Summing amp?

[living sounds]

My ADT console uses an inverting summing amp topology. The channel outputs are 47k. Schematic is wrong regarding the left channel (thanks Jakob!).

The schematic calls for an LT1028, but when I got the console (used) there were NE5534 in the summing position and the designer of the console (who sadly died in 2020) told me that he thought the latter sounded better anyway.

I've also tried LT1028 as well as the 990C+ and they all do seem to work fine without any oscillation to be seen on the scope, and they all do sound quite different...

Now I wonder what op amp might objectively work the best here.

Any suggestions?


(edited to reflect the facts)

 

[gyraf]

V106 drawn correctly, V105 drawn wrong..

would not compute at all otherwise

 

[living sounds]

V106 drawn correctly, V105 drawn wrong..

would not compute at all otherwise

Wow, how did I miss that! It was obivously too late last night when I checked...

So, any ideas on the best op amps for the job?

 

[gyraf]

I always liked NE5534 for this sort of duties - in a well-thought-out design you very rarely need "better" than that

 

[JohnRoberts]

That topology is called "virtual earth" summing. There is a temptation to use an expensive uber op amp in that socket because of the N+1 noise gain. I have killed a lot of brain cells thinking about this. The mile high view is that a decent op amp (like 5534) will not dominate sum bus noise floor.

Keep in mind that you are combining N open channels with mic preamp noise floors and even room noise.

JR

PS: for TMI search my old bloviating discussions.

 

[ruffrecords]

There are two places in a mixer where noise is a real issue. The first is in mic preamps and the second is virtual earth (VE) mixers. So my advice would be to put low noise at the top of your selection criteria. There are also all kinds of stability issues with VE op amps which depend in part on the physical layout of the buses themselves and the VE amps. Very hard to predict what these might be but it might be worth your while reading Steve Dove's excellent mixer design series.

Cheers

ian

 

[JohnRoberts]

There are two places in a mixer where noise is a real issue. The first is in mic preamps

yes

and the second is virtual earth (VE) mixers.

as compared to what? When combining a large number of channels (I had over 100 feeds to my L-R bus in my larger AMR consoles), multiple things matter besides just noise.

A consequence of high noise gain in negative feedback circuits is that the ability of NF to reduce errors (like open loop distortion and open loop phase shift is degraded. ) For a broad discussion a very old article I wrote 1980 console article. and combining thread

I am surely repeating myself but don't over think this, your noise in practice will be dominated by the channels and inputs

So my advice would be to put low noise at the top of your selection criteria. There are also all kinds of stability issues with VE op amps which depend in part on the physical layout of the buses themselves and the VE amps. Very hard to predict what these might be but it might be worth your while reading Steve Dove's excellent mixer design series.

Cheers

ian

+1 Steve Dove is an old friend. If you read the footnotes he gave me credit for finding a mistake in one of his published EQ section discussions when it first ran in Studio Sound Magazine.

JR

 

[john12ax7]

How many channels? An objective view would determine the source impedance and go from there.

Since we are in 5534 land worth looking at more modern bipolar options. LME49710, if you can find them, and then the HA can version is supposed to be even better. OPA1611, although SMT only.

I recall some other LT parts being well regarded for summing amps, was either LT1115 or LT1468.

Edit : fixed typo

 

[JohnRoberts]

I recall some other LT parts being well regarded for stunning amps, was either LT1115 or LT1468.

was that a pun or typo?

JR

 

[john12ax7]

was that a pun or typo?

JR

Ha. Definitely a typo.

 

[user 133392]

In the ADT console does (1) the noise gain change with the number of channels assigned or (2) are the build-out summing resistors back-grounded and the noise gain and source impedance seen by the summing amp remain constant?

If backgrounded and therefor constant what is the Thevenin equivalent of the source?

 

[pucho812]

Can’t beat a 5532/34. They may be long in the tooth but they have held their own after all these years.

