CESSB & COMP--obsolete or not?

K9RX
Posts: 414
Joined: Fri Apr 14, 2017 3:47 pm

Re: Thetis v2.9.0 - Updates and Issues

Postby K9RX » Fri Apr 08, 2022 11:46 am

Since it was mentioned here: what is "wrong" with CESSB? I don't use it because of its bad rep published here - but the guys at Elecraft, having just added it to the K4, are raving about its capabilities! They're saying it is a tremendous increase in average power, as much as 8 db, AND no degradation in the audio quality other than it 'sounds' louder ... is the implementation here not correct possibly, making it bad? [Of course this is not meant as a slight to any/all who worked on its development here - just curious as to why it's branded as "the terrible processor"]

Gary
K9RX
W4WMT
Posts: 325
Joined: Sun Apr 09, 2017 10:12 pm

Re: Thetis v2.9.0 - Updates and Issues

Postby W4WMT » Fri Apr 08, 2022 2:22 pm

Hi Gary,

I wasn't referring to CESSB in my "terrible" remark. I was talking about the old wideband compressor on the console transmit control group. I was thinking it should be deprecated in favor of the CFC. That way users could enable/disable CFC from the console and adjust the pre-comp gain using the old compressor controls there.

73,
Bryan
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

CESSB & COMP--obsolete or not? Also "loudness"

Postby w-u-2-o » Fri Apr 08, 2022 2:41 pm

You can blame me for being negative about CESSB.

Refer to Hershberger's two papers:

http://www.arrl.org/files/file/QEX_Next_Issue/2014/Nov-Dec_2014/Hershberger_QEX_11_14.pdf
http://www.arrl.org/files/file/QEX_Next_Issue/2016/January_February_2016/Hershberger_QEX_1_16.pdf

CESSB is a Rube Goldberg way of solving the problem of undesired peak-to-average ratio (PAPR, aka crest factor) that occurs as a result of overshoots or ringing in either or both of the filtering and SSB modulation process. It is unnecessary in a well designed SDR, such as the one that we enjoy when using Thetis (and PowerSDR mRX, but I will only refer to Thetis hereafter). In comparison to the other methods available in Thetis to achieve loudness CESSB more dramatically distorts the original audio waveform.

In order for CESSB to work effectively the SSB modulation process must of a linear phase type. Hershberger himself shows that CESSB processing can take place either before or after a linear SSB modulator. If the modulation process is of the linear phase type, then any modern, well-designed compressor/limiter algorithm either before or after the linear SSB modulator will work to solve what he calls the overshoot problem, do it more effectively, and do it with less distortion of the original audio waveform.

Refer to the WDSP User Guide:

https://github.com/TAPR/OpenHPSDR-wdsp/blob/master/WDSP%20Guide%2C%20Rev%201.20.pdf

In the case of Thetis, there is a linear SSB modulator of sufficient bandwidth and amplitude flatness, as well as the availability of linear phase filtering. This is followed by a modern, look-ahead limiter design. An example block diagram of such a limiter follows, just substitute I and Q for audio L and R, and of course there is no mixer, the output goes over Ethernet to the direct digital up converter (DUC) firmware in the FPGA:

Compressor_block.png
Compressor_block.png (15.7 KiB) Viewed 7015 times


This limiter we refer to as the ALC. Unlike a simple, primitive clipper circuit or algorithm (such as that found in the old, legacy Flex version of PowerSDR), any "overshoots" are removed by the ALC in a way that keeps transition rates to the limit low enough to prevent easily audible effects under any reasonable conditions. The limiter can do this in a controlled fashion because it has the necessary processing time due to the use of the delay lines, hence the term "look-ahead limiter". These simple and gradual adjustments to the waveform are substantially less disruptive than the summation of the transformed, clipped waveform in the CESSB algorithm.

Thus CESSB is wholly unnecessary in Thetis, and one need not accept the unique audio artifacts associated with the use of CESSB.

In addition to all this, it's important to understand that in terms of winning the "loudness war" in amateur radio, there is no better platform for this than Thetis. Thetis provides no less than three stages of compression: leveler, CFC, and ALC. Each provides their own unique and important opportunity to reduce the dynamic range (PAPR, crest factor) of the transmitted signal. Note that with the advent of CFC, COMP is essentially obsolete. COMP, referred to as the speech processor in the WDSP guide, is a very simple RF speech clipper, and can no longer be recommended for use. Prior to the advent of the CFC, phase rotator, post-EQ and improved DEXP many people relied on external processing, either professional audio rack hardware or digital audio workstation software, to achieve loudness with high fidelity and low distortion. This is no longer strictly necessary, but some people still find it a fun part of the hobby.

When used properly, the leveler, CFC and ALC, together with PureSignal, can achieve ludicrously low levels of PAPR in the transmitted signal while still maintain excellent audio quality and, importantly, very low IMD products. It is quite easy to achieve PAPR of well under 4dB. Such signals are LOUD, indeed most consider them objectionably loud.

