No.. REALLY...

[buzzyreed]

Really, I wasn't kidding...

How DO you clean the contacts of the keyboards on a MOS 600? Really. Where do I start to take it apart just to clean the contacts.

On a side note, even if I don't get the thing MIDIfied for a while, I think in the right acoustic it could sound ok. I have forgotten how some of the notes about middle C sound.

Did they just sample one pipe at or around middle C and then just do some waveform scaling or something for the rest of the pitches because the lower notes really sound awful, despite the fact that some of the upper notes (especially when processed through a simple digital reverb unit) sound pretty fantastic for a first generation digital organ..

[arie v]

Hi Buzzy,

About cleaning the contacts. The Allen MOS organs had contact fingers which shorted together when a key was depressed. To clean, try sliding a business card between the two fingers, press the key down and gently swish the card around. This should get rid of any surface contamination. If the situation doesn't clear up, you will need to get new contact, as the contact points are worn out.

If you MIDI-fy you console, make sure the key contacts work very well. Otherwise you will get miserable results. A MIDI system with de-bounce does improve things, but is not the answer for intermittant contacts.

As to the tone on the early MOS organs, Allen created waveforms for the stops for somewhere around middle F or G, and then pitch shifted that waveform over the whole key compass. In fact the data stored for a stop was only for the positive side of the wave, 16 points across and 8 dynamic levels per point. It amazes me that the sound is anywhere near as good as it was considering how primitive the sound was. Also, they did not filter between points so, there was this "Allen" buzz in the tone, which when you zero in on it, you will always know you are listening to an Allen. Really, the tone especially on manual stops only sounded half decent around the middle of the compass. The sound was hollow in the bass, and stringy at the top.

AV

[davidecasteel]

Arie, I'm going to have to disagree with you a little. I don't know what the standard waveform coding scheme of the MOS instruments was, but I do know how the Alterable Voices were coded, and one would assume that the internal methodology would be at least as complex as that used for the Alterables. The waveforms for the Alterables used a binary code that allowed for values ranging from -64 to +63 (128 values); I don't know what "8 dynamic levels" actually means, but the simplest interpretation would be that only 8 values were possible, and that is not the case (at least for the Alterables). Again thinking of the Alterables, it is true that only 16 points were specified by the coding on the cards (each point able to have the full range of values from -64 to +63), and the full waveform was achieved by using point symmetry (reversing and inverting) to replicate the 16 points and generate an array of 32 values. It is also true that the same waveform was used for the full gamut of pitches controlled by the keys, and that this did not produce great realism at the upper and lower extremes. I cannot comment on the lack of smoothing between the points, as I don't know what the circuitry included. I would have thought that some level of filtration would have been provided, but I just don't know.

David

[don60]

I am most familiar with the ADC series since I own an ADC-7000 and have spent countless hours with the oscilloscope trying to unravel the mysteries of the Allen tone generation system. But I believe that the MOS series is very similar in its essential elements (albeit implemented with completely different electronics technology), so I will share what I have learned for what it is worth to the discussion.

The ADCs use a 1 MHz master oscillator for each "cage" (the equivalent of a "computer" in the MOS organs); the oscillator output, squared and conditioned to TTL logic levels, forms the clock for the entire tone-generation system contained within the cage. Logically, the oscillator pulses are grouped into twelves, so that pulses 1 through 12 occur in succession, over and over, at a 1 MHz rate. A given number pulse occurs at a 1/12 MHz or 83.3 kHz rate.

Each pulse defines a "keying slot" during which interval a keypress can be read by the circuitry and used to generate a tone based on a given pitch (key) and a given stop. Contrary to the practice in the MOS organs, keying slots in the ADCs are universal throughout the instrument, spanning even multiple cages (computers). Thus, only 12 simultaneous keypresses anywhere on the instrument will be considered unique and readable. If a 13th should occur at any time, it is ignored by the key assigner circuitry. Apart from the musical consideration that 12 simultaneous notes would probably be enough to provide a satisfying polyphony, the obvious engineering calculation here is that 10 fingers and two feet will ordinarily generate at most 12 keypresses!

