New AOUT module design?

[seppoman]

Hmm, it has 2 mV/ms slew rate, i.e. if you refresh all 8 channels of each chip in under 1 ms, you have "only" 2 mV ripple which is around 2 cents detune with V/Oct. So for use in a modular synth, the update rate would have to be even faster to prevent a noticeable vibrato on the oscs. Whether this is a preformance issue in the relevant applications, is a question for TK.

Another issue is the price and availability - Farnell, Mouser, Conrad, Reichelt, Schuricht, Segor don´t sell them, RS wants 20 Eur (+ VAT), Digikey wants around 10 Eur incl. VAT per piece. So for the price of a board with 8 SMP08s you´d get 3 additional AD5392s and would have 32 regular AOUT channels without getting the precision and performance issues of S&H.

I don´t want to discourage you about this (sorry ::) ). But how many people will really need more than 32 AOUT channels? And these people will probably have some expectations on performance anyway, so is it really important for them to save a few bucks by sacrificing output resolution?

Seppoman

[stryd_one]

the update rate would have to be even faster to prevent a noticeable vibrato on the oscs. Whether this is a preformance issue in the relevant applications, is a question for TK.

I'll have a go and tell you that it's not much good for my uses... Altthough *some* apps could maybe handle it, I think that is a killer :(

Another issue is the price and availability....So for the price of a board with 8 SMP08s you´d get 3 additional AD5392s and would have 32 regular AOUT channels without getting the precision and performance issues of S&H.

Totally :)

I don´t want to discourage you about this (sorry ::) ). But how many people will really need more than 32 AOUT channels?

Me! hhhehehehe

And these people will probably have some expectations on performance anyway, so is it really important for them to save a few bucks by sacrificing output resolution?

You know me well  ;D

[drsyncenstein]

Thanks for the replies.

I think Seppoman meant the droop rate, not the slew rate.

I had not noticed that, but the performance of the nju is indeed 10 times better.

Never trust a data sheet that says Low Droop Rate!

So you are probably right, it's not a good chip for this.

But as i cannot find the nju part, i need an alternative.

I like the multiple dac idea, but i really need those 64 voltages.

I'm reluctant to go the 4051 route, but maybe that would be the easiest.

I'll try some more part browsing for now.

[seppoman]

Hi guys,

just a little teaser:  ;D

The board size is the MBHP-standard 8x6 cm. All the resistors in the upper half can be left out if bipolar operation is not needed. The rendering of the trimpots is a bit off, but I´m no 3D expert ;)

Seppoman

[moogah]

Excellent work man!

Do you still want some help with design/testing the output stage?  It looks like you've got a good amount of work already done in that area :)

PS.. what is this program you and SmashTV are using to render the boards?

[seppoman]

Thanks :)

The output stage is quite the same as on the old AOUT. I made dozens of OpAmp drawings and had some headache, and in the end I came up with no better solution. But the old AOUT is tested and found working well, so there´s no need to reinvent the wheel :)

The bipolar option is also similar to the old version. Just no trimpots anymore, because the AD5392 can be trimmed via software. With 1% resistors, there will be an offset of around 20 mV that can be corrected via software. This way the option can be included on the board without wasting much space for another 8 trimpots :)

About rendering: There´s an ULP script for Eagle that is called Eagle3D. It creates PovRay files.

Seppoman

[seppoman]

GREAT CATASTROPHY!!! :(

Last time I had checked the availability of the AD5392, it was in stock at Farnell and Digikey. Today I checked again and Farnell only lists the LFCSP package. Digikey marks it "only by request", minimum order for NON ROHS-conform 15 pieces, ROHS conform only trays of 160 pieces. This means, the 15 pieces minimum (which probably would be ok for SmashTV or a group buy) will probably disappear from the listing, and 160 pieces is way to many!

So how should we go on with this project? LFCSP would be DIYable with the "table grill" technique, but I´m sure the ordinary DIYer can´t be bothered with a part that has no normal legs...

Options:

A. keep the module as it is, try to persuade SmashTV to stock up a few and hope that the availability gets better again. AD keeps listing the part as in production.

B. Use the 16 channel version AD5390. Availability is ok (Farnell 47,30€ (more than 10 pieces: 35.70€) Digikey 28.10€). The price per channel is better (actually the price is still lower than for 8 channels of MAX525...), but we´ve also got to keep in mind the people who don´t need more than 8 channels who´d have to spend more money this way. And there´s no chance of putting everything on 10x8 cm including bipolar option for all 16 channels. Probably possible: 16 channels but only 8 with bipolar option.

C. 5390 is pin compatible with the 5392, so I could add only a connector for the additional 8 channels and supply to the current design. Then it would be possible to use either the 5392 (if available) or otherwise use the 5390 and add an external OpAmp module

D. split the AOUT board in two parts: When thinking this idea further, why not do only one small DAC module suitable for both 5390 and 5392, design an 8 channel OpAmp/bipolar module and connect one or two of them to the DAC board. This would probably make the "I want 64 channels" crowd happy and would simplify future changes to other DACs.

