Simple Attenuators - Design And Testing

[_Steve]

Hi - firstly thanks for all the effort that has gone into this. I've found it incredible reading through the thread.

Im going to try to build the M2v from the previous page. Just wondering with the resistors if there is a specific type I need to get? Although ive never bought power resistors before my understanding is that the big ones are usually wire-wound and I assume that will mean they also have more inductance and im wondering if that is going to be a problem? Come to think of it, different inductors will also have varying amounts of parasitic resistance too right!?

Which components are others buying and where from?

 

[JohnH]

Hi @_Steve welcome to our thread

Resistors need to be wire wound and aluminium clad, bolted with thermal grease to a substantial aluminium case or heatsink. You can see them in the various build pics. I work out ratings based on a 50W amp and the M2v post has the ratings on it for that, which include a safety margin of at least 3. I get mine from a Chinese eBay seller, very inexpensive and they seem fine. Or from a local supplier if you want them quicker but more $.

Aircored inductors are trickier, they tend to come from suppliers who sell speaker components.. In the US Madisound has them, or here in Aus, theres Wagneronline. Use 18 Gage or 19 Gage minimum. They do have a bit of resistance but that is figured into the design.

 

[_Steve]

Great thanks so much! I've ordered everything, including case for a grand total of $71 USD.
Shipping is going to take a couple of weeks but I'll post an update and some pics when it's built.

Quick question: the blue wires on your schematic (M2v) - are they are connected to the terminal on the jacks that is closed when there is nothing plugged into it and open when something is? (I think its called a break switch or something?. Sorry for my lack of terminology im fairly new to this)

 

[JohnH]

TThe jacks shown in the digaram for M2v are intended to be stereo types like this:



They are plastic bodied so they don't connect the case, with 6 lugs. Usually these would be used for jobs such as a stereo headphone jack, which cuts out a speaker when you plug in. I use them for all jacks since I find that the ones I get grip the plug very well due to the extra contact.

But in this design and also the 'M', they are also used for a couple of extra things. When you plug a standard mono jack plug into one of these, the barrel of the plug connects the ground terminal to the ring terminal, and this is used to engage an extra connection in the 16 Ohm output and also configuring the input circuitry. The connections used would be on the far side of that picture. These jobs could alternatively be done using extra switches instead.

 

[_Steve]

Thanks for the detailed explanation!

 

[telesto]

Hi John,

I'm planning to build one of your attenuators (not sure which, M or M2). First let me say thanks for all the great research and info! I've been miserably using my home-made L-pad attenuator, and look forward to something that works better And will probably eventually integrate it into an amp I'm designing and planning to build.

I have one question: I see you mentioned that the first -7dB stage with inductor is where the "magic" happens. I also see all the other resistive stages are based on a variant of -7dB (3.5, 7, 14)...you should change the name to the "Lucky 7" attenuator! But my question is, it possible to swap the attenuation in Stage 1 for another value? eg. is it possible to put -3.5dB or -14dB instead of the -7dB? Or will the magic get lost?

For Stage 1 (in the original M attenuator, 8 ohm design), the 15/10 ohm resistors give 6 ohm resistance towards the OT, and if it was replaced by the -3.5dB stage using the 33/5.6 ohm pair from Stage 4, it would give 4.78 ohms towards the OT (and -14dB stage with 4.7/15 gives 3.4 ohms). So I guess these other variants would "ruin" the 6 ohms that the -7dB stage gives? (I see in M2 design the Stage 1 resistance also adds up to 6 ohms)

Since I have mostly smaller amps (5-25W range) would really like to shift the -3.5dB stage into the first stage, since I think -7dB is a bit too much initial attenuation for a small amp. I calculated that if I changed R1/R2 from 15/10 to 35/7 I would get -3.5dB drop and still show the OT 5.83 ohms, which I guess is close enough to the 6 ohms given by the 15/10 resistors in the -7dB stage.

ARCOL has 33/7.5 ohms for 6.11 ohms to the OT ...altho that would be a little more than 3.5dB (maybe 4dB or so), but I guess that doesn't matter? I guess the important part of Stage 1 is DC resistance value presented to the OT? (Together with the inductor, of course!)

I'd appreciate if you could give some feedback on that idea. Thanks!



PS- One more question, if I go with the M2 design, I don't have a .9mH coil available, just 1mH or .82mH, does it matter which I take?

