Author Topic: Femto-verb - Slight Return (Peter Snowberg)  (Read 4848 times)

Mark Hammer

Femto-verb - Slight Return (Peter Snowberg)
« on: August 13, 2005, 10:00:32 AM »
I was fortunate enough to score some "Alesis" chips from Wavefront, that Peter Snowberg had drawn to our attention some time back.  These are SOIP parts that are incompatible with my usual perfing method and etching method by virtue of spacing, so they've been just sitting there.  The decision by a local industrial distributor 5 minutes from my house to start carrying "surf-boards" (SIP and other formats that adapt surface-mount chips to perfboardable form) has rekindled my interest in these chips, though.

Looking over the datasheets for the A/D and D/A convertors, I noticed that they are stereo chips.  The posted schem for the Femto-verb is a mono device, so I was wondering the following:

1) Is there, or would there, be any virtue in building it as a stereo device, given that the two channels would pass through a common DSP chip for effects processing?  Conversely, would there be any problem encountered by doing so? (i.e., stereo processing would require a pair of AL3201 chips)

2) The posted circuit has no means for recirculating delay signal.  For some of the 16 programs on the AL3201, this is moot or even nonproductive, but for some of the reverb and delay programs, it could be useful to apply some regeneration.  *IF* the second set of analog inputs and outputs can be used without harm, is there any virtue in using THEM for the regen path instead of an all analog path back to the input stage of the circuit (which would now become a mixer)?

amz-fx

Femto-verb - Slight Return (Peter Snowberg)
« Reply #1 on: August 13, 2005, 11:41:02 AM »
You can parallel the two audio inputs to the ADC chip and use the reverb chip in stereo, for example.  I would drive the ADC with a low impedance source.

 I don't see much virtue in using the second channel for regen though it might expand the size of the reverb area in some way.  A merely analog regen signal would likely cause some peaks in the output,  and that might or might not prove to be useful, depending on what you were processing.

regards, Jack

Mark Hammer

Femto-verb - Slight Return (Peter Snowberg)
« Reply #2 on: August 13, 2005, 11:57:06 AM »
Quote from: amz-fx
You can parallel the two audio inputs to the ADC chip and use the reverb chip in stereo, for example.  I would drive the ADC with a low impedance source.

 I don't see much virtue in using the second channel for regen though it might expand the size of the reverb area in some way.  A merely analog regen signal would likely cause some peaks in the output,  and that might or might not prove to be useful, depending on what you were processing.

regards, Jack


Thanks for the info.  Would some lowpass filtering increase the likelihood of a more usable regen signal?

Peter Snowberg

Femto-verb - Slight Return (Peter Snowberg)
« Reply #3 on: August 13, 2005, 01:29:11 PM »
Mark and Jack, I'm glad you don't have digital=evil dementia.

I don't see a use for cheap quality digital outside of emulating the sounds of the 80s, but high quality digital can be significantly better than analog at most things.... except for creating 'analogish' distortion. ;)

That ADC & DAC pair have some best in class measurements to brag about. They really are amazingly good chips.... especially considering the price! 8)

The Digital Reverb Engine is a VERY cool chip, but as it stands before you eating ROM, it wants to be placed in a very specific circuit topology. The built in algorithms are based around being placed in a device where a stereo image of the same input signal is buffered, processed, and the device output is taken from a mix pot that dials in the input buffer at one end and the DSP output at the other. The reverbs are meant to be used on stereo signals with an external analog wet/dry mix control and if you don't give them that, the programs become just about useless to me.

You might find something interesting with recirculation if you want a noisemaker. Who knows.... maybe one of the built-in algorithms would yield something amazing, but digital usually doesn't respond to misapplication in the same way analog does. I haven't spent lots of time on abusing the built-ins, but nothing in my experiments was interesting to me.

I'm also not thrilled with the prospect of effects you cannot edit. What use is a flanger, chorus, or echo if you can't adjust it? That's why the LOTUS pedal has consumed so much of my time. You can't change a coefficient like you can change a resistor unless you have access, a map, and a damn good flashlight. Better yet would be a courier that will take that new value there for you and then a translator that would save you from having to know DSP math-speak ;). The built-in programs are wonderful for demoing the capability of the chip but the delays don't even use all the delay memory. I would rate their utility as extremely limited for guitar players. Karaoke machines are a different matter.

I would encourage you to save the chips a little longer.

I've managed to fit 3 DREs, a 1KS DSP, stereo A/D & D/A, 512K of Flash, E2PROM, a 16 bit RISC processor, and 5 channels of 10 bit analog control input into a 1590B.  :twisted:

If you don't want to go that nuts there will be a version in the near future with just a single DRE, but with programmability and adjust ability too. That's going to be a couple of months after LOTUS arrives.

I also have a delay/echo designed using the ADC & DAC, but with a little different way of doing delay memory..... just in case you have a converter pair left over after your LOTUS build. ;) That one will have analog regeneration just like a BBD. Too many projects and too little time.

The only riddle left to solve before completing the LOTUS layout and ordering the first couple of boards is what to do about the input volume? It's a waste to put two volume pots on such a small box. I think the prototypes will have board mounted trim pots accessed from the bottom.


To address your questions a little more directly:

1)  The AL3201 is already a stereo device. The two analog inputs to the ADC are converted and the results are multiplexed on a single digital line. Inside the DSP you have option to keep these as two separate sources and destinations, or you could treat them as a stereo pair. The built-in algorithms assume a stereo pair.

2)  Regeneration should be done inside the DSP via programming. Going external to mix would not be beneficial to noise levels. You could always use the second channel as an effects loop in the regeneration path. ;) (provided you junk the built-in 16)

Hang in there Mark!

As Jack mantioned, it's best to drive the ADC from a low impedance source. I'm using 5532s to do that after a JFET input stage.
Eschew paradigm obfuscation

scratch

Femto-verb - Slight Return (Peter Snowberg)
« Reply #4 on: August 15, 2005, 12:12:22 PM »
Mark, where did you get those 'surf boards'? I'd be interested in checking them out, so many components going surface mount! I'd rather practise on 'cheap' logic gates and op-amps before moving on to things like the AL3201, etc.

Thanks ...
Denis,
Nothing witty yet ...

scratch

Femto-verb - Slight Return (Peter Snowberg)
« Reply #5 on: August 15, 2005, 12:32:18 PM »
Mark, never mind ... saw your post re: 3080 based compressors ...
Denis,
Nothing witty yet ...

Mark Hammer

Femto-verb - Slight Return (Peter Snowberg)
« Reply #6 on: August 15, 2005, 12:36:19 PM »
I was in there on Saturday, and learned that they weren't really "carrying" them.  Turns out they had ordered some in for a client, and these were leftovers.  Quite the supply of leftovers, though, and if the demand seems apparent, they may start carrying them.

Make sure you leave me a 16-pin unit - I only bought 2, and I need 3 for the Femtoverb. :wink: