Simple Attenuators - Design And Testing

[telesto]

Ok, yes, I can see how Excel has more advantages in speed/control/versatility, as you said. I don't know Spice well enough yet, and different versions vary, but yea, might be possible to do something close or similar there as well.

 

[Dwayne Eash]

-also some random things on attenuation-
-i have a EH 12AX7 that one day went south, amp went almost silent, i turned it up and it sounded really good but still quiet, ended up turning every nob on the amp up to 10.... Completely BAD ASS!!! but its a time bomb and i took it out and put it in a marked box, i pull it out every now and then for 15 minutes of the sound in my head

-always use a variac @ 80-90v-

What are the sonic results of this, in terms of volume reduction and quality of tone? Since I am an inventive type, and have been wanting a variac for a long time, so that you say it's a good tool for the jamming guitarist, really interests me.

I have an '82 JCM 800 all tube, vertical inputs. Im trying not to fry my two 25 Watt Scumbacks, but at home jamming, I never get above 4 MV once in a blue moon if the neighbors are gone, so I only rarely go above 3 1/3 on the MV.

I have a newer pair of inexpensive 15" PA speakers for my living room music studio. They are ok sounding, and so my PA can get as loud as band practice, no problem. I can't wait to get my amp up to speed tho, so I don't have to play 4+ on the MV to sound good enough. That's just way too loud at home, so I need some attenuation.

So I am very interested how much quieter my JCM 800 might get, using a variac. I have no clear idea, what happens to the volume, other than going "somewhat" lower.

 

[Dwayne Eash]

After doing some calculations, going from 120V to 80V, would be roughly 40V of reduction. And that amounts to a 1/3 reduction of volts (which "effects" power or watts, but is not the same as watts) or 33.3333% voltage reduction. My amp is the 2204 circuit, 50 watt, '82 JCM 800 4010 but with two speakers. So there's two cabs, for my one amp.

I lucked out making the contact for the purchase, here on this forum about 10 years ago or so.

Hard to believe I've had it so long. I paid like $225-250, just to have it returned to stock, and to take some of the high end out. It was a two visit process, for that to happen.

He asked me which power tube to return it to, I opted for the EL34's, as that's the most notable tube of classic rock fame, at least by the mid 70's. So even if I would end up preferring some other power tube, I first wanted to hear the iconic glory of the original EL34 beast, itself. hehe

I realize a proper JMP Master (Super Lead) is what I should have, if I want the best in vintage and classic rock tone and feel. I played one in Chicago once, and OMG, it nearly melted me on the spot, hehe, I fell in love. It even had something golden with it, like the cloth was golden or the edge corner piping was gold, IDK.

And so, I have the strongest "gold" concept, in my head, because when I plugged into this thing with an SG Standard, it just sounded like gold, so much better and so vintage. The best amp I ever played. I'm for some dirt, so I'd live in the high input channel and just clean it up with the guitar vol.

But costing 2 and 3+ times more than a JCM 800, I decided on the purest old school JCM 800 was the best I could afford. And then put in the best NOS tubes for the first spot, and the best speakers. And I do not cloud up the signal path, keep things simple and pure for the best in organic guitar sonic bliss.

I am a wanna be tone hound, a poor man's version. Gotta love a proper tube amp guitar rig.
Somehow, I figure you can relate.

 

[telesto]

Generally speaking, when you lower the plate voltages to the tubes, either by Variac or VVR, it causes the tubes to break up at lower volumes. EVH was famous for using a Variac on his Marshall on early VH albumns to get his "brown sound". You can also have your amp modded to get a VVR (Variable Voltage Reducer), which has a MOSFET connected to the tubes plate voltage so you can regulate it directly in the amp (the MOSFET soaks up the high voltage and gives a lower one to the tubes). So you can get cranked amp tone at low volumes. I haven't tried a VVR, but reviews are generally good. However, one thing to consider is that it's not pushing the output transformer as a cranked amp would, so the tone may be a little off. If you use an attenuator between the amp and speaker (eg. the one discussed in this thread), then you will be driving the OT hard as well, which would probably give a closer tone to that of the actual cranked amp. The other negatives of VVR is it's a little more complicated/costly, and I also hear the MOSFETs can become unreliable down the road. So in short, I would say skip the Variac and VVR, and just go with a good external attenuator if you want to crank your amp at lower volumes.

