Victor Rubilar

I imagine it with a very big touch screen, some butons, and maybe 2 knobs, the other knobs in an external device like FP8, plus a new fader controller.

thanks Senator, I didnt know the patent had been just published. All my general conclusions were correct since the begining. If people know how it works, then they can make better choices on how to use it to get the best audio. resume of main points in the oficial patent (if I understand it correctly): http://patents.justia.com/patent/8878708 - diference of aproximately 20 dB of input level of the 2 converters through a gain stage or an atenuation stage. - " due to uncompensated differences between the two preamplifier circuits" and the use of the user selectable gain, to match the 2 signals into the digital domain, the system applyes a "pre" digital gain called CCG factor to make a stable diference between both signal levels, This factor is constantly calculated because in analog system the gain varies whith temperature,etc, - the switching can or can not be crossfaded. - the switching from the first to the second signal is at -3 dBfs - there is a small but important error in the text I found. Maybe it is not in the originaly submitted document by zaxcom. This paragraph describes the normal operation. "At 324, process 300 determines whether the value of the Timer variable is greater than zero. If no, process 300 proceeds to 328, at which the ADCComp [ first signal multiplyed by the CCG factor] audio sample calculated at step 318 is output to an outgoing audio circuit... " When a sample over -3 dBfs is detected, the "Timer" is set from zero to 1000 and the second signal must engage... "If, at 324, the value of the Timer variable is not greater than zero, process 300 proceeds to 326, at which the ADC2 [second signal] audio sample is output to an outgoing audio circuit. That is, the outgoing audio is switched to use the sample received from ADC2 rather than the calculated ADCComp sample..." It says " is not " and must say IS for the system to switch the signal. - the release time for switch-back from the second to the first signal is controlled by the "Timer". It is set to "aprox". (?) 1000 every time there is a switch to the second converter, and decremented by "1" every time there is new "sample" , It says that it takes aprox. 10 miliseconds to reach 0 (release time), so 1 Timer unit is not equivalent to a sample?!!! , (if it was as they say, the release time would be aprox 20 miliseconds [see below] . I believe this part is not so well drafted). But then it says that the 10 ms. release time can be substituted. So the Tmer can be set higher in the real implementation. When in the second signal, the Timer is set to 1000 (again and again) every time there is a sample over -6 dBfs. So, the signal must be at least 1000 continuous samples under -6 dBfs for the Timer to reach down to zero, When the Timer reach zero, the system switch back to the first signal. Acording to a fast calculation, defining the switch back to -6 dBfs with this system allows for a release time of 18.75 ms (3/8 of a wave cycle) in order to suport a 20 hz signal input.at 48khzSF, and 22.6 ms. at 44.1khzSF.