E. say goodbye to AD and put one of the other previously discussed DACs on the AOUT. LTC2620 (RS 15€, Digikey 12.70€. INL 4LSB) or TLV5630 (Farnell 13.30€, Digikey 11.40€. INL 6LSB). Accept the lesser quality for easy to source and cheaper parts. (I suppose this option doesn´t sound too great to you as there is/was the opportunity to have a really HQ AOUT?)

I´d like to hear comments/suggestions from as many as possible people on these options to have a good base to decide what´s best for the community!

Thanks,

Seppoman

[NorthernLightX]

Hi Seppo,

I think D is the best option. Modularity is always a great feature, and all the work you did on this project is not lost in this way (assuming it is less work to split the current design than creating a new one). If it's also possible to use MAX525 with your opamp stages that would be even more cool - I've stocked up on 525's in the past and people who've built the old AOUT can upgrade.

Cheers,  Alex.

[seppoman]

I think D is the best option. Modularity is always a great feature, and all the work you did on this project is not lost in this way (assuming it is less work to split the current design than creating a new one).

I also like the modular idea, although I´m not quite sure if (and how much) signal degradation is caused by this. I suppose it would be a good idea to keep these connections as short as possible.

The amount of work doesn´t differ so grossly that it would be a reason to choose one option - except, obviously, option A is least work ;)

If it's also possible to use MAX525 with your opamp stages that would be even more cool - I've stocked up on 525's in the past and people who've built the old AOUT can upgrade.

For people already having a working AOUT, there´d be no need to upgrade. The cirquit is quite the same. Only for new builders, one could think of doing a MAX525 board that in conjunction with the OP board replaces the old version so they have the advantage of a board with bipolar option already included.

Thanks for your comment!

In general: If somebody comes up with an option F or G, just go ahead - this is not only a multiple choice test :)

Seppoman

[stryd_one]

I like option D too... It also has the benefit of being somewhat futureproofed....

I also like the modular idea, although I´m not quite sure if (and how much) signal degradation is caused by this. I suppose it would be a good idea to keep these connections as short as possible.

I would go for a daughterboard kind of arrangement, where one board has male DIL headers and the other has female DIL socket.

[seppoman]

I would go for a daughterboard kind of arrangement, where one board has male DIL headers and the other has female DIL socket.

Nice idea :)

So I see nobody jumps in for option E yet? Of course the modular approach could also be done with other DACs - but it looks like HQ specs are more important than price and easy availability?

Please guys, I need statements from more people :) Whoever reads this and plans to build an AOUT sometime, please tell me your point of view! I need some information on what you need - is excellent precision worth the trouble of using a hard to source DAC? Would you participate in some kind of group buy so that SmashTV would have enough initial buyers to make ordering a pack of these DACs a calculateable risk? Are there a lot of people who want 16 CVs or more with enough precision to drive VCOs or do most people only want a few channels to do a bit of external filtering for MBSID anyway? I need to get a better feeling of what the most important goals for this project are!

Thanks,

Seppoman

[moogah]

C and D seem the best options we have here.  Option E is not so attractive to me because of all the effort that has already gone into selecting a DAC and it seems a shame to take a step backward.  Still tho it does seem like the powers that be don't want this project to succeed! >:(

[jrp]

Since you were asking for opinions, i can add mine.

The only case where a high quality aout channel makes sense to me is to drive vcos.

Now, of all the applications with many cv outs i can imagine there are two that require lots of "good" aouts, the others only some.

And there are so many more applications, the more modular this project is, the better.

1. A polysynth with 8 independant voices. More would be overkill...

The MBCV is well prepared to run them all in polymode, or independant.

The other cv outs can be of lower quality, to controll vcas, etc.

2. A Big midi-cv for a modular synth.

Never know where they might be patched to, but in my oppinion 4 high quality ones for vcos will be enough. But for those who want to controll many vcos, well...

3. A solo synth with complete midi controll.

This interests me a lot.

Imagine this: 3 vcos. If only in mono mode one precission cv would be enough.

Then have many many cheap ones for stuff like resonance, filterfreq, distortion, volume, envelope times, lfo speed and depth, fm amount, pan, fx amount...

Imagine a beast like that controlled by a MBSeq!

If i were to design a synth like that right now, i would split the controll to 2 MBCVs.

One with the max chip, one with the AOut_LC.

This is not the most simple connection, but i would hate to waste one of the high quality cv outs for stuff like driving a simple vca.

Why not make a page with alternative dacs.

It could list their benefits and drawbacks, how to connect them, etc. Not more.

