Yet Another diy Sum Bus

[alexc]

My main diy for the last few months has been a a couple more sum bus units ...

Here's the first one, which I'm just  testing now.

It has 8 mono, switched balanced inputs, plus 5 stereo switched balanced inputs - so that's 13 switchable inputs mixing to one stereo bus.

It's using 'passive'  or 'voltage' summing with mic-pre style make up amps, which are done 'api mic-pre'  style (non-inverting amp). 

The other summing unit I'm working on is a 'virtual earth' or 'current' summing unit with a Lawo dev975-4. More on that later  -  I like to build both at the same time and do some compares like 

I also enjoy doing 'cheap', and so am using recycled parts, cheap racking with my homespun printed transparent film panels.

So unit #26 is up and in test  the APIe Dual Sum Bus 

Input traffos are pikatron rup 219/M in reverse for 1:1.125 step up ratio. They are from an old neuman fader unit.
Something like 300:380 ohms impedance stepup.

Discrete op amps are Yamaha NE80100 modules from an old PM desk.

Output transformers are Tamradio 600:600 from an old Ramsa desk.

I used veroboard for the sum buses. And I did add some cheap VU meters+drivers.

The mic pre boards are from member Mitsos - it's a very flexible board with options for servo as well as input bias trim - 'john hardy' style.

Of course it has the usual mic pre functionality one finds in a well refined pcb : switched phantom and input pad - as well as facilities for an output pad and DI patch point.

The super nice touch is Mitsos included an option for parallel output by way of a drv134 balanced transmitting chip.
It allows me to send off to Fx without having to mult the traffo balanced output.  Very impressive performance!

It's very, very nice job of a pcb    Thanks Mitsos!

 

[alexc]

Here's the back panel ...

 

[alexc]

And inside pic ...

 

[alexc]

I used 4K7 bus resistors, balanced bus and a 390R bus slug resistor.

This one worked pretty great right off the bat - no need to further tune the 'bus loss' by way of the 'bus slug resistor'. No additional 'bus build out resistors' were needed either.

It's showing pretty good results so  far 

Frequency response is ridiculous. Like -0.2dB at 10Hz and -0.1dB at 20KHz.

THD for 1-chn routed with gain makeup for unity, is  <0.001%  at 1KHz, +9dBu

THD+N fo 1-ch routed with gain makeup for unity, is  <0.0047%  at 1KHz, +9dBu with  20Hz..20KHz BW

So far so good   

 

[Rocinante]

Unsurprisingly impressive.
My mitsos m-3124 (4 channel api mic pre) is the go to mic pres to our drum kit. A good amount of tunes have been recorded with them.

 

[alexc]

Funny thing is the hum component of 100Hz is all my Motu 828MK2 interface 

The unit itself has a noise floor approaching 10dB better than I showed above, giving it better than 110dB signal-hum margin whilst still giving 0.00099% THD at 1KHz, +9dB.

The Pikatron RUP 219/M traffos are awesome. The big surprise are the Tamradio output traffos - I expected so-so performance but they are really really good  - super ow distortion at very healthy drive levels.

No sign of any psu traffo noise - no issues with grounding etc.

The funny part is now I have to dive into the Motu 828Mk2 and see if I can improve it's psu filtering to drop that 100Hz somewhat.

I really should get myself a newer Motu ...  now that my builds are highlighting the hum and noise on my measurement channel 

 

[alexc]

Here's a snap of the same thing as above but with the BurrBrown DRV134 chip output.

It has a +6dB shift via the debalancing transmitter chip, over the transformer balanced outputs.

It is connected in parallel to the output transformer  ie. the doa output (unbal) is mult to the output transformer and drv134 chip.

Stunning performance, even when hooked up to a dog and pony show like my sum bus. I am including a couple of module versions using the drv134 into my new 'test bed' 

So, at a very healthy output level of +14.3 dBu, the DRV134 chip is showing 0.00063% THD @ 1KHz and 107.3 dB of signal-hum margin  ....  for 13 sources, 1 assigned.

Not too shabby and getting to be the best I've measured from my own stuff. 

.....

Mind you, I spent most time thd profiling tube stuff. Only really now looking at high performance chips and what not.

Fascinating stuff and driving the desire to set up a proper 'test bed' ...  which I'm doing shortly.

Like I said, this is all pushing me to replace my ageing Motu 828Mk2 audio interface. It has developed a persistent 100Hz spike in the analog balanced loopback which is really giving me the heeby-jeebies.

Also, the 2xfs modes have packed it in sadly. SO - pencil another 600 us clams for a new Motu Ultralite USB interface.