Some additional reading on loudness and PureSignal:

https://apache-labs.com/community/viewtopic.php?f=20&t=2862
https://www.soundguys.com/the-loudness-war-51513/
https://en.wikipedia.org/wiki/Loudness_war

Some excellent tools for measuring loudness (and there is now a whole science behind this, with "Loudness units" being defined in many ways):

https://www.orban.com/meter
https://youlean.co/youlean-loudness-meter/

Finally, the implementation of CESSB in Thetis is perfectly correct. The modulators are all linear. You MUST use the Linear Phase Filters for proper operation, though. And in order for CESSB to be activated you MUST activate the old COMP speech clipper, the two functions are tied at the hip in the code. You can, however, leave the COMP setting at the minimum level.

I'd very much like to read the discussion on CESSB in the K4. Would you have a URL you could share, please?

73,

Scott
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: Thetis v2.9.0 - Updates and Issues

Postby w-u-2-o » Fri Apr 08, 2022 9:02 pm

K9RX wrote:but the guys at Elecraft, having just added it to the K4, are raving about its capabilities! They're saying it is a tremendous increase in average power, as much as 8 db

I see this on their K4 software page, https://elecraft.com/pages/k4-high-perf ... nd-utility.

It seems rather unlikely that they are achieving an 8dB improvement in PAPR unless the basic K4 design is so horribly bad that such an improvement is possible. Even Hershberger doesn't claim more than approx. 3dB, which makes good sense given a competent transmitter design.

One other thing I noticed is that in Hershberger's papers he goes out of his way to dismiss look-ahead limiting. This seems odd given that it is the gold standard in the audio industry. Clearly there is something wrong in Figure 17 in the first paper. No competently designed and operated limiter of any type would be that bad.
W4WMT
Posts: 325
Joined: Sun Apr 09, 2017 10:12 pm

Re: CESSB & COMP--obsolete or not?

Postby W4WMT » Fri Apr 08, 2022 11:11 pm

Okay, so can I put you down as a "NO" vote for CESSB?
:-) :-) :-)
K9RX
Posts: 414
Joined: Fri Apr 14, 2017 3:47 pm

Re: CESSB & COMP--obsolete or not?

Postby K9RX » Sat Apr 09, 2022 1:18 pm

Interesting .... well I had been getting their updates on a daily basis and when first released there was a lot of hoopla on CESSB. The (Elecraft) claimed "no distortion with as much as 8 db increase in average power" was talked about extensively ... and of course, owners would say "it works great". I generally put little credence to those types of discussions as people are easily persuaded to believe/do/go one way or another ... and I do wholeheartedly agree that 8 db seems an outrageous claim.

I tested radios for their compressor over the years - the best one I tested was an Icom 756ProIII, it had a legitimate 6 db increase in average power with the processor on... a Ten-Tec Orion II was by far the worst (and acknowledged in a one hour "lets try to fix this" conversation with one of their design engineers - who said at the end "this is why we introduced our EXTERNAL processor module) ... the FTdx5000 was around 3 db. I've not pulled out that wattmeter (average reading analog, great for this purpose) on the Anan though.

I'll take your word for it Scott. No reason to doubt it.

Gary
K9RX
User avatar
w9ac
Posts: 290
Joined: Sun Apr 09, 2017 4:01 pm

Re: CESSB & COMP--obsolete or not?

Postby w9ac » Sat Apr 09, 2022 2:03 pm

Just to add a point of clarification...

RF peak overshoot that CESSB corrects results from passing a complex waveform through the Hilbert Transform -- and this occurs with any type of SSB generation. CESSB is not designed to correct baseband audio overshoot, for example, from a sloppy baseband peak limiter.

So, to deal with the Hilbert problem and only the Hilbert problem, I am not clear what CFC does to manage Hilbert Transform overshoot -- if it does at all. I've looked at several IEEE papers that deal with CFC processing technique, but they all relate to audio processing in the baseband domain, not the type of control required to pass a complex waveform through the Hilbert Transform.

I do use CESSB but with no other processing internal to Thetis apart from activating COMP to engage CESSB. COMP is set to 0. I use one baseband peak protection device and it's a choice of either a Urei 1176LN or Urei LA-4 audio limiter. When either are used, the intent is to simply control audio peaks into my Presonus 192 sound interface.

Although not germane to CESSB, I am now using a modified ChannelMaster.dll file that bypasses Orion's internal port CODECs with substitution of the Presonus 192. The audio interface jacks on the 7000DLE's front panel are inoperative with the new file.

Paul, W9AC
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Sat Apr 09, 2022 3:01 pm

w9ac wrote:Just to add a point of clarification...

