v2164 linear VCA problem

From circuitbending to homebrew stompboxes & synths, keep the DIY spirit alive!

Moderators: Kent, luketeaford, Joe.

Post Reply
User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Wed Dec 13, 2017 4:20 pm

neil.johnson wrote:
J3RK wrote:I just linearized my test circuit, and it looks very good so far.
Do you know if your model would run in the free demo version?

Neil
I would assume so. While I used the transistors suggested in Hoshuyama's schematic, everything else is generic or a modeling primitive. They're just built in a child-sheet. I'd be happy to post the DSN file.

I still need to set up the mode pin, but I'll post the current version in a few minutes. It's a DSN file with the IC components and child sheet. I haven't made a proper library part yet.

I can also post the full dual linear vca circuit, but it's basically just like the doctor's earlier in the thread right now. I wasn't planning to build a standalone VCA, but rather set up some indexing VCAs for an oscillator I'm working on.

I kind of doubt I can export this to any other useful format directly, but I'll check anyway. Otherwise, maybe I could build it back up in LT or something.
Attachments
J2164.zip
(24.05 KiB) Downloaded 46 times

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Wed Dec 13, 2017 4:51 pm

Thanks. It certainly loads into the demo version, obviously can't save but I can view it. AND you can generate a SPICE netlist (Tool -> Netlist Compiler) :mrgreen:

Neil
Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Wed Dec 13, 2017 4:52 pm

One thought: I played around with something like this a while back and, maybe out of habit, used diode-connected transistors instead of discrete diodes. I wonder if this is what they did on silicon, and what effect it would have on the model.

Neil
Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Wed Dec 13, 2017 5:14 pm

neil.johnson wrote:One thought: I played around with something like this a while back and, maybe out of habit, used diode-connected transistors instead of discrete diodes. I wonder if this is what they did on silicon, and what effect it would have on the model.

Neil
Certainly worth playing around with a bit. I could give that a try. I was also going to replace the "somewhat specialized" choices for transistors with either the generic NPN / PNP as that might be easier if exported.

Nice on the spice netlist. :hihi: I've never needed to simulate outside this software, so this would be the first time I've tried.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Thu Dec 14, 2017 2:06 am

im not too familiar with simulation software, but if they dont let you adjust emitter areas, you can just stack 12 diodes (or diode connected transistors, which is probably whats in there) in parallel on the MODE current side of things. also, the current on the voltage source at the top of the current mirrors is probably 400uA, but if youre just modeling one amplifier, scaling this back to 100uA (as shown) is the right thing to do.
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Thu Dec 14, 2017 10:29 am

I actually just "brute-forced" it, and placed four full instances of the cell into the chip package / sub-circuit, so each one has the 100uA.

User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Thu Dec 14, 2017 11:10 am

neil.johnson wrote:What I do find makes quite a bit of difference is the mode bias current. Ideally the bias current should track the input signal amplitude. This called sliding bias and used to be used in some high-end VCA-automated mixing desks and effects units (compressors).
You've got me curious about this. Do you think it would be sufficient to just make this follow the CV voltage (which in the linear configuration would be proportional to the signal level) or would it be better to actually follow the signal itself? (like a compressor with an RMS or peak detector)

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Thu Dec 14, 2017 2:35 pm

@J3RK, here's an interesting overview: http://www.gyraf.dk/schematics/VCAs_Ben_Duncan.pdf
Pages 3 and 4 discuss the SSM technology, which at the time (1989) was the SSM2014, with the improved SSM2018 coming out soon.

Neil
Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Thu Dec 14, 2017 2:46 pm

neil.johnson wrote:@J3RK, here's an interesting overview: http://www.gyraf.dk/schematics/VCAs_Ben_Duncan.pdf
Pages 3 and 4 discuss the SSM technology, which at the time (1989) was the SSM2014, with the improved SSM2018 coming out soon.

Neil
Thanks! I'll check it out.

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Thu Dec 14, 2017 2:48 pm

Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Thu Dec 14, 2017 3:09 pm

Nice! So the detector is just rectifying the input level, integrating it for the time constant in the spec, and outputting the bias current it looks like. I'm thinking it might be fun to make a little board (similar to an eval board) to try all of this out on.