I seem to be moving away from firewire - it's great when it works, like with my set of 2408Mk3s  but this particular motu has always been difficult. Several different pci cards, several different hosts, several different notebooks - this 828Mk2 has been painful with almost all of them  But somehow I manage to get my audio measurements!

 

[alexc]

Here's a snap of the apie sum bus, from the drv134 outputs, with no channels assigned. Super low noise floor.
That's the noise floor with +23.2 dB of gain applied and it's barely registering at -106 dBu.

So that's an implied noise floor of -129 dBu    I think my system actually hits the noise floor at around -108 dBu.

 

[alexc]

My other build uses the Lawo Sum Bus with DV-975-4 module.

It is a virtual earth summing config (inverting amp) and required a fair bit of optimisation of the bus impedance, bus slug resistor and also a pair of bus build out resistors.

Lawo specs say use a pair of 6K81 bus build out resistors  for the DV975-4 ...  for 140 sources !

I am using  it with 8 stereo unswitched inputs, 2 stereo switched inputs and a further 8 mono switched inputs.

So that's 18x2 ssumming. 

In addition to the Lawo module outputs, I added a  pair of parallel 'Direct' outputs using (haufe) AKM2800 modules for balancing. 

The AKM2800 'balancing module'  is a shielded can with traffo and ne5532 based driver on board.  Has jumper switchable +6dB level shift and uses up to +/- 17V supplies.

I also added the JLH class a headphone amp I described in another thread.

Not to mention some repurposed Yamaha VU meters with JLM drivers  and a switched Mic input for voice over.

Way cool. Just buttoning the last things down now, so I'll wait before posting a few pics ....

......

Now, the virtual earth Lawo beasty was used with up to 140! sources and is specified with a pair of 6K81 bus build out resistors ...  as I like to call them 

In my case, I have a max number of sources of 18  (+1 for the mic!) so I needed to adjust the balanced bus impedance accordingly ..  to find the best signal-noise ratio ....  for my expected range of inputs.

All that lot connects, balanced,  to the primary of the Lawo module input transformer ..  a Pikatron RUP 452-M.

Now the Lawo DV975-4 doesn't use any fancy feedback over the input transformer ..  like in a zero-field transformer setup ... where feedback is used over the input transformer secondary to reduce the 'voltage across the transformer'  ...  and such.

The Lawo unit is a straight up  ....    input transformer debalancing stage  ....    feeding a virtual earth (opamp + pin),  signal inverting (opamp - pin),  ..  gain stage  ..    with variable Rf gain control.

The input traffo is quite low primary impedance which is  common amongst  haufe/pikatron transformers  I've come across.

Currently, the combo of 4K7 bus resistors, 390R slug and a pair of 0.1% 375R build outs gave me my desired gain structure for 18  sources.

Happy to say that the preliminary results are pretty good

Here's a snap of my Lawo sum bus build ...  on the way down with respect to bus 'lossing'    ...  a pair of 820ohm build out resistors and 390R slug.

 

[alexc]

Of course in my investigation, I continued 'way down' .... 

Here's a snap  ...  195R bus slug resistor and straight into the input transformer (no build out resistors)  ...

THD is around 0.0012% at 1KHz and +9dBu level which is reasonably good at this stage of the optimisation.

My target is to get that down to 0.00075%, which is what I can get on my previously described APIe SUm Bus unit. 

The signal-hum margin of 107dB is pretty good. The 100Hz spike is from the Motu not the build    It's actually an easy way of 'dialling in' the gain structure   

'Look for the Motu 100Hz at -118dBFS'  and you are basically 'there'    - it adds something like 17dB of 100Hz into the picture.

With 0 channels assigned, the noise floor drops all the way down and  becomes more like -129 dBFS or  -108 dBu.

 

[alexc]

It may well be that this is the best ..  balance ... between  'low noise floor' and 'low THD' and 'decently high output level'.

I seem to have forgotton why I decided to go with 375R build outs and 390R  - I have to consult my notes  again 

*Anyway* I'm zeroing in on the best 'place' to situate my gain structure so that I can get the best tradeoff of the above.

--> Oh yes  - I'm waiting on my 'Alps' dual 50K pot for the 'gain adjust' network before I continue looking more.

--> Using a dual 50K log pot (stereo fader controlling both channels) with a pair of 10K lin pots (one for each channel)
--> to give a L-R ganged 'Level control with individual L and R 'Trim' controls.

So far, so good. 

.....

Once I stitch up the basic build, I will do some 'chip-rolling' on the virtual-earth-summing-with-gain opamp from the NE5534AN chip in there at present. 

