Page 3 of 4

Re: Nava Last Update: Accent improvement

Posted: Nov 7th, '17, 23:35
by jvq
This is very interesting. :) I've been trying to write my own Nava firmware for fun and the accents have proven challenging.

Here is how I understand what these new boards are changing.

First, some background on a difference between the 909 and the Nava. Each instrument has an accent input which is really the velocity (loudness) of the sound. In the 909, each accent input is connected to a discrete 6-bit DAC. So this means there are 10 DAC circuits just to maintain constant accent voltage levels on all accent inputs in the 909.

The Nava uses a different approach. It has just one DAC (an MCP4822 which has 12-bit resolution) combined with multiplexers and sample&hold circuits. Two multiplexer IC's (IC111 and IC114) are combined into one big 10-way switch between the DAC output and the sample&hold inputs. Each sample&hold circuit consists of an op-amp and a capacitor. So instead of 10 DAC's like in the 909, the Nava has 1 DAC, a 10-way switch (the multiplexers) and 10 sample&hold circuits.

Now in the "previous" Nava 1.0 circuit (before these add-on boards were added), 7 of the required sample&hold circuits were created by adding a capacitor to an existing op-amp. All instruments except the hi-hats and cymbals have op-amp buffers on their accent inputs; the Nava previous design re-used these buffers to implement sample&hold circuits. This is what capacitors like C89 in the bassdrum circuit are for; C89 is combined with the existing input buffer to create a sample&hold.

The sample-based instruments on the other hand do not have op-amps in their accent inputs. So to make a sample&hold circuit for these, the previous Nava circuit has 1 TL074 quad op-amp (IC110, not present in the 909) that provides the three op-amps needed for the three sample&holds of the sample instruments.

My guess is that what changes with these new sample&hold boards is that all 10 sample&holds get dedicated op-amps. This will make the DAC+sample/hold circuit symmetrical in the sense that each accent input uses the same sample&hold circuit, instead of the creative re-purposing of the input buffers of the analog instruments.

Re: Nava Last Update: Accent improvement

Posted: Nov 8th, '17, 00:04
by jvq
It also looks like the new sample&hold circuits use much smaller capacitors (from 470nF to 470pF). I've noticed that the settling time of the previous circuit is quite long; I ended up adding sleep statements in the firmware of about 100us.

More conjecture: It looks like the TL074 (used in the new sample&hold) has a much higher input resistance than the M5218 (the standard 909 op-amps used in most cases in the old sample&hold): 1000000 MOhm for the TL074 vs 5 MOhm for the M5218. Maybe the low input resistance of the sample&hold opamps in the previous design forced the use of the large 470nF s&h capacitors, which made the required sampling time of the previous s&h longer than you'd want.

So apart from making the s&h circuits symmetrical across all accent inputs, the new s&h boards should also be faster because they smaller s&h capacitors.

Re: Nava Last Update: Accent improvement

Posted: Nov 15th, '17, 04:22
by pistrix
superb explanation, thanks jvq!

Re: Nava Last Update: Accent improvement

Posted: Nov 20th, '17, 21:33
by heapish
Added name to sheet! Cheers!

Re: Nava Last Update: Accent improvement

Posted: Nov 30th, '17, 12:36
by peskado
Added name to list

Re: Nava Last Update: Accent improvement

Posted: Dec 5th, '17, 12:07
by istel
Added my name to the list also.

Re: Nava Last Update: Accent improvement

Posted: Dec 6th, '17, 19:02
by jvq
I received my PCB's today. Thanks Vincent!

I have a question about the BOM. It lists this part: https://nl.mouser.com/Search/ProductDet ... 602-404HLF Mouser 649-68602-404HLF

This is a 2x2 block of pins. But the header this is supposed to be soldered on is 2x3. So I think this part on the BOM is wrong.

This looks like the right part https://nl.mouser.com/ProductDetail/FCI ... S9UA%3d%3d 649-68602-406HLF

Something else: if you put the original multiplexer in an IC socket like I did you might want to use these headers to connect the new boards to the existing sockets: https://www.digikey.nl/products/en?keyw ... TS-108-T-A

They have thinner legs than the pins on the BOM.

Re: Nava Last Update: Accent improvement

Posted: Dec 7th, '17, 11:37
by e-licktronic
jvq wrote: I have a question about the BOM. It lists this part: https://nl.mouser.com/Search/ProductDet ... 602-404HLF Mouser 649-68602-404HLF

This is a 2x2 block of pins. But the header this is supposed to be soldered on is 2x3. So I think this part on the BOM is wrong.
Nava BOM updated with the right 2x3 pin header.
Thank you jvq

Re: Nava Last Update: Accent improvement

Posted: Dec 21st, '17, 12:51
by 4teenth
Just added my name...

Re: Nava Last Update: Accent improvement

Posted: Dec 25th, '17, 00:52
by drbach
just added my name to the list