[Elecraft] VP6DX -- the true story ?

[Igor Sokolov]

Eric, thanks for your posting. The description of human hearing mechanics is 
very educational.
OK on DSP issue. Actually one other member of the dxpedition mentioned that 
the effect of DSP that I have described is well known to him and was present 
in K3 as well. He also mentioned that switching off AGC was one of the 
methods to overcome this effect. That was my initial reason to ask you these 
questions. I think you are right. DSP in K3 cannot be completely disabled 
(as well as in most modern DSP based radios) Therefore it is hard to tell if 
the effect should be attributed to DSP proper or DSP derived AGC. Anyway 
several people told me that they have made comparisons between DSP based 
radios and purely analog radios like IC781. In conditions of pile up made up 
of weak signals (not more then S3) analog radios always performed noticeably 
better.

73, Igor UA9CDC

73, Igor
> Igor --
>
>   Your question about DSP is an interesting one.
>
>   In my limited understanding of the K3 signal path, there is always a DSP 
> operation on the signal for filtering.   While noise blanking, noise 
> reduction, and audio effects can all be switched off, I don't know if it 
> is possible to remove ALL digital signal processing from the signal path; 
> i.e., operate only with the crystal filters.
>
>   We certainly had plenty of marginal openings where we were required to 
> work a pileup of signals just at the level of the underlying band/antenna 
> noise.  But... the antenna/band noise were a significant step above the 
> receiver's underlying noise floor (10 dB)... or one turned on the pre-amp 
> to bring the band/antenna noise above the underlying receiver noise floor. 
> In this case, the pileup was not at the receiver's MDS level.
>
>   Furthermore, the operating positions were not equipped with a second, 
> different receiver and associated switching so that the operator could do 
> a real-time A/B comparison between the K3 and "Brand X".
>
>   Yes, signal range could be from S1 (or less) to S9+40 dB, so more than 
> 90 dB.  However, several other aspects of human hearing come into play:
>
>   1.  The most sensitive part of audio spectrum for typical hearing is 2 
> to 5 kHz.  If we take a K3 with very wide filters, and no antenna, in an 
> extremely quiet listening environment, and just gradually advance the 
> audio gain until we can just begin to hear the receiver noise floor, we 
> will be listening to a higher-pitch hiss in this range of 2 to 5 kHz. 
> White noise at lower frequencies won't be perceptible yet until the 
> receiver gain is advanced another 10 dB (at which point frequencies down 
> to 500 Hz are audible) or 20 dB (good for frequencies down to 250 Hz).
>   If we narrow the receiver bandwidth so we are only listening to 100-700 
> Hz, for example, the receiver noise floor will appear about 10 dB louder 
> (relative to the minimum threshold of hearing) at the higher end.
>
>   2.  Another frequency-sensitive aspect of human hearing is the 
> attenuation reflex.  This reflex tightens two muscles in the ear, one of 
> which tightens the ear drum slightly and the other moves the three bones 
> of the middle ear to reduce the transmission to the cochlea (inner ear). 
> This is our own, human protective AGC.
>   The attenuation reflex begins to act at 65-70 dB above the threshold of 
> hearing at 200 Hz... but 80 dB above the threshold of hearing at 700 Hz.
>   The "slope" of the attenuation reflex is about -0.6; i.e., a signal that 
> is 18 dB above the attenuation reflex threshold will be reduced to just 6 
> dB above that threshold (i.e., 12 dB attenuation added) by the time it 
> reaches the inner ear.
>
>   Now let's look at an operator listening to a K3 in a perfectly quiet 
> listening environment (no other local sounds).  If he adjusts the receiver 
> so that antenna/band noise is 5 to 10 dB above his threshold of hearing at 
> a pitch of 400 Hz, and then tunes across a CW signal that is +95 dB above 
> the band/antenna noise floor, that CW signal will be about 100 to 105 dB 
> above the threshold of hearing.
>
>   That signal will also be about 30 dB above the threshold for triggering 
> the attenuation reflex.  At a slope of -0.6, the attenuation reflex will 
> cut that signal down by 20 db... so that it is now 80-90 dB above the 
> threshold of hearing.  -20 dB of attenuation is about the maximum the 
> attenuation reflex can deliver -- but that is only in children and 
> teenagers.  For adults, the maximum attenuation level declines with age, 
> so I (at age 55) can no longer get -20 dB of protective attenuation. 
> Maybe I get 10-15 dB of attenuation, leaving the CW signal at something 
> like 95 dB above the threshold of hearing.  Of course, once this 
> attenuation reflex is activated, that very weak CW signal down near the 
> noise level will be attenuated below the threshold of hearing, so no more 
> copy.
>
>   Even worse, long exposure to signals above the attenuation reflex 
> threshold results in incremental and permanent hearing damage.  So that CW 
> signal, at 95 dB above the threshold of hearing... and 20 dB above the 
> attenuation reflex threshold... represents an important hazard.  The USA 
> National Institute for Occupational Safety and Health has set a limit of 
> about 1 hour per day at this level... and that limit declines quickly at 
> higher levels.
>
>  3.  Fortunately, by setting the receiver gain at these low levels, that 
> loud CW signal is below the threshold of pain (about 110 dB above the 
> threshold of hearing at 400 Hz).  The threshold of pain is where the 
> operator rips off the headphones and says "ouch"!  We want our receivers 
> to limit signals (or static crashes) before they reach this level!