I'm hoping to measure improved performance using some of the TI (formerly National Semi) 'hif def' opamps 

We shall see!  and then hear!

 

[alexc]

Jumping back to my just-forged 'APIe Sum Bus' build .....

Just done a second 'session'    .... this sum bus is for  lounge ....  'Lounge b' to be precise   

Just a simple rack  ..  with a  Yamaha 'integrated'  hifi amp and Vifa 8" boxes

DAW is an M-Audio 1814 firewire and a HP i5 notebook.

....

Good news! is that it sounds good  .. with music  *and* measures good. 

BUT I have a conundrum. *sigh*

The Yamaha amp has a few inputs - 'DAD' for digital audio device (I guess!)  and 'Aux' and 'Tuner' 

NOW, if I use 'DAD' input  for the daw outputs  ..  the  M-Audio 1814 Bal Outs .....  it sounds really great.

When I route that thru the APIe Sum Bus unit, and line up the levels  ....  it sound good but different.

Like a high emphasis - I have to eq in the Yamaha for treble down, bass up a few notches.

Listening levels are quite low .. like 10Wrms max. Intimate levels, with 2  2-way boxes  quite close - a couple of feet only.

..

HOWEVER, I measure with REW and it confirms my previous measurements - the APIe Sum Bus measures flat to within 0.2dB  20-20KHz, has super low noise and very low THD.

SO I'm a little puzzled - the new box tests good, sounds good but the daw 'direct' to the amp sounds ... better !

To be a little more 'subjective' ...    (although I have lined up levels in my comparisons and is very much apples-to-apples)

The Sum Bus sounds very dynamite on individual instruments and small mixes.  Scary good. 

On mixed music as program material, it seems to need  both  bass boost and treble cut ..  to sound the same  .. punch and max powered.

.....

But the measurements are unequivocal, on 2 completely different DAWs and hosts.

There's no malfunctions - the new box is performing very well.
But the 'direct' source sounds better with the same program material.

I have to apply say ....    +3db 100Hz and -3dB at 3KHz  ..  a typical 'tilt' curve
.
.
As I said, a conundrum 

When I set the Yamaha amp flat eq, and use a VSTi plug-in for 'mastering' it replicates what I'm hearing. I set the curve to a bottom tilt and all is right 

...  SO. ...  something to work thru over the next while ..

I'm pretty happy with the result - one and it will be in daily use for a while yet ...

I'll keep on listening ..  and measuring  ....  till I lose interest!

p

 

[alexc]

And I checked already for source  and loading effects and what not. Very little difference.

I'm pretty sure the DAW isn't having any trouble whatever driving the APIe Sum box. And vice versa. I checked the bus operation already .. it is fully balanced, stable  and is not shorting and such.

I think the Yamaha Amp inputs are at least 20K (unbal). Perhaps I could try some series resistance to increase the impedance those medium Z inputs see.  I've used this amp since forever and have a lot of familiarity with it's 'sound'  = flat, neutral.

The Vifa 8" 2way boxes are decently large and capable of decent power - they tend to be a bit emphasised in the highs - I generally want to drop the tweeters a couple of dB.  But they are solid and have no problem with balanced bass repro.

And at < 10Wrms listening levels, perhaps there is some 'fletcher and munchson' (sic) ype of peterbations of the psycho-audio receptors ..

The new build's Tamradio 600-600 output traffos are quite low Z - perhaps some terminations and what  not might be in order.

But so far, my shunts and loads have measured very small differences - everything measures quite damn fine.

Could just be a case of the 'audio subjectiveness es' . 

 

[ruffrecords]

  20-20KHz, has super low noise and very low THD.

SO I'm a little puzzled - the new box tests good, sounds good but the daw 'direct' to the amp sounds ... better !

It is easy to assume that when you swap in a new piece of gear, the only thing that has changed in the chain is that piece of gear; but it loads the source differently and what it feeds is a load as well. The only way to make a definitive measurement is to measure the whole chain with and without the extra piece.

Cheers

Ian

 

[alexc]

You're quite right of course, but I've done that and my measurements say all is very much as expected.

Could be a phase shift thing possibly - I need to keep on testing.

I'm just checking the speakers are wired up properly and making sure there is no wiring mixup.

Anyway - I don't think it's anything obvious.

My instruments sound like magic through it, small track count mix the same - just the DAW sends with mp3s sounds suckier than with the same DAW direct to the amp.

Maybe this is what people mean when they describe sound as 'mid forward' and what not.  ie. an emphasis of the top end in an apparently flat system making me want to tilt some few dBs to the bottom end.

