Just how bad is it? Audio mixer with headphone amp.

[paulca]

And... on the inverting amp.

I put 47R resistors on all the outputs.  So to aim for a gain of 1 of the inverting mix amp I put a 47R on the feedback.  I gather from what you said this probably isn't going to work well.

Do I need resistors between the amps at all?  If I put the OUT of an opamp directly into the IN- of the inverting amp and hard link it to unity will it still function as a unity gain summer or will the virtual ground on the - input kill the upstream opamps output?

[Audioguru]

If the output of the opamp must have a maximum swing of 7V peak then the output current in a 47 ohm feedback resistor needs 7V/47 ohms= 149mA peak! But most opamps can drive 2000 ohms minimum for a current of 3.5mA including the current in the load. Why the extremely low value of 47 ohms? Why not 47000 ohms instead?

Please learn what an inverting opamp is. It has a series resistor to its - input and another resistor fronm its - input to its output. Its voltage gin is the ratio of the two resistors.If one opamp drives the - input of a second opamp that has a negative feedback resistor then the output of the first opamp is almost shorted and the second opamp has a voltage gain of a few hudred thousand so it will be clipping the signal like crazy. Each input needs its own series input resistor whether it comes from a signal source or from the output of another opamp. Yes, the - input of an inverting opamp is a virtual ground, the series input resistor is very important.

[paulca]

I found myself a copy of Small Signal Audio Design.

Interestingly, Chapter 1, page 6, Gain Structures, describes exactly what I am aiming to do.

1. Don’t amplify then attenuate.
2. Don’t attenuate then amplify.
3. The signal should be raised to the nominal internal level as soon as possible to minimize
contamination with circuit noise.

It then goes on to suggest that 1 and 2 are compromises when you have an attenuating gain control, you have to do one or the other.  1 risks clipping, 2 risks signal to noise loses.

As my input stages will not be getting anywhere close to the amps VPP peak of 20V (a gain of about 30-40) I have plenty of headroom to lift it considerably, gain of 10 or 20.  The idea is to lift the level once (rule 3) and then only buffer it or attenuate it for output level.  Not this attenuate, gain, attenuate, gain series stuff, that just adds noise.

It also mentions that active gain controls should be used ideally, although I have not read the chapter on the pros and cons of their use/design yet.  I'm sure there are dragons there.

[paulca]

Well, had another session with this.  Decided to try something different and implemented one of these:

Source: Small Signal Audio Design

The positives are that it's the most stable outputs I have seen on the scope.  Plenty of gain, although when running my headphones the gain fell to 1, which has me a little perplexed.

The downsides are it's an opamp killer on a breadboard.  Two 5532s died.  The first one was most likely because I forgot the GND reference on the second stage amp.  That chip now does nothing but howl with a 13.5V DC offset.

The second chip died when I had a lose connection on the breadboard.  I was (probably foolishly) running the output into my headphones (34 Ohm IIRC), so a total output load of only 134 Ohms.  I had fissling and was poking various connections to find it then WWOOOOOOOOOO!  The opamp broke into full song howling.

If I swap in another 5532 directly everything comes back to working, I conclude both chips are dead.  They seem to work in isolation as voltage followers, but as soon as they go back into this circuit the each behave incorrectly.

I'm not sure what it is about the circuit that makes it easy to kill opamps, maybe the low value resistors and 5532s are not exactly expensive, but I'd rather not have to order more. 

I was using my new PSU an XP Power +-15V 200mA.  It seemed to be pulling around 40mA idle, 66mA running the circuit which was a single 5532 chip + an LED with a 2.2K limiter across the 30 volt rails.  This does not sound like the efficiency in the datasheet of 84%.

Anyway.  I'm a bit concerned about proceeding to create two inputs and a summer from this if I have other issues.

I'll retest the PSU in isolation to make sure it is functioning and I haven't wired something wrong.  40mA to power an LED which should be drawing around 12mA.

The small signal audio design book claims the resistors are small value to keep down the noise, if I take them up a factor of 10 would that should make the circuit safer against breadboard glitches?

[paulca]

Actually checking the datasheet for the XP PSU, it states 25mA idle current.  I believe the 40mA I was measuring included the LED, so 25+12mA = 37mA and not far off the measured 40mA.

That leaves the rest of the circuit pulling 26mA (20mA account from the PSU efficiency).  The 5532 has a max input current of 10mA, but it's hard for me to work out how much is dropping on the various resistors and not going into the opamp.

[Audioguru]

Why does a volume control need two opamps??

The high stray capacitance of your solderless breadboard probably causes the 10MHz opamps to oscillate and draw extra current.
The linear volume control is an on-off switch that does not match the logarithmic sensitivity of our hearing to loudness.

[paulca]

I believe the Baxandall configuration causes the volume control to be an approximation of logarithmic with a steep drop at lower settings.  It forms a nearly straight line on a log graph.