PPS- For the M2 design, to have about 3-4dB attenuation in Stage 1, I calculate R2A and R1A could be 15 ohms each, and R1 at 33 ohms. That would give the same 6.11 ohms to the OT, as the calculation for the original M-attenuator above.

 

[JohnH]

Hi @telesto , Thanks for your post

-7db is a sweet spot in the mathematics, based on having the right balance of input impedance and output impedance. Its possible to have more attenuation in Stage 1 (which would include another resistor) but not less unless values are allowed to drift away from target, which spoils the tonal consistency that this system has. It is this fairly rigorous approach to tone based on the maths that follows through to the end result.

Ive not found a front end for a multistage design with less attenuation than -7db that I wanted to put up. Could maybe stretch this to -6db.

Stage 2 is a resistive copy of stage 1, so it is also at this balance. Stages 3 and 4 have consistent output impedance seen by the speaker, but I allow their input impedances to vary since they are always used after stage 1 at least, and so little of this variation is seen by the amp.

That being said, Stage 4 which is the -3.5db stage can be used on its own as a resistive stage and the full M design on page 1 has this. A similar wiring could work in a version of M2 as well The amp sees about 10 Ohms when used this way which is close enough to 8. For this small initial step, the resistive only -3.5db stage works fine and in fact a good part of the real speaker reactance still reaches the amp. See the charts and sound samples on page 1. But the -3.5db stage is not a good enough basis for the front end for the whole system IMO.

Actually, when you try it, -3.5db is a really small step, more like a fine tuning, and if you need attenuation, you probably want at least -7db. So for most users, that -3.5db first-step option is not needed.

But, if you do need a very small reduction, another good work around is to set the whole box to max attenuation with no speaker and use it as a load box in parallel with a speaker, setting amp to a lower ohm tap.

Finally, just on your maths. When you have stage 1 or 2 with a 15 Ohm parallel and 10 Ohm series resistor, if you want to know what the amp sees, you also need to add the impedance of the speaker or following stages. eg, with stage 1 alone into an 8 ohm speaker, the basic resistive math is 15 in parallel with 10 plus 8, ie 8.2 so close to 8, rather than 6.

The full analysis needs to include the inductive reactances and the speaker reactances, so a SPICE sim is the best simple way to explore ideas, or in my case I put it all into a huge spreadsheet.

 

[telesto]

Hi @telesto , Thanks for your post

-7db is a sweet spot in the mathematics, based on having the right balance of input impedance and output impedance. Its possible to have more attenuation in Stage 1 (which would include another resistor) but not less unless values are allowed to drift away from target, which spoils the tonal consistency that this system has. It is this fairly rigorous approach to tone based on the maths that follows through to the end result.

Ive not found a front end for a multistage design with less attenuation than -7db that I wanted to put up. Could maybe stretch this to -6db.

Stage 2 is a resistive copy of stage 1, so it is also at this balance. Stages 3 and 4 have consistent output impedance seen by the speaker, but I allow their input impedances to vary since they are always used after stage 1 at least, and so little of this variation is seen by the amp.
.....
Finally, just on your maths. When you have stage 1 or 2 with a 15 Ohm parallel and 10 Ohm series resistor, if you want to know what the amp sees, you also need to add the impedance of the speaker or following stages. eg, with stage 1 alone into an 8 ohm speaker, the basic resistive math is 15 in parallel with 10 plus 8, ie 8.2 so close to 8, rather than 6.

The full analysis needs to include the inductive reactances and the speaker reactances, so a SPICE sim is the best simple way to explore ideas, or in my case I put it all into a huge spreadsheet.

Hi John,
Thanks for the reply. I had such high hopes for my back-of-the-napkin idea, but then you had to throw cold water on it with your research and spreadsheets, LOL That's OK, with your explanation, I think I understand now, that the -7dB is needed to "stabilize" things so that following stages don't throw things off too much. So your attenuator is indeed the "Lucky 7"

Ok, I'll have a look at SPICE and play around with that maybe and see what other ideas come up.

Oh, and for the M2 design, I can't get a .9mH inductor, only .82 or 1mH, does it make a difference?
Thanks!

