Hammond clone anyone?

[RJHollins]

Had a quick listen ... man this sounds really nice !

Was taking a look inside ... even the 'internals' are beautifully laid out.

Big Thanks for what you bring to the FS forum, Martin.

[HughBanton]

Had some time at last. It's technically looking great, and definitely ripe for the non-Hammond expansion I've got in mind. Since I got back into FS a year ago I've barely entered the world of blue or packed so it's a real education too. I can immediately see how to play with attack and decay speeds, how to add more harmonics and so on.

Superb stuff Martin

One issue that needs addressing is 'addition'; it currently adds linearly (i.e. 1+1=2) and musically that sounds slightly wrong. It's not always immediately obvious (unless like me you're looking for it ..) but in particular it makes octaves and fifths overloud compared to all other intervals. On a 'scope you can see the higher peaks, but I discovered you can even measure it with a multi-meter; draw 8 & 4 drawbars and play two notes - the reading you get from an octave interval is higher than all the others, because the two waveforms being added are phase locked to each other so presumably produce a higher RMS as far as the meter is concerned. Likewise for ears, unfortunately.

I've faced this issue before with additive organ design, but to double-check I've done some measuring against the two 'clonewheel' keyboards that I currently have, Hammond & Roland, and I also checked out in VB3. Running the above test all three add, I'm guessing, logarithmically (which would be 1+1=1.414), and give approximately the same level, both apparent & measured, on every interval, octaves included. It makes the keyboard feel much more even to play.

Here's an article that relates to this : http://www.engineeringtoolbox.com/addin ... -d_63.html They are talking about physical noise sources - industrial fans for example - which will be completely phase incoherant. I suppose two real musical instruments are going to be 'phase semi-coherant' (whaa?) so maybe the optimum addition coefficient is somewhere elusive, between 1.414 and 2 anyway.

I elsewhere came across an 'Addition Table' for different dB levels (can't locate it again now, sorry ..) which showed: if we call 0dBs "1.0", then adding -12dB (0.25) to 0dB could be ignored altogether, i.e. 1 + 0.25 = 1! Only levels greater than -10dB were significant, amazing. I can vouch for the fact that with long trains of very low level high-harmonics (Compton Organ obviously, not Hammond) linear addition sounds very wrong, 'over-stringy' in the extreme. In this situation you have low level harmonics occupying virtually every note of the scale (way up high) so now every note's harmonics are louder than they should be.

Sorry to ramble on! Finally ... Hammond, Compton, Lowry, all of them really, got away with this because they mixed their signals or voltages using actual keyboard switches (or relays) and resistors. So the more keys pressed the bigger the load, which proportionally drags the summed levels down. Not very scientific but with the right relative resistor values it evidently has the required musical effect.

Mind you, I haven't the first clue how we deal with this in Blue-land! (But I can confirm that way back then, over in Analogue-Multiplex-land, we solved it with a logarithmic amp in the voltage-control chain - does that help?).

H

[Spogg]

Fascinating Hugh!

I played around with one of my synths and used the inverse of the number of voices sounding to reduce the volume. I fiddled with the parameters and got the ratios something like you said. This was just experimental, in green, but I wanted to hear the result.

Now what I noticed is that if you play a lot of notes, then release most but not all of them, the remaining notes still sounding jumped up in volume a bit. Is this what a real Hammond does? From your description of the electronics I would sort of expect that behaviour. When I played “normally”, without this close inspection, it sounded ok but I didn’t really like the idea once I knew it was there.

So, I guess I’m missing something fundamental, because surely a fast compressor could achieve a similar result…?

Cheers

Spogg

[HughBanton]

What you describe is indeed a general compressor, albeit keyed via no of voices rather than signal level; but I guess the compression result would generally be similar. I've heard of this being done elsewhere, and no doubt for reasons that have their origin in 'over-adding', which will plague any system where many phase-coherent* tones are combined and things seems over-loud as a result.

