Using input & interstage transformers in a balanced passive mix bus design

Greetings; this is my first post here, as I've been studying the forum for several months to make sure these questions haven't already been answered.  This forum is a very cool resource, and I hope my newb questions and (hopefully the answers) will be useful to other forum members.

I'm working (in stages) on an analog summing soundstage for a DAW system.
Before boshing something together this time, I thought I'd ask for some advice.

I've seen references here to using transformers for summing instead of resistors, and how transformers can improve a passive summing design, but I haven't been able to find pointers on this approach.

1. From other posts, it seems prudent to have a cap on the input signal to filter DC before hitting the input xfmr; what value/type would be appropriate? and would it be needed for both phases of the signal?

2. The lundahl xfmrs require(request) a 8k resistor and 1.2nF cap to provide the quoted frequency response; how would this zobel network interact with the mix resistors (if any) and calculating their values? Each buss would have 4 inputs.

3. How would the bus xfmr affect the mix bus impedance, and how can this be tuned for optimal source/bus/mixamp loading?

4. Perhaps the xfmrs can be used/wired for additional gain?

5. PRR has made reference to 'mixing coils' and posted a diagram in which several input xfmrs had their secondaries wired in series for summing; are there any advantages to such an approach?

I'm working on a schematic now, and look forward to any informed comments, insights, questions and suggestions.

Many thanks for the consideration and patience with a fledgling DIYer.

cheers

for summing, I'd probably prefer to run the input transformer in "zero-field" mode - see e.g. the NTP ZFT module and/or their 179-160 comp input section

Jakob E.

thanks for the ZFT tip, researching that approach now...

quantyk said:

1. From other posts, it seems prudent to have a cap on the input signal to filter DC before hitting the input TF; what value/type would be appropriate? and would it be needed for both phases of the signal?

Click to expand...

The signals going into your "mixbus" are supposed to be DC-free; if it's not the case, it's a malfunction or a very, very poor design. You should not add components that are not needed. In case of gross malfunction, the bus-injection resistors will provide ample protection.

2. The lundahl TFs require(request) a 8k resistor and 1pf cap to provide the quoted frequency response; how would this interact with the mix resistors and calculating their values? Each buss would have 4 inputs.

Click to expand...

I suspect this is the load the secondary likes to see, when driven with the recommended source impedance. So you'll have to work out the mix resistors value and the shunt resistor value (if needed) and provide the requested load impedance to the secondary. But in fact, as always, there is not one single solution to an equation with so many variables, so you may find that final tweaking is your answer, and simulation software is your friend.

3. How will the bridging TF affect the mix buss impedance, and how can this be tuned for optimal source/buss/mixamp loading?

Click to expand...

The xfmr in itself does not really affect the bus impedance, within certain limits, but the reflected impedance (seen from the primary side) does affect it. So, if you load the secondary with the recommended 8k/1pF (which seems incredibly low to me - 1pF is the capacitance of one inch of wire close to another), the primary will "show" about 5kohms of reflected impedance.
Ratio=1.3
Impedance ratio is squared turns ratio = 1.69
Primary Z with secondary loaded by 8k is 8k/1.69 = 4.8k, to which you have to add DCR's.

It seems one would want the input to be Hi-Z, and the mix buss to be lo-Z

Click to expand...

The input should be what's necessary for correct energy transfer: too much attenuation => not enough signal => poor S/N
Not enough attenuation => cross talk

4. Perhaps the TFs can be used/wired for additional gain?

Click to expand...

It has already about 2dB gain. I don't think you need much gain for a 4 input thingy.

5. PRR has made reference to 'mixing coils' and posted a diagram in which several input TF had their secondaries wired in series for summing; are there any advantages to such an approach?

Click to expand...

Theoretical advantages because mixing is done without voltage loss, but many practical disadvantages, starting with the cost of custom-built multi-winding audio-quality xfmrs, but also cross-talk and HF response problems due to stray capacitance and leakage inductance and lack of flexibility.
I guess PRR mentioned it as an illustration of his "let's have no stone unturned" approach...

quantyk said:

5. PRR has made reference to 'mixing coils' and posted a diagram in which several input TF had their secondaries wired in series for summing; are there any advantages to such an approach?

I'm working on a schematic now, and look forward to any informed comments, insights, questions and suggestions.

Many thanks for the consideration and patience with a fledgling DIYer.

cheers,

AK

Click to expand...

Stacking transformer secondaries together should deliver lossless combining. The largest negative is cost, 2 line level transformers per channel on a stereo bus. The output impedance of all the secondaries in series will increase, but it should be no problem to buffer this sum output.