RF peak overshoot that CESSB corrects results from passing a complex waveform through the Hilbert Transform -- and this occurs with any type of SSB generation. CESSB is not designed to correct baseband audio overshoot, for example, from a sloppy baseband peak limiter.

Paul,

If you read the second paper from Hershberger, the one from Jan 2016, you will see that this is not always true, just as I wrote in my first post above. Thus processing on the baseband audio, prior to the modulator, is sufficient to solve the problem under the right circumstances. Quoting from the paper:

"It is possible to separate CESSB generation from a radio, however, if the radio SSB modulator is designed with this in mind. The SSB modulator must be linear phase, and must have a bandwidth sufficient to pass the CESSB spectrum, including its spectral skirts. If an otherwise conventional SSB modulator meets these requirements, then the peak control obtained by the CESSB process will be preserved. This will make it possible to use external processing to create CESSB."

Since these conditions exist in Thetis one can do it this way, with appropriate and competently used audio processing. In fact, it is worth noting that these conditions existed going all the way back to the legacy version of PowerSDR that was used with the early Flex-1500, 3000 and 5000. The ALC in that version of PowerSDR was a horrible clipper, and there was no CESSB. Those of us who wanted good performance were able to achieve that with proper hardware or software limiters on the baseband audio.

So while that proves the point that, given the right modulator and filtering in the radio, this can be controlled at the baseband audio level prior to the modulator, this is not what happens at all in Thetis.

So, to deal with the Hilbert problem and only the Hilbert problem, I am not clear what CFC does to manage Hilbert Transform overshoot -- if it does at all. I've looked at several IEEE papers that deal with CFC processing technique, but they all relate to audio processing in the baseband domain, not the type of control required to pass a complex waveform through the Hilbert Transform.

As I wrote, this is not what happens in Thetis. CFC does not manage overshoot, and nowhere in any of my posts did I write that, so I'm not sure why you even bring up the question?

Thetis uses a competently designed look-ahead limiter after all processing, including the modulator, but prior to the DAC, to limit overshoot on the I/Q data that goes to the DAC. This limiter is called "ALC" in the code, the documentation and the user interface, but it is a limiter.

Of course, given the linear design of the modulator and filtering (if linear phase filtering is selected) in Thetis, one can use audio processing to significantly limit the burden on the ALC, or even eliminate it altogether.

Although not germane to CESSB, I am now using a modified ChannelMaster.dll file that bypasses the Orion's internal port CODECs with substitution of the Presonus 192. The audio interface jacks on the 7000DLE's front panel are inoperative with the new file.

Is that Bryan's work, or something you developed? Bryan, W4WMT has been working along the same lines.
User avatar
w9ac
Posts: 290
Joined: Sun Apr 09, 2017 4:01 pm

Re: CESSB & COMP--obsolete or not?

Postby w9ac » Sat Apr 09, 2022 5:11 pm

w-u-2-o wrote:
w9ac wrote:Just to add a point of clarification...

RF peak overshoot that CESSB corrects results from passing a complex waveform through the Hilbert Transform -- and this occurs with any type of SSB generation. CESSB is not designed to correct baseband audio overshoot, for example, from a sloppy baseband peak limiter.

Paul,

If you read the second paper from Hershberger, the one from Jan 2016, you will see that this is not always true, just as I wrote in my first post above. Thus processing on the baseband audio, prior to the modulator, is sufficient to solve the problem under the right circumstances. Quoting from the paper:

[i]"It is possible to separate CESSB generation from a radio, however, if the radio SSB modulator is designed with this in mind.


Completely understood and no argument there. I have used W9GR's .WAV files to generate CESSB external to the transceiver. I was referring to analog baseband peak control without going through a pre-processes routine to generate a peak corrected waveform.

So while that proves the point that, given the right modulator and filtering in the radio, this can be controlled at the baseband audio level prior to the modulator, this is not what happens at all in Thetis.


Right, I understand that CFC has the ability to use look ahead processing. I would like to see a detailed analysis of CFC's inner workings within Thetis, not just a block diagram. Anyone know if such an analysis exists in a presentation format similar to what W9GR did with CESSB?

So, to deal with the Hilbert problem and only the Hilbert problem, I am not clear what CFC does to manage Hilbert Transform overshoot -- if it does at all.


You answered that question with the last reply. Thanks!

Paul, W9AC
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Sat Apr 09, 2022 5:30 pm

w9ac wrote:I was referring to analog baseband peak control without going through a pre-processes routine to generate a peak corrected waveform.
Would you please explain/expand on that statement? I don't understand what the difference is between "analog baseband peak control" and "pre-process routine to generate a peak corrected waveform". Or is that another way of saying "limiter" vs. "CESSB"? If it is, then we may be talking at cross purposes. My point is that with the Thetis architecture, a good limiter is more than good enough, the complex CESSB algorithm is unnecessary and causes significantly greater changes to the audio in comparison without achieving a better PAPR.

