Geekslutz DIY Summing Mixer

[Tim Farrant]

Originally Posted by Tim Farrant

For gain you could put a pot between the 2 -ve inputs of the opamps to create gain. Less than x1 gain is how it is drawn. With a pot and limiting resistor between the opamp -ve inputs this becomes the traditional instrumentation input with the transfomer doing the summing/difference passively. Our Elixir mic preamp has the same design idea except it's only 1 input with a lot more gain, but still the siganal is re-balanced by the transformer..

Actually, I'm wrong (too many glasses of wine when I wrote this). The best way to change gain is to vary the feedback resistors around the opamps. You need to change both resistors at the same time and keep them equal in value. You will only be able to achieve a limited range, say +/- 10dB.

[svart]

Hmm I'm not so sure about that..

Your best way to change gain is to use a dual opamp, one side gains a set amount and feeds into a pot. The wiper outputs into the second stage that is configured as a buffer.

That is the best way to do it.

[JohnRoberts]

Without doing a rigorous inspection of this entire thread, in broad strokes varying the feedback resistors of a single gain stage allows you to optimize gain for clipping and/or S/N.

To split this out into two gain stages it is prudent to run the first stage at some nominal lower gain than needed for good headroom and make it up in the second stage. This will add some noise compared to using a single gain stage so not without cost, but that noise increase will be small.

There may be some small benefit from two gain stages in the better control over bus balance if low precision gain pots are used (all pots are low precision) versus fixed precision resistors.

Neither approach is better or worse, just different tradeoffs, as with so much in design.

JR

[svart]

True.

I was thinking in terms of changing effective input impedances over gain range if you use the volume-pot-in-feedback style. Tweaking for S/N and clipping is fine assuming that you aren't tweaking during actual usage.

I suppose in most BJT opamps this isn't a *big* deal but if sensitive circuits designed around Jfet opamps are used, this could be a bad thing. I just don't like changing a circuit's operation over some range as well as the obvious need for tightly matched(read expensive) pots.

My idea was to gain first, then attentuate. In RF land this is required for good S/N to begin with. The secondary buffer was just to have an easily available driver for the following circuit or to drive a bus but is not required by any means.

But after re-reading his description, I initially thought that he was using the pot as the actual feedback resistor. Maybe I'm mistaken and he meant use the pot as the shunt resistor in a voltage divider feedback network? This would be acceptable.

[Tim Farrant]

The important part of my circuit is the transformer because this is doing the input CMR and re-balancing the output. To insert another stage with a pot will bugger this up. Also, to gain first then attenuate will reduce input headroom by the amount of the extra gain. If the feedback resistors are switched using a make before break switch, then it would be possible to vary the gain without disturbing the CMR. However I think only a 10 db range would be sufficent.

[JAY X]

Hi Friends !

Finally i completed the construction of the Geekslutz Summing Mixer, based on the schematics kindly posted by Tim Farrant.

I have to say i learned a lot from this project, but not enough to make it work well.

The pictures i post here represent the full project in detail, and include comments. Also attached is an inaudible sound clip with floor noise or something i can't identify. It is very strange.!!

The very first i turned it on i heard this: Like french fries and Short wave modulation. I did not recorded these sound because i turned off the device, because i thought it could be dangerous.

The sound sample i attached doesn't feature this weird behaviour. I played a track of mine thru the mixer, and recorded to Cubase SX. The result is what you can hear. (inaudible noise). Very strange.!!!

Here are the pictures:



















Up to this point any suggestions , comments, ideas will be much apreciated. We will learn a lot with this project.

Thank you very much

JAY X

[Matt Syson]

Hi
I have not listened to it but if it is making strange noises now and did earlier then it is probably oscillating at high frequencies.
This could well be caused by your construction on the perf board. If the inverting inputs (pin 2 and 7) have too much capacitance to ground then it gives them high gain at high frequencies. Cut (to shorten) the tracks from these 2 pins and you may well find it behaves. I can't see the layout too well but in escence you need to have the MINIMUM amount of copper linked to these pins.
Using adjacent tracks as 'screening' is a BAD idea.
Matt S

[lerone]

Originally Posted by supaheef

I'm in! Me solder quite good, but not make good understand skematik. If you draw it up, specify the bits, and make me a cup of tea while I butcher the circuit board, I'll be fine.