 

[telesto]

Hi John, This is my setup in TINA so far. I imported Mike's Greenback impedance plot (upper graph) and compared it to the measured impedance at the amp out (bottom plot). Two questions: why is the range bigger on the measured output? Mike's speaker is going from about 15 to 85 ohms, and I'm simulating between 20-40. It's the difference between the points being tested? Amp out vs speaker out? Or? And second question, why does the upper frequencies on Mike's speaker tail upwards, and in the Spice plot tail downwards?
Thanks

 

[JohnH]

TINA looks like its a powerful and useful free tool, I might try it.

But Im not sure what is happening in the lower plot, it doesn't look like impedance Ohms. Calculating impedance from first principles, I get the matching graph posted on the previous page. Actually the model looks great for plotting output signals and for adding the attenuator, but not for working out impedances, not sure how best to do that in SPICE. You may get the right shape if you up the amp impedance to a huge value and then plot voltage. Voltage will then be proportional to the circuit impedance.

What is the SP1 speaker component? is it anything different to a 13.5 Ohm resistor? Also, there are upper and lower points that have to be matched if it is plotting impedance. At very low frequency, the impedance has to be 13.5 Ohms, since everything else is shunted by inductors that have no impedance at low frequency. Similarly, at very high frequency the model has to tend to a limit of 13.5+40+2.3 Ohms, being the sum of resistances except for R5 which is bypassed by the cap. That flattening of the upper tail is a correct interpretation of the Spice model and a difference to the real plot, but its at very high frequency above the acoustic response of guitar speakers. It doesn't matter to much to us, this speaker model is just a piece of test kit, not actually part of what we are designing.

When Im tweaking a design to check performance of new arrangements and component values I look most closely a the range up to about 5kHz

 

[telesto]

The Y-axis on both my plots is Ohms. On the bottom plot I made a formula to divide the voltage of TP1 by current thru AM1 to get the impedance. The curves match up pretty good, but I was just wondering why the ohm values are different. Yea, maybe if I change the output power of the amp or something it could have an affect. I'll play around a little. (Actually, voltage of TP1 alone gives the curve shape, so I'm not sure if it's necessary to do the V/I ?)

SP1 is a "speaker" logo, but only contains ohm values, nothing else (it looks nicer than just a resistor )

You said you calculate up to 5kHz, is that the effective range of a guitar speaker (I guess I can Google that, but wondering what your rational is) Looking at Mike's Greenback plot, the "ripples" seem to stop around 5kHz and smooth out after, so yea maybe after that something happens. Human ear can go up to 20kHz tho, so I guess it's still audible (?)

Here's a link to TINA, free download, just register: http://www.ti.com/tool/TINA-TI
I'm still playing with a few different Spice versions in parallel (Micro-Cap, 5Spice, LT, Qucs, TINA), but I find TINA to have a very powerful and intuitive UI. Yea, give it a shot

 

[JohnH]

I think the plot is confused by the 20 ohms amp resistance. You could try exactly what you have done with say a 1000ohm amp resistance, but it may still be off due to phase difference between current and voltage.

Yes guitar speaker acoustic response drops off sharply above 5kHz

 

[telesto]

I think I am also confused by the 20 ohms amp resistance Looking at Aiken's original plot, he doesn't have the amp resistance in the circuit, and is getting the impedance plot in Spice by doing V/I. His actual formula is V(V1:+)/I(R1). I am not using the ":+" in my formula, which may be why my impedance plot has the right curves, but is "shrunk" on the Y-axis (not sure, still looking into this).

I see in your circuit, you don't have an Ammeter attached (un-attenuated circuit), and so you take the voltage after the 20 ohms you defined for the amp, which also gives the right curves, because it still reflects the impedance/current changes of the speaker. (Then I guess you feed that into a spreadsheet and do the rest in there). These are two slightly different ways of doing things, but still getting a similar result. I don't know how much of a difference it makes in the end, would have to make simulations both ways and compare.

 

[telesto]

Hi John, I have a couple of observations/comments:

I've looked at some other attenuator/dummy-load circuits and none of them have the internal resistance of the amp figured into the calculation (aside from the need for nominal impedance matching). The circuits are just calculating the input signal coming from the amp. In your un-attenuated circuit, your 20 ohm internal amp load (R8) is directly in series with the 6.75 ohm speaker (R45). From the point of view of the circuit, it's no different than a 26.75 ohm speaker.

Second strange thing I see, is that in your Stage 1 attenuator (and all other stages as well), you have the series resistor AFTER the shunt resistor. This puts it in series with the speaker instead of the input, which makes for a totally different DCR calculation than what I am used to seeing in speaker attenuators. I didn't go deeper into this yet, but was this intentional?

