Looking for a neutral DOA for summing

[Kingston]

On-board regs in a mixer have often a serious drawback; the noise currents resulting from shunting their output with caps to ground can find their way into the audio ground. I've seen some mixers where replacing the regs with resistors decreased significantly the bus noise.

This just emphasizes the need for proper design of ground currents. Auto-routers aren't good at this.

I've really become fond of PCB designs where local filtering groups are clustered hierarchically with their path entirely separate from all signal ground reference.  They of course eventually meet at the ribbon cable or with fat ribbon cables
(or simple non-mixer applications) all the way at PSU input. This works great at channel strip level in mixers. With more than dual layer boards this is relatively easy to route.

 

[JohnRoberts]

This just emphasizes the need for proper design of ground currents. Auto-routers aren't good at this.

I've really become fond of PCB designs where local filtering groups are clustered hierarchically with their path entirely separate from all signal ground reference.  They of course eventually meet at the ribbon cable or with fat ribbon cables
(or simple non-mixer applications) all the way at PSU input. This works great at channel strip level in mixers. With more than dual layer boards this is relatively easy to route.

Most fun I had was a console installed in the main beam of an AM transmitting tower... The signal ground had a volt of 960kHz on it, relative to power ground.... 

To be low impedance at HF can be a different problem than low impedance at audio frequency.

JR

 

[Andy Peters]

Stability, noise, output impedance. LT3080 and its variants are quite superior to those two. I once managed to kill all spurious oscillations of a console with a PSU based on super regulators (jung regulators - whatever).

Then repeated the feat. with LT3080. I wouldn't build super regulators anymore when they are basically available on a chip now.

The 3080 is a great part. It's our go-to regulator in its current-output class at the day job. The negative complement is the LT3090.

 

[bluebird]

This just emphasizes the need for proper design of ground currents. Auto-routers aren't good at this.

I learned this early building tube guitar amps. If you didn't have the current flow correct in your grounding layout, you were sure to get hum amplified by the first stage.  Especially in high gain amps. Took me years to get a feeling for what worked. Even the radio designers handbook didn't have all the answers for hum and noise. Applying the same methods to solid state designs guarantees a very quiet circuit.

 

[john12ax7]

I've been meaning to try the LT3080. Has anyone used multiple in parallel to increase current capability?

 

[living sounds]

Also if you can do 24v+- the app992 is a great DOA.

Running at +/-17V the APP992 was the second best next to the NOS 5534. Does it sound better at +/- 24V?

 

[Nishmaster]

On-board regs in a mixer have often a serious drawback; the noise currents resulting from shunting their output with caps to ground can find their way into the audio ground. I've seen some mixers where replacing the regs with resistors decreased significantly the bus noise.

If the local card reg and associated caps are right next to the +/- rail input on the card as well as the system ground point (assuming a ground plane on the card), wouldn't this make the reg current loop stay effectively just in that area?

Seems like proper layout is really the answer to these types of issues and not parts swapping, unless of course it isn't your layout, and somebody decided to put the reg and caps miles away from power and ground right next to some front panel controls. Then your choices are of course a bit more limited. 

 

[bluebird]

Running at +/-17V the APP992 was the second best next to the NOS 5534. Does it sound better at +/- 24V?

At some point I had a frequency response problem running a standard 990 at +/-15 volts. Slight high end roll off. I popped a 990 "C" version in the circuit and it fixed the problem. I don't know if the APP992 behaves the same way or not...

 

[abbey road d enfer]

If the local card reg and associated caps are right next to the +/- rail input on the card as well as the system ground point (assuming a ground plane on the card), wouldn't this make the reg current loop stay effectively just in that area?

That simple notion seems to be ignored by some draftsmen; most of them think a ground is a ground is a ground, when a good designer knows a ground is a resistor.
The ground plane can be a serious issue; most software-generated ground planes connect circuit areas that should not be connected directly. It takes some experience and thinking to devise split ground planes that follow hierarchical path.
There is a basic geometry issue there; in most mixers the signal path is from top to bottom, but the power rails are fed from the middle of the PCB. The PCB software will take the shortest path, when good practice requires reorienting the power feed and its concurrent 0V path towards the bottom.

 

[JohnRoberts]

That simple notion seems to be ignored by some draftsmen; most of them think a ground is a ground is a ground, when a good designer knows a ground is a resistor.

or an inductor....