One of the truly elegant features of this approach is that it results in a time-division multiplexing (Allen's own term) in the tone-generation system. That is, a given tone (note and stop) occupies 1/12 of the total time needed for the 12 keying slots to occur; during the other 11/12 of the time, 11 other tones can be generated in their respective slots. Thus, the critical (and at the time, expensive) components such as the frequency and envelope generators and the digital-to-analog converters can potentially be used for as many as 12 tones at a time rather than one. With the improvements embodied in the ADC system, in which keying slots are universal throughout the instrument, inter-division coupling is also made very simple because gating circuitry can be used to make a tone generator respond during two (or more) keying slots rather than one. (A direct result of this improvement is that inter-division coupling does not "use up" keying slots as it does in some cases on the MOS instruments; thus, no matter how many couplers are active, 12 distinct keypresses can still be read.)

The implications of this background to the current discussion: Each tonal waveform is reconstructed using amplitude samples at an 83.3 kHz rate. The ADCs use an eight-bit format, so 2^8 or 256 distinct amplitude values can be encoded for each sample. (Note that it is irrelevant whether the eighth bit is considered a sign or a value bit; the available binary combinations out of eight bits are the same.) Although amplitude encoding with only eight bits is indeed a bit crude, it is sufficient to the task at hand and will generate only a little "quantization noise" that in most cases will be scarcely audible. (If it is audible, it will sound somewhat like inter-station noise on a radio receiver and will be dependent on the pitch of the waveform being reconstructed.) Analog filtering cannot be used to eliminate quantization noise; the only way to reduce it is to use more bits in the sample encoding.

The other primary hazard in reconstructing an analog waveform out of samples is the potential for aliasing.

Simplistically speaking, aliasing occurs when the samples of the reconstructed waveform occur at a rate that is less than twice the highest frequency in the spectrum of the waveform. Thus if the reconstructed waveform has no significant spectral content above 20 kHz, the samples must be generated at a rate of 40 kHz or greater to avoid aliasing. If aliasing does occur, its audible nature and severity are highly dependent on the extent of the aliasing and the nature of the audio signal that suffers from it. As the name implies, though, aliasing will result in audible "artifacts" that are superimposed on (and inseparable from) the desired signal. A key point here is that analog filtering cannot remove aliasing once it occurs.

In real life, the ideal factor-of-two value is never used; rather, a margin of safety is introduced to ensure that aliasing is avoided and to allow some low-pass filtering in the reconstructed waveform. In this example, a low-pass filter would be used to pass everything below 20 kHz and eliminate everything above. Practically speaking, of course, the audio system and the human ears that follow the digital-to-analog conversion process form a pretty good filter on their own, and not much filtering is needed in a system such as Allen's. (I have not yet investigated the voicing filters on the ADC cards to know how much, if any, filtering is actually done.)

I have read a number of allegations that the Allen system suffers from "aliasing" and that the anamolous noises one hears on higher notes is "aliasing." Unless I completely misunderstand their system, I have to say that this nonsense is written by those who have no idea what aliasing really is and have probably never even kicked a book on communication theory and Fourier analysis. (I have.) At least using the ADC series specifications, the Allen system cannot suffer from aliasing on any note, even C5 on the 1' stops, since it uses a sampling rate over four times the highest frequency to be generated by the system.

That is not to imply that the system is perfect, or that the anomolous noises are all in one's imagination! The system does suffer from flaws, the worst in my analysis to date being a rather dramatic phase modulation on certain notes. To explain this phase modulation will require more background on the tone generation system and will have to wait for another day, but I can say that it is directly related to the two points raised in this discussion: storing a single waveform to represent all notes of a given stop, and storing a waveform with a very limited number of time samples (32).

I hope these observations help a little.

Don

[DellAnderson]

..
Thus, the critical (and at the time, expensive) components such as the frequency and envelope generators and the digital-to-analog converters can potentially be used for as many as 12 tones at a time rather than one...

Don,
That was extremely interesting information which I confess to only understanding superficially. If I understood you correctly, the ADC organs allowed a single DAC to create up to twelve notes at one time? That's incredible. Somehow I missed that in my prior readings.

I'm not sure how relevant all this is to the MOS organs as the ADC was a considerable sonic improvement on the MOS (in my opinion), but I recently read Jerome Markowitz book "Triumphs & Trials of an Organ Builder" and there is short chapter on George Watson, the North American Rockwell/ (and thus Allen) engineer and inventor of this time-share technology (Patent #3,160,799 according to the book, but patent number 3,610,799 according to my research: http://www.freepatentsonline.com/3610799.html)* which states that it was first incorporated into the Allen Digital Computer Organ in 1971. This seems to indicate that the MOS/LSI circuitry probably used the same time-share technology. Now I'm not sure how this relates to cleaning key contacts but it does relate to how the notes were generated.