I have never seen this done, and for a modest number of inputs, if you have a pile of transformers laying around unused, why not? That said combining only modest numbers of channels does not normally lead to significant signal degradation using much cheaper approaches.

JR

An alternative for the secondary stacking is a summing transformer.

I have been discussing with Brian Sowter the possibilities for creating a 8-to-1 mixing transformer.
We had a lengthy discussion on the technical part.

This is what he came up with for a 4652 based transformer that is Quadrafilar wound:

I am looking at a two chamber bobbin with 4 primary and one secondary winding in each chamber. The secondaries would be paralleled.  With a 3 section secondary this is 7 coils on each side.  This arrangement should give a good bandwidth and minimal capacitive coupling between primary windings.  We have to be a bit concerned about manufacturing yield with so many coils of thin wire so the optimum so there is a trade off between resistance and cost.

We can propose two options for an 8 channel transformer:

Using a small core where the total cost would be £80.14 UKPounds we can get a loading of 200 Ohms i.e. 100 mA

Using a large core where the total cost would be 114.94 UK Pounds we can get a loading of 600 Ohms i.e. 34 mA

Prices are without Mumetal cans which would be £~10 and £30 respectively.


Best regards

Brian Sowter

Click to expand...

As you can see this can be attractive from a cost perspective.
Attached is the global setup.
Important detail is that the channel-line-buffers must stay connected to the primary windings and have a low output impedance.

Theo

What are the claimed advantages of such an arrangement over resistor mixing?

abbey road d enfer said:

What are the claimed advantages of such an arrangement over resistor mixing?

Click to expand...

complete isolation, balanced summing, character.

grT

Balijon said:

An alternative for the secondary stacking is a summing transformer.
====

As you can see this can be attractive from a cost perspective.
Attached is the global setup.
Important detail is that the channel-line-buffers must stay connected to the primary windings and have a low output impedance.

Theo

Click to expand...

Is that a typo? That price doesn't sound attractive to me to accomplish a modest sized sum bus.

The advantage on paper of stacking secondaries is that you don't have the insertion loss/make up gain of a passive sum, or the noise gain of an active sum.

However this make up gain is trivial compared to the typical mic preamp that most audio paths manage to tolerate.

JR

PS I am not enough of a transformer guy to opine on this but I wonder if there is any mutual coupling between all those primaries on a common magnetic core. With the separate cores I suspect there will be some reflected loading between the stacked secondaries if the whole string is heavily loaded overall, but lightly loaded and buffered they shouldn't mess with each other too much.

JohnRoberts said:

Is that a typo? That price doesn't sound attractive to me to accomplish a modest sized sum bus.

Click to expand...

for 8 channels of summing in one transformer = 10,-/15,- per channel versus 50,- per channel in stacked mode, I think so....
One of the main goals was to come up with a cost competitive alternative for the expensive stacked transformer summing...  

The advantage on paper of stacking secondaries is that you don't have the insertion loss/make up gain of a passive sum, or the noise gain of an active sum.
However this make up gain is trivial compared to the typical mic preamp that most audio paths manage to tolerate.

Click to expand...

if completely passive on the secondary (no make-up amp), the insertion-loss is less than 4db into 600ohm according to Brian.

PS I am not enough of a transformer guy to opine on this but I wonder if there is any mutual coupling between all those primaries on a common magnetic core. With the separate cores I suspect there will be some reflected loading between the stacked secondaries if the whole string is heavily loaded overall, but lightly loaded and buffered they shouldn't mess with each other too much.

Click to expand...

As stated, all primaries must be (and stay) connected to a low output impedance line-driver (preferably balanced out), this will take care of all 'back-loading' effects. Brian did a lot of calculations on this and optimized the winding strategy for this.
Full loading on +28dbu (20V) has been the reference design requirement.

Theo

PRR has made reference to 'mixing coils' and posted a diagram in which several input TF had their secondaries wired in series for summing; are there any advantages to such an approach

Click to expand...

It's not totally lossless.  You can use some of the more common repeat coils to get 4 in, 1 out, so you are limited in that regard.  There are a couple of older threads with some discussion about custom consoles of the past that used coil mixing.  Search under NYDave, tube console.  Audio Cyclopedia shows some actual console schematics using coil mixing.


Some basics of coil mixing are covered here:

http://www.paleoelectronics.com/RDH4/CHAPTR18.PDF

Yup back in the old days tubes and low noise gain was very dear (expensive), and transformers not so much... Today that calculus is reversed.

JR

Balijon said:

abbey road d enfer said:

What are the claimed advantages of such an arrangement over resistor mixing?

Click to expand...

complete isolation,

Click to expand...