Are we talking fixed pannng (switched like the Folcrom) or rotary pans? Curious as to other people's preferences. I like the idea of panning out of the box, but that will make mix recall-ability a bit of a hassle. MHB may argue that point though, he says his recalls take about an hour...

heath.

One thing that REALLY suffers ITB is panning / stereo field definition, especially in the 'nether' regions.

[Tim Farrant]

Those Sowter transformers look different to the 8403X we use. The tags are all on one side whereas our ones have the primary on one side, and the secondary on the other. Are you sure they are wired up correctly?

Triple check your vero board, it's easy to make mistakes on that stuff.

[JAY X]

Hi Tim,

These sowters have two primaries and one secondary. I will rechek the wiring, although in one test without power up, i could hear some sound thru my powered monitors.

The transformers shield is wired to chassis ground. Or should i wire it to 0v at psu.?

The 0v point at the board, ¿should be taken after the bypass capacitors?

¿Can the two 0v point at board be joined together? (i did it so)


The more i get into this project the more questions arise.

Thank you for your help

JAY X

[JAY X]

Hi Tim,

I checked the veroboard and discovered mistakes on the bypassing capacitors, which i wired reverse.

Today i will correct them, and hear what happens.

JAY X

[Matt Syson]

Hi
A comment or two about Tim's design:
You don't need both of the 100uF bipolar output caps (with their poly bypasses) as the 2 outputs of the IC are wired in series through the transformer primary. In effect you are using a 50uF cap in series with the signal which is not that good for LF. It will of course work, especially if driving into 10K Ohms or more but a larger value would be better. If you have built one up, please try it!
Secondly the input common mode being supported by the transformer, this is only partially true since even with one 'input' leg disconnected the signal will only fall 6dB maximum.
In all practical instances this is not relevent since signal levels are high and impedances low.
I don't have the time to 'knock it up' and test it but I would suggest the equivalent of 1000 uF Bipolar.
Assuming it is powered by a clean supply you could fit a 620K resistor from one supply rail to ONE of the inverting inputs which will pull the output about half a volt in the OPPOSITE polarity so that you could then use a 1,000uF or greater polarised cap (with it's bypass). This will affect the maximum headroom VERY slightly, much less than 1dB, but depending on the exact IC used this will not be significant if it is 'biassed' away from the polarity that clips first.
Matt S

[JAY X]

Hi Matt!

Today i tested the circuit and finnaly i got sound. But it sounded like a hi pass filter, with practically no low end. Also the highs, are a bit harsh, like a carbon mic. Asap i will post a sound file.

(Edit) : Maybe the reason it sounds like a hi pass filter is the 47R resistor i used instead of the 100R in the summing node. Iwill change those, and hear what happens.

The main change i did to the lay out is to aproach the two bypass capacitor as much to the OP275 opamp. Also i shortened the legs of the 10pf ceramic capacitors.

I will try to understand your post/ideas. Meanwhile this is the original schematic by Tim Farrant.

[Matt Syson]

Hi
If it sounds that bad you have a 'major' but not necessarily expensive mistake.
Are the transformers wired correctly?
Try it without the transformer.
Matt S

[JAY X]

Hi Matt !

I changed the 47 k summing node resistors for the 100k ones, and the mid-highs improved much. But still no lows.

It seems that the transformers perform the CMR, and after Tim comments are important in the design. Although is possible not to use them. In any case i will re-check the wiring first, as you suggested.

this are two samples of the sound.
The first one as it comes out from the summing mixer (very low)

The second one Compressed a bit for more volume gain.

also i put here the Cubase , Voxengo Span screen captures.





JAY X

[Matt Syson]

Hi
The summing node resistors (connected to pins 2 and 6) which go to the 22K Ohm input resistors are 100 OHMS NOT Kilo Ohms. These are fitted to try to reduce the possibility of oscillation due to capacitive wiring around the mix resistors and not strictly necessary. Others either omit them, or use a choke / resistor combination. If you have fitted 47 K or 100 K then it will be very quiet as it will have a 'loss of about 5 times (about 14 dB) plus another loss due to the 22K resistor network.
If it is still sounding like a filter though, recheck the transformer wiring.
Matt S

[ulysses]

You've inadvertently implemented a high-pass filter someplace. That could be the result of a coupling capacitor that's either much smaller or much more heavily loaded than it's intended to be, or (more likely) you've miswired the transformer so that you aren't getting inductive coupling across the transformer at all, and the HF audio you're hearing is do only to coupling across the parasitic capacitance of the transformer windings.