Right, I understand that CFC has the ability to use look ahead processing. I would like to see a detailed analysis of CFC's inner workings within Thetis, not just a block diagram. Anyone know if such an analysis exists in a presentation format similar to what W9GR did with CESSB?
Again, CFC has no bearing on the overshoot problem that I am aware of, but it will do an excellent job of reducing audio dynamic range. And as far as I know CFC does not use look-ahead processing. Are you continuing to confuse CFC with ALC, or are you really interested in CFC?

The CFC approach is touched on in this video at about the 13 minute mark:

User avatar
w9ac
Posts: 290
Joined: Sun Apr 09, 2017 4:01 pm

Re: CESSB & COMP--obsolete or not?

Postby w9ac » Sat Apr 09, 2022 5:51 pm

w-u-2-o wrote:
w9ac wrote:I was referring to analog baseband peak control without going through a pre-processes routine to generate a peak corrected waveform.
Would you please explain/expand on that statement? I don't understand what the difference is between "analog baseband peak control" and "pre-process routine to generate a peak corrected waveform". Or is that another way of saying "limiter" vs. "CESSB"?

Yes, correct. And, we are talking past each other. It's easy to do with the subject matter! I just looked at Warren's presentation. It answers many questions but generates others. I'll communicate directly with him concerning those questions.

Paul, W9AC
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Sat Apr 09, 2022 6:56 pm

Please post what you find out.
W4WMT
Posts: 325
Joined: Sun Apr 09, 2017 10:12 pm

Re: CESSB & COMP--obsolete or not?

Postby W4WMT » Sat Apr 09, 2022 7:36 pm

Yes please!

In fact, all these tutorials from Dr. Pratt should be in a big wiki somewhere. IMO
N4XD
Posts: 105
Joined: Tue Sep 29, 2020 11:59 pm

Re: CESSB & COMP--obsolete or not?

Postby N4XD » Fri Aug 19, 2022 12:00 pm

With Thetis 2.9.0.6, where do I turn CESSB on or off?

Thanks

Ron
N4XD
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Fri Aug 19, 2022 12:28 pm

The option checkbox is on the Setup > Transmit page. However if COMP is off (as it should be--use CFC) then CESSB is also disabled regardless of the state of the checkbox.
N4XD
Posts: 105
Joined: Tue Sep 29, 2020 11:59 pm

Re: CESSB & COMP--obsolete or not?

Postby N4XD » Fri Aug 19, 2022 3:11 pm

Thanks. Will go look for it now.

Ron
N4XD
Posts: 105
Joined: Tue Sep 29, 2020 11:59 pm

Re: CESSB & COMP--obsolete or not?

Postby N4XD » Fri Aug 19, 2022 3:41 pm

The only choice I see under "Setup- Transmit" that has anything to do with CESSB is the choice "CESSB Overshoot Control" under speech processor. Perhaps that is what turns CESSB on and off?

Thanks

Ron
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Fri Aug 19, 2022 4:08 pm

That is correct. But remember, again, it's only active if that box is checked AND COMP is also on.
User avatar
w9ac
Posts: 290
Joined: Sun Apr 09, 2017 4:01 pm

Re: CESSB & COMP--obsolete or not?

Postby w9ac » Sat Aug 20, 2022 7:26 pm

A while back Dave Hershberger and I exchanged messages concerning CESSB as well as some historical material that we shared when we worked at the same AM/FM broadcast stations while I attended Norther Illinois University in DeKalb, Illinois -- the same station as Chief Engineers but during different periods of time. Dave completed post-graduate studies at the University of Illinois, Urbana. I'll post his comments here that are applicable to Apache/ANAN but also apply to misinformation I've seen on the Elecraft K4 and Flex message boards. Quoting Dave:

"As for look-ahead processing, that's what FlexRadio had for ALC before they implemented CESSB. As I recall they had a 2.1 dB improvement in average power over look-ahead ALC. So more than 60% increased power over look-ahead limiting. Somebody was saying that you don't need CESSB if you do your audio processing correctly and then apply that to a linear phase SSB modulator. Not so. I have not seen an audio processor that controls both the audio and its Hilbert transform simultaneously, which is what CESSB does.

Someone said that look-head audio limiting is all you need. Not so, you still have to deal with the Hilbert transform. I saw myself being misquoted in that forum. Oh, well.

I think that Elecraft is probably combining some Hilbert clipping with CESSB to get their claimed 8 dB increase. And that's fine if they want to do it."


It goes on but that's the part germane to our CESSB discussion. The part concerning the Hilbert transform overshoot is what I was trying to convey above, but probably miserably. No matter the method SSB is generated, the mathematical Hilbert transform creates overshoot with any complex wave, apart from a sine wave. CESSB is simply controlling wasted peak overshoot as a result of the transform -- something not capable with only look-ahead processing, nor CFC unless CFC has its own method of dealing with Hilbert transform overshoot. Look-ahead and CFC are powerful processing tools capable of high modulation density but they don't deal with the wasted peak power overshoot problem.