I'm reading Small Signal Audio Design book and he does not favour log pots very well and while he mentions them, points out their issues, he then avoids using them altogether.   There is a large section on using circuitry around the pot to emulate the log law.  One of which is the Baxandall design. The main problems with them are their tolerances are poor, usually 20% and for stereo pairs they are never matched very well resulting in perceptible balance offsets.

Another reference for this configuration, including a HF filter, that uses 4 op amps is here:
http://www.ti.com/lit/ug/tidu034/tidu034.pdf

However using this configuration with 2 opamps and no filter stage I get very little noise on the scope.  It's down around 70mV which I get anyway no matter what I connect the probes to.

[paulca]

If I get a working module in this prototyping, is it worth perf boarding or strip boarding it and then hooking them together rather than using the breadboard in one mass of mess?

I find it's too easy to nudge a wire or a resistor and upset things... or fry an amp apparently.  I'm fairly sure they are dead.  Definately the one that howls at a 13.5V DC offset.

[Audioguru]

I make all my prototypes soldered together on a carefully planned compact stripboard layout. If a part needs to be changed then unsoldering and replacing it is simple. No stray capacitance to cause oscillation and no messy antenna wires to pickup interference. I cut each copper strip with a drill bit so that each strip is used for a few functions to be compact and a few short jumper wires make the remaining connections. Many of my stripboard prototypes looked good enough to be sold as the finished product when only one at a couple were custom designed by me. 

[paulca]

Thanks.  I think I might made a prototype of a stereo baxandall.  If I create that I can put it inline with my current system and test the extra gain on the pre-amp works as expected.  Then I just need to replicate it twice.

The only thing I don't have is a dual pot.   I suppose that's another RS order then   I can hack in 2 trim pots for now they should match the footprint.

I am going to try and use KiCad for planning the strip board.  I'm also low on strip board, might have a bit big enough, might not.  After that I have IC footprint board (groups of 3) that I could use and if that fails it would be dual sided pad board which works, but is messy creating tracks with fuse wire grid style.

[paulca]

Well the prototype board worked out okay.  In so much that it was laid out okay and soldering went well.  I only did one channel.

On testing it however audio plays through it fine, but it's attenuated, the gain pot does nothing and the opamp pulls 30mA and gets hot after a while.  Sometimes is a-miss somewhere.

I've had 2 beers, so I'm not starting to scope it tonight.  Tomorrow maybe, starting at that gain pot to see whats wrong there.

[paulca]

Got it.  The outer feedback line is coming off the - input, not the output.  That will impress it.  I'll fix it in the morning.

[paulca]

So progress!  At last.  Having fixed the fact that pins 6 and 7 where completely backwards on my perfboard, it now works perfectly.  Still gets a bit warm and still pulls too much current, but at least it's only about 13mA this time for a single channel.

I also made progress with the design, (below).

I'm still a bit baffled with impedance.  For example the LineInputStages all have 100R's on the outputs, which is then followed by a 22K, so that makes it 22.1K and will alter the summing amp gain a little, but more importantly when I split the L/R_MIX to the AUX buffer amps do I need separate resistors on each leg of the split, or is it fine the way I have it?

I haven't done the headphone amp yet and somehow I have to invert the signal in it, due to using another of the Baxandall blocks for the master volume, which inverts phase.  I might change this to make a slightly different master volume block based on the same idea.

A few design issues remain.  1.  The active volume controls do not reduce to zero, I believe due to there being a path through the active gain pot to the output, so the original signal leaks through.  I could be wrong.  This is not a worry on the inputs, but is an issue on the master volume.  I would like "OFF" or "Infinite".  Expensive pots or a proper fader will provide this, but probably ££££.

For the headphone amp, Small Signal Audio Design book actually recommends... wait for it... an array of 14!!!  NE5532s in parallel to drive headphones.  14!  I don't think I need that many, it sounds fine with 2 driving 34Ohm headphones.  I might go to 3, 4 or maybe look at a better op amp for the output.

[Audioguru]

Why do you have so many coupling capacitors? Most of your opamps have a gain of only 1 so their output will have the tiny maximum DC output offset of 0.004V without a coupling capacitor and resistor load to ground. Who cares about an offset of only 0.004V max? The offset voltage might even be nothing.

Why do your coupling capacitors have the huge value of 100uF? Feeding 22k ohms then they pass earthquake frequencies down to 0.07Hz! That is one cycle each 14 seconds. Why not use 1uF then the audio -3dB cutoff frequency is 7Hz then will be flat down to the 20Hz lowest frequency that we can hear. The formula to calculate a coupling capacitor value is very important and is simple.

Can you hear a difference between true phase and reversed phase? Does every radio in your house produce true phase? Do you want to add many other mixers to this mixer?

