MUFF WIGGLER Forum Index
 FAQ & Terms of UseFAQ & Terms Of Use   Wiggler RadioMW Radio   Muff Wiggler TwitterTwitter   Support the site @ PatreonPatreon 
 SearchSearch   RegisterSign up   Log inLog in 
WIGGLING 'LITE' IN GUEST MODE

reducing distortion in OTAs (LM13700, etc)
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Author reducing distortion in OTAs (LM13700, etc)
guest
i just finished a writeup of some work i have been posting here in various places over the past 6 months. here is the link to the paper:

http://www.openmusiclabs.com/files/otadist.pdf

it goes into the fundamental causes of distortion and temperature drift in OTAs, and gives a number of different circuit solutions to the problems. These are all tested and compared, with some comments at the end. my apologies in advance for it being pretty math heavy and long, but if youre interested in the topic, dont let that scare you off. you can look at the schematics and use the circuits without having to read any of it.

and, as always, comments are welcome.
PWM
Thanks for sharing! I'll definitely read this on my (short) holiday.
kvitekp
Great reading, thank you for researching, writing and sharing.

/Peter
nvining
Reading this now, but it looks like a very nice analysis so far!
spacedog
Thanks cool
Scott Willingham
Great! Some old school analog circuit analysis and very well done.
mbartkow
Fantastic job, thank you!
JozeyWhales
Wow, this is great! Such an awesome insight into a really great IC.
neil.johnson
Interesting discussion. Have you read through Don Sauer's pages at http://www.idea2ic.com/
I've had occassional discussions with him over the years so it might be worth sending him a copy of your paper for his thoughts.

Neil
The Real MC
Nice research. There's some pretty interesting schemes in patents too.
guest
neil.johnson wrote:
Interesting discussion. Have you read through Don Sauer's pages at http://www.idea2ic.com/
I've had occassional discussions with him over the years so it might be worth sending him a copy of your paper for his thoughts.


that site is great, and i have learned a lot from it. is the contact info on it still relevant?
neil.johnson
guest wrote:
neil.johnson wrote:
Interesting discussion. Have you read through Don Sauer's pages at http://www.idea2ic.com/
I've had occassional discussions with him over the years so it might be worth sending him a copy of your paper for his thoughts.


that site is great, and i have learned a lot from it. is the contact info on it still relevant?

I believe so - it was back in 2013 when I last discussed something with Don.

Neil
neil.johnson
Another paper worth reading for those interested is Achim's analysis of the OTA:

http://synth.stromeko.net/diy/OTA.pdf

Neil
slow_riot
In a recent thread I was bemoaning the lack of linear 2 quadrant multipliers.

I went back to the Mike Sims VCA updated in this paper with an opamp buffer. I was using a triangle as the input waveform which made distortion more complicated, but I couldn't see any change in the fourier series between the input and output waveforms.

The logging circuit needs a lot of compensation (1nF with 100kohm input impedance placing a 6bB/octave filter pole at 10kHZ), this worried me a bit in the context of designs with multiple interaction paths and feedback (oscillators) but I think this is a pretty manageable level. It might be possible to tweak this to get the pole a bit higher.
neil.johnson
Assuming you mean the EDN paper, cached here:
http://www.teaser.fr/~amajorel/sims/
the performance quoted by Sims is a respectable 0.015% THD and -70dBu noise. It may be possible to improve on that with a better op-amp for IC2B, and you want to minimise noise on the control port as well.

For comparison, the THAT 2180B is typically 0.01% distortion (the 'A' version is half that) at 0dB gain, although it is about 25dB better noise performance than the Sims/13600 circuit.

Now that I've got the big audio analyser up and running it might be fun to try this circuit out.

Neil
guest
i think slow_riot was referring to the updated version in my paper (linked at the top of the thread). a second opamp is employed to eliminate the input errors of the first OTA stage, greatly reducing distortion.

slow_riot: as for the 1nF, i dont think i optimized this at all, although i suspect it cant go much lower than the 330pF originally used.
slow_riot
1nF along with 100k input impedance were the only values I had that quelled the endless oscillations. Also fiddled with bypass capacitance on the LM13700 power pins to no avail.

