[Steve]

Several points I consider for perception of slight loss of mid/highs, explained for ease of understanding.

1) In my experience, I have found that even minute drops in high frequency response (FR) is consistently perceptible. Over the years, I have performed many tests with individuals/auditioners perceiving changes with alterations of 1 part in 180,000 to a resistor in a zobel network across the highs. That is approximately -105db or so from the reference. (We are talking FR change, not SPL change over the entire audio band.) (Perception will depend upon the amount of masking from setup, venue, components etc.

2) The FET output devices (if I assume the wrong type of output devices, please forgive me) operated as source followers; there are several considerations.

     A. The junction capacitances of output FETs are large, but is much lower in Source follower circuits by definition. Now parallel FETs and the capacitance is much larger (add capacitances of each that is paralleled). The voltages across the FET also affect the junction capacitances.

     B. As the load impedance drops, the junction capacitance increases due to lower voltage gain (which is less than 1 anyway). This is by definition.

     C. The input Miller capacitance is also affected by decrease load due to less gain, thus the previous stage's load is affected.   

As one can see, altering the load causes several alterations in the performance. As such, I can understand the perception of slight alteration of the mid/highs. Whether this answers your question AndyR is another matter.

Cheers

Steve         

[twitch54]

wow ........for a minute there I was worried, but hell, I can't hear anything above 15khz anyways !

[Steve]

wow ........for a minute there I was worried, but hell, I can't hear anything above 15khz anyways !

How did you come up with 15khz?   

Olson's work identified that the higher the harmonic the easier to perceive changes in the amplitude.
Thus instruments and voices are critically affected by very small changes.
Rane also performed some research pertaining to FR perception. Pretty interesting stuff.

Cheers

[andyr]

Several points I consider for perception of slight loss of mid/highs, explained for ease of understanding.

1) In my experience, I have found that even minute drops in high frequency response (FR) is consistently perceptible. Over the years, I have performed many tests with individuals/auditioners perceiving changes with alterations of 1 part in 180,000 to a resistor in a zobel network across the highs. That is approximately -105db or so from the reference. (We are talking FR change, not SPL change over the entire audio band.) (Perception will depend upon the amount of masking from setup, venue, components etc.

2) The FET output devices (if I assume the wrong type of output devices, please forgive me) operated as source followers; there are several considerations.

     A. The junction capacitances of output FETs are large, but is much lower in Source follower circuits by definition. Now parallel FETs and the capacitance is much larger (add capacitances of each that is paralleled). The voltages across the FET also affect the junction capacitances.

     B. As the load impedance drops, the junction capacitance increases due to lower voltage gain (which is less than 1 anyway). This is by definition.

     C. The input Miller capacitance is also affected by decrease load due to less gain, thus the previous stage's load is affected.   

As one can see, altering the load causes several alterations in the performance. As such, I can understand the perception of slight alteration of the mid/highs. Whether this answers your question AndyR is another matter.

Cheers

Steve

Thanks for posting, Steve.   

It actually wasn't that I was conscious of some HF roll-off - in fact, I always thought my Maggie ribbons extended a long way and sounded pretty damn good!  As did others who came to listen to my system.     It was simply that I started to do measurements in REW with a calibrated mic - and saw a steep roll-off after 16Khz, which I didn't think should be there.  Now I understand that this kind of roll-off might be standard and due to room absorption ... so I will try out 2 further experiments, to see whether I get different results:
1.  try a different amp on my ribbons - one that is known to handle a 2ohm load with ease.
2.  measure with the mic 3" from the ribbon (instead of at the listening position).

And re. your experiments with minute drops in high frequency response, you might be interested in the following experiment I did a few years ago, during the early development of my 'Muse' JFET-based phono stage.  The initial circuit (which someone sent me with the exhortation "You have to build this - it's the best phono stage I've ever heard.") was judged by several people - when I'd built it - to be slightly rolled off in the HFs.  A friend of mine simulated the circuit in LTspice and found there was a cap between the Gate and the Drain of the 1st gain stage JFET, which was causing a roll-off at 120Khz.  Removing it didn't have any negative affect on the circuit (in LTspice, anyway) - but it meant the circuit now didn't have that 120Khz roll-off.

Sure enough, when people listened to v2 ... they no longer though it was rolled-off in the HFs!   


Regards,
Andy

[Steve]

Thanks for posting, Steve.   

It actually wasn't that I was conscious of some HF roll-off - in fact, I always thought my Maggie ribbons extended a long way and sounded pretty damn good!  As did others who came to listen to my system.     It was simply that I started to do measurements in REW with a calibrated mic - and saw a steep roll-off after 16Khz, which I didn't think should be there.  Now I understand that this kind of roll-off might be standard and due to room absorption ... so I will try out 2 further experiments, to see whether I get different results:
1.  try a different amp on my ribbons - one that is known to handle a 2ohm load with ease.
2.  measure with the mic 3" from the ribbon (instead of at the listening position).

I understand. Looking forward to your experiments Andy.