it is not compression since the level is never attenuated acording a threshold level of a control signal. It is merely a constant attenuation to all the audio before being recorded into a 24 bit word, since the signal can contain higher values than a 24 bit word is able to. If when recording you set the gain so the input signal never goes over 0 dBfs, it is not necessary to use any attenuation. It works like a normal system. But If the level goes over 0 dBfs, it is necessary to use attenuation to disk. The internal 32 bit float mixer is capable of work with higher than 0 dBfs levels, but the recorded audio file is 24 bit and cannot contain values over 0dBFS. with the new available data I conclude this: Never clip in more detail 0 dBfs is the maximum peak voltage a normal system is designed to work with. When connecting "that" voltage level, the normal converter will generate a 24 bit sample word filled with "1" (so any voltage over that value will be coded with the same 24 bit word full of "1" i.e. clipping). In the never clip case, at 0dB fs voltage, the first converter wil generate a 32 bit float word that consist of : -the same 24 bit word full of "1", -and an 8 bit exponent that in this case doesnt count because it represents the number "0". (any voltage over 0 dBfs will be coded with the same 24 bit word full of "1" i.e. clipping). And here comes the "magic" The second converter is conected to a 20 db attenuated signal, so it will generate a 32 bit float word that consist of: - a 24 bit word representig 20 dB less than the first converter, - and an 8 bit exponent representing the number "0". Because the analog input signal is 20 dB lower, the signal /noise gets 20 dB worse than that of the first converter Then, the internal digital 32 bit float mixer aply a 20 dB digital gain to this signal, so the sample becomes approximately "equal" to the sample of the first signal, restoring the attenuation of the analog input. So, for a voltage input of 0 dBfs, the second converter plus the digital mixer will "finally" generate a 32 bit float word that consist of : - a 24 bit word full of "1" and - an 8 bit exponent representing "0" ( None of the never clip "converters" can generate a sample word representing a value over 24 bit,nor any of them can code voltage levels over 0dBfs , so the exponent in the 32 bit float word directly after the converters is allways zero. They are normal 24 bit converters.) When connecting voltages over 0 dBfs, the first converter will always clip. but the second converter analog input is 20 dB attenuated, so the second converter will start cliping (coding a sample word of 24 bit full of "1") when conecting an input voltage of +20 dBfs. Also, because of the digital gain applied after the second converter ,when connecting input voltages over 0 dBfs, there wil be digital signal levels representing voltage levels higher than 0 dBfs, and the internal 32 bit float mixer, will start to use the exponent on the second signal, because when using the exponent it can represent values higher than is possible with the 24 bit word. For coding voltages over 0 dBfs it is necesary to use the exponent, and then the system is really working in 32 bit float. ( I think this is the difference with never clip in wireless, since it should use a 24 bit processor) in the end, the system have 2 nearly identical signals: - the first with the better quality but allways with a 32 bit float sample word with an exponent equal to zero. If the input voltage exceeds 0 dBfs the signal will clip. - the second with 20 dB more noise, and his 32 bit float sample word could include an exponent different from zero if the input voltage exceeds 0 dBfs. Switching: (see my next post for the real switching system taken from the patent ) The system is normaly operating with the first signal. the switching probably uses some crossfade . switching from the first signal to the second: When conecting a voltage over 0 dBfs to a never clip input, the first signal will clip , but when the internal mixer detects a 24 bit sample word very close to full of "1" (signal clipping), it switches to the second unclipped signal,( this signal will be unclipped up to + 20 dBfs). Since the recording format is 24 bit type, when exceeding 0 dBfs at the input voltage, the 32 bit float signal must be attenuated before going to disk so the exponent becomes zero, then , when sending to disk, the zero exponent is ignored. Remenber that this is attenuation, it is diferent than truncation. The less significant bits are there, the noise go down, but te signal does it to. If you truncate (replace whit zeros the LSBs, so loosing them) you lose presicion and increse the noise. The reduced noise level for the quiet sounds with never clip is because the system uses the first (cleaner) signal to record them, but this lower noise is relevant Only if there is a big dinamic range in the acustic sound being captured , i.e. there must be loud sounds too, so the second signal is used, and so then there is a need for the atenuation . The loud sounds are recorded from the second noisier signal, with the extra noise "covered " by the loud sound itself. The quiet sounds are recorded from the first cleaner signal. If the input voltage didnt go over 0 dBfs, then neverclip was not used, and atenuation was not needed, Atenuatting the signal will not improve the recording if never clip was not activated. If digital attenuation is used, there is an increse in noise due to imprecise calulation of the attenuation but this is absolutely inaudible in this case in a 32 bit float system, so aplying digital gain or normalizing a digitaly attenuated signal doesnt have an audible effect . If the input voltage doesnt go over 0 dBfs, then it is not necesary to attenuate the signal, the exponent of the 32 bit float samples will allways be zero and the system will just ignore it before sending to disk. switching from the second signal to the first: to perfectly do this, it is necesary to use a long buffer, but that would introduce to much latency, so it is not possible. The switching from the first to the second signal can be immediate, but switching from the second to the first needs a release time (using the full of "1" 24 bit sample word system for this will lead in a switch about every 90 degree fase change in the input signal, so it is useless). The system ideally needs to analize the second signal for a time enough to cover the longest ( higher "period" ) signal ( 20 Hz signal= 50 miliseconds, this is the minimum but probably the product implemented time is longuer) to detect a real decrease in level under 0 dBfs before switching back. But using such a time can in some scenarios lead to a noticeable residual noise as described in another thread, when recording pulsating or fast decaying sounds. Remember that the second signal has 20 db higher noise floor, so, when recording signal voltages over 0 dBfs and there is a fast decay of the sound, the 20 db extra noise floor remains for at least 50 miliseconds before the system switch back to the first cleaner signal. This is more noticeable when recording very quiet sounds, because normaly you would set the gain very high, and with this you raise the noise floor, and, plus the extra 20 db noise of the second signal, the noise floor becomes audible when there is no sound to cover it (after the fast decay) The release time have been adjusted by zaxcom to minimise this residual noise at least in the 742 plug on tx, I supose this is done by Hipassing the signal to 80 hz (in order to excluding the longer period signals), With this, they could lower the release time to nearly a minimum of 12.5 to 20 miliseconds, So, when recording quiet and pulsating sounds, If you are sure there wil not be loud sounds, it is better to set the gain so never clip does not engages. finaly Never clip is for high dinamic range acoustic sound material. Is for saving you of the sudden scream, the unscripted performance changue, sudden laughs. It replaces the limiter, but is not intended to replace a proper gain setting. it is not for you to constantly peaking over 0 dBfs on purpose.