I also suspect I could go a bit lower 50k and 1nF no good,but maybe 80k/1nF or 100k/700pF

Great circuit though, thanks!
neil.johnson
Oh, right. I did skim through your paper quickly but didn't see a references section at the end with any link to the Sims paper.
guest
i like to reference as i go, rather than put them at the end. so the Sims reference is in the dual ota section.
guest
slow_riot wrote:
1nF along with 100k input impedance were the only values I had that quelled the endless oscillations.


is the 100k here the input to the inverting pin on the opamp, or the Iabc setting resistor on the OTA? i used a 24k as my input to the inverting pin of the opamp, and 100k for the OTA.

im now wondering where the limitations are for speed on this. the opamp and OTA both go into the MHz range. i cant seem to find a phase margin plot for the OTA, but it should vary with Iabc (i think). the feedback path has a hidden lowpass filter in it, which could be causing problems. the voltage divider that feeds the 2 OTA inputs has capacitance at those inputs. but, this should be well in the MHz range as well.

which opamp were you using?
slow_riot
I need unity gain so I was using 100k to the opamp and 100k from a 10V ref to IABC. I'm now doing the maths properly as IABC input is not at 0V and ohms law means I need to get out a pencil and paper and think about temperature dependence of diode drops.

I think the current gain of the feedback network that includes the OTA is significant, higher gain increases the propensity towards oscillation.

I cut out the resistor divider which is probably exacerbating the problems:



I think it's the phase compensation cap that needs to work against the active and inductive feedback path that is affecting speed.
slow_riot
I have protoboard phobia right now, but some simulations have revealed that removing the voltage divider had exacerbated the oscillation problems.

I hadn't realised that negative feedback kept the output of the logging stage at unity even with the voltage divider. All hail the mighty op-amp!
guest
i still had the circuit on my breadboard, so i powered it back up and ran some tests. i found that it was primarily slew rate limited, so i put in a faster opamp (originally had an op297:0.15v/us, then went lt1364:1000v/us, then tl072:15v/us). the tl072 was fast enough, and didnt have the horrendous bias current of the lt1364.

i removed the feedback caps, and didnt notice much difference in performance. well, there was more high frequency noise, and the rolloff moved up to 200kHz or so, but that was to be expected.

some things to note: i used a 3.9k/36ohm combo for the attenuation in the feedback loop. this helps keep the gain down. using a 10k at the output stage instead of a 100k helped with the slew rate issues.

ultimately, it did not oscillate without the stabilization caps, but it did have a lot more "instability". there was a lot of noise in the 50kHz to 300kHz range that would come through. a simple filter at the end got rid of it.
guest
i just ended up coming back to this, as i had some new ideas for compensating the dual ota scheme after looking at how the CEM3340 and SSM2164 do their compensation. it seems to work much better:



the issue with the old version, was that it used a second capacitor to feed forward some of the distorted signal, to cancel out the rolloff due to the feedback capacitor. this worked fine for distortion, but added a bunch of noise, that just got worse and worse as you went up in frequency. this method doesnt need any of that business. increasing the 1.8k resistor helps, but limits the full amplitude you can reach. this was done with a TL082, so faster or slower opamps might need different values. i didnt see too much difference in either distortion or noise floor by altering the 1k/100pF combination to 560ohm/1nF, so probably anything in that range is fine. the circuit needed less than 2k to eliminate oscillation, and more than 50pF to get the cutoff frequency below the oscillation.
slow_riot
Nice, I didn't like the feedforward method. I really want to use this circuit as I am sitting on a bunch of LM13700s but the board real estate is huge.
MUFF WIGGLER Forum Index -> Music Tech DIY Goto page 1, 2  Next [all]
Page 1 of 2
Powered by phpBB © phpBB Group