An audiophool would use 14 NE5532 opamps in parallel to drive headphones through solid gold Monster cables. An audiophile uses only one fairly high power NE5532 opamp through ordinary copper cables to drive 600 ohm Sennheiser headphones.

[paulca]

I'm borrowing building blocks from that book!  Now, granted the book is basically his experience in designing real pro-audio 16/24/32 channel mixing consoles, but... a 100uF cap and 22K block is about 10p, I might as well.  Removing them would then require I understand much more possible have to redo the maths of the various parts.

On the coupling caps, there is a reason for the larger uF caps after the summing mixer.  I only scan read it but basically the summing portion has a LF roll off, so the coupling caps should be kept large to prevent further LF damage.

On the gain.  The Baxandall stages have a gain of around -70db to +17db.  The mix sum amp, I believe has a gain of 2.  The Aux buffer is driving un-known external inputs, so I figured decoupling it would be wise.

On the 5532 headphone amp, re-reading it, he shows THD graphs for a single 5532 driving 5Vrms into 200Ohms and how adding a second parallel 5532 drops that THD to almost nothing.   4 can drive 100Ohms with virtually no distortion.  12 are used for 25Ohms.

I doubt I would notice the distortion talked about with 4, driving a 34Ohm load.  So I'll maybe aim there.

On phase.. you are right.  Unless I ever intend to mix the output with another, phase will not matter.

[BrianHG]

Why do you have so many coupling capacitors? Most of your opamps have a gain of only 1 so their output will have the tiny maximum DC output offset of 0.004V without a coupling capacitor and resistor load to ground. Who cares about an offset of only 0.004V max? The offset voltage might even be nothing.

Why do your coupling capacitors have the huge value of 100uF? Feeding 22k ohms then they pass earthquake frequencies down to 0.07Hz! That is one cycle each 14 seconds. Why not use 1uF then the audio -3dB cutoff frequency is 7Hz then will be flat down to the 20Hz lowest frequency that we can hear. The formula to calculate a coupling capacitor value is very important and is simple.

Can you hear a difference between true phase and reversed phase? Does every radio in your house produce true phase? Do you want to add many other mixers to this mixer?

An audiophool would use 14 NE5532 opamps in parallel to drive headphones through solid gold Monster cables. An audiophile uses only one fairly high power NE5532 opamp through ordinary copper cables to drive 600 ohm Sennheiser headphones.

Now I know I like low frequencies, but I must agree with you Audioguru.  Paulca has gone off the rails in all respects when his first schematic would have worked fine beyond his ability to hear have any better.

However, Paulca may be doing this just for creating an over electronically glorified mixer, not that all of this may actually improve the audio.  In fact, it will most likely lower the sound quality just having the audio passing through so many transistors in so many op-amps with so many feed-backs and gain stages.  (I can almost guarantee a proper laid out board of the earlier design will have a better sound especially in the high frequency range...)

As for driving headphones, I stand by my earlier suggestion of using 1 op-amp, driving a 2N3904/6 emitter-follower setup.  Too many op-amps will kill sound fidelity.

[paulca]

I'm borrowing from the design of pro-audio mixing desks.  I found it a little odd that I was hearing two different things from two different places.

Yes the original might have worked fine and no I probably wouldn't notice the difference, but I have at least a small part of my background and family working in pro-audio.  Things are done differently.

An example was pre-amp gain.  Everything audio I know prior to electronics told me to add gain at the input.  But all the amatuer audio stuff in EE forums use pots on the input to attenuate then boost which is a pointless noise generator circuit.

A the same time I still need you guys to help me as borrowing blocks from a book can lead to trouble until I learn the details of how to calculate my way out of a pickle.

Out of interest, if you put all the blocks together from the book and built a 24 channel desk with full inserts, parametric EQ, sends, returns, channel groups, PFLs, AFLs etc. etc. it would probably have in the order of 500-1000 op amps.  Do your CDs sound distorted or noisy?

[BrianHG]

On the gain.  The Baxandall stages have a gain of around -70db...

With headphones especially, you wont be able to mute the volume with only -70db.  Better add a mute button to each stage.

Be careful when signal mixing at lower impedance with higher uF caps.  These caps begin to exhibit some funny responses with their internal ESR and using cheap electrolytics or ceramics here may do funny thing to your signals below 500hz.  In really high quality audio, at the line level, we tend to go for 4.7uF or 10uF film type cap, with only a 50k or 100k load to avoid this issue.  On rare occasion, I've seen 22uF used, but, it was in a 24k/12k ohm XLR balanced load.

[paulca]

As for driving headphones, I stand by my earlier suggestion of using 1 op-amp, driving a 2N3904/6 emitter-follower setup.  Too many op-amps will kill sound fidelity.

This is the books suggestion for a cheap, simple headphone amp:

[paulca]

As someone on here said, if an audiophool spends £400 on a set of speaker cables it WILL sound better to him.