For example, take a typical L-Pad calculation from Weber's website:



In this case, R2 and the speaker are in parallel, so you would take the parallel calculation of them, and then add the R1 series circuit and get the total DCR. If you move R1 after R2 (as in your circuit), then R1 is directly in series with the speaker, so then you would add R1 plus the speaker resistance together, and then take that value in parallel with R2 to get the total resistance.

 

[JohnH]

it's all intentional. Getting the attenuator to feed the speaker a correct and consistenr output impedance is hardly written about at all, but it is the first and most important key idea in this design.

Nearly all basic passive attenuators use the normal L-pad diagram, including Weber. As you turn it down, output impedance drops which flattens the bass peak and treble rise.

 

[telesto]

Ok, it works out, when I plug in 6.5 ohm speaker to your Stage 1, I get about 8.1 ohms. Whew, I was worried there for a second
Actually, your post above mirrors what I just read on Aiken's site about L-Pads

There are two ways to make an L-type attenuator:

(1) Match the impedance of the L-pad in the direction of the series arm. This is the "traditional" L-pad configuration. In this case, the input impedance stays constant, and the output impedance gets lower as you increase the attenuation, down to a theoretical minimum of zero ohms. This configuration sounds like crap for a guitar amp, because the decreasing output impedance increases the damping factor and removes all "tubeyness" from the tone at lower volumes, because the amp no longer reacts to the variations in speaker impedance.

(2) Match the impedance of the L-pad in the direction of the shunt arm. In this case, the input impedance also stays constant, but the output impedance gets higher as you increase the attenuation. This is great for guitar amps, up to a point. The increasing output impedance lowers the damping factor, which enhances the interaction between the amp and the speaker, giving a natural bass and treble boost as you increase the attenuation (sort of a "built-in" Fletcher-Munson effect compensation!). The problem is that it gets to be too much, and you end up with too much bass and treble boost and "hangover effects" as it was called in the old days of audio, and you get a flubby, fizzy tone at high levels of attenuation. This can be alleviated by limiting the maximum output impedance with an additional shunt resistor at the output.

This is an eye-opener for me, and need to get my head wrapped more around this. I've got a L-Pad wired into my amp in the first way Aiken describes (and the way Weber shows), and it indeed sounds like dog poop (but at least keeps my wife and neighbors from killing me ) So from what I see, you have taken the second implementation that Aiken describes? My dislike for L-Pads are suddenly changing. By coincidence, I also saw Elektra has an L-Pad wired in together with some fixed value resistors as the MV on their 185 amp (have a closer look here). This looks really interesting to me, and something I want to explore as I design my own amp (eventually).

...altho I noticed the Elektra doesn't have any inductor (that I see), and Aiken mentions that his earlier amps had reactive attenuators, his next generation had only resistive, so makes me start to wonder what benefits the inductor really brings...

 

[JohnH]

RA's points in item (2) is one reason why each of our stages is fixed, and optimised to provide the intended consistent output impedance. The -7db stages also provide close to the right input impedance too. The other stages don't do this, but they are always after Stage 1 so their tendency to throw the input Z off is diluted by stage 1 and the amp sees an adequately consistent input Z.

The inductor is the second key idea, and the way it is provided partly in series partly in parallel is unlike any other design that I know of. If this is not correct, it can throw the tone off balance again. As designed either in design M, or M2, it shows the amp the right inductance so the amp reacts to it, then it balances itself out so the signal passes through the resistive parts with very little phase shift, ready to be recreated at lower volume when it gets to the real speaker. The inductor makes a difference when the amp gets driven hard and also helps to adapt to different amps. See my last series of plots.

 

[telesto]

Thanks for the explanation. While I'm digesting all of that, I'm also playing with Spice in parallel. I've put in a simple resistive only L-type attenuator, but no matter what R values I put, I keep getting the same waveform output, the only difference is it's shifted down on the Y-axis (ie: attenuated). I found the diagram below explaining impedance plots, and it also seems to imply that the DCR is not affecting the curve of the impedance plot, but just moving it up/down on the Y-axis, and it's the reactive inductance of the speaker shaping the signal. If this is correct, then I'm confused, as you are always tweaking to make input/output waveforms to match, so I was expecting to see some "bad" output plots when I threw in some crap values for the resistors. Am I missing some variable in Spice? (Changing power source settings also doesn't seem to have an effect.)

 

[telesto]

it's all intentional. Getting the attenuator to feed the speaker a correct and consistenr output impedance is hardly written about at all, but it is the first and most important key idea in this design.

