[jacobmj]

I'd only shorted the diodes for one channel.  Shorting all four changes nothing.  Traces of the 295V and 305V points are below, this time with a 1X probe, so believe the 2mV/div on the screen.  These traces are before shorting the diodes, and are with the Mullard 5AR4 rectifier.  They look identical after shorting the diodes.

295V


305V


I notice that the 295V trace is rather wide, on the same order as the noise at the output.  I couldn't tease any signal out of it at shorter timescales.

~Jacob

[poty]

The traces are normal. For 295V there is some noise. I'm not sure where it is from. Ideally the noise should be blocked by rather high value of C411 in high frequencies and C400 on low, but the absolute value for the noise is around 0,3mV which is 72dB comparing to the output signal of the stage. I don't think it can pose a problem, but who knows... I'm still for trying to shunt the existing caps with a spare temporalily and listen.

[jacobmj]

Okay, adding 0.65uF across C411 doesn't change the noise in level or sound.  I do believe it reduces the width of the trace some.

Out of curiosity, I also tried tying each of the tube side of R428, R427, and R429 to ground across this capacitor.  Only that of R429 eliminates the noise, as I had previously tested.

I'd wondered if might have a bad ground on that side of the board.  Running a separate wire doesn't make a difference.

~Jacob

[jacobmj]

Hello again.

I had a chance to substitute a few more components to try.  I have temporarily substituted many of the components around V400.  All of of the following, together, and in various combinations, were tried as I progressed through them.

R437, R429, R430, R428

I also added 1uF across C411, without removing C411. I tried grounding the grid of V400, before and after R427.

All of this was done on the left channel, without any change to the right.  There was no change in the noise.

Thanks. ~Jacob

[poty]

All your testing shows that the noise is generated by V400 tube, the socket or the PCB traces to the socket.
You can temporarily substitute the tube with the resistors across the anodes and cathodes (watch the heating voltage!) pulling out the tube and insert the resistors straight into the socket. If the test eliminates noise - the problem is with the tube. If not - you should look at experimenting furter.

[jacobmj]

Hi.

I've reverted the changes from yesterday, and tried without the tube.  I've substituted 18ohms for the heater, which gets me to about 6.15V--close enough, I figured. To maintain the same voltage at the grid of V401, I calculate that I need around 127K in place of V400.  I used 120K between the plate and cathode of each of L and R.

I can report that the noise decreased by at least 5dB (~55-58dB), measured as before at the tweeter.  My spl meter doesn't accurately measure lower.  The noise is of the same character by ear.

I'll admit that this noise is now only barely audible at 1m.  I could definitely live with a tube that behaved the same, especially if I keep attenuating -10db after the Chime.  I've ordered an old GE blackplate 5751 to try.

~Jacob

[jacobmj]

I don't have a sense of how relevant this is, though I'll mention it anyway...

I'm able to increase the sensitivity of my measurements to 400uV/div by cascading the scope channels.  Once I do this, I'm able to see the noise a bit more accurately, and can trigger on mhz components of it.  I've looked at the 295V and output.  This requires a bit more careful probe usage, as using the ground clip or tip hook picks up quite a lot of noise.

These signals are always present (Chime off, plugged in or not), though increase by 5-10x in magnitude when it is on.  They continue for 1-2 seconds when the Chime is unplugged from the wall during the test.

There is a dominant 94.5mhz signal of the same amplitude as the "noise", if I look at longer timescales and apply a 20mhz bandwidth filter.  I assume this is the local 94.5 radio station.  I also see strong signal at about 14.5mhz.  I can't seem to reliably scope anything much slower, but I suspect there are slower components because the trace becomes progressively more blurred and wider as the timescale is increased.

I suppose this would be much better if I had a slower bandwidth filter.

~Jacob

[jacobmj]

I thought I'd attach a few traces of the B+.  Before R425L, I notice considerably less noise than anywhere after.  I've done my best to capture this in a way that can be compared with static pictures.  These use a 10x probe, so are 50mV/div.

Between R424L and R425L:


Between R438L and R425L (305V):


Adding 1uF to the 305V portion doesn't change the trace.

