Haible Vokoder?

janvanvolt · Joined: Nov 24, 2005 Posts: 285 Location: Mainz, Germany

Can't wait to get the PCB Smile

I got most of the chips now and also finished measuring out the 1% ones out the 10% bag (for the 47n's).

Regarding the SMD: It's not too hard, with a small solder tip it's easy to do.

/mr · Posted: Thu Jul 15, 2010 4:16 am Post subject:

whomper · Posted: Thu Jul 15, 2010 5:15 am Post subject:

Where can I find %1 caps? I will probably not go into selecting caps from a big bag.

Is there perhaps a part number for both of them that I can use?

jhaible · Posted: Thu Jul 15, 2010 5:30 am Post subject:

guitarsandsynths · Joined: Jul 13, 2010 Posts: 7 Location: Los Angeles

JoeMorris · Joined: Apr 26, 2009 Posts: 161 Location: Brighton

I agree with the above method of wetting one pad with solder before you start, but I use a small flat-tip screwdriver to maneouvre and hold the upper surface of the smt cap into place and then use the iron in my other hand to re-melt the solder on the pad. I remove the iron whilst i keep pressing down with the screwdriver until the solder sets again. then obviously the other pad is easy to solder as the cap is already firmly in place. once you get in the rhythm its actually pretty easy.

diablojoy · Posted: Thu Jul 15, 2010 5:12 pm Post subject:

yes jurgens boards seem better than some for SMT they wet very well
if the board is new and has not been stored for too long and kept clean
no finger prints! . I have not had to use a flux paste or pre wet any of the pads so far on any of jurgens boards
this is my method, it is quite simple
I found by holding all the parts in place with a tiny bit of blue tack on one end and soldering the other side first it is goes very smoothly note: the blue tack must only be on top of the component and pushed down onto the board so as to hold the component in place, aligning the component correctly is not difficult then as it allows small adjustments to be made
you must also allow the blue tack to cool to the touch before removing it before soldering the other side, this happens fairly quickly, but don't rush this part if you do don't panic what residue remains can be removed by dabbing it with more blue tack it just adds more time to the process
the use of blue tack does not seem to affect the solderability in any way
once visually removed and i have had no faults caused by this method so far
this way for 143 SMT caps you may only need to pick up the iron twice
as i said previously this method is very quick, much quicker than thru hole.
cheers
denis

diablojoy · Posted: Thu Jul 15, 2010 5:23 pm Post subject:

zthee · Joined: Feb 20, 2008 Posts: 414 Location: Stockholm

A quick question...

http://www.jhaible.de/vocoder/schematics/channel_5.pdf

Inputs and outputs are confusing me - It probably means that the bottom pin goes INTO the patchbay (The pin is an output, but labled as an input?)?, and the output is what comes OUT of the patchbay (The pin is an input, but labled as an output?)? Very Happy

Just needed it clarified! Thanks!

Edit: What's throwing me off is the

jhaible · Posted: Fri Jul 16, 2010 12:59 am Post subject:

Sorry for the confusion.

From the PCB viewpoint, the lowest pin is the analyzing filter output, and the middle pin is the envelope detector input.

But if you look at the matrix as a module, that can (optionally) be inserted at this break point, the matrix' input signal will come from the lowest pin of the PCB, and the matrix' output will go to the middle pin.

And if you look at channel potentiometer(s) as another (optional) module, the matrix output will go to the potentiometer's "input" (cw end), and the potentiometer's "output" (wiper) then goes to the env detector input (middle pin on PCB).

JH.

zthee · Joined: Feb 20, 2008 Posts: 414 Location: Stockholm

Thanks!

That's the way I thought it was - now it's all clear! Very Happy

OH - And while I have the possibility to beat everyone to it; 10K log pots will be OK instead of the 5K?

jhaible · Posted: Fri Jul 16, 2010 1:38 am Post subject:

robotmakers · Joined: Mar 12, 2008 Posts: 7 Location: NY, USA

Please excuse my returning to my earlier point, and I promise I'll drop this if it's annoying to the forum. I'm posed with a dilemma that may be relevant to these discussions. I don't have the time or desire to build the board myself (which is of course my problem). I've had conversations with folks who would build one for me and have found the cost to be prohibitive. For example, the comparison I made is that I could purchase a vintage Moog vocoder for about 2/3rds the cost of having the Haible Vokoder built for me. Yes, there are performance differences, but the Moog unit in the long run will likely have more investment value. Just my personal calculus.

Which then begs the question for me, could this board (with the same circuitry) be designed such that it would cost less to construct (after taking into account the value of the builder's time)? Could a layout employing far more SMT components (rather than through-the-hole) be prepared for automated stuffing? Would this be more economical for the builder? Could this also be more profitable for Mr. Haible? I think so, as it would dramatically increase the number of units he could sell and with more profit margin. Maybe this could be subcontracted to someone who has experience with automated SMT board runs?

Thanks for your time,
Roger

jhaible · Posted: Mon Jul 19, 2010 11:40 am Post subject:

robotmakers · Joined: Mar 12, 2008 Posts: 7 Location: NY, USA

Thanks for your thoughtful reply. I'm glad that you're open to the possibility of a manufactured SMT version. Perhaps someone will see this post and it will start a ball rolling somewhere. Until then, unfortunately I don't have a deal in hand for you. But, who knows?

