Nevey Sum Bus

[alexc]

I've been working on a neve styled sum bus. 

I recently retired my trusty Soundcraft Spirit Auto desk - a fantastic desk which has always come thru for me and previous owner who actually had a profitable small studio with it.

Literally 10yrs of service every day 12hrs a day. Not one thing wrong. No noise, crackles or anything. I had a little PSU fuglies a few years back which took 30mins to repair.

Anyway - I know this desk and how it sounds, which is  pretty good.
And quiet. 72 inputs active and around -74dBu noise floor I measured.

So I retired it, for reason that it's too big, uses a bunch of joos, and with all the DIY I have now, as well as DAW automation and what not, it doesn't get much used except as a fancy summing box 

While I was building my (first) sum bus, I setup a patch bay with multing of all my instrument and diy inputs to the Motu converters of my DAW and to some connectors for the sum bus(es!) to come.

--------

So I've been using the Motu hardware mixer for 'instrument/DIY monitor mixing' and headphone mixes. It works, but not so convenient. For someone used to buttons and faders, not much fun.

That aside, the sound quality is pretty good. Very good in fact.

I do believe I have more clarity in my inputs, I assume because I have removed oodles of Soundcraft circuitry (nice and utilitarian as it is!)

There's not *that* much in it, and I could be biased - the ear hears what the heart wants, after all.

Main problem is the latency (48KHz, 256 sample buffer - not sure what the ms figure is for hardware monitoring thru cuemix)

Even with hardware monitoring, to my guitar sensitive ears it sounds like a cheap and nasty noise gate - a kind of a digital quack on the attack of plucked notes.

My lounge room Yamaha 03D digital desk doesn't really suffer from this - the latency on those is quite low. And the gates are fantastic too.

The 1176 fast settings certainly don't do the 'latency quack' - they piss blue rings around the motu hardware monitoring in terms of picking up the transient.

Funnily enough, I'm collaborating with a girl singer/guitarist with a lot of natural talent but no experience of the recording process (till now!). She complained alot regarding headphone mix delay of my music room Motu hardware.

Hence my 'please be hurrying to be finish please'.

-------

Anyway, I am finishing up this unit - a neve styled sum bus for use in summing my instruments/diy on a mult with the same going to daw inputs.

It will be the main summing bus for now, feeding a finalizer and speakers with a digital feed back to the daw for the master mix recording. The expansion stereo input is for monitoring the daw 'in the box' mix.

Will also feed the headphone mix so no delay on the instruments and diy. 'No way' to Motu hardware monitoring!

I've started *another* sum bus which will be the 'one bus to rule them all'.

This guy will feed into it, along with 24 daw outputs and a control room mic! And leave behind mixing 'itb'.

Will be very cool - a Lawo dv975/4 with a pair of transformer driving pico compressors built in 

------------

So this neve styled instrument monitoring sum bus  has :

12 inputs with Left/Off and Right/Off switches, 4 stereo inputs with no switches and 1 stereo expansion input with On/Off switches. All on 1/4" balanced sockets.

So thats 17 inputs to each L, R bus worst case. The inputs without on/off switches have switched shorting of + and - wires when nothing is connected.

2 outputs with Level attenuators, 12 pos grayhill switches and xlr balanced sockets.

External PSU 24Vdc at 200ma or so.

It is balanced bus with 4K7 resistors.

Uses Haufe 1:2  summing transformer (500R : 2K) so far.

Neve style BA283 line amps (no first stage) for gain make up feeding gapped Carnhill traffos.

Parts has been maybe 450us or so. Maybe a little less.

----------

Not finished yet - just completed wiring and testing the passive sum bus, but I thought I would write up what I am finding.

Here's a progress pic of the insides.

 

[alexc]

So happy - the passive bus works perfectly first go

Tested all the inputs, switches and outputs - all test good .

Noise of my Motu unit's mic preamps with nothing connected and no gain dialled in shows as -78.7dBu on my RTAS - that's an aggregated figure from 10Hz to 48KHz and my unit is calibrated for the dBu scale - absolute voltages using CRO, multimeter, 600R loads and iso transformers to the best of my capabiity.