 

[JohnH]

For the inductor in M2, around the range in question it really makes very little difference to a point where you'd never hear it and maybe couldn't measure it either. But just based on my OCD analysis, 0.82mH would be my choice, the optimum being somewhere between that and 0.9mH. This is to target the closest performance to match typical Celestion speakers in terms of what the amp sees. I think some US speakers may be a bit more inductive. It doesn't have much influence on the basic clean tone, but begins to have an effect as the power amp starts to overdrive.

 

[telesto]

That being said, Stage 4 which is the -3.5db stage can be used on its own as a resistive stage and the full M design on page 1 has this. A similar wiring could work in a version of M2 as well The amp sees about 10 Ohms when used this way which is close enough to 8. For this small initial step, the resistive only -3.5db stage works fine and in fact a good part of the real speaker reactance still reaches the amp. See the charts and sound samples on page 1. But the -3.5db stage is not a good enough basis for the front end for the whole system IMO.

I was thinking about this...so theoretically, I could bypass the first 3 stages and go straight to Stage 4 to have only the -3.5dB resistive attenuation. Then to go lower, I could switch in Stage 1 (with inductor) and then have -10.5dB (and then Stage 2 and 3 in succession).
...oh wait, this is what the dashed line and switch is doing in the original diagram...ah, ok, sorry, I have this habit of doing things first and then reading directions later

For the inductor in M2, around the range in question it really makes very little difference to a point where you'd never hear it and maybe couldn't measure it either. But just based on my OCD analysis, 0.82mH would be my choice, the optimum being somewhere between that and 0.9mH.

Ok, thanks, I was leaning towards the .82mH myself

 

[Gene Ballzz]

@telesto
Operationally, the most logical and gradual progression for your intended/desired scenario and usage, from minimum atttenuation to maximim would be: -3.5db resistive only > switch off the -3.5db which automatically activates the -7db reactive only (except when the whole unit is in full "bypass" mode) > then activate the -3.5db along with the -7db reactive (for -10.5db) > switch off the -3.5db and activate the second -7db (for -14db) > then switch in the -3.5db (for -17.5db) > then switch off both the second -7db & -3.5db and activate the -14db (for -21db) > then switch in the -3.5db (for -24.5db) > then switch off the -3.5db and activate the second -7db, along with the -14db (for -28db) >and then finally add in the -3.5db (for a total of -31.5db)! Remembering of course, that the -7db reactive stage is always active in all scenarios past the -3.5db only!

Where this all gets kinda tricky is in the wiring of the switching system, in that any stage that would get used by itself MUST use the bigger 100 watt & 50 watt resistors, as it needs to handle the full wattage of the amp. This is why once the -3.5db is deactivated, the wiring/switching of "Design M" in the first post of this thread, "forces" the activation of that first -7db reactive stage at all times to bear the full brunt of the amp's full wattage. This is also why both that -7db reactive and -3.5db resistive only stages have the bigger/higher wattage rated resistors and why either one of theose two MUST be activated at all times, except when the unit is in full bypass.

Through multiple builds, testings and usage by multiple folks it seems to fairly well point to the idea that the -3.5db only option is inconsequential enough to not really warrant the extra complications and expense of it's implementation! Also, with the use of the -3.5db alone, the speaker impedance comes more impotantly into play.

And while @JohnH has redesigned in a number of upgrades, to mildly tweak the tonal responses and operational variabilities, it has become my considered opinion that the most practical, simple, convenient, economical, easiest to build and useful design is the "Design M-Lite" in that same first post, in either dedicated 2 ohm, 4 ohm, 8 ohm or 16 ohm amp tap configuration. One of the great benefits of this design is that when using it with it's designed for amp tap, it doesn't much care what the speaker impedance is (within intelligent reason), except that if the speaker doesn't match the amp tap, it must get the amp tap changed to match the speaker when fully bypassing the unit. Even though there has been a redesign to accommodate multiple amp tap impedances and speaker impedances and incorporating a pile of additional features, that simplest "Design M-Lite" is so simple, easy and inexpensive to build, putting together an additional unit or two for different impedance amp taps seems a no brainer to me!

I use a 16 ohm "Design M-Lite" installed in every amp I own and couldn't feel more happy, pleased and liberated by the results! And although I've not yet built an 8 ohm, or other impedance unit (as I don't have any amp that is 8 ohm only), all reports of the 8 ohm variants seem right in line with what I've experienced. FWIW, the redesigns to accommodate 2 ohm and 4 ohm amp taps are in fairly recent posts. And yes, the "Design M-Lite" preserves the tone, feel and response better than any of the vast multitude of passive attenuators that I've tried since the early '80s, as well as providing consistently safe operation of your 50 watt or less tube amp. The only down side is that power tubes will not likely last as long, given the ability to run your amp "fully cranked into it's sweet spot" at all times!