Best wishes,
Roger

StephenGiles · Joined: Apr 17, 2006 Posts: 507 Location: England

Many constructors like myself with what are best described as "tired eyes" curse the very existence of SMD - I've been used to "normal" sized components for over 30 years and I won't change now. Hopefully, thanks to TI, my stocks will last me!!!

For me, normal sized components look much prettier Laughing

magman · Joined: Feb 04, 2009 Posts: 363 Location: Liverpool, UK

TekniK · Joined: Aug 10, 2008 Posts: 1059

robotmakers · Joined: Mar 12, 2008 Posts: 7 Location: NY, USA

Just thought I'd mention that I brought the recent discussion in this thread to the attention of two of the major US modular synth makers. Unfortunately, neither are currently interested in pursuing partial or complete manufacture of the vokoder board.

Anyone in Europe interested in contacting someone like http://www.diatron.co.uk/ and running the BOM and board info by them? At the very least, it would be interesting to see how much money one saves by doing all the soldering work by hand.

Cheers,
Roger

Funky40 · Posted: Thu Jul 22, 2010 8:54 am Post subject:

Jürgen,
when do you think are you ready to sell the boards ?
must plan my finances.....
( i'm not in hurry, springtime would be cool )

jhaible · Posted: Thu Jul 22, 2010 9:02 am Post subject:

Funky40 · Posted: Sat Jul 24, 2010 4:22 am Post subject:

thanks, it's as hoped
( i mixed up sring with fall in my post. Fall is good ...... winter is perfect )

/mr · Posted: Sat Jul 24, 2010 9:13 am Post subject:

I suddenly checked out the channel schematics a bit more than before, and realize that the (optional) crosspatching between analyzer channels and syntesizer channels is done before the envelope detectors - i.e., with the bandpass-filtered audio signals, not with the envelope-followed analyzed CV versions of them. Confused

Am I understanding this correctly?

(There are envelope outputs as well, after the envelope detectors, but the CV inputs of the bands are before them, so they can't be used for crosspatching.)

To me it feels like a strange design choice... do other vocoders employ it?

I was expecting CV patching and see no obvious benefits with audio patching, rather the opposite. I was planning to pick out the handy low-frequent CVs and feed in externally treated versions of them (shifted bands etc, without a need for a physical matrix), but when the signals are in the audio range it will not be done so easily.

jhaible · Posted: Sat Jul 24, 2010 11:26 am Post subject:

It's another EMS trick. Surprising at first, it avoids a lot of trouble with offset voltages. The way it is, you only have one CV offset adjustment per channel. (Still, these are 20 Pots that must be carefully adjusted, or you're ruining the Vocoders dynamic range completely.)
If you'd do the crosspatching adter th eenvelope detectors (on DC rather than AC), you'd need two offset trimmers per channel, one for the DC path before the patchpoint, and a second for the DC path after the patch point.

JH.

/mr · Posted: Sat Jul 24, 2010 2:59 pm Post subject:

Oh my, those British chaps again. Wink

Yes, I expect the "making of proper CV level" after the analyzer filter would need a trimpot to adjust for amplitude errors in the analyzer filter, and then the VCA in the syntesizer filter would need another trimpot to adjust for amplitude errors there. With your design these two amplitude error corrections combine into one.

Or... do they really? In a hardwired vocoder, yes - but if the channels are patched differently, don't you need both trimmers anyway? Confused

Anyway - with AC patching instead of DC patching we can forget all the free CV connectivity with the outer world a la Doepfer Vocoder. Almost no use putting telejacks at the 20+20 patchpoints (unless you want something quite weird and know how to use them), and perhaps it's not much to cry for. The more introvert and internal solution of a pin matrix would still make sense though... if the channels allow it well enough.

Do they?

The single trimmer solution mentioned above is the first thing that makes me wonder, and the other thing is the time constants of the envelope followers (EF). With AC crosspatching (before the EF) between different channels, the signal gets EF'ed by "the wrong channel", and higher channels have shorter EF time constants in order to allow for the more rapidly changing envelopes of those channel analyzers. For example, if I read the schematics correctly, with channel 2 patched to channel 19, we have a bandpass signal around 250 Hz from channel 2 analyzer, which goes to the 19th channel EF with its time constant R*C = 0.0002 s which translates to a lowpass cutoff frequency of roughly 800 Hz. This frequency works well for the 6.3 kHz frequencies of channel 19, but is way too high for the 250 Hz channel 2 signal, letting it all through instead of following its envelope! Shocked

I hope I haven't confused it all into something completely gibberish...
Having shorter time constants in higher-channel EF's is at least something expected, since the higher-channel envelopes can vary more rapidly, and it should be used to avoid a slow vocoder response for higher channels. But for proper performance I expect each channel EF to always take care of its own analyzer signal, otherwise we end up with the problem just described. To do this, the patching should be after the EF's, with the DC signals and not the AC ones... I'm afraid. Rolling Eyes

(Hoping I'm all wrong.) Smile