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

[BrianHG]

So to be sure, you want to lower the dynamic resolution of your music, not protect you amp and ears.

The cleanest noise free way, with optional independent channel limiting is to use 2 vactrols, or, a stereo vactrol.

Any IC you use will most likely introduce noise and distortion, + you will need +5v, or 12v, + such ICs for some reason like a DC offset you might have to deal with.  Otherwise, you will need to go with a full 2x high quality 4 quadrant multiplier IC from analog devices, 8$ each (You need 2 unless you have gone mono), + a additional set of opamps to drive the gain of each of them.  But, in the audio band, the vactrols are an old, but kinda known around to be a perfect voltage controlled resistor exceedingly high quality for audio apps.

Now thanks to the loudness wars, and the music demo you left to me on another thread, note that unless you redesign these limiter circuits to interpret how the human ears responds to loudness, any circuit you try will still make some recordings sound louder than others since the source already has been clipped and will as one volume already, but be louder than that because of the embedded square wave nature.  So, to solve this problem, once again add a few more op-amps into the gain control circuitry if you wan to fix this.

My green leds at the inputs of your final stage driving opamps controls your upper limit, yet, isn't quite as harsh as lowering your power supply stripping the dynamic functionality of the opamps & the rest of your stages.  And all it requires is 2led + 1 resistor / channel.  Around 20 cents compared to 16$ of analog multiplier, or vactrols, and supply filter caps and around 6 additional opamps + signal diodes+ 4 trimpots, 2 for max signal level, 2 for recovery speed, + diodes or transistors & more caps.

[paulca]

Thanks Brian.

I am keeping the LED solution as a fall back, but there are dedicated stereo audio limiter ICs out there.  Bassman linked the THAT devices which are an option, but I wanted to explore if there are others out there.  Basically they are audio processor ICs which provide limiting, compression, companding or expanding.  So they are, one would assume sensitive to the dynamic range and physco-acoustics of audio.

The THAT ones seem hard to find are expensive etc.  So I was asking if anyone had experience of different chips or even what kind of parametric search I should be running to find them.  I tried looking on Farnell and RS but didn't find anything that looked right, so I'm probably not searching for the right thing.

On the LED clamp circuits, I have found quite a few variants, some even designed to progressively limit as the overlimit increases.  Do the LEDs actually light up when they clip due to the voltage exceeding their Vf?  I would assume so, which would be cool.  It gives me that clip light.

[BrianHG]

Thanks Brian.

I am keeping the LED solution as a fall back, but there are dedicated stereo audio limiter ICs out there.  Bassman linked the THAT devices which are an option, but I wanted to explore if there are others out there.  Basically they are audio processor ICs which provide limiting, compression, companding or expanding.  So they are, one would assume sensitive to the dynamic range and physco-acoustics of audio.

The THAT ones seem hard to find are expensive etc.  So I was asking if anyone had experience of different chips or even what kind of parametric search I should be running to find them.  I tried looking on Farnell and RS but didn't find anything that looked right, so I'm probably not searching for the right thing.

On the LED clamp circuits, I have found quite a few variants, some even designed to progressively limit as the overlimit increases.  Do the LEDs actually light up when they clip due to the voltage exceeding their Vf?  I would assume so, which would be cool.  It gives me that clip light.

Yes the LEDs light up.  On my Subwoff board, since the LEDs are only treating the signals below 25hz, especially the sub 10hz-2hz signals because of the massive VLF gain, it has a cool effect of showing you the breeze, or wind direction, or, impact direction when watching movies.  Since 1 LED handles the positive blow, then the other LED handles the negative.
I should video it since it is slow enough to cleanly capture at 60fps.

The analog devices part number is an AD633.  In the data sheet, they have example schematics on audio volume limiting.