On a side note, some of Cameron Carpenter's pedal work involves 4 notes at a time, so perhaps there is a reason he is building Excalibur instead of an off-the-shelf Allen!

*I obtained the original US Patent Number from p. 96 of Jerome's book, but alas, it appears to be a typo as that patent is owned by someone else and does not really appear relevant. An brief but intensive search found a number of Allen Organ related patents, of which 3,610,799 appeared to be the most relevant. It is well worth opening the scanned image PDF on the above link. Perhaps the more technologically savvy can explain what it says, but it sure looks like it would have given a big head start to any oscilloscopic probings.

[michaelhoddy]

Don,

I for one appreciate very much the detailed schooling you just gave those of us who have for years wondered about the nature of the MOS/ADC/MDS tone generation system. Thank you for taking the time to write and post it.

I still remember me and my father trying to reconstruct Allen voices from Alterable Voice tone cards using additive synthesis back in the 80's. It didn't work too well, since, as we learned, the cards weren't just a simple harmonic reconstruction. But that's another story!

[DellAnderson]

Well...I spent too much time researching this to keep it to myself, so here are some of the more interesting and possibly quite informative Allen/Rockwell patents I found this morning. Truly interesting reading and if you view the scanned PDF versions, nice waveform images that explain exactly how this stuff works (or so one is led to believe). I threw in an early (1962), non-computer organ patent for the swell pedal resistor because Allen was still using it in 1989 when they made my organ. I added the inventor's names to some. The patents are an amazing source of insight into how the organs are designed, a sort of 'How things work for Allen Digital Computer Organs' if you will, and all the more fascinating because they are so public yet Allen will not provide technical manuals to owners. So we have both extremes - a user manual which basically says nada, and the patents which tell you how to make your own organ, but nothing to maintain or voice the one you have. NOTE: I made no distinction between Allen & Rockwell because I believe Allen had right of first refusal on all Rockwell organ inventions which they subsidized (mainly by way of Ralph Deutch)

Here are is the best of the best:

1962- Light sensitive resister (swell pedal)
http://www.freepatentsonline.com/3045522.html

1970- Digital Organ - Waveform Storage etc
http://www.freepatentsonline.com/3515792.html

1971- Multiplexing Note Selection
http://www.freepatentsonline.com/3610799.html

1971- Chiff
http://www.freepatentsonline.com/3740450.html

1972- Attack & Delay
http://www.freepatentsonline.com/3610805.html

1972- Memory Addressing
http://www.freepatentsonline.com/3639913.html

1973- Alternate Voice
http://www.freepatentsonline.com/3755608.html

1974- Frequency Modulation For Sampled Amplitude
http://www.freepatentsonline.com/3794748.html

1977- Touch Response - Deutch
http://www.freepatentsonline.com/4033219.html

1978- ADSR - Attack Decay Release - Deutch
http://www.freepatentsonline.com/4079650.html

1979- Demultiplexing Audio Waveshape
http://www.freepatentsonline.com/4134321.html

1980- Attack & Delay -Woron
http://www.freepatentsonline.com/4212221.html

1980- Timbre Modulation
http://www.freepatentsonline.com/4189970.html

1982- Mixture Tones
http://www.freepatentsonline.com/4357851.html

[don60]

Thanks, Dell. About a year ago, I actually started printing out and reading as many of the original patents Allen patents as I could; like you, I was astonished at the amount of information that they give the savvy reader about the functioning of the production tone-generation systems that Allen used through the MDS era. A few of the patents are essentially worthless because their ideas were never implemented, but others are extremely close in concept to what one would find in the actual circuitry.

I recognize good old "799" in your first post as one of the key patents and, I believe, the one that Kimball ultimately broke after an extended court fight. (For all the good it did them, right?) Others, however, are equally useful, and I do not yet have a roadmap or suggested order for reading through them.

Like many other of my projects, this one has been on hiatus. I do mean to get back to it sometime soon. Meanwhile, I will offer an observation about tracking down the patents that were actually applicable to a given organ: Look on the builder's plate inside the console--it lists dozens of patent numbers that protect the ideas embodied in that model.