THere are so many 2ch dacs out there, for many users a single board with two good cvs might be enough. Then chain some other board with a cheap quad dac...

To start with, this is a bit against the way these projects are presented, as it will miss the excellent ducumentation that is part of the philosophy of this page.

But it should change... As soon as i design a board i would be glad to offer layout and part lists. And i´m sure many others will be too.

One advantage of this should be, that if a user has some dac, or is about to sample one, or getting some good price, he could look up first if this dac is compatible.

Then it´s up to him to do a physical design with the chip and present it to this comunity.

[Sebo]

Hi:

I'm waiting for build a MIDIbox CV. I didn't build yet because I can't

find the MAX525.

I think option D is the better for me, so I could build a little 4 channel

MIDIbox CV and start to add channels later.

Thanks for all your work Seppoman.

[seppoman]

Hi,

(Sebo: option D was not meant to support less than 8 channels, this would give too many tiny boards and connections for people needing lots of channels.)

UPDATE:

I´ve discussed the matter extensively with Thorsten, and decision was made that I will use the TLV5630. Actually, the original reason for a new design was not to get better specs than the MAX525, but to have an alternative board with an easy-to-get and cheap DAC. The big majority of future CV users in context with Midiboxes will be people wanting to add some filters/VCAs/external effects etc. to synthesizers like MBSID or MBFM. For this purpose, the TLV is perfect. A lot of shops sell it (Digikey, Mouser, Jameco, Newark, Farnell, to name a few), it´s cheap and the specs are absolutely sufficient for this purpose, and SOIC is easier to solder than LQFP. So this decision puts the needs of many over the needs of a few. I´m really sorry for the modular guy(s?) - actually, the response to my "poll" looks like stryd_one is nearly alone with near-future modular plans, so sorry, Todd  :-\

It´s really sad having to give up the 5392, but putting emotions and perfectionism aside I understand that using a nearly unavailable DAC for an official MBHP module just doesn´t make sense - the problems with the MAX525 are not to be forgotten...

But as I´ve already got 2 AD5392s lying around here and the PCB layout is kind of completed, I´m thinking of ordering two PCBs anyway, just to play around a bit and to have a souvenir from all the wasted time ;) So if someone feels able and willing to write a driver for it and use it in a custom project, I´m offering to send one PCB with presoldered DAC for PCB+shipping costs (probably around 20 EUR - contact me via PM) - Todd? :)

Anyway, the TLV board is in progress and I´ll keep you all posted :)

Seppoman

[stryd_one]

Hahahahaha it's all good mate, I think you've made the right decision. I think the rarity of the AOUT ICs was a real setback for quite a few people, so a widely available IC is of paramount importance.

Plus, I can't get samples of AD stuff here. So those chips suck.  ;D

I'll take a shot at a driver for sure!

[seppoman]

So those chips suck.  ;D

Actually they don´t - this is what makes it so hard to drop them ;)

Seppoman

[Sebo]

Cheap and easy to find?

I'm in.  ;)

I going to end building an 8 channel MIDIbox CV anyway.

Thanks Seppoman again.

[seppoman]

Hi,

it´s preview time again :)

As the TLV has no way of adjusting the offset via software, 8 trimpots were added. Also a 74HC595 to provide 8 gate outputs :)

New board size is 8x8 cm.

Seppoman

[TK.]

What a beauty! :)

What happened with the green power-on LED?

Best Regards, Thorsten.

[sneakthief]

why do you need 8 gates on the board when the core can provide 8 gate outs?

unless of course you want to use J5 for something else.

[seppoman]

why do you need 8 gates on the board when the core can provide 8 gate outs?

unless of course you want to use J5 for something else.

yes, good example. Thorsten ordered them so I´ve put them there. Any more questions? ;)

@Thorsten: just press Reload ;)

Seppoman

[moogah]

the response to my "poll" looks like stryd_one is nearly alone with near-future modular plans, so sorry, Todd

Actually, I'm in this boat too, but I realize that my needs are best taken care of myself without dragging down the larger project (I want to fully modernize my Oberheim OB1 which will take about 24 CV's, at least 2 of which need to be VCO precision).

The new chip looks good to me :)  Certainly enough to start a modular control box.

Since I've got a couple samples of the AD chip I would be interested in getting my hands of some of those old PCB's too, same with writing a driver... however as the 808 project continues to go into extra innings I'm not sure when I will have time to really dig in  :-\

[TK.]

why do you need 8 gates on the board when the core can provide 8 gate outs?

unless of course you want to use J5 for something else.

It's a more generic solution. The usage of J5 was a workaround for the MAX525 based module, which only provides two gates.

And there is another reason: J5 are allocated by MBSID V2 for analog inputs

Best Regards, Thorsten.

[drsyncenstein]

Very Nice.

Any old prophet or oberheim would wish it had such a good dac.

Would be nice if they could be daisy chained though.