>
>   So, we can't use a receiver that is perfectly linear over a 130 dB 
> range -- it would destroy our hearing!  But we need to listen to signals 
> in a very quiet listening environment, as quiet as we can get... and set 
> the gain levels appropriately... and use some form of signal limiting to 
> keep signals well below the pain threshold.
>
>   And we should recognize that hearing varies from person to person.  As a 
> result, one person with poor hearing range, listening in a noisier 
> environment and having his attenuation reflex triggered often, will have 
> receiver AGC and his own attenuation reflex interacting to reduce signal 
> strengths... eliminating weaker signals... while another operator 
> listening to the same radio with good hearing (big dynamic range between 
> his threshold of hearing and attenuation reflex trigger point), with 
> minimal receiver AGC, will find a rich range of signals in the pileup.
>
>   The psycho-acoustic phenomenon of "masking" further complicates the 
> management of a pileup.  But that's a subject for another time...
>
> -- Eric K3NA
>
> on 08 Mar 13 Thu 14:57 Igor Sokolov said the following:
>> Thanks Eric,
>> I understand you think AGC in high dynamic range radio like K3 is not a 
>> necessity for the op with high dynamic range ears.
>> Actually the range should be close to 90-95 db on nowadays bands where 
>> 59+40 signals co-exist with 1S unit signals very often. I also figured 
>> from your post that other ops from  VP6DX team choose not to use AGC not 
>> because of some sort of problem existed in that version of the firmware 
>> but purely because they prefer adjusting RF gain manually to squeeze out 
>> the most from the pile up.
>> What about the DSP then. I experienced that myself operating from 8Q7 in 
>> the past and many others confirmed my impressions that modern DSP based 
>> radios are no good when pulling call signs from the pile up of many 
>> stations when the average level of that pileup is close to the MDS of the 
>> radio. For example we could not use Orion in the expedition on a weak 
>> pile up on 10 and 15 meters while the same operators could easily pick 
>> out calls using IC775. Therefore I would be happy to have a radio where 
>> DSP could be switched on or off depending on the situation. Did you 
>> notice the above mentioned effect while using K3?
>>
>> 73, Igor UA9CDC
>>
>>
>>> Privet Igor --
>>>
>>>   As best I can remember, we used MCU v1.66 and DSP 1.52.
>>>
>>>   I didn't use the AGC because I did not need it.  The point of 
>>> automatic gain control is to adjust the gain of the various stages of 
>>> the receiver in order to
>>>   a) avoid distortion/overdriving a stage, and
>>>   b) bring signals up/down to a comfortable listening level.
>>>
>>>   (b) is not relevant to an operator who is working a pileup.  (b) is 
>>> relevant to an operator who is monitoring a frequency with one station 
>>> transmitting.
>>>
>>>   For operating a pileup, there are a variety of tools the brain uses to 
>>> distinguish the many signals:
>>>   -- pitch (CW)
>>>   -- style of speaking (speech) or keying (CW)
>>>   -- artifacts; e.g., auroral flutter, chirp, etc.
>>>   -- strength (all).
>>> AGC tends to reduce the difference in signal strength, and so removed 
>>> valuable information.
>>>
>>>   In situations where static crashes interfere with reception, AGC hang 
>>> time on a loud static crash also increases the length of time that a 
>>> specific static crash interferes with reception.
>>>
>>>   I used headphones with good audio isolation between my ears and the 
>>> rest of the world around me.  That allows me to set receiver gain levels 
>>> with the underlying antenna/band noise just above my threshold of 
>>> hearing... and to use at least 80 dB of my hearing range for listening. 
>>> In this quiet listening environment, I don't need AGC.
>>>
>>>   Even in a less-than-quiet listening environment, if a band is just 
>>> open weakly (e.g., 12m to Europe), the range of signal strengths in the 
>>> pileup can be smaller: maybe less than 30 dB between band noise and the 
>>> strongest signal.  So AGC isn't needed here either.
>>>
>>>   My ideal AGC in these situations is one that only makes changes in 
>>> receiver gain when a stage in the receiver is about to be over-driven 
>>> (e.g., the A/D converter)... and removes those changes relatively 
>>> quickly. Even then, it might be fine to allow the receiver to be 
>>> over-driven (a static crash contains no information).  If a signal I 
>>> want to copy is over-driving the receiver, the best solution often is to 
>>> reduce the RF gain manually during the duration of the time when I want 
>>> to copy that station.  If that station is just "interference" (e.g., a 
>>> loud USA station on 80m CW calling VP6DX, when I want to work northern 
>>> Scandinavia and northwest Russia/western Asia during the brief opening), 
>>> I have other controls (filter bandwidth, notch) than might be better to 
>>> use that gain reduction (automatic or manual) that could suppress the 
>>> desired weak signals.
>>>
>>>   So, almost any AGC system is inappropriate for a DXpedition or content 
>>> environment... as long as the receiver and one's own ears have enough 
>>> dynamic range to handle all the signals presented to it.  The K3 has 
>>> more dynamic range than other receivers.
>>>
>>> 73,
>>>   -- Eric K3NA
>>>
>>> on 08 Mar 13 Thu 02:26 Igor Sokolov said the following:
>>>>>   Almost all of the operators ran the K3 with AGC off (all modes).
>>>>>
>>>>> -- Eric
>>>>
>>>> Eric, can you explain what did you not like about AGC in K3?
>>>> Did you notice any problems with DSP being permanently on in K3 when 
>>>> listening to heavy pile ups?
>>>>
>>>> 73, Igor UA9CDC
>>>>
>>
>>