And again, multiple thanks to @JohnH for his generous donation of time and efforts towards this amazing project design, along with detailed testing and sharing of the results!

My Apologies For The Long Novel!
Gene

 

[assaf110]

Hi John!
I need a stripped down version of the attenuator -
50w max
No need for switching stages - a constant 20db ish is fine.
One 8 ohm input, one 8 ohm out.
No compromise on tone.
Compact build.
How will you approach it? M2 design front end with a single 14db stage?

 

[JohnH]

Hi @assaf110 , yes that would work fine, to get -21db. If getting specifically -20 is needed, could make a small adjustment.

What's it for?

 

[assaf110]

Thanks for the Uber quick response.
I have the 16 ohm M design, use it on the Marshall SV. Works great.
I need to attenuate an incoming 35w amp MV (8 ohm 112 cab).
A grab & go small box would be great. I’ve noticed that I really don’t use the switches at all, once I settle on a certain attenuation level, I keep it more or less constant. The MV on this one should handle small changes, I will either plug the attenuator or unplug it if attenuation is not needed. -21 is great.

 

[JohnH]

Attenuators M and M2 compared - Analysis

At this time there are quite a considerable number of Design M attenuators built in various forms, and now a few M2 versions too. They are generally similar in many ways and both rely of putting inductance partly in series and partly in parallel with the input. But M uses two inductors while M2 uses only one. Although they have very similar performance they are not quite identical in how they work. So this is to compare them by analysis in a number of situations.

The benchmark is the tone of a full cab directly connected to an amp. An impedance model based on measured results of a G12M 4x12 is the basis.

In these charts, various attenuation levels are engaged but volumes are equalized to compare tone. The Red line is the signal at the fully connected real speaker, and also at the amp. Blue is the signal at a speaker after the attenuator, and green is the signal as seen by the amp. The ideal would be all three lines coincide.

The first two charts assume the effective output impedance of the amp is 20 Ohms, as measured on my VM from the 8 Ohm tap. -7db and -28db plots are shown.


Both designs show good tracking with red and blue lines very close. M2 tracks a bit better at very high frequencies, and the signal seen at the amp is a bit better too. Not much in it though.

But one of the interesting aspects of these designs is how they adapt to different amps, and to the same amps performing differently when overdriven. Amps with no NFB have very high output impedance at normal levels, while those with more NFB have less. Also, as an amp gets driven its output impedance drops dynamically. So here are three further tests, all at -14db, with the amp output impedance changing from 20 ohms, to 50, then 8 then 2 ohms:







See how in both designs, the blue output curves follow the red at frequencies above low mids, even at these very different amp values from those assumed in the design. Both are good but M2 seems to be a little better at the tracking.

Actually, there is some magic happening here. The output response curves are developed by the real speaker interacting with an output impedance, and the lower that impedance the flatter the response.. And in all these calcs, the attenuated speaker is always seeing the same impedance, and yet even through the attenuator the differences in response due to different amp impedance are feeding through fairly closely.

The Bass peak does not adapt in this way, but it does stay within about 3 db of the fjll-volume red curvs except for the extreme low output ohms.

That's it for now. M and M2 both seem fine and quite similar, M2 is my favorite at the moment (haven't built it myself yet though)

 

[Mcentee2]

Attenuators M and M2 compared - Analysis

At this time there are quite a considerable number of Design M attenuators built in various forms, and now a few M2 versions too. They are generally similar in many ways and both rely of putting inductance partly in series and partly in parallel with the input. But M uses two inductors while M2 uses only one. Although they have very similar performance they are not quite identical in how they work. So this is to compare them by analysis in a number of situations.

The benchmark is the tone of a full cab directly connected to an amp. An impedance model based on measured results of a G12M 4x12 is the basis.

In these charts, various attenuation levels are engaged but volumes are equalized to compare tone. The Red line is the signal at the fully connected real speaker, and also at the amp. Blue is the signal at a speaker after the attenuator, and green is the signal as seen by the amp. The ideal would be all three lines coincide.