Thanks again for the help.
~Jacob

[poty]

I'm now traveling so can't answer quickly.
First of all, the source is outside IMHO, but with some "help" inside. You should find the source of the noise or try to shield the input stage as carefully as possble.
Second - the capacitors in the PS filter are there for shorting the power line noise too, but in your case - they are not doing their job. Such big frequency should be definitely eliminated by the design. Try to decide what causes such behaviour.
I don't think changing tube helps in this case, maybe only slightly if the tube would have less gain.

[jacobmj]

Hi.  Thanks again for the assistance.  I have a few more data
points to discuss.

My overall strategy has been to work forward from the transformer
toward the plates, disconnecting everything after a certain point
(e.g., after R424R/L).  For much of this, I disconnected the 8V
transformer from the mains, and the heater rectifier from the main
transformer.  (It turns out doing so doesn't change the results.)
Since I'm unsure of the audibility of everything I'm measuring by
oscilloscope, I've also been connecting various points directly to the
coupling capacitor to listen to them.  I've only done so only when I
have enough load to bring the voltage under 400V.

Is there a risk of damage to the transformer or rectifier tube if
I let B+ float with no load?  It reaches about 415V.

Working forward with only one channel connected to power:

 - Disconnected R438 and added 36k load to ground: 305V point to
   output capacitor. Obvious hum on the speakers, and a clear ripple
   on the scope.  No white noise, however.

 - All connections through R437.  36k load at 295V point.  305V point
   to output capacitor.  Still hum.  No white noise.

 - All connections through R437.  36k load at 295V point.  295V point
   to output capacitor.  Much less hum.  The voltage ripple certainly
   decreases.  No white noise.

 - As above, but using a 150k/120k voltage divider for about 140V
   before the output capacitor.  (This is as with the typical grid
   voltage on V401.)  I didn't use R429, or the trace on the board
   connecting it and R430.  No white noise.

 - V401 present, with enough load on 295V point for appropriate
   voltage.  305V or 295V points connected to output capacitor (not
   the tube).  No white noise.

 - Voltage divider connected to V401 grid, through R430.  Output
   connected to V401 as on schematic.  The level of hum present on the
   295V point remains audible.  Now, there is white noise.  Adding
   0.68uF to ground at R430 stops the noise.

 - Ground from near V401 to output.  No noise.

 - Ground from near V400 to output.  No noise.

I have not yet substituted the path from R432, Q402, and R431 to
ground.

Am I likely filtering the noise when I connect the rail voltages to
the output capacitor, in a way that doesn't occur when being driven
from the tube?  That is, is my test appropriate?

For the hum, my take is that this can be addressed by adding more
capacitance at the 305 point.  Is this correct, and the most effective
means?

I'd read about connecting the grey wire in the transformer to the
chassis ground point.  I've done this.  I've also twisted most of the
pairs of leads to and from the transformer windings.  I can't say that
there's any difference.

Also, I did receve a 5751 tube.  It happens to sound great and does
have less gain.  Interestingly, I do notice that the left channel (R
on board) has exactly the same increase in noise over the right
channel as with the Telefunken 12AX7.  The Electro Harmonix tube has
equal levels of noise.  I find it unbelievable that both tubes could
have the same defect, on the same plate.  Is there some other possible
explanation?  I don't yet have a spare socket to reverse the wiring to
swap the channels at the tube.

~Jacob

[jacobmj]

I can report that bypassing the cathode CCS on the left channel with a resistor to ground, holding the cathode at about 145V, does not affect the white noise.

~Jacob

[jacobmj]

I had a chance to sit down for a few more tests.  This time, I continued to test various points through the output capacitor.  I used a pair of cheap (so as to not be sad if I destroy them) headphones to listen.  I could hear nothing when connected directly to the output--not surprising--, but could when using the amp's speaker level outputs.  Of course, this gives me plenty of signal to hear!

With this connectivity, I am able to hear noise on the grid of V401, with or without the tube installed.  There is a hum that is much louder, but some audible white noise.  The white noise is considerably (10 times?) louder at the cathode of V401.  There isn't as much hum after V401. 

I can, maybe, see a factor of about 2x the noise amplitude between these points by oscilloscope, measuring at the OUT RCA.  Here's a comparison:

(All:  Top trace is 400uV/div, bottom trace 2mV/div.  The scope channel 2 is cascaded into 1.)

C409 connected to ground at V401.   This shows the cleanest signal my measurements might be expected to produce.


