Simple Attenuators - Design And Testing

emann

New Member
Joined
Aug 24, 2022
Messages
26
Reaction score
12
Hi again. thanks for that pdf Gene however the box I opted for a slightly larger box to have more spacing as well as maybe some more ventilation since the room at the roof tends to get pretty warm during summer. The circuit I am doind is also a bit different then yours for my fender 65 deluxe reverb reissue as per below.

The grid below is the one I came up with for the ventilation holes - 8mm diameter. Do you think 8mm is an overkill and maybe 6mm will be better? Holes will be vertically on top of each other on box and lid.

Attenuator Vents.jpg

curcuit I am building:

M2 layout and schematic211231.gif


What do you think of this arrangement please?

Also re my previous question on the paste, is it just a matter of applying some paste on the base of the components and then screw them in place or is there some other precaution to take please.

Thanks to all.
 

SwedishWings

Member
Joined
Oct 9, 2009
Messages
52
Reaction score
11
Ive been updating the M2 schematic, to catch up with basic tweaks over the last couple of years.
Thanks John!

Looking at the power ratings, it looks like most resistors are over-dimensioned (with 50W amp, basic version, 8Ω).

If i understand correct, worst case dissipation would be ruffly R1=25W, R2A=12W, R2B=10.2W, R5=6.5W, R6=0.9W, R3=5.2W, R4=2.4W, R7=2.8W and R8=2.8W.

Edit: corrected, I mixed up order of stages.

In free air with ambient temp up to 70C, a derating factor of 0.8 is common. What is the reason for the large derating, in particular for R2x through R8?
 
Last edited:

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Thanks John!

Looking at the power ratings, it looks like most resistors are over-dimensioned (with 50W amp, basic version, 8Ω).

If i understand correct, worst case dissipation would be ruffly R1=25W, R2A=5.8W, R2B=6W, R5=4.5W, R6=3.5W, R3=6.5W, R4=1W, R7=2.8W and R8=2.8W.

In free air with ambient temp up to 70C, a derating factor of 0.8 is common. What is the reason for the large derating, in particular for R2x through R8?

Well I'm not an electrical engineer, but I am an engineer in other fields, so I have a tendency to analyse stuff and to tend to be conservative! I work out the values for power rating by calculating power in every component at every switch setting. I run the numbers using the nominal amp power, then I add 50% for amp overdrive and x2 for safety and temperature control, so basically x3 total on nominal amp power.

So take R1, on that basis based on 50W in, it dissipates 28W max, so x3 and the next common power rating is 100W. R2A and R2B are trickier to work out because their response is frequency dependent. So 50W for them is on the safe side. Of the others, some can get up to around 20W with that x3 factor included, so 25W is a good common value. Once we are at a 25W rating, it's a fairly compact unit so they might as well all be similar.

I also note that Arcol rates their values using a standard heatsink, which is 995cm.sq for the 100W. We don't get as much surface as that, but if I read it right, unmounted you just get 30W rating. Also, many of us buy the resistors unbranded from Asia, where those labelled as 100W are smaller. They work well, but don't want to push them too hard!

Screenshot_20221024-215719_Samsung Internet.jpg

I'm pretty happy with the ratings that I put, but the units still get hot, especially at 50W or more, and fans are needed by the time you get to 100W amps.
 

ellipsis

New Member
Joined
Oct 18, 2022
Messages
15
Reaction score
7
Location
London, UK and Riga, Latvia
Hi @ellipsis , very close, but swap the tip and ring lugs! Currently it shows the main hot connection going to the ring.

Thanks @JohnH , that's just me drawing the jack symbol incorrectly. I always assume tip is on the outside and ring in the middle. Just need to go back and redraw the position of the "little v things"
 

ellipsis

New Member
Joined
Oct 18, 2022
Messages
15
Reaction score
7
Location
London, UK and Riga, Latvia
Ive been updating the M2 schematic, to catch up with basic tweaks over the last couple of years
This is great, many thanks @JohnH !