The first two charts assume the effective output impedance of the amp is 20 Ohms, as measured on my VM from the 8 Ohm tap. -7db and -28db plots are shown.
View attachment 72876
View attachment 72875
Both designs show good tracking with red and blue lines very close. M2 tracks a bit better at very high frequencies, and the signal seen at the amp is a bit better too. Not much in it though.

But one of the interesting aspects of these designs is how they adapt to different amps, and to the same amps performing differently when overdriven. Amps with no NFB have very high output impedance at normal levels, while those with more NFB have less. Also, as an amp gets driven its output impedance drops dynamically. So here are three further tests, all at -14db, with the amp output impedance changing from 20 ohms, to 50, then 8 then 2 ohms:

View attachment 72872

View attachment 72874

View attachment 72873

See how in both designs, the blue output curves follow the red at frequencies above low mids, even at these very different amp values from those assumed in the design. Both are good but M2 seems to be a little better at the tracking.

Actually, there is some magic happening here. The output response curves are developed by the real speaker interacting with an output impedance, and the lower that impedance the flatter the response.. And in all these calcs, the attenuated speaker is always seeing the same impedance, and yet even through the attenuator the differences in response due to different amp impedance are feeding through fairly closely.

The Bass peak does not adapt in this way, but it does stay within about 3 db of the fjll-volume red curvs except for the extreme low output ohms.

That's it for now. M and M2 both seem fine and quite similar, M2 is my favorite at the moment (haven't built it myself yet though)

Well, my M2 works so well I haven't felt the need to contribute more to the thread Lol!!!

Mind you, I am constantly looking at tweaks to make it more useful for me outside of the attenuation.

I added a line out, and am also thinking of adding a head-swap function, ie two inputs, one speaker out, with the attenuation (or none) on one side, and a simple 8r 100w load on the other.

I need to look at safe ways of switching before I attempt that, I think

 

[telesto]

Hi @telesto , Thanks for your post

-7db is a sweet spot in the mathematics, based on having the right balance of input impedance and output impedance. Its possible to have more attenuation in Stage 1 (which would include another resistor) but not less unless values are allowed to drift away from target

Hi John, coming back to this, I'm curious why would another resistor be needed for more attenuation in Stage 1? It's not possible to just adjust the values of the resistors without adding an additional one? For example, just replace the -7dB stage with the -14dB?

 

[JohnH]

hi @telesto
Stage 1 needs to have particular input and output impedances. As a resistive (ie no inductor) network, this generally requires three resistors of defined values to achieve a given attenuation level. But at around -7db, with the resistance values I was targeting, one of those resistors gets to zero, leaving two. This is how Stage 1 works at low frequency.

If you consider the M2 Stage 1 at low frequencies, the inductor has no effect and R2A and R2B can be considered combined to one value of about 10 Ohms. That, and R1 at 15 Ohms do the-7db attenuation.

 

[telesto]

hi @telesto
Stage 1 needs to have particular input and output impedances. As a resistive (ie no inductor) network, this generally requires three resistors of defined values to achieve a given attenuation level. But at around -7db, with the resistance values I was targeting, one of those resistors gets to zero, leaving two. This is how Stage 1 works at low frequency.

If you consider the M2 Stage 1 at low frequencies, the inductor has no effect and R2A and R2B can be considered combined to one value of about 10 Ohms. That, and R1 at 15 Ohms do the-7db attenuation.

Hi John, Thanks for the explanation, but I'm afraid I don't quite get it. In the original design R9 is shorted out by the inductor, is that what you mean by "going to zero"?
But then looking at the M2, do the resistor values need to be in any special formula, ie: -7dB and adding up the ohms to give the OT 8 ohms to see?
For example, in M2, if R2A and R2B are 60 and 68 ohms, and R1 is 10 ohms, the OT will see 8 ohms, and attenuation will be above -7dB (somewhere between -7dB and -14dB). Will that work?

 

[JohnH]

R9 was for a different job, at high frequencies,

In your example, the net result may be safe enough but will not sound right with many amps because it doesn't have the right output impedance. Gotta look at the whole system all at once, amp, resistors, inductors and speaker, which all interact. Changing a value or two in isolation can give a random tonal result. Personally that would be a something I wouldn't want to do, having done a lot of work and testing to keep tone consistent.