This is C409 connected to the grid:


Here's to the cathode:


And, to the plate:


I am tempted to assume that the difference in amplitude between the grid and cathode is due to the low impedance after V401, and a higher impedance before.  Is this a reasonable assumption?

The noise is still not present on the high voltage rail, or the ground, with or without the tubes installed.  Hum, of course, is.

Listening after R429, I can hear the noise only when V400 is present.  I also removed V401 to avoid applying a large voltage to its grid.  Substituting R429 doesn't change the noise.

Thanks again.
~Jacob

[poty]

Sorry for long delay in answering - Being on the move is not very pleasant way to analyse something! Another sorry for long answer.
Let's sum your testing:
- voltages mostly OK.
- tubes presumably OK (replaced without effect), many experiments with sockets allows to presume the contacts are not oxidized.
- all places that can have colder joints retouched.
- shorting the serial diodes in the power lines tried.
- grounding the grid of V400 - no effect, then - the noise is not from inputs (removed HagDAC and all input cables, shorted the inputs - just different ways of achieving the same result).
- grounding the grid of V401 - eliminate the noise. The stage has no gain (=1) so it may be "noise at very low level" option: the noise is the same just there is no amplification. This thought supported by the fact that "equivalent V400" - resistor drop the noise (no amplification by the resistor stage), but not eliminate it.
- disconnecting the device from mains with no effect - mainly the source of the noise is outside, but may be a bad part (strange for both channels - then only small probability).
- the noise is outside ("These signals are always present (Chime off, plugged in or not)") and amplified by the V400 when it is on. The detected frequency of the noise in the power line is very high though and cannot be the direct source of the noise.

Some answers to your questions:
"Is there a risk of damage to the transformer or rectifier tube if I let B+ float with no load?  It reaches about 415V."
No.

1)
"- As above, but using a 150k/120k voltage divider for about 140V before the output capacitor.  (This is as with the typical grid voltage on V401.)  I didn't use R429, or the trace on the board connecting it and R430.  No white noise."
vs.
"- Voltage divider connected to V401 grid, through R430. Output connected to V401 as on schematic. The level of hum present on the 295V point remains audible. Now, there is white noise. Adding 0.68uF to ground at R430 stops the noise."
2)
"Am I likely filtering the noise when I connect the rail voltages to the output capacitor, in a way that doesn't occur when being driven from the tube?  That is, is my test appropriate?"
3)
"I am tempted to assume that the difference in amplitude between the grid and cathode is due to the low impedance after V401, and a higher impedance before.  Is this a reasonable assumption?"
IMHO the first two points have been explained by the third sentence. But it can be something unusual in R430s placements (sort of puzzle as far as we have 2 channels with the same effect).

"For the hum, my take is that this can be addressed by adding more capacitance at the 305 point.  Is this correct, and the most effective means?"
As soon as you are not suffering from the hum it's not the first problem you should think about. Increasing the capacitance is a good thing and helps in many cases, but you should consider the value of the capacitor to not harm the rectifier tube and to balance the resulting good income from it.

"I find it unbelievable that both tubes could have the same defect, on the same plate.  Is there some other possible explanation?  I don't yet have a spare socket to reverse the wiring to swap the channels at the tube."
Why defect? The difference between halves of the dual tubes is not defect (to some degree) and is not unusual at all. You don't have to have a spare socket - just swapping the connections of R430 to the grids of V401 would be enough.

We have already huge set of data. In my opinion the source of the noise is outside (with small probability that it is injected by bad trace/contact in the area of R429-R430-grid of V401). The gain of V400 together with very high input impedance of the stage on V401 aggravates the influence. Lets do a test: Having the sort of external voltage divider you use in the previous tests connect the middle point of it directly to the grid of V401 (should be in place). Listen to the noise. If the noise is gone - the problem should be inside IMHO. If not - the only way of mitigating the noise without modifying the schematic is to shield carefully the Chime.
You can try also to connect C409 directly to the plate of the V400 (without V401 in place) just in case I overlook something about the CCS in the buffer stage.

[jacobmj]

Thanks much for the assistance!

I finally received enough parts to separately build the output section on an external PCB.  I figured this would be the best chance of determining what is at fault. 

I am taking power from after R424, and have constructed the identical circuit as the Chime schematic between R424 and R438. I've used a TIP50 for the CCS.  I'm using a 150k/~120k voltage divider as input to the grid resistor (R430) of the 12AU7.  Heater power comes from the Chime, in which I now have a 6.3V regulator in lieu of R423. 