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Thu Dec 14, 2017 3:30 pm

in case it helps, here is how distortion changes with mode resistance. the lines on the left are where the noise floor comes up on it, so you can see the difference in noise floor.

Image

sorry this is so huge, i need to fix the font so its readable at a smaller scale.

EDIT: the X-axis is probably better labeled: input level [dBu], 0dB gain, 30k I/O resistors.
openmusiclabs.com

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Thu Dec 14, 2017 4:49 pm

Thanks! Although, it would be better to identify the mode current rather than resistor, especially as you have not stated your supply rails so it is difficult to properly interpret your data.

Neil
Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Thu Dec 14, 2017 4:53 pm

15V. i tried to keep things matched to the datasheet so comparisons could be made.
openmusiclabs.com

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Thu Dec 14, 2017 4:58 pm

Your curves are consistent with my own measurements. I interpret the results as showing that the underlying noise floor is the basically the same (the left-hand side of the curves), all you really do with the mode bias current is shift the point (the dip) at which the crossover distortion becomes the dominant factor (orange is class AB, green is class A). The jump at the far right is due to overloading causing instability.

Neil
Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Thu Dec 14, 2017 7:01 pm

It was just suggested to me to parallel all four cells and use the sliding mode to make a really nice low distortion single VCA. I was planning an LME497xx based output module using the 49660 current amplifier and DC servo, so perhaps this would be a good way to add voltage control to that. Probably overkill for synth applications, but...

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Fri Dec 15, 2017 4:46 am

J3RK wrote:It was just suggested to me to parallel all four cells and use the sliding mode to make a really nice low distortion single VCA.
The noise level would certainly be lower, although crossover distortion level won't change. Why not use a THAT2180 though?
J3RK wrote:I was planning an LME497xx based output module using the 49660 current amplifier and DC servo, so perhaps this would be a good way to add voltage control to that. Probably overkill for synth applications, but...
The LME49600 is certainly an interesting device, with vanishingly low distortion. But when you have a VCA that introduces more than 0.001% THD+N (compared to the LME49600's quoted 0.00003%) I don't believe there's much point going much better than that. Although of course if you convince yourself that it must sound better then of course it will sound better.

Neil
Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Fri Dec 15, 2017 8:42 am

here is the noise floor versus gain and mode resistor. basically, the noise floor tracks with the square root of the current in the output stage.

Image

the rounding off on the lower gains is due to the resistor noise of the 30k resistor. if you were to subtract this out, those lines would be much straighter.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Fri Dec 15, 2017 10:23 am

neil.johnson wrote:The jump at the far right is due to overloading causing instability.
im wondering what you mean here, as i havent noticed any instability at these levels (nothing out of the ordinary up to 100kHz - maybe beyond that?). when i get to the clipping point, i definintely notice instability, but this is below that region (that is the third region on figure 6 in the datasheet, where it really goes steep).

i noticed distortion due to the limited gain of the input stage when i was testing out the 2164 as an exponential converter. i had to drive the input stage with a current source to keep it linear. but these levels dont seem large enough to be causing that effect:

Image

EDIT: the left hand side rise is the noise floor of my spectrum analyzer.
openmusiclabs.com

User avatar
J3RK
Super Deluxe Wiggler
Posts: 6733
Joined: Tue Dec 22, 2009 12:46 pm
Location: Seattle

Post by J3RK » Fri Dec 15, 2017 10:30 am

neil.johnson wrote:
J3RK wrote:It was just suggested to me to parallel all four cells and use the sliding mode to make a really nice low distortion single VCA.
The noise level would certainly be lower, although crossover distortion level won't change. Why not use a THAT2180 though?
J3RK wrote:I was planning an LME497xx based output module using the 49660 current amplifier and DC servo, so perhaps this would be a good way to add voltage control to that. Probably overkill for synth applications, but...
The LME49600 is certainly an interesting device, with vanishingly low distortion. But when you have a VCA that introduces more than 0.001% THD+N (compared to the LME49600's quoted 0.00003%) I don't believe there's much point going much better than that. Although of course if you convince yourself that it must sound better then of course it will sound better.

Neil
Good points! My reasoning for the current amplifier was more for being able to drive pretty much any load. The servo was more for educational purposes as I haven't done that before. I've always used capacitors for my AC output modules. I'm not sure one method is necessarily better. I've never felt like the capacitor distortion was unpleasant for example.