 

[Matt Syson]

There are many other sources of noise and distortion way beyond the 'summing amp' and even the mic amp 9depending how rigorous you want to be and layout and component values play a large part.
at around 30 channels of summing the decision virtual ground or 'plain resistive (voltage) summing are around 'equal but VE can benefit on noise IF you can drop the unnecessary input feeds. Reengineering for really low noise gets very involved and complicated (expensive) because you need loads of 'current boosters' to allow low resistance values all over the place. Shortage of 'loop gain' in a summing stage can be an issue easily addressed by a pair of transistors for example as true 'symmetry' is not necessary. Whether a fancy 'new' chip can really outperform a 5532/4 when boosted by transistors will be moot because you are needing to consider several other factors not just the 'headline' gain figures.

 

[ruffrecords]

In the ADT console does (1) the noise gain change with the number of channels assigned or (2) are the build-out summing resistors back-grounded and the noise gain and source impedance seen by the summing amp remain constant?

If backgrounded and therefor constant what is the Thevenin equivalent of the source?

I think the OP already said the bus feed resistors are 47K
I don't know of any ADT console that does not ground bus resistors when unassigned.

Typical ADT consoles have 12 or 24 channels so the equivalent source resistance will likely be in the 2K to 4K region.

Cheers

Ian

 

[abbey road d enfer]

Actually, with an average of 16 channels and 47k feed resistors, the bus impedance is about 3kohms, which is close enough to the OSI of a 5534 (7kohms). In that case the noise voltage density is dominant over the noise current density.
An ultra low noise opamp such as OPA1611/12 has an OSI of about 400 ohms, which results in the noise current being dominant over the noise voltage.
The overall noise performance would be in favour of the 5534.
In order to take advantage of VLN opamps, the bus impedance must be lowered, that's why many mixers use feed resistors around 10 kohms.
Using a "standard" opamp in conjunction with a transistor front end has the same constraints, since operating transistors at high current results in lower OSI.

 

[user 133392]

Thanks Ian and Abbey.

I agree that the 5532/34 are "stunning" op amps for this application.

The reason for my question regarding the source impedance is that introducing an identical value resistor to the source impedance in the non-inverting input will lower distortion considerably regardless of what op amp is used. I realize that this is an an inverter but my 5532 voltage follower tests produced a 6-8 dB distortion reduction when the source impedances at the inputs are balanced. Obviously, if the source impedance is high-ish, the added noise of the resistor could be an issue but I don't see that here with R-Thevenin in the 1-2K range. As JR has pointed out other noise sources dominate. For the "cite your sources crowd" see: Balancing Resistors Reduce Voltage Follower Common Mode Distortion - Pro Audio Design Forum

Assuming 24 back-grounded 47KΩ resistors then the Thevenin equivalent seen at the inverting input is 1k95||22K||47pF.

Put a 1K78||47pF in series with the non-inverting input (when drawn correctly in the OP) and it will likely measure lower distortion with whatever op amp you try.

In situations where the bus feed resistors are not back-grounded R-Thevenin isn't constant and the approach will provide limited -if any- benefit.

An ultra low noise opamp such as OPA1611/12 has an OSI of about 400 ohms, which results in the noise current being dominant over the noise voltage.

And it's bias current-compensated which limits its use in some LN applications. The 5532/34 is one of the few remaining audio op amps which is not Ib-compensated. A lot of the PNP input op amps such as the LM4558 derivatives (LM833, MC33078 etc.) are being discontinued. The 1612 is a great part but in higher source impedance applications the 5532/34's lower current noise will often outperform it.

 

[abbey road d enfer]

The reason for my question regarding the source impedance is that introducing an identical value resistor to the source impedance in the non-inverting input will lower distortion considerably regardless of what op amp is used. I realize that this is an an inverter but my 5532 voltage follower tests produced a 6-8 dB distortion reduction when the source impedances at the inputs are balanced.

I'm not sure you can extrapolate your conclusions regarding voltage-followers to summing amps, because the inputs are more or less grounded.