It's not beautiful, and I expected some hum.  Here are a few snapshots:



Additionally, I have build a 60dB (switchable to 40dB) op-amp based measurement pre-amplifier.  (The design is here: http://tangentsoft.net/elec/lnmp/).  This has allowed me to see the noise much more clearly by oscilloscope, but also works great with a pair of headphones in parallel to the scope.  It has extremely low self-noise, even at +60dB gain.  It also gives me an easier way to test than using my hifi amp and speakers.

Notably, with most anything (say, an unterminated RCA cable) connected to the measurement amp, I am able to perfectly listen to an AM station of 1250khz.  This station has a 5000W transmitter 2.5 miles from my house.  When connected to the Chime output, the AM station is obvious only when powered off.  Once powered on, it goes away. 

This (AM reception) is also true of my external 12AU7 setup.  With it, I do get hum, as might be expected given my shoddy construction and long power leads.  The most disappointing aspect is that I can hear and measure the same noise as when using the Chime output.  This appears to exclude all of those components as the source of the noise.  I did use the same Electro Harmonix tube, so it could be at fault.  The only other one I have is a Telefunken, and the result is the same.  It is not dependent upon the grid voltage, which I adjusted with a 25k pot in series with 100k to ground.  Running the heater from a SLA battery changes nothing.

If I disconnect B+ from my external board, I believe the noise persists for the second or so it takes to lose power.  I'll have to add much more capacitance to it to get more runtime to be sure.

I'm not entirely sure what to make of this.  Should I be ordering a third 12AU7 tube to try?  Should I be shielding with lead?

Another aspect I've read of is heater to cathode leakage.  I believe I saw mention in Morgan Jones' book that this could lead to 1/f noise.  I haven't yet tried biasing the heater voltage, which I could easily do with my battery.

Now that I have a more portable way to test, I can take it elsewhere and try as soon as I can.

~Jacob

[poty]

Heater power comes from the Chime, in which I now have a 6.3V regulator in lieu of R423.

I'm glad it works for you too!

... I am able to perfectly listen to an AM station of 1250khz.  This station has a 5000W transmitter 2.5 miles from my house.  When connected to the Chime output, the AM station is obvious only when powered off.  Once powered on, it goes away. 
This (AM reception) is also true of my external 12AU7 setup.  With it, I do get hum, as might be expected given my shoddy construction and long power leads.  The most disappointing aspect is that I can hear and measure the same noise as when using the Chime output.  This appears to exclude all of those components as the source of the noise.  I did use the same Electro Harmonix tube, so it could be at fault.  The only other one I have is a Telefunken, and the result is the same.

The only cause of the noise I can think of now is external. It may be possible that the only SS device in the signal circuit brings about detection of radio station and so adds the noise, because usually there are no such wide-band source in the air.

It is not dependent upon the grid voltage, which I adjusted with a 25k pot in series with 100k to ground.

You almost cannot change the grid voltage because of the CCS in the cathode.

Running the heater from a SLA battery changes nothing. If I disconnect B+ from my external board, I believe the noise persists for the second or so it takes to lose power.  I'll have to add much more capacitance to it to get more runtime to be sure.

OK, you'd done this before and the test says that the noise is not from the mains (or at least - the filtering eliminates the noise from the mains).

I'm not entirely sure what to make of this.  Should I be ordering a third 12AU7 tube to try?  Should I be shielding with lead?

I don't believe it is tubes, but who knows... I'm almost definite in view that the last stage doesn't have any relation to the noise. Having high-impedance amplifier now - could you try to feed it directly with the divider you use for the grid biasing?
Shielding - the only way of eliminating the noise IMHO.

Another aspect I've read of is heater to cathode leakage.  I believe I saw mention in Morgan Jones' book that this could lead to 1/f noise.  I haven't yet tried biasing the heater voltage, which I could easily do with my battery.

Yes, the noise is possible. I believe I read several complaints about that even in this forum. But you've said the noise is gone when you ground the grid with the help of a capacitor. As soon as the leakage is between the heater and the cathode the noise is injected to this part of the circuit (cathode-ground) and should not be affected by grid-ground manipulation. But it is definitely worth to try - just connect one of the heater pin to cathode while using a battery for heating.