Also a friend of mine has some SSL schematics and has been telling me about some of the tricks they use to get good performance, and while they may not be particularly useful in the modular synth environment, I've wanted to try some of them out for fun. :party:

User avatar
neil.johnson
Super Deluxe Wiggler
Posts: 1692
Joined: Thu Apr 08, 2010 3:51 pm
Location: Cambridge, UK
Contact:

Post by neil.johnson » Fri Dec 15, 2017 10:38 am

guest wrote:
neil.johnson wrote:The jump at the far right is due to overloading causing instability.
im wondering what you mean here, as i havent noticed any instability at these levels (nothing out of the ordinary up to 100kHz - maybe beyond that?). when i get to the clipping point, i definintely notice instability, but this is below that region (that is the third region on figure 6 in the datasheet, where it really goes steep).
Here:
Image

And yes, it is rather high frequency. If you consider the values in the input compensation network (500 Ohms, 560p) you get a 3dB point of 568kHz.

I'll run up a test circuit on the AP tonight and get some plots at high amplitude.
EDIT: the left hand side rise is the noise floor of my spectrum analyzer.
Which model are you using? Also, what are you using as your source?

Neil
Random ramblings :: http://www.njohnson.co.uk
SSI parts and more :: https://www.cesyg.co.uk

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Fri Dec 15, 2017 10:54 am

im using an HP3561A. i have a higher frequency one for RF that i can check it out on. for a source, i have 2 i use. i have an HP 33120A which is good for dealing with things that have a fair bit of distortion (it has -76dB 2nd and -82dB 3rd, so its not good below that). for the lower distortion stuff i have a home made thing with -87dB on both 2nd and 3rd.

the thing about my graphs, though, is that i rolled off everything above 20kHz, so any high frequency oscillation wouldnt make it in. those points up there are just the 2nd and 3rd harmonic distortion components, nothing else.
openmusiclabs.com

v8pete
Common Wiggler
Posts: 90
Joined: Thu Jul 18, 2013 5:32 pm

Post by v8pete » Fri Dec 15, 2017 11:04 am

guest wrote:here is the noise floor versus gain and mode resistor. basically, the noise floor tracks with the square root of the current in the output stage.

Image

the rounding off on the lower gains is due to the resistor noise of the 30k resistor. if you were to subtract this out, those lines would be much straighter.
Useful plots! - I just compared your 0dB curve to the the LTSpice model, and the model shows about a flat 110nV/rt. Hz into 30k (30k input resistor also) out to 10kHz, then it takes off exponentially ( @ 100uA MODE current). Interestingly, at 0dB gain, the Bode plot just shows the gain tailing off gradually after 10k, but it's a different story for control voltages much above 1V -here the gain starts to ramp back up quite dramatically (eg. at 2V control voltage the gain is ramping back up again from 2.5kHz). Also the transient response to a pulse input looks as clean as a whistle at 0dB, but starts to show a bit of ringing at 2V control voltage, but nothing that looks like it would be a big issue in practice.

So, if I take your 95dB re. 1V and divide by rt. 20k, then I get -138dB/rt. Hz, which is quite a bit quieter than the model, if I've interpreted the measured data correctly?

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Fri Dec 15, 2017 11:31 am

v8pete wrote:So, if I take your 95dB re. 1V and divide by rt. 20k, then I get -138dB/rt. Hz, which is quite a bit quieter than the model, if I've interpreted the measured data correctly?
-138dB/rtHz = 126nV/rtHz, so its pretty close. did you divide down the mode current for the actual diffpair drives? its somewhere around a 1:12 ratio, which will make the output 10dB quieter in the simulation.
openmusiclabs.com

User avatar
guest
Super Deluxe Wiggler
Posts: 5852
Joined: Mon Aug 19, 2013 11:49 am

Post by guest » Fri Dec 15, 2017 11:37 am

here is my noise analysis for varying input resistance:

Image

this was measured with a 30k output resistor. the low frequency "noise" is actually oscillations and amplification of line noise. i had to clean up my setup to get those values down to that level. the high frequency noise doesnt seem to have any structure.
openmusiclabs.com

Post Reply

Return to “Music Tech DIY”