I see that this version now has the post-reactive stages in descending order of attenuation, -14, -7, -3.5. The previous versions had -7, -14, -3.5, and I seem to recall you saying somewhere on this thread that this previous order was slightly electrically better, but suggested to at least mount the switches on the panel in descending order. Any particular reason for this change, or does it simply make wiring cleaner and explanations easier, with no/minimal tonal difference?
 

SwedishWings

Member
Joined
Oct 9, 2009
Messages
52
Reaction score
11
Well I'm not an electrical engineer, but I am an engineer in other fields, so I have a tendency to analyse stuff and to tend to be conservative! I work out the values for power rating by calculating power in every component at every switch setting. I run the numbers using the nominal amp power, then I add 50% for amp overdrive and x2 for safety and temperature control, so basically x3 total on nominal amp power.

So take R1, on that basis based on 50W in, it dissipates 28W max, so x3 and the next common power rating is 100W. R2A and R2B are trickier to work out because their response is frequency dependent. So 50W for them is on the safe side. Of the others, some can get up to around 20W with that x3 factor included, so 25W is a good common value. Once we are at a 25W rating, it's a fairly compact unit so they might as well all be similar.

I also note that Arcol rates their values using a standard heatsink, which is 995cm.sq for the 100W. We don't get as much surface as that, but if I read it right, unmounted you just get 30W rating. Also, many of us buy the resistors unbranded from Asia, where those labelled as 100W are smaller. They work well, but don't want to push them too hard!

View attachment 117823

I'm pretty happy with the ratings that I put, but the units still get hot, especially at 50W or more, and fans are needed by the time you get to 100W amps.

Agreed that being conservative is mandatory. But I suspect derating for several resistors might be a waste of money. In the graph below: typical derating for wire wound aluminum housed resistors with limited heat sink gives a hint (ref is 900cm^2). Curve B is for Vishdale RHA series 25W size. Note that derating is 50% without heat sink. For R5, R6, R3, R4, R7 and R8 10W would suffice even in a small box.

Screen Shot 2022-10-24 at 7.41.09 PM.png

I just ordered parts to wet theory in practice. To be sure, I ordered the cheapest Chinese stuff i could find :) R5, R6, R3, R4, R7 and R8 are all 10W cement type that will us direct convection cooling only.
 

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Hi @SwedishWings , thanks for your assessments. It sounds like you are confident in this area and clear on what you are doing.

Ill just make a couple more points:
6 or 7 watts max in the smaller resistors, with 50W coming in, gets to around 10W if a 50W amp is driven hard, as some may wish to do Derating on a 25W resistor, assuming (conservatively) that heatsink is much less than the standard (so zero), could be 12.5W, which is a good spec in comparison to 10W and the heatsinking that is available helps. A 10W air-cooled ceramic inside a warm box is getting very hot!.
These resistors from china are just a dollar or two, so nothing really to save unless smallest size is important
When all the resistors are case mounted, theres a good amount of mass and everything heats up slowly, which averages out peaks very nicely over about 20-30 minutes. Air mounted resistors will heat quicker.

Just my thoughts and I wish you all the best for your build!
 
Last edited:

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Load boxes and using M3 as a loadbox

Hi @ellipsis


For a pure loadbox, no speaker, I dug into it again to explore what the best design could be.

The M3 attenuator works well as a loadbox, and as an attenuator. But if its just a loadbox, it can be simplified, and also optimized to make it a bit better for that one purpose.

So I came up with this just following my nose to get the best curve that I could. It ended up a bit different:

loadbox 221025.gif

The curves are dB vs frequency plots. The blue one is from a more complex model that matches real 4x12 measurements very closely - and I use it for these calcs. The red is the circuit above, also allowing for losses in the coils and cap. Its a very close match. I let the width of the bass peak be a bit wider on the low side to cover a bit more range.


TGP Loadbox Thread - Freeman and Aiken

This kind of circuit has been posted before, and there is a classic thread on TGP by James Freeman, building on designs by Randall Aiken:


He comes up with something very similar and has used it for years.