You are correct that the time-division multiplexing was first used in the MOS organs, and you are correct that it allows up to twelve simultaneous notes to be generated by one subsystem. In our ADC organs, then, a single tone generator card (one of the TG series) together with the ancillary envelope generator and frequency generator cards (EGs and FGs) would be capable of generating 12 distinct notes, voiced according to the selected stop(s) stored on that TG card. However, if all 12 notes were played in that division of the organ, no further distinct notes could be played on that or any other division because of the limitation on keying slots. Again (CC aside), most organists would be pressing no more than 12 notes simultaneously anyway.

Don

[DellAnderson]

I recognize good old "799" in your first post as one of the key patents and, I believe, the one that Kimball ultimately broke after an extended court fight. (For all the good it did them, right?) Others, however, are equally useful, and I do not yet have a roadmap or suggested order for reading through them.

Thanks, Don. You are correct, according to Markowitz book's chapter "Losing a Patent", although Allen had obtained the rights to the Rockwell patent on "799", Kimball won their case apparently over the issue of whether the demonstrator organ had these features (whether or not the demonstrator was ever on sale itself was contested, but I can see how a judge could have interpreted a travelling demonstrator as a sales tool, even if that organ itself was never formally "for sale" itself (p. 156). Apparently, there were many years (and presumably many dollars) wasted over this issue to no avail on either side. Naturally, Markowitz took a very dim view of Deutch for this and other reasons, but it would sure be interesting to read a book written by Deutch had their been one.

As for finding time to study all these patents, I fear that technology is passing this by and no engineer in his right mind would want to build a duplicate today, but it is still fascinating for those of us who still have ADC organs and are interested in how they tick.

Thanks for the tip on the Patent numbers. My ADC4900 lists 35 patent numbers, including the "799" and some I mentioned earlier (I haven't done extensive cross checking - it looks like a huge job!)

[buzzyreed]

That's all well and good...

As an elementary kid I used to build robots and program in assembly and machine language. As I became further educated, I think I lost all of my technical ability (which is why I have to have students help me even put up a sound system). I have forgotten how complicated the whole waveform synthesis is.. uff..


But the question remains.. UM.. WHAT DO I TAKE APART TO GET TO THE AREAS I NEED TO CLEAN??

I mean, literally do I come in from the screws in the front where the piston buttons are or do I come in from the back. I am sure some of you probably think I'm joking, but I really can't figure this out.

Anyone else see that Harrison Labs is selling some sort of kit to make your own MOS cards? I think I'll still with Hauptwerk or even analog tonal generation. I understand oscillators.

So, um, if someone could give me a simple hints as to how to start.. it would be great.

[don60]

Oops--I guess we got sidetracked.

If you have a D series stop tab console (roll top, squared-off sides, 35" deep), I can help:

Raise the lid on its hinges; a chain should prevent it from traveling much beyond vertical.

Reach in through the top and release the two twist latches on the top of the back panel. Tilt the panel out at the top and lift it out of the opening.

Close the roll top.

Locate the two screws (one on either side) driven through the roll top tracks at the back of the console. (They limit the travel of the top when it is in the open position.) Remove these screws.

Locate the two filler blocks at the top of the console just above the roll top tracks. Each block is held by two screws; remove the screws and the filler blocks to expose the tops of the tracks.

Carefully slide the roll top to the open position. The front rail should just extend into the openings created by the removal of the filler blocks. Preferably working with a helper, very carefully lift the front rail up enough to clear the tracks and slide the roll top up and forward, removing it from the tracks.

Locate two bronze-colored screws and washers at the bottom of the stop tab panel. (They are driven into the side cheeks of the top manual.) Remove these screws and hinge the stop panel upward. (It is much heavier than it looks because of the numerous steel magnet cores in the stop brackets.) Allow it to rest on the lid of the console.

(The contacts for the top manual will now be accessible.)

To gain access to the bottom manual, hinge the top manual out of the way.

I apologize if I have omitted or mangled a step here or there. My console has been apart for so long that I have forgotten all of the details of its disassembly! These directions ought to get you close, though. If you have another series of console, please advise and perhaps someone else on the forum will be able to help.

[circa1949]

Wow - thanks Don!

