THAT VCA high 2nd order distortion

[Kingston]

Hello,

I'm playing with Behringer MDX2600 compressor (schematic here https://groupdiy.com/index.php?topic=44991.msg564977#msg564977). The original has a THAT2159 VCA clone by the brand of cool audio ie. Behringer. It has too much distortion even when optimally trimmed. I changed it to THAT2180A (factory pre-trimmed) and removed the trim and its scaffolding completely.

Here are the THD specs before and after.

Cool audio "THAT2159", 1khz THD 0.048%

THAT2180A, 1khz THD 0.179%

I managed to clear most crap above the 2nd order distortion. But now 2nd order distortion dominates and is way too high. I have used THAT2180A in GSSL variants and with DBX166 modding and it usually performs with something like THD 0.01%.

The biggest difference I can see comparing MDX2600 to GSSL or DBX166 is that input resistor to the VCA is way lower in MDX2600 (R17/R28). Can I fix the distortion simply by changing this to a way higher value - to something like 27K of GSSL?

(I of course need to compensate this with the post-VCA opamp stage)

THAT2180A datasheet also talks about this resistor as a trade-off between noise and THD.
http://www.thatcorp.com/datashts/THAT_2180-Series_Datasheet.pdf

 

[JohnRoberts]

Yup, you can think of smaller value resister which feeds higher current into the VCA, like driving a conventional linear circuit with higher voltage swing. 

Indeed no free lunch so linearity, vs noise floor trade off. Pick your poison.

JR

PS: That thd still looks a little high, but i don't have that memorized. Are both EC ports terminated with same low impedance?

 

[Kingston]

That thd still looks a little high, but i don't have that memorized. Are both EC ports terminated with same low impedance?

EC- (pin3) to GND. EC+ control signal terminated with 1k to GND. pin4 is floating due to pre-trimmed VCA.

I removed VR2, R19, R20. But perhaps it was also acting as symmetry trim for control signal or something? It's certainly routed to control signal input through R20 and I somehow missed this earlier.

 

[JohnRoberts]

If EC+ is pin 2(?) that is driven from a low Z op amp output. 

If that ap note says float pin 4,  all good..

JR

PS: It has been a very long time since I messed with VCAs

 

[PRR]

> removed the trim

Why?

There's always some mV of pair offset. This directly causes 2nd order distortion. While some of the THAT-made parts are so close to no offset that they can be fine un-trimmed, if a trimmer is in there already I sure would not take it out.

 

[Harpo]

Maybe increase R10 to 5K9 for the from datasheet recomended 2.4mA bias current of the THAT2180 part ?    (-2.85V - -17V)/2.4mA=5K896.  (the 4K7 set this to 3.0 mA)

 

[abbey road d enfer]

Have you tried reinstalling R20?

 

[Kingston]

> removed the trim

Why?

There's always some mV of pair offset. This directly causes 2nd order distortion. While some of the THAT-made parts are so close to no offset that they can be fine un-trimmed, if a trimmer is in there already I sure would not take it out.

The datasheet is quite clear to leave this floating with the factory-pretrimmed parts. This has worked perfectly fine in other projects.

Have you tried reinstalling R20?

Since this is connected to pin4 that the datasheet says to float, no.

But the trim does also bias VCA EC+ control input, which is something that I missed earlier. I will put them back as a quick experiment.

Maybe increase R10 to 5K9 for the from datasheet recomended 2.4mA bias current of the THAT2180 part ?    (-2.85V - -17V)/2.4mA=5K896.  (the 4K7 set this to 3.0 mA)

Good catch, will try.

 

[JohnRoberts]

I would expect reducing the operating current to perhaps lower noise but increase distortion. 

Lower nominal operating current makes the audio signal current modulation a larger fraction of the total.

or not... (remember no free lunch).

JR

 

[Kingston]

I would expect reducing the operating current to perhaps lower noise but increase distortion.

DBX166XL - that has great THD spec - also has the operating current set with 4K99 which is close enough to this Behringer. But it won't take long for a quick experiment.

