa Martikainen]

And the 1.8mH coil is this "air core". It is huge having 100mm diameter (mounting case is 130) and 55mm thick. Wire is 2mm and resistance 0.25 ohms and it did cost 33 euros.

https://audio.nl/Aprodview.asp?type=product&Product=3604

That previous post 15mH Torobar is rated for 600W/8ohms and is 0.29ohms but it has a ferrite core and it is toroid wound diam 95mm and 50mm thick. It did cost 46 euros.

For capacitors I found a damaged and trashed circuit board which was used on a powerful "switching power supply" and there are eight 68uF capacitors which I beliece should handle current very well. I think that current handling comes better when more smaller capacitors are used than one or two big. But does it cause "stray noise" i don't know.

Esa

 

[genuinemrjay]

Hi John. First off, I want to say your project is amazing! I have read through all sixty some pages of it in the last two days, and can't say enough about your knowledge, capabilities, and creativeness in this arena. I am going to build your 16 Ohm M2 attenuator for my AC30 (I know, it's a Marshall forum so please don't boo me everyone). With a 30 watt/ 16 ohm amp would the following values below suffice (the number after the - is Ohms)? Thank you for your time and effort John!

 

[JohnH]

Hi John. First off, I want to say your project is amazing! I have read through all sixty some pages of it in the last two days, and can't say enough about your knowledge, capabilities, and creativeness in this arena. I am going to build your 16 Ohm M2 attenuator for my AC30 (I know, it's a Marshall forum so please don't boo me everyone). With a 30 watt/ 16 ohm amp would the following values below suffice (the number after the - is Ohms)? Thank you for your time and effort John!
View attachment 77408

Thanks for your message. Nice drawing and the values look fine. I see you scaled down the power ratings a bit, for your AC30. Should be OK if you get them bolted onto a thick aluminium chassis with thermal grease and plenty of vents. I've not see the 75, 30 and 15W resistors though where I've shopped. For the inductor, I'd stick with 18 (or 19) gage wire.

My other amp is a DSL401, and like an AC30, it also has 4xEL84 with no negative feedback. The tone of this type of output circuit is very sensitive to what its driving, so I'm glad you picked this attenuator design since it works well with it! Good luck with your project and we look forward to hearing how it works out.

 

[JohnH]

Hi @Esa Martikainen , thanks for your links. That looks like a good source for parts in the EU. Have you built your project yet?

 

[genuinemrjay]

Thanks for your message. Nice drawing and the values look fine. I see you scaled down the power ratings a bit, for your AC30. Should be OK if you get them bolted onto a thick aluminium chassis with thermal grease and plenty of vents. I've not see the 75, 30 and 15W resistors though where I've shopped. For the inductor, I'd stick with 18 (or 19) gage wire.

My other amp is a DSL401, and like an AC30, it also has 4xEL84 with no negative feedback. The tone of this type of output circuit is very sensitive to what its driving, so I'm glad you picked this attenuator design since it works well with it! Good luck with your project and we look forward to hearing how it works out.

Thanks for the reply John! I was able to find all the resistors except the 30W for R2A and R2B online at Mouser, so those will have to be changed to 50W. I also used draw.io for the diagram which was pretty handy. Prior to discovering your thread, I planned on buying a Two Notes torpedo. Downside was that it only has one fixed dB cut, and several other features I really don't need. I just need to tame the wildness that is 4xEL84, and your attenuator design looks like a perfect fit. I'll be sure to document the build and post in detail here in the near future. Thanks again!

 

[Esa Martikainen]

Hi @Esa Martikainen , thanks for your links. That looks like a good source for parts in the EU. Have you built your project yet?

Not yet! I just received those coils but soon I put it together.

I also thank You for all the research you have done and inspiration I have read here!

Is it easy for you to explain why is it better to use half the capacitor and double the inductance and resistance values when changing from 8 ohm to 16 ohm attenuator?

I understand how resonant frequenzy depends of capacitance and inductance but not how they and the loudspeaker impedance or attenuator go together.

If I use 15mH which I already have and 200uF (which I build using at least 4 caps) resonant circuit it is going to have about 92 Hz resonance but the peak seems to come quite narrow, narrower than majority of Jensen loudspeakers have. Jensen webpage contains best graphs and data of all manufacturers!

My 1.8mH series inductance has 3,2ohms around 280Hz where 15mH/200uF also has 3,2ohms which is the "low point" of whole circuit if I have understood it right?

My plan now is to measure the resonance peak of the loudspeaker I am going to use and tune the circuit resonant frequuency about the same switching more or less capacitors to the resonant circuit.

Esa

 

[JohnH]

Hi @Esa Martikainen

I'm assuming you're building a 16Ohm version of the M3 design? with the resonant circuit?