However here I'm referring to the levels of individual harmonics within each note's spectrum, and how they are influenced (ideally not at all), by other notes that are also playing. Only applies to additive harmonic schemes. Consistency of tone rather than level.

A real Hammond behaves remarkably smoothly btw, you never notice any irregularitries at all; one of the reasons for its elite reputation (But is this maybe just a bit of good fortune??)

H

(* oh look - he CAN spell!)

[martinvicanek]

Thanks for the nice words and detailed comments.
Hugh, your point on adding partials intensities is spot on: they should be added incoherently, that is energies not amplitudes add. Energies are basically amplitudes squared.
Fixed in the download below.

[HughBanton]

Gosh, yes, that does seem to fix it. Major refinement.

But embarrassingly I just can't figure out what you've done here. I've even put bits of the new/old code side-by-side but confess I still can't see it and the hours are ticking by. Looking in the wrong place no doubt. I'm still way over my head in "blue & pack" but nevertheless I'm very keen to learn what's going on here .. cos I have big plans!

Please put us out of our miseries and reveal. (That's always assuming I'm not the only stupid person here )

Hey Martin .. when are you going to solve nuclear fusion, time travel, perpetual motion etc .. you must be getting pretty close

Thanks again

H

[tulamide]

(That's always assuming I'm not the only stupid person here )

Don't worry! If Martin is the sun, you might see yourself as an ant, but be assured I am a microbe...

[martinvicanek]

Sorry about that. There are two modifications there:
1. Squaring the drawbar outputs before summation. This is implemented in the - uhm - Drawbars module.
2. Taking the sqrt after summation. I have implemented that bit in a somewhat hacky way inside the Tonewheel modules in the Click Suppression modules. Admitedly not the first place you would look for it, but that way I could take advantage of hop(32) which was already there.

[Spogg]

(That's always assuming I'm not the only stupid person here )

Don't worry! If Martin is the sun, you might see yourself as an ant, but be assured I am a microbe...

In that case I'm a molecule.

[HughBanton]

Sounds like we're all shrinking rapidly .. no time to lose, team!!

Let me briefly describe what I'm hoping to achieve with this -- once I start actually contributing, that is ..

I'd personally like to emulate what Compton (and later Makin) did with this arrangement, originally using spinning electrostatic disks, and then during the 1980s with 'solid state'. Compton actually used to build huge 4-manual and pedal 'pipeless' organs with a only a single set of 12 spinning disks! Not ideal, but ...

Firstly you need way more harmonics than a Hammond, of course, but around 60 suffices for the vast majority of organ stops at various pitches. As regularly noted here, a single set of generators based on equal temperament is most inaccurate for many harmonics, so the Makin arrangement was to have a 'normal' generator set to provide harmonics 1, 2, 3, and all multiples (4, 6, 8, 9, 12 etc); a second set of gens tuned to the 5th harmonic (slightly flat), which is used for 5, 10, 15, 20 etc, and a third set tuned to the 7th harmonic, flatter still, which is used for all the remaining really awkward ones - 7, 11, 13, 14, 17 etc. With MV's current creation being pretty CPU light I'm sure that three like this could easily work in parallel. (And I reckon I can manage this much!)

No drawbars as such .. organ stops become virtual-preset-drawbar-levels (using the 60-odd virtual drawbars), and when you draw more than one stop it simply adds the settings together. It works. It's not the greatest, or most realistic, way to generate 'pipe organ' sound, but does have the major advantage of theoretically enabling a limitless stop-list, manuals and pedals, with negligible extra CPU hit. (Unlike HB3, my other organ emulation ... ) Ah .. def needs a rotary-speaker, I should add; Compton devised a device they called a 'Rotofon'.

Attack & Release durations need to be increased progressively for low notes, but I've already discovered this is easily achievable extending MV's 'click supressor' numbers.

Other details to consider as well but I'll leave it there for now. (Before we shrink any more )

I'll see if I can come up with such a triple generator expansion soon ... somewhere in my cupboard I have lists of all the numbers, harmonics and stuff!

H