Clock buffer chips. Are these chips buffered or not?

[DrAlx]

A while back I took a whole bunch of measurements to see how using using clock buffers affected the gain of a BBD.
I tried a bunch of different buffer chips in the EM3207 electric mistress clone and compared them against using just an unbuffered CD4013.

I tried two types of inverting buffers (CD4049B and CD4049UB) and also tried non-inverting hex buffers (CD4050B).

Now my understanding was this...

Each inverter on the 4049UB chips consists of a single CMOS NOT gate.  The data sheet makes that clear.
The (non)-inverter 4050B chips consist of 2 CMOS NOT gates in series.  Also from the data sheet.
Each inverter on the 4049B chips consists of a 3 CMOS NOT gates in series.  At least that's what I seem to remember but I cant find the data sheet.

The UBE chips I found were easily available.  The 4049BE not so much. 

Anyway I tested all three over a range of clock frequencies to see how they affected BBD gain.
For each clock frequency I took an average over audio frequencies from 100Hz to 1kHz.

I was expecting the most gain from the 4049B chips due to 3 inverter stages and the least from the UBE chips which have just one.
What I found was that there was no measurable difference between the 4049B and 4049UB chips.
The 4050s gave the strongest gain.

So what is the deal with the 4049B vs the 4049UB chips? 

[Fender3D]

If these are real numbers (500MHz?), I guess ~0.5dB (max difference) may be lost for whatever reason...

[anotherjim]

I think you are more likely to spot a difference if you could study the rise and fall times. Buffered should give a sharper response as the input swing gets more gain from the extra stages. With good sharp input swings, low source impedance and low stray capacitance, you maybe won't see much difference between B and UB.

Paradoxically, buffered can cause havoc with slow changing inputs leading to full swing jitters on the output, but you won't be going anywhere near that slow in this application.

I'm not how much BB gain limit is down to frequency or rise/fall times or a bit of both. I think they only spec against frequency but surely switching speed matters too?

[DrAlx]

When I put my scope on the clock lines and read the dBm level on the scope, the 4050B chips were giving a power reading that was higher than the inverting chips, in agreement with the curves above.
I did actually look into all the rise, fall, and response times of these chips because the MN3207 datasheet has minimum requirements on how complementary clock voltages cross each other. I wanted to see if I could improve the BBD gain by "waveshaping" which I did by using some small caps between some of the inverter stages to make sure that one clock line would not start to go high till the other was well over 90% of the way down.  It didn't help though, and I got best results without doing any such tweaks.

[ElectricDruid]

Those results say "they all work" to me. I mean, you've got a drop of a *decibel* between 50KHz and 500KHz?! That's pretty insignificant. Ok, it'll change the depth of the notch in your flanger, but hey - what with everything else it's doing, you won't notice/care. You can hide that under the carpet labelled "character" and groove on.

The rise/fall thing is interesting. The small gap is known as a "deadband". I've found it does help a bit at the very highest frequencies, but doesn't make much odds until you get up there. These experiments were done when I was developing my PIC-Based flanger chip, and on that chip, you can control the deadband directly without affecting the rise or fall time.

Tom