 

[JohnRoberts]

without getting too esoteric the VCA operating current density needs to be coordinated with the audio input resistor. If this units 12k audio input resistor is lower (higher current) than in the pro forma app note, the current density needs to be proportionately higher to scale.

or not... you'll see, or maybe not.

JR

 

[abbey road d enfer]

without getting too esoteric the VCA operating current density needs to be coordinated with the audio input resistor. If this units 12k audio input resistor is lower (higher current) than in the pro forma app note, the current density needs to be proportionately higher to scale.

or not... you'll see, or maybe not.

JR
[/quote Yes, the input stage introduces about 2.5dB attenuation, which accounts for about 30% reduction in input current.

 

[Kingston]

I put the trimmers and the scaffolding back, even if the chips are pre-trimmed. Performance is now pretty much equal between THAT chips and Cool audio clones. Not so good even with the lowest THD trim setting I could coax for the THAT chips..

Cool audio "THAT2159", 1khz THD 0.048%

THAT2180A, 1khz THD 0.050%

Next up experiments with re-gain staging the VCA's.

 

[Rocinante]

Years ago I believe it was Volker that tested the THAT2180 vs the THAT2159 in the GSSL and found that the THAT2159 had better distortion figures for that circuit. A THAT engineer chimed in and explained why. I can't remember his explanation exactly and can't ATM find the thread but it's very interesting.
In an experiment I took apart a mdx4600 and used the IO for a THAT4301 dual comp I built. After tweaking the MDX4600 IO (trimmers located after the THAT2159 chips) it sounded surprisingly transparent. It sounded surprisingly good. I changed the THAT2159 for 2180 and found distortion occurred unless I trimmed it a lot.

 

[Monte McGuire]

I'm just grasping at straws, but given that DC offsets will shift the VCA out of balance and increase the 2nd HD numbers, it makes sense to check that C52, the 220µF 4V input coupling electrolytic, is not leaky.  That cap is well oversized, since it only has to couple into a 12KΩ load, but in my experience, small, low voltage caps can get leaky much more easily than larger, higher voltage parts. So, while this cap might be completely fine, if it leaks DC, this will directly bias the VCA out of balance, and increase 2HD.

I'd consider a quality replacement that has an optimal combination of low leakage, high capacitance, and low impedance. I haven't done any research for you, but the typical suspects of Panasonic, Nichicon and United ChemiCon should be able to provide some low leakage, low impedance, long life caps.

 

[Kingston]

I'd consider a quality replacement.

I did this before even switching this on the first time. Elna Silmic caps all over. And yes 220µF is pointlessly oversized, replaced with 100uF.

 

[Kingston]

I replaced the VCA current and gain staging pair with pair of 30k instead of 12k1. Left channel performs especially well: THD 0.01% exactly as promised by THAT specification. Overall THD did not go down as much I'd like but a lot of high frequency THD crap is cleared and it's a better shape - tonally speaking. 3rd order distortion is gone on both channels, even 2nd order gone for left channel (not visible in the below image. I'm lazy to post the two channels as separate images).

THAT2180A, 1khz THD 0.036%

Starting point in comparison.
Cool audio "THAT2159", 1khz THD 0.048%

 

[JohnRoberts]

That is a classic distortion vs noise floor trade off... you could see roughly 10 dB higher  (30k/12k) noise floor, which may or may not be a concern.

I have done some experiments for cut only VCA application where the VCA output is subtracted from a dry unity gain path.

For unity gain the subtract VCA is cut off so performance is same as dry path (very good).

-6dB cut has VCA noise contribution roughly -6dB. For deep attenuation this loses to simple forward VCA topology, for lowest noise floor.

I know this was looked at for use in a simple limiter by a friend (now RIP).

Back in the 80s, I used the subtractive process to make decent sounding limiter/ noise gates using cheap OTA gain elements.

JR

 

[abbey road d enfer]

I replaced the VCA current and gain staging pair with pair of 30k instead of 12k1.

What about S/N?

 

[Kingston]

What about S/N?

Should be apparent from the two images in comparison. Noise increased perhaps 2-3dB. But noise floor is so very low the difference is negligible.