Paul, W9AC
W4WMT
Posts: 325
Joined: Sun Apr 09, 2017 10:12 pm

Re: CESSB & COMP--obsolete or not?

Postby W4WMT » Sat Aug 20, 2022 8:26 pm

I've read Dave's paper I don't know how many times. Toward the end of his paper, he cautions us to not drive the CESSB algorithm into compression. But rather to use multiband baseband processors (CFC passes that bar) and then use CESSB at the very end to only remove the Hilbert artifacts.

While I love the idea of CESSB, I can't ignore the fact that it sounds terrible on the air. But why ??!!??

73, Bryan W4WMT
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Sat Aug 20, 2022 9:33 pm

w9ac wrote:No matter the method SSB is generated, the mathematical Hilbert transform creates overshoot with any complex wave, apart from a sine wave.

Agreed.

CESSB is simply controlling wasted peak overshoot as a result of the transform

Agreed.

-- something not capable with only look-ahead processing

Respectfully, not correct.

If you want to get into semantics, and modify your statement to read "Something that look-ahead processing does after the fact rather than as part of the modulation process", then I would agree with you. And just to put a super-fine point on it: I'm discussing PAPR at the antenna port, not where or how the PAPR is reduced. I.e. I'm making no discrimination about what peaks are removed/limited/modified where. Nevertheless, mathematically, any, any, clipper or limiter, no matter how simple or advanced, and CESSB processing, all of these methods reduce peak to average power ratio (PAPR) at the antenna port.

The questions are: what do these PAPR reduction methods do to both the RF characteristics of the transmitted signal, and what do these methods do to the information content of the transmitted signal? Generally we are discussing audio here, a voice waveform, so what do these methods do to the voice waveform?

Clippers and simple limiters are obviously responsible for substantial distortions of both the RF and voice signals. Their use creates substantial intermodulation distortion (IMD) and therefore these methods are best avoided.

More advanced limiters, particularly those using look-ahead algorithms, produce very little IMD and other distortions. That is why they are the method of choice for when limiting is desirable. Nowhere is this more evident than in the professional audio industry, but it carries over to any waveform from DC to daylight that needs dynamic range reduction against an absolute maximum level.

CESSB does, indeed it must, distort the baseband and transmitted RF waveforms. Otherwise a reduction in PAPR would not be possible. The argument for CESSB is that it removes unwanted peaks (another form of distortion) caused by the Hilbert transform. Thus one would expect it to sound good subjectively since it is apparently not messing with information content outside of those peaks (maybe). However, also subjectively, the vast majority of operators have abandoned it in favor of other methods because a) the other methods sound better (contain less distortion), and b) can achieve higher levels of PAPR reduction without sounding bad.

I honestly have not studied IMD levels of CESSB vs. WDSP ALC (in Thetis) in either the baseband or RF domains. If someone has the time (I don't at present, maybe in the later Fall time period), it would be interesting to get some real numbers. Signal levels throughout the system would have to be carefully adjusted to achieve the zero ALC action in the CESSB case, and the correct amount of ALC action to equal the PAPR reduction achieved using CESSB.

To bottom line this: there are a variety of methods to achieve PAPR reduction, from the sublime (argue CESSB vs. look-ahead limiters), to the mundane (simple limiters, clippers) to the ridiculous (saturating your amp--yes, that will do it, too!) These various methods cause varying amounts of RF IMD and baseband audio IMD and other distortion products. CESSB is limited in what it can do in terms of PAPR reduction, other methods are effectively unlimited at the expense of increasing distortion. We need an experiment to a) objectively measure the distortion vs. PAPR reduction of the various methods, and b) make a subjective determination of what sounds better.

Anecdotally it would seem most people prefer the leveler/CFC/ALC combination in Thetis above all other methods, but I am willing to be shown to be misguided in that statement ;)
User avatar
w9ac
Posts: 290
Joined: Sun Apr 09, 2017 4:01 pm

Re: CESSB & COMP--obsolete or not?

Postby w9ac » Sun Aug 21, 2022 1:46 am

W4WMT wrote:I've read Dave's paper I don't know how many times. Toward the end of his paper, he cautions us to not drive the CESSB algorithm into compression. But rather to use multiband baseband processors (CFC passes that bar) and then use CESSB at the very end to only remove the Hilbert artifacts.

While I love the idea of CESSB, I can't ignore the fact that it sounds terrible on the air. But why ??!!??