That's great, you should know that I think you will be the first to try this one though there are quite a few M2's now as well as many M's

With your parts, the 1.8mH wound with 2mm wire is great, obviously thicker than we've been using so far, no problem there. With the resonant circuit, 15mH and 200uF, I agree that leads to 92 hz as you noted, which is close to many open back cabs, and less than most closed back ones. If you have this circuit, then it affects what the amp responds to, but at the speaker, the output will be more determined by the actual resonance of the real speaker. Ideally these will match. But this is why the other designs M and M2 still work very well even with no resonant circuit, the speaker provides it itself.

Ive done sound clips with my M design (no bass circuit) , comparing tone from the full unattenuated speaker with that from reduced volume, trying to drive all the bass notes very hard, and really I couldn't hear a tone difference at all.

When you go from an 8 Ohm build to 16, all resistances and impedances have to double, so resistors and inductors x2. But impedance of a capacitor is dependent on 1/C actually (1/(2Pi.f.C), so it halves, and with L x2 and C x 1/2, resonant frequency stays the same.

With those caps, note that the capacitor gets current through it both directions, so needs to be bipolar. Most large caps are polarized electrolytic so you should check out what you have. If buying, I would get the big polypropylene ones or bipolar electrolytic but we need to make sure it can handle high ripple current. Also, almost the full amp power passes through the cap. I have read however that if you put two polarized electrolytics in series, pointing in opposite directions, its equivalent to a bipolar cap of 1/2 the value

 

[Boysenman]

Hi @JohnH. I finally got around to building the M2 attenuator (with line out) for the 10" speaker. It works like a charm!

I look forward to building more variations of these designs.

Thank you!

 

[littlewyan]

I'm getting parts together for your M2 build and am wondering about the rating of R2A and R2B. Is 50W necessary? Does R2A get warm at all in your build John? And how do R2A and R2B work together? I worked out that in parallel they give you 10Ohms, but I'm perplexed as to how they work with the inductor...

 

[JohnH]

Hi @Boysenman , very neat build it looks great.

@littlewyan R2A and R2B may be a bit over-specced at 50W but its better than under. Together they take up to about 20W if 50W is coming in, and I go x3 as a factor. How much power they each take depends on the frequencies running through, and how you play, how much distortion etc. More goes into R2A at high frequency. These two are different from the other resistors where the power can be assessed easily by proportion.

So im sure 50W is plenty for each but I wasn't sure if 25W would be enough. My own build is the earlier M type, which has the same performance but needs two coils. M2 is more economical but was harder maths!

In both M and M2, the inductance is partly in series and partly in parallel with the circuit downstream. Thats the key to both versions.If it was only in series the speaker would get a dull sound, while if only in parallel it would be too bright.

 

[littlewyan]

If they're already over spec then actually it may not be an issue. Reason I ask is, I'm going to use a Hammond 1444-972 chassis which isn't very thick and won't be suitable as a heatsink. So if I use these chassis mount resistors then I need to really go by the 'Watts with no heatsink' rating to be safe. Luckily for the other resistors I can use these https://www.mouser.co.uk/datasheet/2/418/5/NG_DS_9-1773453-2_C-727458.pdf
I may use 75W chassis mount resistors instead for R2A and R2B just to be safe.

 

[peefnik]

Hey all! I've been investigating various load boxes/attenuators for a while now. I'm currently using mesa boogie mark Vs ... 35W, and the big boy. I also have a Helix HX stomp ... which is pretty sahweet. I get great sounds from that thing. Regarding the Mesas ... I say they are the Fender amps I always wanted. I've been looking at ox, waza, and recently the captor x.

I like the captor x (in theory) but feel like two attenuated levels would probably fall short in most/different situations.

Anyway, since I'm into the engineering aspect of this stuff I've done a deep dive into the spice modelling. I figured I'd start with the M3 8ohm version. All real tube amps are 8 ohms

I'm using LTSpice but this is really applicable to most spice programs.

I created an M3 subckt (w/symbol) and also an M3 schematic with an identical (to the subckt symbol) 'block' symbol which just pulls in a schematic as a subckt under the hood. The good thing about the latter is that you don't have to edit a subckt manually to make changes on the fly. You can just modify the schematic.

I have run sims using just a voltage source and an actual deluxe reverb output section (transformer and all) into the great JH attenuator circuit. Its way too much fun, or I'm way too much of a nerd.

The thing with the resonance impedance facing the amp is you have to model it depending on the speaker you are using, or you get some funky multiple peaks facing the amp. I also have a complete fender deluxe simulation which I'm playing with ... thats where I pulled the output section from.

Anyway, here are some fine images of the results of these shenanigans I apologize if these images don't come thru ... I'm unfamiliar with how this interface (image/attachments) really works.

 

[peefnik]

Heres a grab with the full deluxe reverb showing impedance @ the amp and output dB at the speaker.

 

[JohnH]

Hi @peefnik
Thanks very much for this study! Getting the attenuator modelled at the end of a full output stage model, is not something I've done myself.

I'll need to look on a proper screen, I'm currently just on a phone. I'm likely to ask more about it, but here's a few initial Q's:

Does it include a speaker model? and if so what's in it?