in my TV, receiving from the open AIR transmission, some times the sound apears out of sync. Changing to other channel, and then returning to original channel solves the problem.

to mister Larry: would you tell us why your wireless tx dont have an internal back up recorder? or is it to much an intromission? It is an exelent feature. I just saw that only the tascam lav beltpack is not for sell in USA, but the XLR version seems to be available, at least at tascam web site, I think the only diference between both units is the type of mic suport.

since a patent can specify a solution for a problem, (like "wearable recorder",etc), I think zaxcom may have pantented something like " system/solution for a belt pack recorder for lav mics" . This can include the TRX and specialy in this case the ZFR recorders.

it was an old technic, trying to fighft the feedback unity gain. Phase reversed signal from a second mic(for monitors) that is gaf-taped to the main mic (for front of house). I have never did that, and dont know if it realy works. You can see it in old live rock videos. It is not in use anymore, replaced by graphic eqs.

maybe a case of amplitude versus peak to peak measurement?

Yes, bias signal is for correction of the hysteresis ( a type of non linearity) when magnetizing a material.

I think is the compresor what is making the noise apear louder than a normal NC with ISO atenuation, making it like having 40 db higher noise instead of the normal 20 db extra for the second converter. I Think zaxcom needs to offer the atenuation prior to recording and transmiting, and leave the compresor as an option.

i have just got the redirection from google search to that URLshort .info site. I recently deleted the internet cache of Chrome, and then the redirection hapens again, but just once. I have tried several google searches for jwsoundgroup but can not get the rediredction .

"The H6 Omni Headset, with its new watertight connector, is submersible to one meter. It's ready to take on dust, sweat and make-up on stage and screen."

i apologize to you, ill try to write more carefully, I readed to quick and mixed some things in my response. I got into this thread because some people didnt bilieve that never clip was capable of doing what it states. I have explained a way of obtain the same results.(I believe that it is the same methot that neverclip is using), All the known information at this moment agree with my theory.