Vactrol compressors: (You can change the Vactrol with a opto-photo-fet H11F3SR2M, you can also drive this guy right into your active volume control.)
https://www.google.ca/search?q=vactrol+compressor+schematic&client=firefox-b&sa=X&dcr=0&tbm=isch&tbo=u&source=univ&ved=0ahUKEwj_9qaY147aAhUE74MKHbcRD28QsAQIKA&biw=1317&bih=714

[BrianHG]

I was wrong about the price of vactrols, they are dirt cheap:
https://www.aliexpress.com/item/5PCS-VTL5C1-DIP4-VTL5C/32714837501.html?spm=2114.search0104.3.2.1234651743yxny&ws_ab_test=searchweb0_0,searchweb201602_3_10152_10151_10065_10344_10130_10068_10547_10342_10343_10340_10548_10341_10696_10084_10083_10618_10139_10307_5711215_10313_10059_10534_100031_10103_10624_10623_10622_10621_10620,searchweb201603_50,ppcSwitch_5&algo_expid=125acb79-4ed6-4b4c-8f1a-1a4c693c72fb-0&algo_pvid=125acb79-4ed6-4b4c-8f1a-1a4c693c72fb&priceBeautifyAB=0

Easy to breadboard with...

[Audioguru]

I hate the way an audio compressor causes the level of the background music to pump up and down with the beat of the music or with the speech of the DJ. Radio stations replaced the sound man's hand on the volume control with a compressor so that their average and peak sounds can be louder than other radio stations.

[paulca]

I hate the way an audio compressor causes the level of the background music to pump up and down with the beat of the music or with the speech of the DJ. Radio stations replaced the sound man's hand on the volume control with a compressor so that their average and peak sounds can be louder than other radio stations.

Not all compressors are created equal.

If you are listening to anything recorded in the last 30 or 40 years it has been compressed.  All commercially recorded digital music is compressed.  Heavily compressed.  Simply because you cannot clip it ever.  Your choices are to run the volume very low to give you plenty of headroom and thus quiet on domestic audio equipment or compress it. Most is multi-band compressed.

What you are describing is a simple compressor with too short a release time. 

If you put Radio1 on and put it into a level meter you will find the meter barely moves.  This is extreme compression due to trying to absolutely maximise the dynamic range of the medium.

[paulca]

Actually maybe "dynamic range" was the wrong word in the last statement.

Anyway.  A few questions...

These LED limiters.  I like the fact the light lights up, so I got to thinking.  How can I split off a signal for the pure purpose of lighting an LED at a set peak voltage without altering the signal?  Would I need to buffer the signal off with another op amp?  Is there a simpler way?   I'm thinking of an LED to show "Signal" and the limiter LED showing "Clip", so I can put them on the preamps and just the clip on the master as a limiter.  This will allow to tune the inputs visually with signal being tuned to be lit at 0db line level and clip at about +10db or something I'll calculate later.

On ICs I can find companders, but it looks like they are "all or nothing".  What I want is a compressor with a threshold.

When I did digital DJ mixing to CD with an absolute hard limit of 0dB I used two compressors in series at the final stage of my master filter chain.

Compressor one was set to -6dB and had a short rise and long fall time with limited gain.  So it was soft and non-intrusive and basically compressed the signal over -6dB lightly.  This allows loud music to compress up into the top volume area.  Compressor two had a -1dB threshold a minimal rise and short fall time and infinite gain authority.  These were the default settings for the "Hard limiter" settings.  The reason it was set to -1dB was, under test it still breached the 0dB limit occasionally.

While I realise this would be very complex, expensive and probably noisey to do in analogue electronics it is what I'm after.  Signals well below the limit should be left alone.  Signals approaching the limit should be compressed, signals above the limit should be suppressed, ideally by attenuation rather than clipping.

However I'm getting the feeling this will be tricky to achieve and I'll end up with some form of LED clamp limiter.

[Audioguru]

AliExpress is Chinese like ebay and Amazon. Don't they also sell cheap Chinese fakes and manufacturer's rejects?
Years ago they made slow compressors with an incandescent light bulb eventually lighting an LDR that took a long time to react.