73, Bryan W4WMT

Right, and I believe that is still Dave's current position. I'm not sure how others have been using CESSB (i.e., with or without additional processing), but my only pre-processing at the moment is a UREI 1176LN peak limiter ahead of a Presonus Studio 192 audio interface. It's there to ensure that the input level to the Presonus doesn't reach clipping if I close-mic the audio. Apart from that, CESSB is active and the COMP setting is normally at 0, but I've used it with a setting as high as 4 dB without any ill effect during recorded A/B comparisons with and without the CESSB box checked. For monitoring, I use either an ICOM IC-7610 or EE SunSDR2, both in wideband filter modes.

At least to me, Flex's implementation of CESSB sounds equally as good. I've not yet heard an Elecraft K4 on the air, but I do have one on order to compare CW and SSB performance with the ANAN 7000 MKII. Whether I keep it will depend on its performance not relative to the ANAN, but with the IC-7610.

The 7610 and K4 are very close in terms of measured performance and operating features (compare QST charts side-by-side). But, the K4D with optional tuner is 2x the price of the 7610. One benefit of the 7610 that's not addressed in the K4 design is a tracking preselector. However, the K4HD version is expected to have top-of-the-chart performance. Time will tell.

Paul, W9AC
User avatar
w9ac
Posts: 290
Joined: Sun Apr 09, 2017 4:01 pm

Re: CESSB & COMP--obsolete or not?

Postby w9ac » Sun Aug 21, 2022 2:16 pm

This morning, I tried the CFC-CESSB settings as suggested in the latest Thetis operating manual. Following this advice, I found that harshness is greatly diminished whether or not CESSB is engaged. So, it's possible that some ops are not taking this advice because at least in my case, it does require a substantial reduction of the PRE-COMP slider to achieve a 0 dB peak reading. After the reduction, there's still a significant overall loudness change when CFC is engaged.

Here's the relevant portion of text from the Thetis operating manual:

Summary: Remember that while you are operating you can manually toggle COMP on or off to add a hard limiting effect to your transmit audio if desired. If you have enabled CESSB its hard limiting effect will also be present each time COMP is enabled. Remember, if COMP is enabled when you save your Transmit Profile these features will be on by default. For those who wish to enable the console COMP button and CESSB, excessive output from the CFC components may make your transmitted audio sound somewhat harsh. To minimize this, try reducing the PRE-COMP slider so that the CFC meter displays maximum peaks of 0dB. Note that when COMP and CESSB are enabled, the output is hard-limited at 0dB as shown with the “ALC Comp” meter. A new adjustment for COMP and CESSB users is available that allows you to exceed 0dB of ALC compression with COMP and CESSB enabled to make it possible to use the lookahead algorithm at the ALC level to incorporate soft-limiting in the final stage. You can try this new feature by moving to the Menu > Setup > DSP > AGC/ALC form and using the new ALC Max Gain setting to increase ALC Comp in 1dB steps from 1dB to 10dB. Several dB of ALC Comp should increase your over-all loudness without added harshness. Note that these CFC adjustment steps should be considered a starting point for optimizing your transmitted audio. When you have become comfortable with the interface, you might wish to experiment with changing the frequency points for the CFC sliders so they span the transmitted bandwidth of each transmit profile you are working on. As an example, for a 3.0k sideband profile try the following values: 50, 150, 300, 500, 750, 1250, 1750, 2000, 2500, 3000. There’s nothing magical about those numbers so experiment with values that give you the best tonal control for your intended bandwidth.

Paul, W9AC
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Sun Aug 21, 2022 6:11 pm

w9ac wrote:This morning, I tried the CFC-CESSB settings as suggested in the latest Thetis operating manual. Following this advice, I found that harshness is greatly diminished whether or not CESSB is engaged. So, it's possible that some ops are not taking this advice because at least in my case, it does require a substantial reduction of the PRE-COMP slider to achieve a 0 dB peak reading.

No meter should ever exceed 0dB at any point in the chain with the exception of ALC-Comp (or ALC-Group, they measure the same thing but in different ways).

After the reduction, there's still a significant overall loudness change when CFC is engaged.

Of course. CFC reduces the dynamic range of the audio by exactly the same amount regardless of where the pre-comp slider is set. All you are doing is moving the span of that dynamic range up or down in the absolute audio level regime with pre-comp.

Remember that while you are operating you can manually toggle COMP on or off to add a hard limiting effect to your transmit audio if desired. If you have enabled CESSB its hard limiting effect will also be present each time COMP is enabled.

I don't know if these functions exhibit a hard limiting characteristic, but if they do then you certainly don't want them.

For those who wish to enable the console COMP button and CESSB, excessive output from the CFC components may make your transmitted audio sound somewhat harsh.

This is true with or without CESSB.

To minimize this, try reducing the PRE-COMP slider so that the CFC meter displays maximum peaks of 0dB.

Yes, and this same approach should be taken all along the audio processing chain. For instance, select the EQ meter mode and then adjust the EQ preamp setting to keep levels below 0dB.