And re. your experiments with minute drops in high frequency response, you might be interested in the following experiment I did a few years ago, during the early development of my 'Muse' JFET-based phono stage.  The initial circuit (which someone sent me with the exhortation "You have to build this - it's the best phono stage I've ever heard.") was judged by several people - when I'd built it - to be slightly rolled off in the HFs.  A friend of mine simulated the circuit in LTspice and found there was a cap between the Gate and the Drain of the 1st gain stage JFET, which was causing a roll-off at 120Khz.  Removing it didn't have any negative affect on the circuit (in LTspice, anyway) - but it meant the circuit now didn't have that 120Khz roll-off.

Sure enough, when people listened to v2 ... they no longer though it was rolled-off in the HFs!   
Regards,
Andy

Your conclusion reinforces all my past work, and more, Andy. Back in the late 80s, I began research in the high frequency arena by performing an experiment where I altered the -1db point at 200khz to -1db at 150khz and visa versa, which alters the rise time and fall time. That change was clearly perceived over and over by all who have participated. Our work also agrees with research performed by several scientists, the medical community etc. (By the way, the small capacitor that was removed was to prevent any possible RF oscillations under unusual conditions.)

Cheers and thanks for your information Andy.

Steve

ps. Please forgive me for the updates to this post.

[Speedskater]

I understand. Looking forward to your experiments Andy.
Your conclusion reinforces all my past work, and more, Andy. Back in the late 80s, I began research in the high frequency arena by performing an experiment where I altered the -1db point at 200khz to -1db at 150khz and visa versa, which alters the rise time and fall time. That change was clearly perceived over and over by all who have participated. Our work also agrees with research performed by several scientists, the medical community etc. (By the way, the small capacitor that was removed was to prevent any possible RF oscillations under unusual conditions.)
Cheers and thanks for your information Andy.
Steve
ps. Please forgive me for the updates to this post.

Were there any visible differences in your high frequency O-scope inspection?
Any differences in your high frequency IM tests?
How about audio band frequency response?

[andyr]

Were there any visible differences in your high frequency O-scope inspection?
Any differences in your high frequency IM tests?
How about audio band frequency response?

No, S - I didn't measure IM.

I was only interested in (solving) the frequency roll-off.

Now that I have removed the cap which was providing a LP filter @ 120Khz, my audio band FR is pretty 'flat' (taking into account the RIAA correction).

Andy

[Steve]

Were there any visible differences in your high frequency O-scope inspection?
Any differences in your high frequency IM tests?
How about audio band frequency response?

Time was back in the 80s, will attempt to remember what I can. As you know, the frequency response changes (FR) become very small as one goes towards mid band. As such, using a scope on a sinewave with very small FR changes is rather difficult. I would rather use a rectangular waveform, especially with a higher fundamental frequency as one can check rise and fall times, ringing etc easier.

Never checked IM, but the HD distortion spec on the pre was at least -78db down. Changing the FR, Rise and fall times would be affected, which would be perceived according to several studies, including Medical inclusion in studies and Jneutron's reference which I do not remember.
(Jneutron is an avid audiophile who teaches science classes to other scientists, has worked at FermiLab, CERN, and last time I heard at Brookhaven National Laboratory.)

With -1db down at 150khz, 75khz would be very approximately -0,4db down. At 37khz response would be down approximately -0,15db. From going lower would require my computer simulation, but that computer has recently acquired video problems. These are general approximate figures when halving frequency; except as the frequency approaches mid band the minus db figures tends to really "flatten out" of course.

With -1db down at 200khz, 100khz would again be very approximately -0.4db down, and etc etc.

The roll off with the circuit I used would be 1st order, so approx. 6db per octave.

Another recent experiment has been my test speakers. For the past 4 years I have been adjusting the zobel and xover networks.
As such, one one experiment involves adjusting a parallel resistor across the full range driver. Over and over, adjusting the 9,000 ohm (9K ohm) resistor by very easily less than 0,1 ohms is clearly perceived by individuals. That is not only 1 part in 90,000, but is approaching 1 part in 180,000, equating to a frequency response change so small, in the neighborhood of -105db. (This is not an spl change across the entire audio band.)

Previously, over several years, in another experiment, bass response has been altered by a parallel adjustment on the back of my preamplifier. This adjustment compensates for the differences in amplifier input impedance. Over and over, with a single cello, we could perceive differences in the 32udb (32 micro db range) at 20hz. Yet a cello's lowest fundamental is in the range around 60 or 65 hz.

Each of the different methods I have used and still use, and by other studies, resulted in nearly the same results, that extremely small changes in frequency response, at both ends of the audio spectrum, are perceived. Of course, venue, musical selections, component quality etc would make for different results.

Olson's work included the effects of perception as the harmonic number rises (2nd harmonic, 3rd harmonic, ...... 9th harmonic etc). The higher the harmonic number, the more perceptible it is when that harmonic is altered in relation to the fundamental. It is incredible the sensitivity of the "ear" when it comes to tonal balance.

Cheers

Steve