Obviously, if the source impedance is high-ish, the added noise of the resistor could be an issue but I don't see that here with R-Thevenin in the 1-2K range. As JR has pointed out other noise sources dominate. For the "cite your sources crowd" see: Balancing Resistors Reduce Voltage Follower Common Mode Distortion - Pro Audio Design Forum

I must admit I haven(t read it ... yet. Will do ASAP. But I suspect the principal cause of distortion there is due to the unbalance of dynamic offsets, which wouldn't happen in an inverter.
This is just a guess, I can't prove it, I don't have the equipment anymore.

The 1612 is a great part but in higher source impedance applications the 5532/34's lower current noise will outperform it.

Indeed. Particularly in the OP's case, it would be a tad noisier tha a 5534. Apparently the OPA1611/12 is not bias-compensated, as any hybrid BJT/opamp wouldn't.

 

[user 133392]

@abbey road d enfer "Apparently the OPA1611/12 is not bias-compensated"

The typical OPA1612 Ib spec is ± 60 nA suggesting that, due to the indeterminate sign, it IS Ib-compensated.

@abbey road d enfer "I'm not sure you can extrapolate your conclusions regarding voltage-followers to summing amps, because the inputs are more or less grounded."

Perhaps. I don't think the improvement is necessarily related to common mode distortion alone.

I've never measured a 5532/34 inverter directly with and without, but in my ULDO-Nacho low distortion oscillator (which has a dual NPN/PNP differential helper pairs ahead of the 5532 inverters) introducing a balancing resistor in the composite op amp's non-inverting inputs makes a huge difference. With the added resistance it's an Ultra Low Distortion Oscillator and, without, it its just an oscillator.

Whether a "naked" 5532/34 benefits is easy enough to find out when you have an oscillator capable of THD at -140 dBc. Jung and Jensen have both written about balancing resistors (Jung) and added input resistance for input degeneration (Jensen). Jung argues for distortion reduction and Jensen stability. Thomas Barefoot has posted here about adding equalizing resistors though IIRC the context was Sallen-Key filters which are followers.

 

[abbey road d enfer]

The typical OPA1612 Ib spec is ± 60 nA suggesting that, due to the indeterminate sign, it IS Ib-compensated.

OK, I just had looked at the stripped-down diagram.

Perhaps. I don't think the improvement is necessarily related to common mode distortion alone.

That was just a guess.

I've never measured a 5532/34 inverter directly with and without, but in my ULDO-Nacho low distortion oscillator (which has a dual NPN/PNP differential helper pairs ahead of the 5532 inverters) introducing a balancing resistor in the composite op amp's non-inverting inputs makes a huge difference. With the added resistance it's an Ultra Low Distortion Oscillator and, without, it its just an oscillator.

Whether a "naked" 5532/34 benefits is easy enough to find out when you have an oscillator capable of THD at -140 dBc. Jung and Jensen have both written about balancing resistors (Jung) and added input resistance for input degeneration (Jensen). Jung argues for distortion reduction and Jensen stability. Thomas Barefoot has posted here about adding equalizing resistors though IIRC the context was Sallen-Key filters which are followers.

The big question is choosing between HF distortion and noise.
I know we've been differing with John about it, I still maintain that making the summing amps as quiet as can be is a good thing, IF ONLY to make punters happy.
Now I believe that using truly balanced (i.e. driven from in-phase and out-of-phase signals) bus would reconcile both clans.
Not that it has any practical value in the context of our digital audio era...

 

[living sounds]

In the ADT console does (1) the noise gain change with the number of channels assigned or (2) are the build-out summing resistors back-grounded and the noise gain and source impedance seen by the summing amp remain constant?

If backgrounded and therefor constant what is the Thevenin equivalent of the source?

The resistors are always connected to the mix bus input, but are shorted to ground if the output is turned off. My console has at least 66 channels going to the master bus, counting the inline channels.