Note that when COMP and CESSB are enabled, the output is hard-limited at 0dB as shown with the “ALC Comp” meter.

This is incorrect. COMP and CESSB, or CFC for that matter, are not the arbiters of what happens on any of the ALC meter modes. It is the ALC that controls this. The default setting for ALC is 0dB, and since ALC will peak limit at that threshold. Without adjustment of that threshold what you see is what you get. Note I wrote "peak limit" to distinguish what ALC does from a hard limiter, since ALC is not a hard limiter.

A new adjustment for COMP and CESSB users is available that allows you to exceed 0dB of ALC compression with COMP and CESSB enabled to make it possible to use the look-ahead algorithm at the ALC level to incorporate soft-limiting in the final stage.

This adjustment is not new at all. It was introduced in PowerSDR mRX 3.4.1 in 2017, so it's over 5 years old. Savvy users have been doing this for a long time with or without CESSB as it effectively provides a third stage of compression (leveler, CFC, and ALC--four stages if you count COMP).

You can try this new feature by moving to the Menu > Setup > DSP > AGC/ALC form and using the new ALC Max Gain setting to increase ALC Comp in 1dB steps from 1dB to 10dB. Several dB of ALC Comp should increase your over-all loudness without added harshness.

Very true, and only because ALC is a look-ahead, soft limiter. However, because ALC is a limiter, even though it is an advanced limiter one should not look for more than about 3dB of additional compression at this stage lest things start sounding harsh.

Note that I'd strongly recommend using the ALC Group meter mode rather than the ALC Comp meter mode. ALC metering modes in Thetis and PowerSDR mRX before it are bizarre. The ALC metering mode shows what the signal level is doing below the ALC max. level threshold. ALC Comp shows what the signal is doing if that threshold were not in place, but does not show any levels below the threshold. ALC Group shows both of those indications on a single meter, which is much more convenient.

Paul--what you did not do is measure PAPR with and without CESSB enabled while keeping all other things equal (and COMP at 0dB).
User avatar
w9ac
Posts: 290
Joined: Sun Apr 09, 2017 4:01 pm

Re: CESSB & COMP--obsolete or not?

Postby w9ac » Sun Aug 21, 2022 6:59 pm

This morning's test wasn't meant to be a thorough analysis. I wanted to determine if CESSB and CFC can coexist and not create excessive grunge of the kind that's been reported. In fact, most of the distortion heard results when COMP and CFC are engaged with default CFC values, without CESSB active. Bringing the PRE-COMP slider down made a significant difference.

Others may be activating CFC with COMP engaged to activate CESSB, while not paying attention to Thetis metering and not realizing how hot the default CFC PRE-COMP level may be. To that point, it may be best to default PRE-COMP to a lower level, then allow the user to advance it as needed.

Regarding any inaccuracies in the Thetis User Manual or reference to any "newness" of a function or feature, that should be taken up with Laurence Barker, G8NJJ who edited V1.1. I don't see anything newer on Github. If there is, perhaps attach a direct link.

Paul, W9AC
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Sun Aug 21, 2022 8:25 pm

w9ac wrote:This morning's test wasn't meant to be a thorough analysis. I wanted to determine if CESSB and CFC can coexist and not create excessive grunge of the kind that's been reported. In fact, most of the distortion heard results when COMP and CFC are engaged with default CFC values, without CESSB active. Bringing the PRE-COMP slider down made a significant difference.

This is a problem even without COMP or CESSB, but merely CFC by itself. You can also do yourself in by having the EQ preamp level too high, post EQ gain too high, leveler gain too high, leveler decay too low, and the same for ALC.

Regarding any inaccuracies in the Thetis User Manual or reference to any "newness" of a function or feature, that should be taken up with Laurence Barker, G8NJJ who edited V1.1. I don't see anything newer on Github. If there is, perhaps attach a direct link.

I wasn't saying any of those inaccuracies were in any way associated with your work, just pointing them out for what they are.

AFAIK that is the latest version. Unfortunately it contains a lot of mistakes and misperceptions. So many, in fact, that when it first came out I could not read more than ten pages into it without wanting to make voluminous corrections. This is not meant as a complaint, BTW. One can't complain about such efforts. Even with its flaws it is still head and shoulders above what came before it. Nevertheless, sometimes there is the opportunity for clarification ;)
wx7y
Posts: 25
Joined: Tue Mar 01, 2022 5:01 pm
Location: Castle Dale
Contact:

Re: CESSB & COMP--obsolete or not?

Postby wx7y » Wed Sep 21, 2022 4:10 am

This post is not much to do with the topic BUT I have a question.
If I run Both the COMP and EQ my TX audio is really badly distorted and choppy, If I turn one or the other OFF the Audio works and sounds really good.