Don your write-up is quite fascinating and obviously your expertise in digital signal theory far exceeds that of anybody else to have posted here.*
I'm sure I've used the term aliasing to describe the sonic issues with an ADC organ being played on very high definition speakers compared to the stock models. (tannoy dual concentrics with a bryston amp, as was the case with the ADC-1140 I sold) Some notes, particularly E&Fs throughout the compass, sounded really rough depending on the stop. Certain flutes were the worst. The military fife 2' alt. card had some notes sound like a kid going crazy on Roland D50 synth. Fine that it wasn't truly aliasing...this term is generically applied, I suppose, to any digital seeming audible artifact. Fine that it really is a phase modulation distortion. My concern is with describing the system as an 88kHz sampling system with an eight bit bit depth. I don't have time tonight to do experiments with my trusty old copy of Cooledit 2000, but I just don't think it sounds as good as such a system would. I can't quite articulate it without sounding like an idiot, but it has something to do with the fact that an 88k 8 bit LPCM recording system would be able to record sounds without the same issues that having to recreate a waveform at multiple pitches at the same time in a synthesizing system would cause...I'm sure if I'm even close to correct on this you will be able to help me. Let me try to approach the problem from the opposite side...the ADC system stores 256 sample points IIRC. I can't remember if that's full or half...let's be optimistic and assume half, so 512 sample per full wave. So to produce a C1 note, no matter how fast a system clock is available, the recreation of that note will be 32 X 512=16K. Whereas if you recorded said sound with 88K LPCM recording you'd have 88kbit/32bps=2750 effective sample points. In other words roughly 5X as much info.
I could be totally croking smack on this...but indeed to my ears the pedal 16' bottom octave almost, but not quite, sounded as bad as something generated with a 1980s COVOX soundcard. Not that it really mattered in the grand scheme of the pedal notes being overlaid with 8' & 4' frequencies. And supporting manual notes. Whereas the C1-C2 compass sounded much better. And the top 3 octaves of any 8' rank sounded great, with the exception of some flute ranks where the aforementioned phase issue (?) on e&f was bad.
OK...just my 2 cents. Look forward to hearing from you about this.

*- with the exception of one of our good friends...some will know who I'm talking about...if he's left this Earth, he's playing some great organs in the heavens, I'm sure. We haven't heard from him in a while and we knew he had insurmountable health problems.

[DellAnderson]

But the question remains.. UM.. WHAT DO I TAKE APART TO GET TO THE AREAS I NEED TO CLEAN??
I mean, literally do I come in from the screws in the front where the piston buttons are or do I come in from the back. I am sure some of you probably think I'm joking, but I really can't figure this out.

Anyone else see that Harrison Labs is selling some sort of kit to make your own MOS cards? I think I'll still with Hauptwerk or even analog tonal generation. I understand oscillators.

Buzzy - you sound like me. I had to pick up a junky Rodgers to practice on because everything (especially around the keyboards) is not entirely self evident. I hope one of the experts is reading this. EDITED: Doh! Should have done a screen refresh.. Thanks Don for the great explanations - both electronic and mechanical.

And yes, I just saw the Harrison Labs 'kit' yesterday on Ebay - $15 for what looks like a set of instructions. Someone on the forum already has figured this out and probably would be willing to give you the information for free however - do a search on MOS voice cards. I have it if you can't find it, but I think it would be more appropriate to get from the source. I haven't explored too much yet because my ADC Alternate Voice card is not (yet!) functional.

To avoid hijacking the thread further, I'll just refer to another where I looked up every one of the 35 patents listed under the lid of my ADC4900 and indexed them in the ADC4900 thread
http://www.organforum.com/forums/sho...hlight=ADC4900
(the latest post at the moment).
Some are a bit mystifying: Like the keyboard instruction method that appears to output to a TV, or the touch sensitivity (didn't know I had it) and tracker touch (ditto).

Let us know what you find. On the old Rodgers I have, the keyboards are screwed down to the base, but you don't need to loosen them because they hinge up one manual at a time (starting with the top manual of course). I don't remember if you have to remove the stop tab rail to do that though (I literally had the whole organ in pieces at one time - now it's back together).