What isolation? This thing relies on buffers for correct operation. Mixing is obtained by brute force, based on resistive losses in the windings.

balanced summing,

Click to expand...

resistor balanced summing is pretty common...

character.

Click to expand...

I doubt there is much "character" in a transformer operating at mW level. I agree that something must happen in your proposed schematic, though. Not necessarily very nice, being stressful for the buffers and the windings.

Balijon said:

if completely passive on the secondary (no make-up amp), the insertion-loss is less than 4db into 600ohm according to Brian.

Click to expand...

There's something very wrong here; due to the sources being connected in parallels AC-wise (all primaries are almost 100% coupled), the insertion loss is equal to the number of sources: 4 sources=>12dB, 8 sources=>18dB... without secondary load. If the secondary is loaded, the additional loss will be dependant on the number of sources too.

  As stated, all primaries must be (and stay) connected to a low output impedance line-driver (preferably balanced out), this will take care of all 'back-loading' effects. 

Click to expand...

  What I don't like here is that each buffer sees the reflected impedance of the other buffers augmented by the DCR of the windings. Unless Sowter deliberately makes the windings with very thin wire to increase the DCR, the buffers will be loaded by a few tens of ohms, which puts a lot of stress on them, unless they are small power amps.

abbey road d enfer said:

Balijon said:

if completely passive on the secondary (no make-up amp), the insertion-loss is less than 4db into 600ohm according to Brian.

Click to expand...

There's something very wrong here; due to the sources being connected in parallels AC-wise (all primaries are almost 100% coupled), the insertion loss is equal to the number of sources: 4 sources=>12dB, 8 sources=>18dB... without secondary load. If the secondary is loaded, the additional loss will be dependant on the number of sources too.

Click to expand...

I discussed this with Brian over the phone, because this was a concern to me too. Brian stated that with the way he wound the spools, he would expect to stay below 4db loss any channel to secondary.

  As stated, all primaries must be (and stay) connected to a low output impedance line-driver (preferably balanced out), this will take care of all 'back-loading' effects. 

Click to expand...

  What I don't like here is that each buffer sees the reflected impedance of the other buffers augmented by the DCR of the windings. Unless Sowter deliberately makes the windings with very thin wire to increase the DCR, the buffers will be loaded by a few tens of ohms, which puts a lot of stress on them, unless they are small power amps.

Click to expand...

We had a lengthy discussion on this. Brian is planning for using thin wire (this was also his concern on making this and potential manufacturing yield).
Here is a snip from one of our emails:

I made a mistake in the current!
The load on one channel assuming no signal on the other 7 is made up of the
dcr of one winding in series with the dcr of the other 7 in parallel,  So if
the dcr of one winding is 800 Ohms  It will be 800 plus 114 Ohms = 914 Ohms
so the current at + 28 dBu (20V) will be 20/914= 20 mA
Sorry to scare you....

Click to expand...

By the way I also discussed the concept for higher count summing than 8 channels.
Stacking and cascading would be both an option.

Theo

Sowter's one-core summing transformer requires that sources (or windngs) have significant output impedance. This is not hard to arrange. It does give galvanic isolation, which avoids garbage in a shared ground.

A different concept uses a separate core for each source. Secondaries are stacked. Just like stacked batteries, the voltages add-up. Either 100% voltage transfer or 100% power transfer is possible, compared with 12% voltage or <2% power for one input of an 8-in resisty mixer.

Major drawbacks. Core per channel works only if you can pad your profit with lots of extra irons. Stray capacitance makes this a wacky network with strange top-end response problems.

As a practical sub-set: eight 600r transformer-output channels may be stacked for a ~~5K output impedance and perhaps tolerable top-end response. Eight 75r outs can be stacked directly to 600r line level, no makeup amp,

Thanks PRR,

I now understand some confusion I may have caused on the secondary winding and related circuit.
The design is intended to have a secondary buffer-amp, exactly as it is in the drawing (with a make-up gain of +/-4db).
The secondary resistance is high (if I remember correctly 10k), the loading of the primary line-drivers was not to exceed 600ohm.

The starting-point of the discussion with Brian was to find a more cost attractive alternative for stacked-transformer-summing which is expensive, takes a lot of real-estate and weight. I like the secondary isolation that both concepts provide.
The intent was not to beat (balanced) resistor based virtual-earth summing in cost...

I will gladly relay any questions and concerns to Brian.

grT

i believe that once you get the core going with the magnetizing current, the insertion loss would be the same regardless of  how many input windings there are,

So did a 50-60, can be expanded up to 8 channels.

get a chance to use it yet? how is the sound?

like smooth creamery butter,  with all that fine iron would you expect anything else?