Use of an M3 Attenuator as a load box

Now turning to the M3, when it's used as a load box, its also good, almost as close:

M3 loadbox 221025.gif


The same reference curve is now the orange one and green is M3 switched to full attenuation with no speaker. To optimise it for this use, I changed L1 slightly from 0.9mH to 1mH (will make negligible difference to the attenuator use) Its also a really close match except at the bass peak where its about a further dB down than the design shown above. Thats because as an attenuator, energy needs to be drawn off from the front end to feed the speaker, adding a bit more damping and supressing the peak. But this effect is localized just to when playing frequencies at the peak (around low A), over a one or two semitones range, and I dont think its a big issue.

I hope that's interesting or useful!
 
Last edited:

ellipsis

New Member
Joined
Oct 18, 2022
Messages
15
Reaction score
7
Location
London, UK and Riga, Latvia
So I came up with this just following my nose to get the best curve that I could.

That's great, @JohnH , thanks so much for continuing to share your impressive work.

On most or all of this thread, the "benchmark" impedance curve is that of a closed-back 4x12 cabinet, right? And you optimize your circuits to match that curve as closely as possible. Do you have any idea what the impedance curve for something on the other end of the spectrum, say an open-backed 1x10 cab, might be, and how robust the attenuator designs here might be for non-4x12 scenarios?

there is a classic thread on TGP by James Freeman, building on designs by Randall Aiken:

Oh great, another deep rabbit hole to go down! ;-)

I hope that's interesting or useful!

Very much so!
 

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Hi @ellipsis

With regard to different cabs, and also different amps, I was surprised early on to find that having designed the attenuator circuit based on measurements from just my main amp (a VM combo), and tweaking a little based on my DSL401, it worked very well for many others as well. Quite honestly I haven't heard of a single build out of somewhere between 50 and 100 that I've read, that hasn't been reported as sounding good. So most of those will be with 12" speakers, some open and some closed back, but smaller combos like the Studio series have 10" open back, and there are several generically different output circuits out there with different tubes, different NFB and biasing arrangements.

So there's something lucky about this attenuator circuit to make the planets align. I can see it coming through in the maths but its still hard to rationalise. What I think is as follows:

There's two aspects to the performance that are key.

One is how the amp reacts to the attenuator, which affects dynamics and how the tone develops as you dig in hard, and as a note decays. This is all about the reactive Stage 1. The most important part of this is the treble rise determined by L1. This controls the feel and tone at the amp end, and if its right, then it covers all frequencies from low mids upwards Any 12" speaker seems to be well covered by the L1 value listed, whether its Celestion or others, and its not affected by open/closed back. I expect that for a 10", it might be a bit lower, but that turns out not to be very important for the attenuator use (but maybe in a load box scenario).

The bass rise is not represented at the amp in the M2 design, and that turns out to be totally fine for attenuator use because the real speaker does it itself. But it is important when used as a loadbox, unless some EQ adjustment is made downstream. An open backed cab will have a lower resonant frequency, close to that of a speaker in free air (which is what speaker data tells us). To adapt for that, L2 and C1 would be larger values. The difference might be around 80hZ instead of around 110hZ with closed back

The second key point is how the speaker reacts to the attenuator. One of the design parameters is the output impedance of the attenuator which needs to be kept consistent and fairly high, like a real amp. Then the speaker reacts as itself, so the sound you get is the resonance and treble rise of the real speaker, which will adapt to whether it is open or closed back etc a different speaker will sound different

When you investigate the signal when there are several stages of attenuation involved, you see the amp reacting to the reactive front end, but mid-way through the circuit, the response is almost flat in terms of tone, though it contains the dynamics from the amp. Then at the speaker, the speaker does its part to provide its own bass resonance and treble lift. This is why its not critical to change the attenuator design for different speakers

There's one more trick: All of the design assumptions are based around starting with the effective output impedance of the amp, which varies with different amps and with frequency, and is dependent on NFB and resonance and presence controls, if they exist. I based the design on around 20 Ohms for an 8 ohm output, which I measured, but its just an assumption. But it turns out that with different amp output impedances, the reactive front end shapes the tone appropriately, so that you still get a really close match at the speaker.