Loopback with the line in/out is -80dBu.

Rightmark would probably report this as -120dB (below FSD!) Everything is -120dB with that.

The sum bus L, R connected with no gain dialled is the same.

When I dial enough mic gain for 1 typical line level input connected, it remains the same

There are NO measurable 50Hz or 100Hz components  (for the first time in my life!)
The noise figure I quoted is mostly due to the rising high frequency curve reported by the RTAS.

I'm pretty sure this is an artifact of the RTAS. Other RTAS apps for example don't show this.
My RTAS does irrespective of which interface I use (3 different types used so far)


---------------

That's as far as I got for now - passive bus all working and noise floor is what I expect.

 

[alexc]

Finally - the sound is GOOD!

Next - some more measuring of the passive sum bus as I add sources.

Then, wiring up the neve gain stage and reedo all the tests.

Finally to go in my music room for use with this great new talent I've found 

 

[joaquins]

Nice project... I'm planing to make a passive summing with carnhill for a friend, waiting for them yet.

JS

 

[alexc]

I wired up  the nevey line amps and did some basic tests today.

All seems good. Running with 75mA idle current for now.

Am working out what makeup gain I need so I can figure the total resistance and steps for the attenuators.

Also trying a few different input transformer types for different ratios.

Should be able to button it up in a day or two after I had a chance to do some more testing.

 

[alexc]

Some more  observations :

Please bear in mind I'm just writing what I see and try to make sense of it later .
Summing busses can a little confusing in practice.

I need to compare my theoretical knowledge (which *feels* considerable) with practical (very limited   )

So I'm trying to better understand the operation by looking at the voltage waveforms with a cro (high impedance input) and driving with a signal generator (low impedance output)

- When looking at the bus (primary of the input transformer), because it is balanced, I'm looking at each phase with respect to ground.

- When looking at the input to the make up amp (secondary of the input transformer), because it is unbalanced (one leg grounded), I am looking between phases.

So trying to understand  the effect of :

- 1 or many inputs connected with front panel switch shorting of + and - in play
- 1 or many inputs not connected with jack field shorting of + and - in play
- different bus shunt resistances
- different input transformers  ie.  1:1, 1:2, 1:4 with varying quoted impedances

sig gen output is 2Vpp measured with no load and 1KHz
17 inputs to bus, 4K7 bus resistors
no bus shunt
input transformer is 1:4.5 turns ratio open circuit



1 source connected and switched ON
=======================

2Vpp sig gen input to 1 source

- rest front panel switches set to OFF and rear panel unplugged (all short + and - inputs to bus)
      -> bus voltage 1.11Vpp
      -> line amp voltage 0.505Vpp (x  0.485 voltage gain)

- switching front and rear panel unconnected sources from OFF  to ON (floating) causes :
    -> not much change in the bus voltage for up to 10 or so inputs (total increase 0.07Vpp)
    -> increase in the bus voltage 11 ..16 inputs  to 1.63Vpp (total 0.52Vpp)
  and
    -> increases  in the make up amp input voltage to 5.6Vpp  (x3.45 voltage gain)

- switching front panel unconnected sources from ON to OFF (short circuit) causes :
    -> decrease in the bus voltage 1..6 inputs to 1.11Vpp (total  0.52Vpp )
    -> not much change in the bus voltage for last 10 inputs  (total decrease 0.07Vpp)
  and
    -> decreases in the make up amp input voltage to 0.505Vpp (x  0.485 voltage gain)

So as inputs are short circuited (disconnected or switched off) :

-  some decrease of the bus voltage due to the increased loading caused by progressive shortings
-  very dramatic decrease in the make up amp voltage due to reflected increased loading

So, the first thing is that there is some loading effect on the bus voltage of disconnecting/switching off channels but a very dramatic effect on the make up amp input voltage.

-------------------------

Each shorting adds (4K7+4K7) approx 10K in parallel with the primary of the input transformer.
So 1 input is 10K is parallel. 6 additional inputs lowers this to is 1.4K. Remaining to 17 total takes the parallel resistance 588R. So that causes the voltage to decrease.