With the M3 circuit, do you see a difference in how closely the attenuated signal tracks the shape of a full-volume one with the ougput stage model, as compared to just a voltage source/impedance?

Can this provide any insights to the benefits or not, of including the resonant M3 circuit as compared to the M2 which has only one coil?

Anyway, thanks again.

 

[peefnik]

@JohnH
I'm using a generic 8ohm speaker model ... as you have posted here (from aiken?).
The signal tracks pretty well ... but you have to tweak the resonance values to match the speaker or you can get multiple/offset peaks at the different resonance frequencies.

I tweaked the resonance inductor in the attenuator to 18mH to bring the peak down to the frequency when bypassing the attenuator (e.g. just the speaker load).

Here is the speaker schematic and a plot with the attenuator bypassed, then with only first stage. The top green line is both the amp and speaker when bypassed. The top blue and lower green are at the amp and speaker respectively when the attenuator is not bypassed.

Now that I can see the amp model working, I'll probably just use the output stage w/transformer model which eliminates all of the preamp filtering, etc.

NOTE: This is with bass and treble both at 5. Fenders, at least the classic ones, are mid scooped as the tone stack has a fixed mid resistance. They only approach being flat when bass and treble are at 0. It is also on the vib channel which is considerably brighter than the normal channel.

 

[JohnH]

next morning and Im looking at it more clearly on a 24" screen.....

very impressed....

I just saw your latest post that answers a lot more. it looks like a good match.

But with the output-stage-only model, it would be worth building into it the NFB circuit from the full model. NFB totally changes the effective output impedance of the amp. Its where my modelling with a fixed 20Ohm output Z, based on one of my amps, will differ in comparison to various real amps. With just the M2 circuit, most of the frequency response seems to track with different values for this 20 Ohms. But the bass response tends to stay constant. With M3, my models suggest that it will all track better including the bass resonance. But if you are able to make an output stage with the NFB in it, then this can be analysed properly, and tested with different NFB amounts.

With the speakers, my values have been targeting measured response of a closed back cab, which raises the resonant frequency. I agree that if the bass res circuit goes in, and is intended to be correct, it needs to match the cab. Whether it matter in listening tests is up for consideration.

If you like this stuff, Ill float past you the following issues as subjects for investigation:

1. These attenuators have been well liked by folk using very different amps to those that I designed it with. This issue of response to different output stages with different NFB may be key and worth exploring
2. Also, is it really worth it to put in the resonance circuit in real builds? Builds so far have not had it
3. Given a model of a real output stage, what happens if you drive into non linearity? That's actually the point of having an attenuator, to drive the amp hard but quietly! This would probably need a transient analysis using a single frequency and the plotting of waveforms. Could the models be wrangled to do that?

 

[peefnik]

I'll throw the PI (long tail pair) into the tube output only sim. Then I can have the NFB as well.

Once I get that in I'll see what happens when I drive it harder. Not sure what the models do in that case to be honest.

Even for clean signals at the low end the resonance being present would certainly present a more realistic condition to the amp. Is the effect enough for people to hear or feel in some way as the player ... I don't know.

 

[JohnH]

I think from the bits I know about Spice calcs, with the plots so far, all signals are based on them being small signals staying linear, even if numerically they are large. So to explore large signals going into the tubes, its a different type of run but the same model. I use a simpler Spice version called 5Spice (cant handle this tube model though, I need to get into LT) and its called a transient analysis there, where you feed in a waveform such as a sine at a chosen frequency, and you get an output waveform like on an oscilloscope.

 

[JohnH]

Even for clean signals at the low end the resonance being present would certainly present a more realistic condition to the amp. Is the effect enough for people to hear or feel in some way as the player ... I don't know.

That's the thing! does it make a difference? here was my attempt to explore it a couple of months ago:

Have a listen to these, which try to show bass response. It's my attempt to play the start of Cinnamon Girl. First, attenuator bypassed, then switched to -21db, both from the same loop.

Full volume (attenuator bypassed), then -21db:
https://voca.ro/jwLCSg3wC9x

Here it is with the attenuated second part normalized in software back up to the same volume:
https://voca.ro/ncne6xsR8JS

Its miced off the VM2266c, Master volume at about 6.

I picked this since the dropped lower D gets frequencies down to 73 hz for low D, and has some notes at E#, G and A above that, so it should be in the range of the bass peak for comparison.

I hope that helps.

ps. another small nugget of info, this test was done using the -7 and -14db stages to make -21db. Audacity, which I record with lets me check that, and the actual difference in levels as miced, is -20.7db, so numbers are good.

With those clips, need to ignore the choice of music, the playing, the recording, the hiss when the attenuated signal is normalized and the actual tone! Just listen for the differences in the low bass notes as I was trying to provoke them to make things happen on the A and drop D lowest strings, right around where the resonance would be. This was with an M attenuator without the specific resonant circuit.