Digital gain is only a multiplication operation, I would not call it a technology. This post have lot of confussion, mixing diferent and unrelated cases. Sorry, I can not give you an answer to that. 153 dB of dinamic range is by faaaaaaar more dinamic range than any microphone will have ever. regarding the input voltage of the never clip inputs, there is a video where Glenn S. says: " ...the microphone output wil distort before the input of the recorder..."

lot of confussion, it seems you dont understand my posts. I suppose it is because you lack the necesary knowledge, sorry, I am not intending to be rude. "I always thought of the advantages of NeverClip to be in the quiet part of audio.." "The related advantage of NeverClip is that when you have gained the input to record a quiet scene, but then suddenly an actor starts to scream or whatever, there is enough headroom to not clip the signal" My posts explain exacly this topics, I have been accepting and I have been promoting this same caracteristics of never clip. and I have been trying to explain how this is posible. My explanation works, because it have been working for years with some extra manual work.

modern mic preamps can acept at least 24 dBu of input voltage, about 17 Volts of amplitud, that is more than any microphone can output without distort, so it is not necesary a new preamp. "...two different inputs with a difference of at least 20dB" Wow, that confirms my assumption , regarding the noise, it seems that you are confused, the noise have two components, 1- the noise added by all the analog electronic components before the cuantizer. that includes the preamp and the external analog circuitry necesary to setup the converter chip. The noise floor for a very good preamp at minimum gain seting is about -130dB of it`s maximun input level without distortion. 2- the cuantizing noise. this have 2 parts: a) error in the measurement of the sampled signal cause the 24 bit scale can not represent all the posible values. The measurements need to be rounded. This noise is mathematicaly calculable and is at -144dB, b ) the induced error in the cuantization produced by the real/non perfect components, and the timing errors in the previous sampling stage, like the jitter. so all this ends with a noise floor about -117 dBFS in the final product. and never clip adds 20 dB of headroom . That ends with 117+20=137 dB of dinamic range. Now, you only need to put a -20dB atenuator before the low converter to make the 20 dB diference, as I said before, and your diagram will be correct

yes, ultrasound capable gear for ultrasound recording. 20hz-20khz sound do not need 96khz recording, 48khz is perfect apart from extending the audio bandwidt, the other benefit of higher sample rates that I can think of is that with proper design, it is posible to reduce the latency.

If you want to record ultrasound to investigate , or if you want to hear a bat`s sonar by reproducing it at a low sample rate , it is necesary to use high sample rate recording. If you only cares about sub 20 khz content, in order to facilitate interpolation for time stretch, upsample gives the same result for the audible bandwidt.

we disagree because conversion is not an example of integral calculus. Sampling is about modulation, and cuantizing is about discrete value measurement. Conversion is to sampling and then cuantizing. Conversion is not directly related to area under the curve calculus nor primitivisation of mathematical functions. All the asumptions on better representing is just a missunderstandig on how converters do really work.

1 . No 2. like you said, no intermodulation in linear systems. there is a psicoacustical efect of intermodulation, sum and difference tones, discovered by the violinist and acustician Giuseppe Tartini, We had a demonstration with a violinist at the university. The sound can be listened. but it doest not exist in the acoustic realm, only in the brain. 3. yes, It adds frecuencies not present in the original sound, it is a type of non linear distortion. 4. yes , it is the same. If ultrasonics are not required, I repeat that it seems that in general, people dont understand the conversion process. The comon view of a converter as a device that fill voltage values thruogh time as seen visualy in the waveforms of a computer editor make people to think that with more samples the signal is better represented, and better sounding. That is incorrect, digital audio samples are not exactly the same than pixels in pictures. Into an editor, a 1 khz frecuency waveform looks uglyer sampled at 3 khzsf but when reproduced thruogh a properly designed converter for that sample frecuency , it sounds the same as it would sound if sampled at 96 khzsf. If there are tiny details between samples in a signal, that are not captured at 48 khz sample frecuency, is because that details are made of frecuencies higher than 24 khz the reason for CDs to use 44.1 khzsf has nothing to do with aliasing, is because of in the beginig of digital audio, the conversion of the masters to digital was made into U-matic analog videotapes with a special code of analog patterns. from wikipedia: " with the famous compact disc 44.1 kHz sampling rate based on a best-fit calculation for the U-matic's video horizontal-sync rate. " http://en.wikipedia.org/wiki/U-matic Also, I have noted a confusion in the terms "conversion, sampling and cuantizing". Sampling is an analog precess, the first part of the conversion process, Cuantizing is the measurement of the result of sampling. Only when the result of cuantizing is buffered, then the signal has been converted.