Now the transformer reflects that impedance according to 1:19 impedance ratio (being 1:4.5 voltage/turns ratio squared). So the reflected impedance goes from quite high (10K*19) to moderate (588R*19). 

Seems that I have more reduction in voltage at the secondary than I would expect.
Which is why I started looking at this

Seems to be the case with the make up amp in circuit as well (as opposed the transformer being unloaded).  I think it is around 56K input impedance or so, so should be a relatively light load on the transformer secondary.

Now to think more about what I've done

----------------------------

OK - it seems that driving the input from an unbalanced sig gen output (50R Z) is not so good.

Changed this to driving the input from the sig gen thru a balancing iso transformer (600:600).

Is better

Now I have in the worst case of 1 input driving, 16 inputs shorted :

input jack + -> 2Vpp wrt to ground
input jack -  -> 2Vpp wrt to ground

bus + -> 0.828Vpp wrt ground
bus  - -> 0.828Vpp wrt ground

traffo secondary -> 1.15Vpp with one leg grounded, open circuit

So thats  better : gain of traffo is x1.39

With everything ON (floating) -> bus is 1.85Vpp each, traffo sec is 13.1Vpp, gain  x7.08
Looking at the output traffo with one leg not grounded, each phase is 5.25Vpp, gain which is close to the 1:4.35 turns ratio.

(you get 6dB increase in voltage wrt ground running a traffo single ended vs balanced)

So - drive the thing properly first! Then, what I described still holds with the worst case

But the best case (floating) matches expectation.
 
Makes more sense now. 

----------------------

By way of additional comparison, driving input with sig gen -> active opamp balancing box 600R output Z :

Now I have in the worst case of 1 input driving, 16 inputs shorted :

input jack + -> 2Vpp wrt to ground
input jack -  -> 2Vpp wrt to ground

With everything ON (floating) -> bus is 1.33Vpp each, single ended traffo sec is 11.1Vpp, gain  x8.35

---------

That's it for today. Tomorrow : adding Rshunt across the bus and then checking the variations, followed by loading of the input transformer secondaries and rechecking. After that, reconnect the make up amps and checking.

 

[dandeurloo]

I think I may have at least a pair of the vintage transformers that came out of my neve console if anyone is interested in them.  Maybe 2 pairs?  PM if you need some.

This is cool project BTW! 

 

[alexc]

Thanks!

I should say I'm using neve line amp pcbs from 'Bauman' from several years back.

I believe distribution of Bauman pcbs was taken over by 'The DIY Pill'  http://www.thediypill.com/store

I don't know if they are still available. I've done a number of Bauman pcbs and found them to be very interesting. They also have one called 'cheapsum' which could be really interesting to those wanting to try this kind of thing.

Also, please feel free to improve my understanding here!  Any comments appreciated 

Cheers

 

[alexc]

More thinking and re-reading later ....

The Rshunt is the thing which 'permanently' knocks down the bus impedance, and thereby the bus  voltage, such that switching sources in and out makes little difference to the  voltage on the secondary of the input transformer.

So I need to decide what bus loss is appropriate for me to run and Rshunt accordingly, assuming I have the gain to makeup. I have some 10-40dB available from the nevey line stage and some 1:1, 1:2, 1:4.5 that I can use.

17 sources connected and unshunted is around 25dB loss and bus impedance of 552R with my 4K7 resistors.

I'll Rshunt that for say an additional 6dB loss, 31dB loss all up, initially. That would be an Rshunt of 552R I guess.

Giving me a bus impedance of 276R or thereabouts all sources connected and/or disconnected/shorted.

The output Z of the sources is low compared to the 4K7+4K7 bus resistors so the difference between connected and disconnected/shorted should be quite small.  Even without shorting unconnected inputs, once the bus has Rshunt, the differences would be small.

OK. I'll do that and check the situation with a bunch of things connected.

Just takes some practical and a bunch of re-reading to get the mind to 'assume the position' and be learnt!

Cheers

 

[alexc]

Another test day ...

Now I have put in 1:1 transformers and connected up the make up amp as well as added some 620R bus 'slug'. (more of a 'half a slug' really)

My numbers for bus loss are all now in the ball park and the variation in bus levels due to switching OFF/shorting and switching ON/connected/floating is quite small.