Any easy and very explicable version of that scenario would be if you've got everything built, wired, and working perfectly; but you're feeding the output into an unbalanced load that takes its audio between pins 1 and 2 of the XLR connector, leaving pin 3 floating.

[JAY X]

Hi Matt, Ulysses,

You were right. The output wires entering the transformers primary were reverse connected. Now i can hear the bass frequencies.

But now i have another problem. It sounds like little cheap speakers.
The highs sounded better when it was wrong wired.(Hi pass filterF)

asap i will put a sound file and window capture.

I will re-chek the caps value i used.

Maybe the problem resides in the output capacitors. Tim recommended
100uf bypolar caps, but not Electrolytic. The only alternative i found, is to use tantalum capacitors.

¿what do you think?

JAY X

[Matt Syson]

Hi
I think you still have the transformer wired incorrectly or the load on the output is too great (low resistance) for the output caps you have used (see earlier). If you check pin 1 and 7 of the op amp and they are less than say 50mV away from ground then you don't actually need the output capacitors at all.. Are all your resistor values correct? the output resistors (from pin 1 and 7) should be 75 Ohms.
Matt S

[Jim Williams]

I noticed that 10 pf across 22k ohms sets the bandwidth very high, so high that stray capacitance from the sum resistors could cause ringing or even oscillations with the sum opamp. I would increase that cap to either 47 or 68 pf to maintain stability. That would also eliminate the need for the 100 ohm buffer resistors which will lower crosstalk. You can fine trim it by feeding a 2 k square wave and adjusting that feedback cap for best symetry.

If you want lower noise and THD, sub the OP-275 with a National LME49720NA or a AD8599 on an adaptor. Those are about 20 db cleaner. They are also precision devices that can be used without the 100 uf coupling caps.

Jim Williams
Audio Upgrades

[JAY X]

Hi Matt, Jim,

The last test i did was with the 100 ohm summing resistors.

Today i will try with the 47 ohm resistors again. After reading all posts, it seems that Tim agreed with 47 ohm value.

I will try without the transformers. ¿should i bypass the output condenser/resistor ? ¿what is the purpose of these?

The capacitors values i used in this circuit (2 channels) are:

4 units polyester 100 ohm 63v
4 units polyester 1uf 125v
2 units ceramic 470pf
4 units electrolytic 100uf 63v
4 units ceramic 10pf

Resistors: 22ohm, 47 ohm, 75 ohm, and 100ohm (to be changed for 47 ohm)

I hope tonight or tomorrow i will post the results.

JAY X

[Matt Syson]

Hi
The output capacitors prevent the DC offset from the op amp from magnetising the transformer core. The offset should however only be a matter of millivolts so is not normally a problem and as Jim pointed out, chips with even lower offset don't need them at all.
I said that you only need ONE cap because the transformer is wired from one out to another, this is only really precautionary and in any case should be much larger than 100uF (or as you have 2 in series yopu only have 50uF).
As jim pointed out 10pF matched with 22K ohms is rather a high bandwidth so these should be increased and may well be the root of the problem with YOUR layout. I presume that Tim breadboarded it in the past but his physical layout will have been different.
I presume from your list you meant 22 K Ohms not 22 Ohms? That would mess it up big time!
Matt S

[Tim Farrant]

Matt, because the output of the opamps is 180 degrees out of phase (assuming the inputs are) then the 100uF capacitors are not really in series. Imagine if the primary of the output TX had a grounded centre tap, this would be the same thing as the circuit stands. The caps should remain in the case an opamp fails and goes to +15V DC, which the TX would not like. If desired, one trick is to place a low value resistor across the output capacitors making the connection a little more DC coupled. I'd say 220ohms would still provide protection for the transformer.

I do not think the 100uF need to be any bigger. The calculated -3dB point into a 10k load would currently be 0.318Hz !!

Tim.