Recently I made an LED blinker that alternates lighting groups of LEDs quickly then slowly then quickly then slowly. instead of abrupt switching I used an LED lighting up an LDR and when it switches to make the alternating go slowly you can see that it takes some time to change its resistance.

[Bassman59]

Limiting supply voltage is not limiting it's clipping.  I'd like to explore the option of using an actual limiter/compressor IC.

The THAT 4301 I mentioned has everything you need to implement a limiter. Add resistors/pots for threshold, output, ratio and you're set. Read THAT's app notes, it's all there.

You will find this to be a frustrating challenge, taking into account varying source input levels and different headphones with different impedances and sensitivities. There is no one-size-fits-all solution.

But here is the deal. A peak limiter is not really all that different from a clipper. They sound bad.  My Genelec 1031A studio monitors have vactrol-based limiters. Detectors monitor the levels going to the treble and bass amplifiers (after the internal active crossover) and when they exceed the fixed threshold the vactrol turns on and reduces the gain of the input buffer circuit. They bark at you when you hit them. That means "turn it down."

A perhaps better approach is a compressor with a high ratio, a fast attack time, a long hold time and a long slow release time. A peak is hit, the gain is quickly reduced and then the gain stays down for a while. The long hold time helps keep modulation (pumping and breathing) to a minimum. As long as the input level goes above threshold during the hold time, the gain shouldn't change. "Modern" low-dynamic-range-music will probably guarantee this. But, invariably, the input will go down, and the gain will go up.(Or, more correctly, the attenuation reduces.) But this will still sound different from the uncompressed source, so the gain-reduction indicator should tell the user, "turn it down!" You don't want to actually have the thing working in gain-reduction mode.

Or, you could do what most headphone amp designs do, and skip the limiter entirely. Be aware of your levels and what you're listening to.

[paulca]

Yes, I know I keep getting accused of trying to drive a Marshall stack through my headphones, melt the drivers and perforate my ear drums that is not my intention.

However I kind of want that option, if I choose to.  I never want the volume control to get to 11.  10 should be painful for me and the headphones and probably never used, but 11 should be available.

The limiter is more for mistakes.  Say I'm listening to a really rubbish rip of an audio book and the gain is way up, the media player files are out of order and a digitally remastered Slayer track jumps on giving me 20Vpp on the outputs.  Not the volume at 11, but more the volume at 27!  POP goes my headphone drivers and the input amp on my desktop speakers, plus extreme pain in the ears.

This scenario has happened to me, listening to a really quietly ripped audio book at max gain and the media player selected a really loud modern recording next.  Nearly fell out of my seat.

Similar funny story... listening to music on headphones drunk one night at about 4am, I leaned back in the office chair, promptly unbalanced it, fell over backwards spilling beer all over my face and to make matters worse the headphones got pulled out of the amp which promptly lit up the speakers at full tilt and caused a glass lamp to fall off the corner unit and smash into a flower pot.  Instant devastation, total comedy sketch... and the neighbours asked me what the **** I thought I was doing the next morning.

I prefer switched amps where you can have both headphone and speaker or one or other at your choice.  This little box gives me that... if I add a mute on the aux output to the speakers.  Nothing more embarrassing that listening to something dubious on headphones only to realise the speakers are still on!

On the later... how do opamps take to having their outputs switched to floating?  Is it a better idea to switch it to ground through a 1meg?

The THAT ...  how much "compressor" like control does it yield?  You mentioned "threshold" does that limit the actual intervention point or will it compress the output even at lower levels?  Does it provide attack and sustain control?  I did scan read the datasheet in work today, but not in details.  You are leading me to think I should.

[Bassman59]

Yes, I know I keep getting accused of trying to drive a Marshall stack through my headphones, melt the drivers and perforate my ear drums that is not my intention.