That is now operating as expected.

1 input connected, driving with sig gen-> 600R iso box balanced out -> balanced input, 1KHz sine
2Vpp at each + - input jack wrt ground
88.4mVpp at each + - bus wrt ground    (-27.1dB)
128mVpp at make up amp input single ended, equivalent to 64mVpp balanced (-2.8dB)

So numbers expected are :

17 inputs all connected/shorted, 4K7+4K7 bus resistors balanced
- bus loss  -24.6dB, bus impedance 553R

bus slugged with 620R
- bus loss -30.1 dB, bus impedance 553R//620R -> 292R

I'm getting -27.1dB (-30.1dB expected) at bus, -29.9dB (-30.1dB expected) at the makeup amp.

A few mV difference at the bus when switching in/out sources

Close enough  However, I do think it is still seems to be quite driving source Z dependant.

---------------------

I have max gain of 35dB using the nevey amp gain setting resistor+pot of 100R+2K.
Thats a range of 18dB .. 35dB expected. I can go to 40dB max I think.

I see 3.35Vpp at the XLR output balanced, so that is 3.35/0.064 = 34.4dB.
That is for a 2Vpp balanced input to 1 source, all others shorted

When I increase the input drive to 4.6Vpp, I get the max useful output to the Motu of  around 9Vpp.

So thats fine for now - I will check in the intended application to determine the final gain requirement then select the attenuator resistors for the range I want. For now I have a 2K pot in there.

Noise still very low at -75dBu (see earlier post about measuring method), at full 34dB gain but now with a 50Hz component at -85dBu where before was none. I'm using a cheap wall wart for now, until I build a proper supply for this and my next sum bus. The wall wart is spot on the voltage but perhaps I can clean it up some more.

This figure is actually pretty close to what I'd expect for a simple, low current in-the-rack psu.
So may not so much advantage as I thought. But I'll never know

I will wait a while before doing freq response measurements. Suffice to say I expect the top end to be fine, but the bottom end could be rolling off due to my choice of input transformer.

------------------

So now to have at cross talk and what not.

No est bueno  :-[

I am getting major crosstalk between left and right channels.

I have checked my wiring of my jack field, grounds and front panel switches - does not appear to be an error there.  I've removed the input transformer and line amps. No effect on the problem.
Not sure what's going on and am trying to figure it out.

And it gets fully weird - transmission line weird 

Depending on what I drive this 1 connected input, the coupling is increased or decreased.
I can drive it with 2Vpp to the jack field by :

a) sig gen (50R out Z) unbalanced -> 600R iso traffo balanced -> input jack balanced

b) active balanced fet +chip preamp balanced (? out Z)-> input jack balanced

c) Motu line outs balanced (? out Z) -> input jack balanced

I also can drive it sig gen unbalanced -> input jack unbalanced but that is even weirder, so I'll say no more about it!

And they all give different bus voltages and left/right coupled voltage.

So I think it may be coupling due to proximity of the wiring, switches (happens also on the stereo inputs - minimal wiring and no switches), jacks or bus. (hey I narrowed down some!)

    This is how I used to feel doing transmission line theory. But at least then my maths was good!

Oh well - trying to understand.

Have to pull something out of my .. bag .. because it is not usable like this.

-------------

Seems like it is also one of those quantum relativity things too.
Depending on where you ..put .. your .. probe .., you affect the .. situation.

So - measuring at the single ended make up amp input, with the unit powered up, the problem much is reduced.

Using my sig gen unbalanced -> 600 iso traffo bal -> balanced input jack as most representative of how I'll be using this :

- 2vpp at the input jack, each phase wrt ground
- one mono channel ON and rest OFF(shorted)

left  input to make up amp ->  93mVpp
right input to make up amp -> noise approx 3.5mVpp  (28.5dB below)

and same in reverse

YAY! its better  

Still, after x56 gain makeup (35dB), still ends up 162mV or so - my RTAS confirms this.

It says for those same conditions left XLR -> +1.1dBu, right XLR -> -83.3dBu

OK - maybe it's not a problem after all!