To summarize for your question - M2 and M3 are tonally very robust in terms of being used as attenuators and no changes are really needed for different speakers. As a load box with no speaker, M3 is better than M2 (or the dedicated load box as above), and the tone shape will be based on the design assumption for the front end, ie a closed back 12", unless its adapted.
 

Marcomel79

Well-Known Member
Joined
May 3, 2021
Messages
974
Reaction score
1,312
Location
Oslo
Hi @JohnH, i just wanted to ask you a question and i apologize if this has been asked before, but with more than 150 pages it would be impossible to find. I have built your M2 50W attenuator which works beautifully, again thank you so much for your work. Can i use it as a dummy load with all the attenuating stages engaged?

Marco
 

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Hi @JohnH, i just wanted to ask you a question and i apologize if this has been asked before, but with more than 150 pages it would be impossible to find. I have built your M2 50W attenuator which works beautifully, again thank you so much for your work. Can i use it as a dummy load with all the attenuating stages engaged?

Marco
Yes you can and it will give the amp a safe load to run width. It will pick up the main treble rise, but since theres no bass circuit, it wont have the bass peak which only affects a few low guitar notes. Still useful though. If you are feeding into a mixer and cabsim or IR loader , you might want to add a touch of extra bass EQ. I get a good result via a mixer with some added bass then cut below 100 Hz
 

SwedishWings

Member
Joined
Oct 9, 2009
Messages
52
Reaction score
11
Hi @ellipsis

There's one more trick: All of the design assumptions are based around starting with the effective output impedance of the amp, which varies with different amps and with frequency, and is dependent on NFB and resonance and presence controls, if they exist. I based the design on around 20 Ohms for an 8 ohm output, which I measured, but its just an assumption. But it turns out that with different amp output impedances, the reactive front end shapes the tone appropriately, so that you still get a really close match at the speaker.

Thanks, again, for your efforts. I'm still waiting for components and am truly excited to hear the results!

Meanwhile, an observation on output impedance from tube amps. I have measured this before on HiFi and guitar amps, and found that the output impedance is lower than what you found in your amps. I took the time to simulate the power amp of my home brew 50W (2xEL34, 8Ω output) guitar amp which i have valid data on (transformer inductance and DCR, etc).

The output impedance varies as expected depending on NFB. I also tested several other parameters like plate voltage, transformer inductance etc, but output impedance variance was small in comparison.

The presence control affect the feedback which affects the output impedance. The first graph is presence at zero and the second presence at max. The circuit for presence is from the Marshal plexi.

zout.png

zout presence max.png

Not sure if this is of any interest, but figured someone might be interested.
 

jmerwin62

New Member
Joined
Sep 26, 2022
Messages
6
Reaction score
1
I have a totally nieve question.

I've used an L-Pad attenuator before. It worked fine on my low power tube amp (Epiphone Valve Jr- I sold it a few years ago). It was limited to a specific impedence (8 or 16 Ohm I don't remember), but that's OK, just limits the amps that it can be used on, and there are other L-Pads for other impedences, so I could make an 8ohm and a 16ohm attenuator. The 50 Watt version of the L-Pad is $20 on amazon (see link below). There's a few DIY versions of this idea out there. Some add bright caps an other options if you're willing to add complexity and components.

It seems like the complexity of the M2 design is brought on by trying to vary the attenuation and hold(or select) the impedence. The design of the L-Pad seems to handle setting the impedence.

The other part of the M2 secret sauce is the inductor.

So, couldn't I just place an inductor across the input on the L-Pad and see improved results over the L-Pad only design with significantly simpler design? I'll admit inductors and chokes still sort of baffle me.

Or maybe a slightly more complex design that has R1, R2a, R2b, L1 and the L-pad?



 

Gene Ballzz

Well-Known Member
Joined
Feb 19, 2013
Messages
3,952
Reaction score
4,881
Location
Las Vegas, NV
I have a totally nieve question.