[ulysses]

Originally Posted by Tim Farrant

Matt, because the output of the opamps is 180 degrees out of phase (assuming the inputs are) then the 100uF capacitors are not really in series. Imagine if the primary of the output TX had a grounded centre tap, this would be the same thing as the circuit stands.

I don't think so. The transformer does not have a grounded center tap on the primary, which means the only audio circuit path from one amp's output is through the other amp's output. When one amp is sourcing current, the other is sinking, and vice versa. Therefore, the output caps ARE in series, and their effective capacitance is halved.

[Matt Syson]

Hi
The caps are DEFINATELY in series, disconnect one and there is no signal.
I proposed one capacitor which would provide DC protection for the transformer in the event of IC failure. I also suggested a way of polarising it, rather than using a bipolar cap which is in effect 2 caps in series internally.
While 100 uF (actually 50uF) might have a -3dB point at 0.318 Hz (calculation not checked) there are a couple of other issues.
There is the possibility of in band resonance with such a low capacitance, and the fact that the transformer requires power to operate, particularly at low frequencies, and the load even if 10K resistive, will be coupled by cables with significant capacitance, both effects increasing the load on the output.
Matt S

[Matt Syson]

Hi
OK I'll concede that 50uF is adequate if driving 10 K Ohms but what may be of greater interest is the HF behaviour with different cables.
This is actually transformer related rather than the coupling caps.
With 1 metre of twin screen cable on the output there is a 11 dB 'peak' at about 160KHz, with a small rise on the nominal 20K area.
If you increase the cable to 14 Metres then you get a 7dB peak at about 85 KHz with a smaller rise at 20KHz.
These peaks are much better controlled if the transformer is loaded with about 2 - 3 K Ohms resistive.
If the output is loaded with 600 Ohms the response is flatter '20 - 20K' but then 50uF is not sufficient at the LF end.
Since the sending impedance is approximately 150 Ohms (2 X 75) the capacitor values have a reduced effect.
Thus the general excitement that are transformers, juggling the send impedance, load impedance and load capacitance.
Matt S

[Jim Williams]

The new BurrBrown OPA2211A is another great sub for the OP-275. It's a 27 v/us slew rate, very low noise, 1.1 nv/hz/sq and extremely low offset at 125 uv. S0-8 only, they need a dip adaptor. It also can be direct coupled. These are also rail to rail devices for greater headroom for a given supply. Run them at + - 18 volts and they will output more than other devices. I wouldn't worry about needing those blocking caps for this part. Even a blown opamp won't damage a quality transformer, transformers are used for DC all the time as well as AC. Users could disconnect the output transformer and just run the opamp signal out differentially.
That way you can place the transformer in a box and use it selectivly or add different models.

Order free samples at
Analog, Embedded Processing, Semiconductor Company, Texas Instruments

Jim Williams
Audio Upgrades

[Matt Syson]

Hi
Aha!
Due to the excessive sun the last few days I had the transformer primary and secondaries swapped over (in was out).
Now it gets more interesting.
With a 10K load we have a 4 dB peak at about 120KHz. Now the interesting bit.
Without the capacitors (so the source impedance is 150 Ohms resistive) the response at LF is showing a slight fall at 10Hz (about 0.25dB). This is expected.
With 100uF we get a 1dB RISE at 10Hz.
With 50uF (as proposed by Tim 2 X 100uF in series) we get a rise of nearly 2dB at 10Hz.
The transformer is resonating.
Matt S

[Matt Syson]

Hi
With 10K load and replacing the electrolytics with one 6u8 polyester cap you get a lovely 6dB peak at about 15 hz.
Matt S

[JAY X]

Hi All,

Today i tested the summing mixer without the transformers: Sounds Very Good, With low end, very detailed sound. (no power up!!)

I tested with my old Roland CR8000, which has 6 separate and quite noisy outputs, and thru the mixer sounds very clear.

After power up, sound is good. I notice an adrift to right channel, but maybe this is because i have not yet labelled the input jacks.

The sound thru transformers (before today) sounded harsh and resonating. I will read attentively the last posts this week-end.

I have the intention to go further with this project, and add a buffered input balanced line receiver :Analog devices op-amp applications handbook page 6.40., with a transformer line driver, page 6-72.

Or simply put 2 ssl9k preamps with their balanced line drivers.


JAY X