I'll need to wait until I test with more active inputs at more realistic bus levels I think.

So - I still have cross talk between L/R bus but depending on where you poke about, the 'problem' appears worse and does affect the performance of the bus too.
And it worsens with all non driven inputs ON and floating.

Appears to be plain old proximity coupling of the busses.


Some usage and listening tests show theres basically no cross talk left and right busses.
So all good.

----------------------

Now to be looking at channel to channel crosstalk.

 

[alexc]

Measure frequency response with my first pick Haufe 1:1 was not good.
Big rolloff in lows. Highs extended quite well.

Tried my next pick - Pikatron RUP 349M  1:1 - the inputs from a neumann fader unit.

Amazingly perfect - ruler flat with -2.5dB 20Hz and -0.7dB at 40KHz in this circuit.
No sign of clipping or crapping out of bass at lower input levels.
Best .. response .. ever!

Should do

 

[alexc]

Frequency response is great -2.8dB at 20Hz  ... -1.8dB at 44.5KHz for a 0dBu sweep amplitude
Remains very similar for -10dBu sweep as well as +10dBu sweep.

Distortion is very very low.

For a 1KHz sine wave in at 0dB on my RTAS, I see virtually no harmonics at all.

Something like 2nd harmonic -> 92dB below fundamental
                      3rd harmonic -> 89dB below fundamental
                      5th harmonic -> 91dB below fundamental

Looking at noise floor, it is now a little higher due to the non-drooping low end response of these transformers.

-74.5dBu (loopback is -80dBu) at 34dB make up amp gain.

It appears to not be dependant on gain applied.

Seems mostly 50Hz hash on the psu.

I need to remember the calc for THD but it is the lowest I've ever seen in my diy.

---------------------

looking at grounding ...

- jack field has a solid copper wire some 1.9mm thick connected to chassis
- jack grounds are connected directly to copper ground bus
- jack wiring shield are connected directly to copper bus
- switch field wiring has shield floating
- transformer cases are grounded to chassis

and

- psu dc +/- from a 2-prong wall wart, so is not connected to earth
- psu dc +/- feeds gain make up amps and power led

and

- make up amps have a common ground trace
- gain attenuator is not shunt signal to ground type

and

- gain attenuator body, knobs and switch toggles  are not isolated from chassis

I see around 5mV of hash on the psu +ve  with the psu -ve wire also connected to ground.
Makes no difference to noise floor  if psu -ve wire is lifted from chassis.

------------------

OK - now buttoning up and taking for some studio tests to determine final gain settings 

 

[alexc]

After integrating into my music room and testing ...

Everything works exactly as expected  ;D ;D  Bonus points!

Noise floor has remained low with no weird buzzes or hiss or anything at all. I have 12 sources connected in mono/stereo/mute configuration, 4 stereo sources in always-on config and 1 stereo aux mono/stereo/mute config.

So all my instruments and diy channels are connected, along with the daw itb mix and it feeds balanced to a Finalizer.

Noise is a mere 2dB worse with everything connected and in typical usage

Not as quiet as my Soundcraft by around 3dB or so I would say (17 sources)
But pretty good as that desk is very quiet. And it was that quiet with 72 sources connected and active.

Not noticing any cross talk issues channel-channel or left-right or switched on-switched off  8)

Level shifts when switching other channels in and out are quite minimal.

Sound is superb. Powerful and hifi. Bold! and all the other adjectives you want to use.

Much better now that I can monitor in analog and not use the audio interface. No delay, no quack  8)

Gain setting is pretty good - no shortage of gain even with only a few instruments banging away.

Can handle all my big, powerful diy output stages with 600R outputs without batting an eyelid.
This always would cause me a problem - my gtr tube pres always having to run almost turned off so as not to blow the input stage of the motu and even the soundcraft as well. Now they are breathing properly and I can really hear the more natural crunch coming through.  BIG plus.

Very high headroom without blowing anything!

--------------

Couple of observations

-> mute switching is a must for anything that tends to be not-absolutely-quiet.

I'm very glad I included these for my instruments and diy channels. Makes keeping noise under control much easier.