Maybe this is just the way it is BUT would be good to know.
Thanks a lot

73's
Bret
WX7Y
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Wed Sep 21, 2022 10:44 am

Short answer: don't use COMP, it's obsolete with the advent of CFC.

Longer answer: read these two topics, watch the videos, get your levels right:

viewtopic.php?f=6&t=2550
viewtopic.php?f=6&t=2508
rdwing
Posts: 92
Joined: Sat May 21, 2022 2:05 am

Re: CESSB & COMP--obsolete or not?

Postby rdwing » Mon Nov 14, 2022 9:55 am

w-u-2-o wrote:
After the reduction, there's still a significant overall loudness change when CFC is engaged.

Of course. CFC reduces the dynamic range of the audio by exactly the same amount regardless of where the pre-comp slider is set. All you are doing is moving the span of that dynamic range up or down in the absolute audio level regime with pre-comp.

A new adjustment for COMP and CESSB users is available that allows you to exceed 0dB of ALC compression with COMP and CESSB enabled to make it possible to use the look-ahead algorithm at the ALC level to incorporate soft-limiting in the final stage.

This adjustment is not new at all. It was introduced in PowerSDR mRX 3.4.1 in 2017, so it's over 5 years old. Savvy users have been doing this for a long time with or without CESSB as it effectively provides a third stage of compression (leveler, CFC, and ALC--four stages if you count COMP).

You can try this new feature by moving to the Menu > Setup > DSP > AGC/ALC form and using the new ALC Max Gain setting to increase ALC Comp in 1dB steps from 1dB to 10dB. Several dB of ALC Comp should increase your over-all loudness without added harshness.

Very true, and only because ALC is a look-ahead, soft limiter. However, because ALC is a limiter, even though it is an advanced limiter one should not look for more than about 3dB of additional compression at this stage lest things start sounding harsh.


I've read all the literature, the forum posts, watched the videos, and I'm still having a hard time totally wrapping my noggin around what you're saying regarding Pre-Comp. Is this just an overall compression slider? I haven't found that to be shown on the CFC Comp meter. Can you explain more regarding CFC Pre-Comp and the COMP sliders? They appear to work differently from other multiband compression solutions in the audio world.

Also, I've noticed that with CESSB enabled, the transmit bandpass is cleaner than without it. Also still trying to understand exactly how setting the ALC Gain above 0 dB interacts with CESSB. I understand that you can adjust the max alc gain to gain some additional soft compression, but how these two functions interact is unclear.
User avatar
w-u-2-o
Posts: 5540
Joined: Fri Mar 10, 2017 1:47 pm

Re: CESSB & COMP--obsolete or not?

Postby w-u-2-o » Mon Nov 14, 2022 12:37 pm

rdwing wrote: I've read all the literature, the forum posts, watched the videos, and I'm still having a hard time totally wrapping my noggin around what you're saying regarding Pre-Comp. Is this just an overall compression slider?

No.

I haven't found that to be shown on the CFC Comp meter. Can you explain more regarding CFC Pre-Comp and the COMP sliders?

Pre-comp is merely an input gain adjustment to the CFC itself, just like you might find an input and output gain knob on a regular, rack-mounted, multi-band compressor. The other sliders in the CFC section adjust the compression gain bands centered on the frequencies assigned to the sliders.

Similarly, the post-EQ gain slider is both the CFC output gain adjustment and input gain adjustment to the post-EQ stage.

They appear to work differently from other multiband compression solutions in the audio world.

That is true, it is different. The original request of Warren was for a conventional multi-band compressor, whereby there were some number of passband filters of some shape and compression gain was adjusted individually in each one. However, from a code perspective, it was much easier for Warren to simply re-use the existing TX EQ code in a different way. It also turned out to be relatively trivial to add a post-EQ section the same way all as part of the same processing block. As it happens, the code that Warren developed for all three of these stages is very different compared to a multiplicity of bandpass filters, either for EQ or compression. Instead of a bunch of filters acting in the time domain, the calculation is performed in the frequency domain and, given the large number of FFT bins involved, the compression is continuously varying across the array of bins. Hence the name "continuous frequency compressor". While I was admittedly initially skeptical of this approach it turned out to work really well and it allowed for significant code reuse and a simplified development cycle. A much better explanation of this occurs in one of Warren's presentations, the second one posted here, https://community.apache-labs.com/viewtopic.php?f=9&t=4018, starting a little over 13 minutes in.

Also, I've noticed that with CESSB enabled, the transmit bandpass is cleaner than without it.

Funny, I've observed exactly the opposite.

Also still trying to understand exactly how setting the ALC Gain above 0 dB interacts with CESSB. I understand that you can adjust the max alc gain to gain some additional soft compression, but how these two functions interact is unclear.

They don't interact per se, other than the fact that any IMD or other distortion products created by CESSB (or any other type of processing) will be further accentuated by compression at the ALC stage.

Return to “Thetis”