Regards,
Vladislav

[jacobmj]

The only cause of the noise I can think of now is external. It
may be possible that the only SS device in the signal circuit
brings about detection of radio station and so adds the noise,
because usually there are no such wide-band source in the air.

Last week, I did substitute a resistor for the CCS transistor.
The noise remained.  I haven't tried this with my external
circuit.  Incidentally, none of my other electronics in the house
show evidence of broad noise.  I'd think if it were present in
the air, and so overwhelming, I would notice in other equipment.

I'll see about rigging some shielding.  I don't think I've got a
metal box large enough to house a UPS for isolated power, the
circuits, possibly a scope, and me with the headphones.  So, I'll
necessarily have some wiring in and out.  Will this fatally
compromise any test?

You almost cannot change the grid voltage because of the CCS in
the cathode.

Correct.  I'm not able to change it much at all.

I don't believe it is tubes, but who knows... I'm almost definite
in view that the last stage doesn't have any relation to the
noise. Having high-impedance amplifier now - could you try to
feed it directly with the divider you use for the grid biasing?

Actually, the pre-amplifier isn't all that high-impedance.  I
built it at 600ohms.  I also didn't put a high voltage input cap
in it to be able to handle such a high DC offset.  Through the
coupling capacitor for the Chime, I don't hear anywhere near the
same order of noise at the 150k voltage divider. (At 60dB gain,
everything is somewhat noisy.) 

I can probably pass the signal through my oscilloscope first to
get very high impedance, and listen to it.  I'll report back...

Yes, the noise is possible. I believe I read several complaints
about that even in this forum. But you've said the noise is gone
when you ground the grid with the help of a capacitor. As soon as
the leakage is between the heater and the cathode the noise is
injected to this part of the circuit (cathode-ground) and should
not be affected by grid-ground manipulation. But it is definitely
worth to try - just connect one of the heater pin to cathode
while using a battery for heating.

Okay, we can exclude this as a factor.  Biasing the battery to
140V with 100k to the cathode makes no difference.

~Jacob

[poty]

Incidentally, none of my other electronics in the house show evidence of broad noise.  I'd think if it were present in the air, and so overwhelming, I would notice in other equipment.

DIY stuff has obvious limitations in design and parts, so it can't be compared to professionally built equipment in the field. I'd not disregard some system incompatibility and hidden design errors in the equipment either DIY and industrial.

I'll see about rigging some shielding.  I don't think I've got a metal box large enough to house a UPS for isolated power, the circuits, possibly a scope, and me with the headphones.  So, I'll necessarily have some wiring in and out.  Will this fatally compromise any test?

I see your joke here. Actually in your testing the only part of the circuit which adds to the noise is the "output side of the V400 - input side of the V401" as far as blocking this part returns the noise level to acceptable margin regardless of any wiring, scope, UPS, parts of your body wearing the headphones ...

Actually, the pre-amplifier isn't all that high-impedance.  I built it at 600ohms.  I also didn't put a high voltage input cap in it to be able to handle such a high DC offset.  Through the coupling capacitor for the Chime, I don't hear anywhere near the same order of noise at the 150k voltage divider. (At 60dB gain, everything is somewhat noisy.)

I don't see how you got to input impedance of 600 Ohms... Anyway - the capacitor must be high voltage - good point! Your trying added limited input impedance of your amplifier to the ~120k of "output impedance" of your divider. The input impedance of the V401 is almost infinity, so it can pick up noise. Usually it doesn't happen (in my replica circuit for example), so something around your build is wrong. If all you've done to the time is right - the only thing left - the external influence of some nature.

[jacobmj]

Thanks again for your assistance.

I see your joke here.

I'm glad. Only half a joke though...  I haven't tried further with this, but I
do have some other measurements, below.

I don't see how you got to input impedance of 600 Ohms...

From my understanding of op-amps and its circuit, the "LNMP" I build
should have an input impedance equal to the first resistor.  In my
case, I used 600ohms there.  For any of my high voltage measurements,
I'm using C409 and R435 in the chime connected by a jumper to an
insulated probe.  I'm not able to say what the resulting load would
be.

(For all measurements of the external trial board, I always measure at
the output, and haven't changed its schematic.)