 

[alexc]

A couple of weeks later ...  with my newly forged API SUm Bus unit in my ..  lounge room .. rack 

It's held up really well to extended daily use - I sorted my issues out and have things really humming along  8)

ALso during this time, I recalibrated my 'measurement channel'  to avoid a problem on one channel ..  100Hz spiking.

......

SO - here's the new baseline for the unit (13 into 2) with 1 channel assigned, with a +4dBu balanced send.

I use REW with 32K fft, Hahn window 50% overlap at a measurement BW  20Hz ..22KHz and 24bit/44.1KHz sampling

;D

 

[alexc]

As the spectra show, it's really good performance - not much more thd, noise or noise floor than the Motu 828Mk2 balanced analog loopback.

THD is 'triple 0' 65 for the APIe SUm Bus vs 'triple 0'  27 for the Motu loopback.
Hum floor for the unit is just 2dB worse than the Motu loopback, at -102.5 dBu

 

[alexc]

And looking at the 'onset of Motu clip' test, again it fares perdy darn good too.

thd is better, at 'triple 0' 45 ..  thanks to the higher output level,  and the hum floor is mostly unchanged 

It can do a LOT more level, but I have to test it with an attenuator to adequately 'map' the wide output swing to the rather limited Motu measurement channel input-before-clip.

 

[alexc]

And that's that 

In use it sounds really fabulous - totally clean and clear .. like crystal. 

I have, at present, a bunch of my DIY units feeding it : a couple of 'vintage channel strips', my indispensible 'fender pre+ la2a' channel and my 'Neotek Channel Strip' which is a series 3 strip racked with an opto compressor and discrete traffo I/O.

I use a bunch of pedals integrated into this rack and all that goes to the sum bus unit, along with some audio interface I/O.

So - it's a wrap : cheap, high performance with super minimal audio path!  Just what the lounge was wanting  8)

 

[alexc]

Flipping now to the other sum bus unit, the one based on the Lawo dv975/4 ...

I've pretty much finished it up now .... it's an 8 mono switched+5 stereo switched + 3 stereo unswitched  inputs -> 18 into 2+2.

Settled on 4K7+4K7 balanced bus, 390ohm slug and a pair of 375ohm 0.1% build out resistors - that gives me my desired 'loss' and the makeup gain structure is well 'placed' with respect to the noise floor and so on.

It's measuring great on 'signal-hum' margin, noise floor, freq resp and phase  BUT my THD is stubbornly stuck at 'double 0' 18  (0.0018%). 

Can't seem to shift it lower : I've tried a whole range of bus resistances etc  but that's the best I can get.

In contrast, my 'api mic pre' style sum bus unit (described above) does around 1/3 of the THD at the same nominal +4dBu levels.

Of course, I can't 'hear' any problem with the higher THD - it sounds really good with a very strong bottom end

.....

I have the feeling that the lawo unit, with it's virtual earth (inverting) gain stage and very low impedance input traffo is a lot more particular about .. stuff .. than the voltage summing (non inverting) gain stage.

The bus arrangements in each are similar, with the exception of the 'build out' resistors in  the Lawo unit - these feed into the Lawo's input transformer after the bus slug resistor across the bus + and - lines.

Lawo specifies 6K81 resistors, which is fine for 140+ sources to the bus, but for a measly 18, it drops the signal way down close to the noise floor and so degrades the 'signal-to-hum' margin really badly when the gain make up is applied.

SO I need to do some more testing    to determine whether it's my bus arrangement or the Lawo module which is the culprit. I've eliminated the obvious possibilities and I don't think it's the summing gain stage.

So for now, it looks like this is as good as it gets with the Lawo unit - great for 140+ sources with very little gain makeup, but not so good for just 18 sources with significant gain makeup, THD wise.  17V

One other thing is that I haven't done measurements on the effect on THD of using a +24V unipolar supply arrangement vs +/-17V bipolar supplies.

I don't think it makes a difference but .. you know .. any little thing .. when you are striving for really low THD.

....

My unit has a second set of 'direct out' connectors, with their own debalancing stage which gives identical results to the Lawo module 'main outs'. I checked performance before and after connecting up that section - no change.

ps      - I removed most of the 'broadcast' crud components from the Lawo, but I still have the 'line amp stage T filter'
pps    - I replaced the 'red can' electros
ppps - I added a coupling cap electro after the sum amp stage for dc blocking on it's output

....

I hope I can improve it, otherwise I'll have to really take a pipe to the Lawo module!

I'm certain I can get better performance from my tried-and-true voltage summing method if I have to  8)

On the upside, the installed JLH headphone amp is working pretty well as are the Haufe/AKM balancing modules

I've still to add the 'talkback mic' preamp module

[pics to come]