'T' attenuator response interactions at higher attenuations

[moamps]

Thank you so much. So, the contribution of the noise and system measuring errors is very small at HF.  Next step I would do is replacing a each section of the attenuator with a fixed resistor one by one (middle section of the attenuator with a small resistor (1,5ohm metal film))  and measuring the FR in the -56dB position. I'm not sure, however, does it worth the time and hassle.

 

[Ilya]

Thank you so much. So, the contribution of the noise and system measuring errors is very small at HF.  Next step I would do is replacing a each section of the attenuator with a fixed resistor one by one (middle section of the attenuator with a small resistor (1,5ohm metal film))  and measuring the FR in the -56dB position. I'm not sure, however, does it worth the time and hassle.

Well, yes. This is kind of a PITA to do. If it turns out that the culprit is the middle section, I wonder if sticking in a permanent 1R resistor would help?

 

[emrr]

Here's the same test as original post, unit wired directly into Audio Precision ATS-1 with 150Ω source and 600Ω load.
You can see a similar slight boost before it starts moving more dramatically at the end of the throw.  Sweeping the cursor on the plot confirms the visual. 

5:00 on the knob (fully CW, minimum atten)

9:00

8:00

7:30

7:00 (fully CCW, max atten)

Back to MOTU 16A/Spectrafoo, here's 4 of the same amplifier unit compared with same gain, and maximum output attenuation into 10K. 

I don’t see this as a problem in this case, after all, I don’t expect anyone is running a preamp with -30 dB or greater of OUTPUT attenuation, and if they are it’s for overload effect.


Next is the same amp as the OP, built in ’T’ set for minimum attenuation (1K load), connected to external Daven decade ’T’ loaded with 600Ω except highest trace (-40, light blue) loaded with 10KΩ. 
Purple is -40.  The -40 into 600 and -40 into 10K plots are the same shape when visually aligned. 
Green is -50 from -40 setting on 10dB/step control plus -10 on 1dB/step control.  This is two pads in series. 
Yellow is -50 from -50 setting on 10dB/step control.

Note this type ’T’ behaves to greater attenuation, and you can also compare to the post above showing the same ’T’ by itself showing a treble loss due to different loading conditions.

full size pic HERE

In the case of the 1176 one might be running that much attenuation on the input, and should be aware of response changes that are blended with compression artifacts. 

 

[Ilya]

I found that shorting the hot to ground at the input (before the T-pad) makes the HF bump go away. Shorting cold to ground doesn't have any effect on the bump though.
Also, inserting a small resistor between the shunt  section and gnd has no effect. I have no simple way of replacing the other sections with fixed resistors unfortunately.

 

[CJ]

do side by side test with a Daven and see what happens

 

[emrr]

do side by side test with a Daven and see what happens

I did, see above.

 

[craigmorris74]

This thread has made me paranoid about using a box I built with two attenuators that I bought from a forum member here.  Just tested those using Voxengo SPAN, annd it looks like there isn't really any difference in response when attenuating 12 dB.  I would never be doing 30 dB of attenuation (I don't think), so I don't imagine the more extreme cuts we see here apply to my workflow, but wanted to make sure I'm testing properly.

Thanks,
Craig

 

[Ilya]

This thread has made me paranoid about using a box I built with two attenuators that I bought from a forum member here.  Just tested those using Voxengo SPAN, annd it looks like there isn't really any difference in response when attenuating 12 dB.  I would never be doing 30 dB of attenuation (I don't think), so I don't imagine the more extreme cuts we see here apply to my workflow, but wanted to make sure I'm testing properly.

You have nothing to worry about. This behavior is valid for deep attenuation only. 30dB and more. 12dB is perfectly fine.