However I kind of want that option, if I choose to.  I never want the volume control to get to 11.  10 should be painful for me and the headphones and probably never used, but 11 should be available.

Well, then, what you want seems to be at cross-purposes with what you want.

The limiter is more for mistakes.  Say I'm listening to a really rubbish rip of an audio book and the gain is way up, the media player files are out of order and a digitally remastered Slayer track jumps on giving me 20Vpp on the outputs.  Not the volume at 11, but more the volume at 27!  POP goes my headphone drivers and the input amp on my desktop speakers, plus extreme pain in the ears.

This scenario has happened to me, listening to a really quietly ripped audio book at max gain and the media player selected a really loud modern recording next.  Nearly fell out of my seat.

iTunes has a feature called "Sound Check." When you import an audio file into it, it scans the file and looks for basically its "loudness," and it keeps track of that. On playback, the output volume is adjusted so that all songs you play are at the same "loudness," to counteract exactly what happened to you. It's not compression, it's just level normalization. The level is never changed dynamically during a song. Instead of you adjusting the volume for each song (to compensate for loudness differences), iTunes does that for you.

It works well, usually. There's not much it can do with music that has high dynamic range, such as classical, which might start a piece at piano and then go to fortissimo. iTunes will normalize against the fortissimo, but when you listen you might still have to reach for the volume knob during the piano parts.

I prefer switched amps where you can have both headphone and speaker or one or other at your choice.  This little box gives me that... if I add a mute on the aux output to the speakers.  Nothing more embarrassing that listening to something dubious on headphones only to realise the speakers are still on!

My studio monitor controller has such a feature. When you switch to a new output, the others are muted. It has a "override latch" feature, which lets you have more than one of the speaker outputs on at once.

On the later... how do opamps take to having their outputs switched to floating?  Is it a better idea to switch it to ground through a 1meg?

Short answer: op-amp outputs are never truly floating.

Longer: Look at any op-amp schematic. For the non-inverting configuration, the output connects to ground through the feedback network.

For the inverting configuration, the output is connected to the input via the feedback network. [img]https://www.electronics-tutorials.ws/wp-content/uploads/2013/08/opamp6.gif[/url]

The THAT ...  how much "compressor" like control does it yield?  You mentioned "threshold" does that limit the actual intervention point or will it compress the output even at lower levels?  Does it provide attack and sustain control?  I did scan read the datasheet in work today, but not in details.  You are leading me to think I should.

Yes, you must RTFDS. Remember, THAT is a company that specializes in audio circuits, so their parts are intended to be use in the standard dynamics-control devices (compressors, limiters, gates, expanders).

Threshold sets the point above which attenuation occurs. It does not attenuate when the signal is below the threshold. (If it did, it would be a downward expander.)

It does provide attack, hold and release control. (Not sustain, you are thinking of analog synthesis.) Well, you have to design in the pots to set those controls, which affect the detector operation. It is the output of the detector which drives the VCA's gain-control input.

[BrianHG]

Just use Microsoft Windows built in multi-band loudness equalization feature in your 'Speakers' enhancements.  And highight the filter and click on settings, select 'Short'.  Everything will come out from your PC at 1 volume to your ear, set by the system master volume.  Great for all those stupid Youtube and system audio notifications.  Everything will have a sane volume output.

To replicate the effect of this particular very useful digital loudness equalizer, you need a multi-band limiter in your IC.  It would need to be sensitive and adjusting the individual volume of multiple frequency ranges separately based on each range's volume, not just use the voice band as a level reference and adjust the entire audio band based on that selective level.

[Audioguru]

You never disconnect the load on a vacuum tubes amplifier because the inductance of the output transformer will make huge voltage spikes. A solid state amplifier does not have that problem because it does not have an output transformer.

Some solid state amplifiers have a single positive supply and a capacitor coupling the speaker and blocking the DC. They need a load to charge the capacitor when the speaker is disconnected when powering up to avoid a huge POP when the speaker is connected to the amplifier.