Just to clarify.  I do not mean I want to throw money at the thing and hope that it sounds better because of it.  I believe what I am doing is bypassing the marketing and mark up and building something that would cost 10 times what I could afford to buy already built.

Audio gear falls into three categories. 

* Domestic stuff which will have a quoted figure of 1000 when it only delivers 10.
* Audiophool stuff which quotes 1000, delivers 2000, but only 5 was needed at 100th the price.
* Pro-audio stuff which quotes 100, provides 100 and 50 was needed, but you pay a shed load more.

It's emulating the later I am aiming for.

Of course, most of my gains may be lost in my poor board design, crap cheap components and poor source cabling etc. etc. etc.

[BrianHG]

I'm borrowing from the design of pro-audio mixing desks.  I found it a little odd that I was hearing two different things from two different places.

Yes the original might have worked fine and no I probably wouldn't notice the difference, but I have at least a small part of my background and family working in pro-audio.  Things are done differently.

An example was pre-amp gain.  Everything audio I know prior to electronics told me to add gain at the input.  But all the amatuer audio stuff in EE forums use pots on the input to attenuate then boost which is a pointless noise generator circuit.

A the same time I still need you guys to help me as borrowing blocks from a book can lead to trouble until I learn the details of how to calculate my way out of a pickle.

Out of interest, if you put all the blocks together from the book and built a 24 channel desk with full inserts, parametric EQ, sends, returns, channel groups, PFLs, AFLs etc. etc. it would probably have in the order of 500-1000 op amps.  Do your CDs sound distorted or noisy?

No longer, everything today is all digital.

But, back to older analog.  You need to count the op-amps through 1 source to the central mixing point.  Then that mixing point point to the output jacks.  And yes, after all that, there is some loss and change in sound & the EQs had bypass buttons on the better units consoles for true flat.  Remember, they take all the sources & they only want to go through that mixing console once.  Take that output and pass it through another mixing console, and the sound will be affected.  I'm not saying these consoles were junk, far from it, but, there is a reason we have gone all digital today.  Pass a digital out from 1 digital mixing console into the next digital mixing console and remix with flat EQ settings and I bet the sound will go through perfectly clean if the code was written properly.

[paulca]

I think we agree, however do not be fooled into thinking that "everything is digital today".

In the past few years I have probably seen about 20 mixing desks in person,  I have yet to see a digital one (that wasn't a PC or Mac based thing).

Small 8 to 16 channel mixing desks you would use as an audio semi-pro running a few gigs and recording a few demo tapes would be like buying a higher grade oscilloscope.  Priced, today, around £1000.  A fully digital mixer would cost about ten times that.

PC and Mac based solutions are plagued with problems, not least the thing rebooting or installing updates live on stage, but they are used.  Very often you would have multiple programs producing audio on a single laptop, these would be mixed digitally with an external remote pyhsical control unit, but the sum output of those will end up going through an analogue mixing desk.

If you are going to a pro-audio studio and want a fully digital desk, expect to pay £500 a hour+.

[BrianHG]

As someone on here said, if an audiophool spends £400 on a set of speaker cables it WILL sound better to him.

I spent 300$ CDN (around 250$us) on my 10 feet speaker cables 100% pure silver and measured them to copper with my scope.  Mixing a 40hz sine wave with a smaller 10Khz square wave.  With the thick solid core copper, the peaks on the 40hz squished the squareness and amplitude of the smaller 10Khz signal inside.  The silver cable were almost perfect, though, I could not head the difference.  I still prefer that the signal out of my amp is reaching the speaker where a high current low frequency isn't mushing the fidelity above 10Khz.

[BrianHG]

As for driving headphones, I stand by my earlier suggestion of using 1 op-amp, driving a 2N3904/6 emitter-follower setup.  Too many op-amps will kill sound fidelity.

This is the books suggestion for a cheap, simple headphone amp:

Yes, that's the basic design and you will get a whole lot more mileage out of that circuit instead of parallel op-amps....
In fact, that circuit will ROAST the parallel opamps...
Just double C1 (if you keep the 10k load) and C2.  Make C2 a low ESR cap.  Only take it that far if you have 20hz headphones or are using my Subwoff...

[paulca]

Yes, that's the basic design and you will get a whole lot more mileage out of that circuit instead of parallel op-amps....
In fact, that circuit will ROAST the parallel opamps...

I do hate safety nanny limiting and ... the 5532 has a current limiter.

I know I like my headphones and they are the most expensive I have owned, but they have too easy a life.  I haven't heard them clip or distort very often.  Too soft.

So... adding in the real point of all this is to learn and to play with electronics.... maybe I should do a transistor output stage for the experience.

It can't been anywhere near as frustrating as debugging opamp circuits.... oh wait.  It probably can.