[circa1949]

"Like the keyboard instruction method that appears to output to a TV"
When I visited Allen in the late 1980s, as a preteen, their director of church marketing or whatnot (Sally, I believe) showed me a primitive system for training that showed the notes being played on an ATARI-ish looking video display. We didn't spend much time on it but I took it to be something like a automated organ teaching system? Or at least a very rough start of something like that. I only have the vaguest memory, but I do remember her showing me a computer/TV screen and somehow the notes played on the organ showing up on a monitor.
They probably had one standard placard to encompass any technology that could be found in the various models and/or customized organs. After all the metal placards were preprinted but then stamped with the model & serial number, right?

[don60]

Don your write-up is quite fascinating and obviously your expertise in digital signal theory far exceeds that of anybody else to have posted here.*
I'm sure I've used the term aliasing to describe the sonic issues with an ADC organ being played on very high definition speakers compared to the stock models. (tannoy dual concentrics with a bryston amp, as was the case with the ADC-1140 I sold) Some notes, particularly E&Fs throughout the compass, sounded really rough depending on the stop. Certain flutes were the worst. The military fife 2' alt. card had some notes sound like a kid going crazy on Roland D50 synth. Fine that it wasn't truly aliasing...this term is generically applied, I suppose, to any digital seeming audible artifact. Fine that it really is a phase modulation distortion. My concern is with describing the system as an 88kHz sampling system with an eight bit bit depth. I don't have time tonight to do experiments with my trusty old copy of Cooledit 2000, but I just don't think it sounds as good as such a system would. I can't quite articulate it without sounding like an idiot, but it has something to do with the fact that an 88k 8 bit LPCM recording system would be able to record sounds without the same issues that having to recreate a waveform at multiple pitches at the same time in a synthesizing system would cause...I'm sure if I'm even close to correct on this you will be able to help me. Let me try to approach the problem from the opposite side...the ADC system stores 256 sample points IIRC. I can't remember if that's full or half...let's be optimistic and assume half, so 512 sample per full wave. So to produce a C1 note, no matter how fast a system clock is available, the recreation of that note will be 32 X 512=16K. Whereas if you recorded said sound with 88K LPCM recording you'd have 88kbit/32bps=2750 effective sample points. In other words roughly 5X as much info.
I could be totally croking smack on this...but indeed to my ears the pedal 16' bottom octave almost, but not quite, sounded as bad as something generated with a 1980s COVOX soundcard. Not that it really mattered in the grand scheme of the pedal notes being overlaid with 8' & 4' frequencies. And supporting manual notes. Whereas the C1-C2 compass sounded much better. And the top 3 octaves of any 8' rank sounded great, with the exception of some flute ranks where the aforementioned phase issue (?) on e&f was bad.
OK...just my 2 cents. Look forward to hearing from you about this.

*- with the exception of one of our good friends...some will know who I'm talking about...if he's left this Earth, he's playing some great organs in the heavens, I'm sure. We haven't heard from him in a while and we knew he had insurmountable health problems.

Circa, the anomolies in the E's and F's on the high-pitched stops are indeed manifestations of the phase modulation that I alluded to. Back in the fall of 2009, jbird and I exchanged a series of posts and private emails on this very topic as I struggled to find the cause of what I assumed to be a defect in my own organ. After he played a similar ADC at his church and reproduced the "problem" there, I realized that whatever was causing the issue had to be a design limitation, not an electronic fault. After much thought, I hit upon the explanation--at least I believe that I did. You are exactly correct that the heart of the matter is the need to synthesize a waveform of arbitary frequency out of stored samples, as opposed to simply recreating a waveform out of transmitted samples (a much easier task). I believe that the problem is much worse than your calculations indicate, though. As far as I know, the ADC stores the same number of sample points in EPROM as it does on the punch cards: 32. So every waveform to be synthesized must be recreated from just 32 samples per half cycle--as you observed, a rather paltry number for the lower frequencies. Again, you are correct that the audio quality resulting from the Allen system is not nearly as good as it would be were they simply sampling an analog signal at 83 kHz, transmitting the samples digitally, and performing a D/A conversion to recover the analog signal. The need to generate samples at precise instants of time according to the system clock conflicts with the fact that the stored waveform (with its small number of sample points) may not have a sample corresponding to that instant of time.

I'll try to put together a fuller explanation of my thoughts on this phase modulation problem. It's interesting that you are the first to mention specifically the E's and F's of the higher pitched stops as being particularly offensive. I noticed the same thing and suspect that others would too if they listened carefully.

Don