I've used an L-Pad attenuator before. It worked fine on my low power tube amp (Epiphone Valve Jr- I sold it a few years ago). It was limited to a specific impedence (8 or 16 Ohm I don't remember), but that's OK, just limits the amps that it can be used on, and there are other L-Pads for other impedences, so I could make an 8ohm and a 16ohm attenuator. The 50 Watt version of the L-Pad is $20 on amazon (see link below). There's a few DIY versions of this idea out there. Some add bright caps an other options if you're willing to add complexity and components.

It seems like the complexity of the M2 design is brought on by trying to vary the attenuation and hold(or select) the impedence. The design of the L-Pad seems to handle setting the impedence.

The other part of the M2 secret sauce is the inductor.

So, couldn't I just place an inductor across the input on the L-Pad and see improved results over the L-Pad only design with significantly simpler design? I'll admit inductors and chokes still sort of baffle me.

Or maybe a slightly more complex design that has R1, R2a, R2b, L1 and the L-pad?




FIRST> :welcome: to the forum!

THEN> One of the more important aspects of the @JohnH design is a specific and critical balance/ratio of series and parallel resistance, that simply can't be easily achieved with an L-Pad type design. While it could likely be accomplished, to do so would be far more complicated (and probably costly) than the M2 design! Throughout this thread there have been multiple endeavors to add a variable component to cover the "in between" gaps of the -3.5db steps. These attempts have yield varying results and been generally deemed as not worth the effort. The base version of the M2 works amazingly well with minimal loss of tone, dynamic and tactile feel and response. This design is by far the best passive design on the net, until possibly getting to the much higher priced units, like the full sized Tone King IRONMAN II, at +/- $800.

I will leave the technical discussion to @JohnH , who will likely chime in at some time soon.

Happy Attenuatin'
 

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Hi @jmerwin62 , thankyou for your post.

Most of the lower cost attenuators on the market use L-pads or similar, and as you note, they are usually very simple and there are several DIY versions out there. They may work acceptably depending on the uses and expectactations, but in general they can't fully preserve the tone and dynamics of the amp as they turn down. Typically they become to some extent, dull and flat sounding. It's not possible to properly control all the required parameters in these units. If it was, we would not have this thread.

Our designs aim higher, and the target is to fully match the amp tone and feel across a range of optimised steps, in a design that is still a simple DIY build. The switched steps allow optimisation at each stage, which is not possible in a single swept control.
 

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Thanks, again, for your efforts. I'm still waiting for components and am truly excited to hear the results!

Meanwhile, an observation on output impedance from tube amps. I have measured this before on HiFi and guitar amps, and found that the output impedance is lower than what you found in your amps. I took the time to simulate the power amp of my home brew 50W (2xEL34, 8Ω output) guitar amp which i have valid data on (transformer inductance and DCR, etc).

The output impedance varies as expected depending on NFB. I also tested several other parameters like plate voltage, transformer inductance etc, but output impedance variance was small in comparison.

The presence control affect the feedback which affects the output impedance. The first graph is presence at zero and the second presence at max. The circuit for presence is from the Marshal plexi.

View attachment 117943

View attachment 117944

Not sure if this is of any interest, but figured someone might be interested.

These are interesting results. I've seen numbers like these quoted from measurements on some Plexi or EL34 amps, but not modelled across such a sweep of frequencies and control positions. Is your model able to adapt to represent other amps? It would be very interesting to see results with say, the NFB taken out, or maybe a JTM45 circuit with KT66 tubes (it's the basis of my VM), or one with no NFB such as an AC30.
 

pietro.castelli

Active Member
Joined
Aug 18, 2021
Messages
62
Reaction score
114
Question for @JhonH and whoever built the attenuator so far:
Should/could the sleeves of the input/output jacks be grounded to the metal enclosure?

Thanks!
 

JohnH

Well-Known Member
Joined
May 31, 2008
Messages
5,177
Reaction score
3,449
Location
Wilton NSW
Question for @JhonH and whoever built the attenuator so far:
Should/could the sleeves of the input/output jacks be grounded to the metal enclosure?

Thanks!

Hi pietro, I reckon the case should not be grounded, and with the plastic Cliff jacks that we recommend, then the case is not in circuit at all.
 

Latest posts



Top