I continue to second-guess my measurements, since they strike me as
incredibly unlikely.  To help, I've now got a good (USB Audigy) sound
card for performing FFT in the audio band.  This does reveal that
there is actually a substantial difference in the spectrum of the
external board.  The following is a comparison of using the same half
of the Telefunken 12AU7 either in the chime, with all normal
connections (fushia), or in my external test board (green). 

This is measured at the output, and amplified by 40dB before input
into the computer.  I'm not positive of the absolute value of dB on
the axis, though relative values are correct.  For reference, the
fushia line measured about 75mV peak-peak on my oscilloscope.

Ignoring the increase in 60hz harmonics, there is a large difference
at higher frequencies.  By ear, I still contend that these don't sound
very different.



This next plot has the same fushia line in addition to the following,
described bottom to top at the 10khz position:

* NC: No connection.  This is my 3 inch jumper floating.
* v401gridvol0: V401 grid.  Volume control is at nil.
* V400gridvol0: V400 grid.  Volume control is at nil.
* 295V: 295V rail, at the top of R429.
* V401platevol0: V401 plate. Volume control is at nil.
* V401cathodevol0: Normal output connectivity.  Volume control is at nil.



There does seem to be some volume dependence of the noise.  At nil or
full volume, 60hz harmonics are more audible.  They are reduced at 50%
volume.  The noise sounds lower at 100% volume, though I can't
rationalize this with the following plot.

Here, from bottom to top at the 10khz position:

* chimeoutvol100% (red): Volume at 100%.  Note that this is identical
  to 0% at high frequencies.  To ear, the 60hz harmonics sound louder
  than 0%, despite what this plot indicates.  The noise sounds lower.
* chimeoutvol0 (golden): Volume at 0%.  This is identical to the
  fushia line I described above.
* chomeoutvol33%: Volume at approx. 33%.  The noise does sound louder.
* chimeoutvol66%: Volume at approx. 66%.  This is the maximum loudness
  of the noise that I can achieve.

This plot was made with a complete Chime.  I did try removing the
HagDAC.  It didn't make any difference in the measurements.



Incidentally, these plots seem to be reduced in quality once I upload
them to the forum.  The originals are at
http://jjoseph.org/photos/projects/hagerman_chime/.

Thanks again.
~Jacob

[poty]

Just thoughts, no suggestions yet.

From my understanding of op-amps and its circuit, the "LNMP" I build should have an input impedance equal to the first resistor.  In my case, I used 600ohms there.

I often get mixed up the matter, but if you are shure - then - it should be so... It's strange that measurement amplifier has such low input impedance though, in most cases affecting measurement accuracy...

Ignoring the increase in 60hz harmonics, there is a large differenceat higher frequencies.  By ear, I still contend that these don't sound very different.

In my opinion the fushia line shows signs of V400 circuit influence - that is why there are differences. The n*60Hz frequencies are more prominent on "open" case because of lack of shielding.

This next plot has the same fushia line in addition to the following,
described bottom to top at the 10khz position:
* NC: No connection.  This is my 3 inch jumper floating.
* v401gridvol0: V401 grid.  Volume control is at nil.
* V400gridvol0: V400 grid.  Volume control is at nil.
* 295V: 295V rail, at the top of R429.
* V401platevol0: V401 plate. Volume control is at nil.
* V401cathodevol0: Normal output connectivity.  Volume control is at nil.

While it is not exactly clear - your "air" environment (noise) is almost exactly repeated in the circuit. It's a pity - I don't have the Chime, so can't say anything about external EMI shielding capabilities of the build, but as far as I have read in the forum - there are many users of the Chime in JH-recommended case and only small numbers with problems. Lets assume it is not the cause of the noise. I ask you to check once more if your mains outlet have "the ground wire" connected to the house ground bus somewhere? Then I'm not sure the legend is correct. For example, according to the legend the amount and "picture" of noise at grounded by the volume control grid of V400 is the same as in 295V rail.

There does seem to be some volume dependence of the noise.  At nil or full volume, 60hz harmonics are more audible.  They are reduced at 50% volume.  The noise sounds lower at 100% volume, though I can't rationalize this with the following plot.

Have you short-circuited inputs? Then it explain the dependence: at nil the V400's grid is grounded by the volume control, at 100% - by the short-circuiting jumper, the volume control has logarithmic characterictic, so the max noise should be not the middle point. The fact that you do not hear any difference is for very small differences in response in the margins of frequences where our ears have maximum sensitivity.