Would not want to have always-on configuration for any of my instruments or diy channels.
OK for things like good quality electronic instruments, fx and daw returns

-> I may do some 'hotting up' or 'cooling down'  of some channels with the bus resistors to accomodate some of my quieter instruments and some of my more over the top output stages.

-> maybe a  good idea to have a mute on the outputs

-> maybe a good idea to have a make up gain 'range' switch to cater for the case where your range is too limited
    ie. all instruments blaring, may need to lower gain beyond that which you can. Depends on your setup

-> I wish I had added that extra couple of stereo inputs I had room for. Noise doesn't seem to be a problem.

That said, this has been one of the more rewarding builds.

Now - to get the 'master bus' on the path to building!

Cheers
AlexC

 

[alexc]

Few hours of studio noodling later ..

OK - I'm going to say it.  The nevey sum bus sounds a lot better than my desk did !

It's like I'm hearing a whole new sound - more detailed, less veiled and more vibrant with saturated color!
My guitars suddenly give me the sound without fighting.
That's the main thing. I would have to fiddle more to get happy. Here it's just - OK. There it is!

Wow! Best few hundred dollars and time I ever spent!

Completed pic ..

 

[kpearsall]

Great to hear Alexc!

Been kind of following this, hoping to have the time to sit down and REALLY read through it. Just bought the Neve Tech docs from Geoff Tanner and hoping to put one together myself.

Great work!

 

[alexc]

Thanks !

You won't regret making a neve style sum bus. Can be as simple or comprehensive as you want.

Sounds  stonkingly great and shouldn't break the bank.

Good luck

 

[Scrappersa]

Oh man, it's beautiful! Where did you get those BA283 cards? I might have to do something similar on my summing mixer project I'm working on!

And what made you go with 4.7k resistors for summing? Do the old neve console use that value? I know old API consoles use 47k.

 

[alexc]

Thanks!

nevey line amp cards were from a 'Bauman' group buy years ago. Pretty simple stuff - 2 voltage gain transistors + 1 current transistor driving a gapped transformer in class A. Standard nevey stuff.

I chose 4K7 after some audio alchemy ... value depends on

-  'balanced bus' or 'unbalanced bus'
    -> balanced bus are 1/2 the value of unbalanced bus 'cause there are two of them
    -> one for the + phase, one for the - phase

- various bits and pieces of 'claculating'

    -> 10K  balanced  with 250R terminating resistance is pretty standard in 16 channel balanced passive summers
    -> mic pre gain make up is also pretty standard

This one was different in that I was not using a mic pre but a neve style line amp  (less gain, higher input impedance).
That affected my decision to go with lower value bus resistors.

So far it seems to be working
  ->  stable performance,  low noise floor, wide bandwidth, miniscule distortion, huge headroom, cheap ...
;D

It's no 'Neve Capricorn'    but from the wide variety of 'semi-pro' gear I'm familiar with, I'd rather have this.  8)

............

There's surely an art to blending of bus resistors, bus terminating resistance, summing transformers and gain make up amps.

Not to mention pan pots, aux sends, mutes, solos and bus sends. And crosstalk and noise coupling too!

After that, it gets *really gnarly* - plenty of seasoned designers around here can attest to that!

All one can do is do the research (ie. reading), do your figuring and go with that.
Build it, measure and use it for music.

Then do it another time, but better.


I do feel you can't go too wrong if you keep it relatively simple.


 

 

[lotus]

Not sure if I understand the requirements for this,but if I was building a 8 input summer with left and right switches,mute,etc feeding a pair of neve input transformers feeding 1272 line in amps and a gappped output transformer would I need to use 4.7k resistors and what about the resistors for shunt noise what values would these be?
Thanks Eric

 

[Scrappersa]

Hmm.  I'm looking to do 16 channels with individual level pot and panning pots for each channel but I'm stuck a bit on what values I should get for everything. On my api I was gonna do 1k level pots, 10k panning pots, and 47 k summing resistors. Panning pots were gonna have 2k2 resistors slugged to them. But since the neve circuit is so different im a bit concerned about my values.

What would you do with that in mind?