Audio RMS versus Peak

Construction information for DIY projects, including the MS Mid Side Matrix, Elliptic Equalizer, Mastering Console, Phono Transfer System, Insert Switcher and the Dual Class-A Amplifier. You can post your baby pictures here.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Audio RMS versus Peak

Post by mediatechnology »

I am not sure what exactly a balanced VCA configuration is. Using two (one for positive zig, and another for negative zag) will sum to +6dB more signal, while noise should combine incoherently to raise noise floor +3dB. So net 3dB S/N improvement for 2x cost.
People don't seem to care about mass-production costs here John.
I don't know if terkio was going to actually make his own VCAs - he didn't say that.
My answer was related to turbo-charging THAT VCAs.

I did a cost no object high headroom VCA for Greg Hanks that used a pair of 2180s operating differentially.
He didn't care about the cost so I figured it was wise that I didn't either.
At high levels there is some really effective even-order THD cancellation provided by a differential topology in addition to the modest 3 dB noise improvement.

I used differential in/out on the audio ports which were I-V-converted by two OA stages and then followed by a cross-coupled dual diff amp.
The VCA was balanced in/ balanced out with common mode rejection in-between.
The Ec control voltage was also differential drive giving it high "off-ness" and low control feed-through.

THAT2180's work best.
Though the 2181 can be trimmed for lower THD, the THD tends to drift more with temperature than the 2180.
One factor is the tempco of the external trim versus the internal closely-matched tempcos of the 2180's internal trim.
Nulling multiple VCAs also seemed silly.

The VCAs being current in and current out aren't voltage-limited but the I-V op amp outputs limit swing.
Having cross-coupled output doubles voltage headroom if you want or need a balanced output.

I agree with Duke not liking rail-to-rail op amps in audio.
The two exceptions might be the OPA1612 (bipolar) and OPA1642 (FET) which have nearly-identical "rail-to-rail" common emitter outputs.
The one odd thing about the 1612/1642 outputs is that the output impedance actually rises below 100 kHz.
Zo open loop at 100 kHz is about 1Ω but at 10 Hz its 2KΩ. Why?
That's weird - is it some thermal phenomena?
Never seen that before.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Audio RMS versus Peak

Post by terkio »

Thanks all, for your inputs, I had typed many answers but lost all cuz a timeout.😣
To make it short, I think of using MDAC like DAC8811 at the VCA stage ( unbalanced )
Balanced XLR output using OPA454 with high voltage rails.
Stay with usual voltage rails at the XLR input stage and VCA stage.
The requirements is a very high headroom and very small gain steps, so noise and stability are issues that eliminate analogue VCAs.
Last edited by terkio on Sat Sep 14, 2019 2:19 pm, edited 1 time in total.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Audio RMS versus Peak

Post by mediatechnology »

I never get timeouts.
How does that happen?
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Audio RMS versus Peak

Post by terkio »

mediatechnology wrote: Sat Sep 14, 2019 2:08 pm I never get timeouts.
How does that happen?
I am extremely slow at typing on a smartphone and checking stuff. I got logged out.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Audio RMS versus Peak

Post by terkio »

I looked at a fully balanced configuration, using two VCA circuits as suggested. Compared with balanced to unbalanced input stage, one VCA, unbalanced to balanced output stage.
I like the fully balanced for a nicer output stage and the harmonic 2 cancellation and a better S/N.
The drawbacks: Cost of course and a more difficult and larger layout prone to pick up more hum and rf.

Using MDACs rather than analog VCAs, I expect more stability and accuracy. The I/V op amp is critical, since it needs all good characteristics: Low THD, high Slew rate ( power bandwith ) hight accuracy ( Vos Ios ).
I do need monotonicity.
I had looked at precisons, rail to rail amps with a not too shabby speed
OPA 197 192 140
and OPA 828 OPA 1612
It seems rail to rail is a bad idea, except
OPA 1612. It is all good but Ios Vos might not be good enough. OPA 1642 FET input has too poor Vos.
The datasheet of the DAC8811 says a precision op amp should be used ( obviously ) and recommends OP97.
I rather would know what is required to guarantee 16 bits monotonicity.
The DAC8811 is guaranteed monotonic, but the vagueness about the I/V op amp, makes me think, it is border line monotonic, asking for a perfectly accurate unavailable op amp.
What can I expect using OP1611 with DAC8811 Guaranteed monotonic 16 bits ?15 bits ? 14 bits ?
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Audio RMS versus Peak

Post by JR. »

mediatechnology wrote: Sat Sep 14, 2019 1:58 pm
I am not sure what exactly a balanced VCA configuration is. Using two (one for positive zig, and another for negative zag) will sum to +6dB more signal, while noise should combine incoherently to raise noise floor +3dB. So net 3dB S/N improvement for 2x cost.
People don't seem to care about mass-production costs here John.
I don't know if terkio was going to actually make his own VCAs - he didn't say that.
My answer was related to turbo-charging THAT VCAs.