Nearly all basic passive attenuators use the normal L-pad diagram, including Weber. As you turn it down, output impedance drops which flattens the bass peak and treble rise.

Ok, now I see. You're not just looking left-to-right (presepctive of the amp), but also looking right-to-left (perspective of the speaker. You started out with a U-shaped attenuator on Page 1, and progressed from there. Ok. I was also looking at U-attenuators, and found them used on some microphone attenuators, but true, don't hear it much talked about when it comes to guitar attenuators. Weber's L-pad design gives the correct 8 ohms to the amp, but the speaker is only seeing about 2 ohms, which can't be good. Ok, I will now read the technical summary (Page 1-46) with that in mind

 

[Dwayne Eash]

...

However, one thing to consider is that it's not pushing the output transformer as a cranked amp would, so the tone may be a little off. If you use an attenuator between the amp and speaker (eg. the one discussed in this thread), then you will be driving the OT hard as well, which would probably give a closer tone to that of the actual cranked amp. The other negatives of VVR is it's a little more complicated/costly, and I also hear the MOSFETs can become unreliable down the road. So in short, I would say skip the Variac and VVR, and just go with a good external attenuator if you want to crank your amp at lower volumes.

Seems to me, using a VVR and a power attenuator, both, is the better solution. Using the VVR lets the power tubes break up earlier, which in my case is the main thing sonically lacking.

So the VVR get's me most of what I need. But also using an attenuator, pushes the OT in a more natural way than just VVR.

And since I'm also using VVR, I don't need to use quite as much power attenuation, and using less power attenuation, generally equates to better transparency to the amp, or, less disturbing the tone.

 

[Dwayne Eash]

Generally speaking, when you lower the plate voltages to the tubes, either by Variac or VVR, it causes the tubes to break up at lower volumes. EVH was famous for using a Variac on his Marshall on early VH albumns to get his "brown sound".

If you can reduce volts before the power tubes, how about also increase the voltage before the OT, like we do in raising the bias to the power tubes, so they wont fail so quickly when we drop the volts so low, as was suggested to be done when using a variac to lower the volts?

 

[telesto]

Seems to me, using a VVR and a power attenuator, both, is the better solution. Using the VVR lets the power tubes break up earlier, which in my case is the main thing sonically lacking.

So the VVR get's me most of what I need. But also using an attenuator, pushes the OT in a more natural way than just VVR.

And since I'm also using VVR, I don't need to use quite as much power attenuation, and using less power attenuation, generally equates to better transparency to the amp, or, less disturbing the tone.

VVR gets earlier tube breakup, but doesn't push the OT. Attenuator will push your OT, but puts a metal wall between your OT and speaker so they don't interact naturally. Pick your poison. Or take both and get the worst of both worlds Well, actually, I don't know, anythings possible, beauty is in the ear of the beholder. So play around and find what works for you.

If you can reduce volts before the power tubes, how about also increase the voltage before the OT, like we do in raising the bias to the power tubes, so they wont fail so quickly when we drop the volts so low, as was suggested to be done when using a variac to lower the volts?

VVR reduces the power AT the power tubes, and the power tubes annode is connected to the OT primary. The VVR or Variac will raise/lower them together.

 

[Dwayne Eash]

VVR gets earlier tube breakup, but doesn't push the OT. Attenuator will push your OT, but puts a metal wall between your OT and speaker so they don't interact naturally. Pick your poison. Or take both and get the worst of both worlds Well, actually, I don't know, anythings possible, beauty is in the ear of the beholder. So play around and find what works for you.

...

But of course, I would use some sort of reactive circuit, so that the amp is fooled into thinking it's all good and speaker like. @@ I mean, we are chatting with the speaker reactive emulation man, hehe JohnH, after all. DIY economy, and we're good.

Can't be very hard to install a VVR, but where to get a good one?

 

[telesto]

But of course, I would use some sort of reactive circuit, so that the amp is fooled into thinking it's all good and speaker like. @@ I mean, we are chatting with the speaker reactive emulation man, hehe JohnH, after all. DIY economy, and we're good.

Can't be very hard to install a VVR, but where to get a good one?

True, John has done alot of work to find a way to blaze a trail over or thru the metal wall of attenuation
A VVR is basically just a pot and MOSFET, but you have to know exactly where to put it, and make sure it is done well, as you are tapping into the highest voltage in the amp, 300-500 VDC range.
Someone on this forum installed one, you can read here: http://www.marshallforum.com/threads/18w-vvr-installation.71431/