The ground plane can be a serious issue; most software-generated ground planes connect circuit areas that should not be connected directly. It takes some experience and thinking to devise split ground planes that follow hierarchical path.
There is a basic geometry issue there; in most mixers the signal path is from top to bottom, but the power rails are fed from the middle of the PCB. The PCB software will take the shortest path, when good practice requires reorienting the power feed and its concurrent 0V path towards the bottom.

With cad you sometimes have to name multiple local ground (0V) regions to keep them separate.... then you have to remember to connect them all.  :

JR

 

[abbey road d enfer]

At some point I had a frequency response problem running a standard 990 at +/-15 volts. Slight high end roll off. I popped a 990 "C" version in the circuit and it fixed the problem.

According to the John Hardy litt, the 990 and 990A were available in 3 versions, a +/-15, a +/-18 and a+/-24V. The C version is supposed to work across the range of rail voltages.
If your "standard" 990 was one of those optimized for higher rails, it is not surprizing that performance was not right.

The 992 is supposed to work from +/-16 to +/-24.

 

[alexc]

I did a pair of summing boxes recently ... 

Measured up ne5534, lme48860, capi 2520 and yamaha 88100.  They differed a little in terms of THD but not so much.

I consider all of them to be fairly neutral   

My next summing box is using some tube make up - of course, I would expect significantly more THD than any of the above. 

Once again, I'll use passive balanced summing network, this time into step-up transformer,feeding a  tube gain make up amp + line amp with output transformer.

I'll be interested in a couple of different possible nfb schemes - perhaps with feedback around the input transformer ... or maybe more conventional around the output stage.

Anyway my point is how 'neutral' is neutral ? 

I was measuring 20 sources summing passively and getting some low 'triple 0' THD+N figures off the solid state units  ..  like 0.00045% THD+N at around +10dBu into 600ohm loads.

Virtually nothing at all in  the way of coloration of the source signal I could measure, let alone hear.

No way I could tell the SS apart blind; the tube 'maybe'.  I'll know when I build it.

ps - surely everything is a great deal  more challenging in the 100+ sources category, and especially so when 'facilities' are added to basic summing functionality ...

 

[abbey road d enfer]

Anyway my point is how 'neutral' is neutral ?

https://groupdiy.com/index.php?topic=67150.msg853534#msg853534

 

[alexc]

Yep, you are right of course.  Summing ITB,  is *very*  much 'neutral' than even good performing solid state summing, given a decent  audio interface and DAW.

I've done some measures with REW and using all digital ports and it shows some totally great figures.

I've yet to measure the really good stuff, mind you - but even the moderate cost AD/DA interfaces  do summing with super super low artifacts in terms of THD+N etc.

I can only imagine what some of the really good boxes can do, like the mastering quality comverter types.

I sort of miss having a lot of 'facilities' on my summing but do appreciate the qualities 'bare-bones' summing brings.

I do measure better figures on my DIY chains going thru minimal summing vs with a desk

I use a Yamaha 03D in my 'live' room, which is a good performer for synths/drum machines/effects  and converting them to/from digital.

I used to have a decent Soundcraft which I got rid of a few years back - I still kind of have residual desires for a large analog surface, but nowadays it's all small holder stuff

The passive sum + makeup amp units sound better to me for the DIY analog stuff, benefiting mostly from the bare-bones aspect.

 

[abbey road d enfer]

I do measure better figures on my DIY chains going thru minimal summing vs with a desk

The passive sum + makeup amp units sound better to me for the DIY analog stuff, benefiting mostly from the bare-bones aspect.

There are two major factors that make a desk perform not as good as a summing box:
- Mixers usually have more channels so the summing amp has to work with a higher noise gain, resulting in less gain and phase margin
-Mixers suffer from longitudinal noise build-up resulting from currents circulating along the reference ("ground").

 

[living sounds]

I'm not sure the 5534s were really neutral, but the mixes I made with those on the master bus didn't turn out the way I wanted. The individual tracks sounded good and close to the source (maybe a little hard in the high end), but it was very hard to get a coherent and overall pleasing mix.  I made more tests and surprisingly ultimately ended up with APP's germanium op amp on both the summing and the fader amp.

They add a great definition to the sound across the spectrum, a kind of gut punch in the low end as well as a relaxed presence and snappyness in the high end, and allow for much more EQ without things sounding unnatural. In a mix individual tracks can be heard easily and it still sounds coherent. I found it hard to believe at first.

It is definitely not neutral, but it really helps getting the job done and ending up with a subjectively superior result. And the way the coloration works is sufficiently inobstrusive not to create a genre-specific sound. Very unlike an all 2520 or 1731 mix bus, that works well for hard rock or metal, but not for acoustic mixes. I'm pretty happy now.  8)