1. In some way, yes. You can conect an analog signal 20 dB higher than with a normal input, The reason for call it never clip is for, at common pre amp setings, and with high acoustic level sound, the output of a microphone will distort before the input of the recorder reach its maximum capacity. So the input will "never clip". But this is a tricky theme if you dont understand the way never clip works. In reality, the converters receive the same "normal" maximum voltage level, they can not convert voltages over the normal 0 dBFS. The essential is to have diferent gain settings before the converters. The amount of difference is what specifies the maximun extra headroom that you can add. The converter in the low path ( with a lower pre amp gain) is what make posible to add into the digital domain the signal portion of the higher than 0 dBFS voltage that the microphone outputs. 2. to the first,yes, but only for the loudest part of the signal, so the noise is totaly masked and inaudible. This is somehow tricky too. in reality what makes the low path (low gain) noisier is the normal theory. You are recording this signal with a sub amplification of 20 dB so the signal gets recorded 20 dB nearer the noise level compared to that of the high path converter. this under level recording is used to restore the cliped portions in the high level path, To reconstruct the cliped parts the low level signal needs to be digitaly amplified to match the level of the high level signal (raising the signal level and the recorded noise level), then, the clipped sections (loudest audio segments) reconstructed with the low level signal has 20 dB more noise, but this is into the digital realm. Once you atenuate the signal, the loudest part`s noise go down to the normal level, but the noise of the parts with a sound level under the normal clip level go down 20 dB lower than the normal noise level.

it is exactly what I stated in the other post "never clip in simple pictures", is a known process, The new thing is making the thing all automatic. Is called "high gain and low gain path", or "safety track". the conbiner replaces the clipped parts with the unclipped low gain path, but with a 20 dB digital gain to match levels, all this into the 32 bit float mixer capable of handling over 144 dB of encoding. finally someone who understand

yes ,tere is no such thing over 0 dBFS. what people still dont understand is that the trick of going "over 0 dBFS" is made into the digital mixer. Into the digital 32 bit Float mixer, the maximum posible encoding level is more than 1000 dB ( tecnicaly limitless for audio purposes) so the164 dB reconstructed signal from the 2 converters is perfectly encodable into 32 bit float. But the resulting signal is biger than a 24 bit system can handle, that is the reason for the atenuation prior the recording, and that is the reason cause the noise floor go down 20 dB compared to a normal 24 bit system

164 dB is not a dinamic range spec, It is the maximun posible encoding digital level, It corresponds to the whole resultant recording, this is : the cuantization noise summed with the imput signal. the cuantization noise (floor noise) at the input can be 38 or 58 dB. the dinamic range concept sais that it is the range between the minimum posible recording signal and the maximum posible recording signal. when the input signal is low, the noise floor is 38 db. when the input signal is very hot (over the clipping level of the first converter) the signal path switches to the 20 dB padded signal path, and aplyes to it a 20 dB digital gain. then the noise floor is rised 20 dB over the normal 38 dB (actually it is a little more noise because of the loose of resolution, cause this higher than 144 dB signal needs to be encoded in 32 bit Float using an exponent, multiple of 10 ) into the 32 bit float mixer the dinamic range is : 164-38=126 so the final dinamic range of wireless never clip after a 20 db atenuation is is: 144-18=126 without never clip the dinamic range is : 144-38=106 the extra 20 dB of never clip is due to the second converter extending the posible analog voltage encoding 20 dB over the normal 0 dBFS.