I did a cost no object high headroom VCA for Greg Hanks that used a pair of 2180s operating differentially.
He didn't care about the cost so I figured it was wise that I didn't either.
At high levels there is some really effective even-order THD cancellation provided by a differential topology in addition to the modest 3 dB noise improvement.
I wasn't aware of THD cancellation (at what gain?). Does that allow you to hit the VCA harder (more signal current) for even better S/N? Usually the trade off is THD rise with hotter level.
I used differential in/out on the audio ports which were I-V-converted by two OA stages and then followed by a cross-coupled dual diff amp.
The VCA was balanced in/ balanced out with common mode rejection in-between.
so CM rejection after the VCAs? should be OK for most apps.
The Ec control voltage was also differential drive giving it high "off-ness" and low control feed-through.

THAT2180's work best.
Though the 2181 can be trimmed for lower THD, the THD tends to drift more with temperature than the 2180.
One factor is the tempco of the external trim versus the internal closely-matched tempcos of the 2180's internal trim.
Nulling multiple VCAs also seemed silly.

The VCAs being current in and current out aren't voltage-limited but the I-V op amp outputs limit swing.
Having cross-coupled output doubles voltage headroom if you want or need a balanced output.

I agree with Duke not liking rail-to-rail op amps in audio.
The two exceptions might be the OPA1612 (bipolar) and OPA1642 (FET) which have nearly-identical "rail-to-rail" common emitter outputs.
The one odd thing about the 1612/1642 outputs is that the output impedance actually rises below 100 kHz.
Zo open loop at 100 kHz is about 1Ω but at 10 Hz its 2KΩ. Why?
That's weird - is it some thermal phenomena?
Never seen that before.
I don't want to see that... :lol:
======
I guess the old strategy for price no object designs to parallel VCAs doesn't work with modern VCA ICs (asking I don't know)?

I soaked a few VCA design ideas in beer to use VCAs (added/subtracted) from a dry path to generate mostly cut, with modest boost limited to +6dB in an alternate configuration. In subtractive mode the unity gain performance is only limited by the dry signal path (VCA cut off), so very good. In deep attenuation you get noise contribution of unity gain VCA so not bad but not ideal for console fader automation wanting deep (and quiet) fader kill, but probably fine for high performance comp/limiters.

I shared this idea with JP (RIP) and he confirmed on the bench that it works as predicted. IIRC he was going to use it in an internal limiter for some mixer design he was working on.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Audio RMS versus Peak

Post by terkio »

THD cancellation.
With a differential topology, if the two legs are exactly the same then you have cancellation of all even harmonics. This is from theory and applies at any level.
This theoretical result I have seen effective in bridged power amps.
Cancellation of even harmonics does decrease THD, you are stuck with the odd harmonics though.
User avatar
JR.
Posts: 3700
Joined: Sat May 24, 2008 7:21 pm
Contact:

Re: Audio RMS versus Peak

Post by JR. »

terkio wrote: Tue Sep 17, 2019 2:28 pm THD cancellation.
With a differential topology, if the two legs are exactly the same then you have cancellation of all even harmonics. This is from theory and applies at any level.
This theoretical result I have seen effective in bridged power amps.
Cancellation of even harmonics does decrease THD, you are stuck with the odd harmonics though.
I used to be a product manager for power amps and they typically measure higher distortion in bridge mode due to driving 1/2 the load impedance for 2x current.

I don't know if the trick works to lower distortion for VCAs but I will eagerly wait for your results.

JR
Cancel the "cancel culture", do not support mob hatred.
User avatar
terkio
Posts: 304
Joined: Thu Feb 07, 2019 3:17 am
Contact:

Re: Audio RMS versus Peak

Post by terkio »

I think it is a good trick for analog VCA. Their exponential transfer curve mostly makes H2 harmonics, H2/H1going higher with signal amplitude. The trick zeroes H2/H1 lowering the THD figure, but unfortunately does nothing about H3/H1.

The results about bridged amps was not from commercial PA, this was from a simple DIYaudio amp based on an op amp with a 2EF output stage and massive feed back. Bridged because not using high voltage op amps.

When bridging to get double the power ( doubling the voltage and keeping the same current ) should give a lower distortion thanks to even harmonics cancellation.

I will not use the trick of differential VCA because I will use Multiplying DACs, the H2 removal would not give a worthwhile benefit using MDACs.
User avatar
mediatechnology
Posts: 5442
Joined: Sat Aug 11, 2007 2:34 pm
Location: Oak Cliff, Texas
Contact:

Re: Audio RMS versus Peak

Post by mediatechnology »

This thread is now hopelessly mis-titled.

For the record:
Their exponential transfer curve mostly makes H2 harmonics, H2/H1going higher with signal amplitude. The trick zeroes H2/H1 lowering the THD figure, but unfortunately does nothing about H3/H1.
The dominate source of even-order harmonics in THAT VCAs is due to Vbe-mismatch of the gain cell transistors.

The THD-even reduction from fully-balanced operation is mostly realized at elevated levels and is only as good as the pairs of VCAs match.
Usually they match pretty good and the reduction is significant.

With the title "Audio RMS versus Peak" on this thread I'm not sure I want to continue a VCA discussion here.
Maybe in another thread some day.
I am asked to design a VCA with very high headroom
So why did we start with VCAs and veer to MDACs?
I would suggest starting a new thread on your MDAC project and let this on die since your original question got answered in post #2.
Post Reply