[BrianHG]

Limiting supply voltage is not limiting it's clipping.  I'd like to explore the option of using an actual limiter/compressor IC.

The THAT 4301 I mentioned has everything you need to implement a limiter. Add resistors/pots for threshold, output, ratio and you're set. Read THAT's app notes, it's all there.

That THAT IC's data sheet has some nice specs in the data sheet.  3 things I would have like to have addressed in the data sheet would have been scope shot, IC-IC gain matching and a better way of addressing stereo, both with a common focused ALC, or independent channel rather than having so many separate controls.

Example, with 2 ICs, and the voltage control input, each IC has a gain of around -95db at +0.6v in and a gain of 0db at 0v in and a gain of 30db at -0.17v in.  How precisely matched is this voltage gain setting from IC to IC?  The spec of +/-15mv isn't too convincing since that can be around +/-3db difference per IC.  The data sheet also says these figures are affected by temperature since the spec has a listed temp-coefficient.  If you want a perfect clone stereo volume control, do you need to match ICs + heat-sink them together?

(Wish I new about this chip 18 years ago for my CATV converter, would have been great for the mono volume control...)

[paulca]

Yes That4301 looks like the right bit of kit.

Here's the crunch however.  To do a stereo master volume control with it, I need two of them.  I would need to buy 2 in DIP package to breadboard it out, then buy another two in SSOP for the actual PCB.  That's £60 all in, plus a few quid for the correct trimmers and pots.

So the question comes down to, is it worth £60 for the output to be compressed limited or do I just go for the LED clipper.

I'm leaning towards the LED clipper.  I know it will sound harsh, but maybe that's good.  I should be set to a level beyond which the circuit should never be running, so I shouldn't care what it sounds like.

[paulca]

So I have to make some decisions.  Too many things in the project, now I am considering metering and the headphone amp stage etc. all require me to know my maximum peak voltage.

A few secondary design limitations steer me towards 5Vpk, giving 10Vpp.  These would be notably using an ADC in an MCU on a 5V rail for metering.

I would like to translate that 5Vpk to the SPL of my headphones to determine if it's too much, too little or Goldielocks.  That then allows me to pick output series resistance and the final h/phone amp gain.  I also allows me to validate if the "nominal" level of 0.615Vrms will work.

It will also allow me to design and test LED/Diode clip circuits to enforce the absolute maximum peak voltages to protect the lower voltage parts of the circuit, such as the MCU and downstream connected equipment, such as my desktop powered speakers.  It will also allow me to consider dropping the rails.  If I can get below +-8V I can probably go for some form of linear rail spilt solution and still get my 9-15VDC input range.  Although I still believe that requires a switch mode voltage inverter IC.

I've made a spreadsheet showing various voltages rms, pk, vpp against dB based off a reference "nominal" internal voltage.  The bit I can't seem to get the spread sheet to calculate is the effective SPL at the headphones based on their sensitivity.

Of course I can find calculations based on dB per mW, but I have specifications only for dB/V.  I tried using the same equation and it gave me non-sense.

Can anyone help with that?

The equation I was using was:

sdBV+10*LOG10(Vpk)

sdBV is Sensitivity dB / V.  Which has the figure 114.6.  Straight off I can see that 1Vpk is going to be quite loud.  But is 5Vpk too loud?  Is it even db/Vpk or db/Vrms?

[paulca]

I'm wondering if it was as simple as changing the 10x to 20x in the equations.  It gives me much more realistic figures.

It also now confers with this test report:
http://reference-audio-analyzer.pro/en/report/hp/audio-technica-ath-m50x.php

It gives 96dB as 0.093Vrms and 0.233mW

My spreadsheet (at -17.39db) gives 95.999dbSPL @ 0.083Vrms and 0.182mW.

These are internal amp levels assuming the master and headphone volume are to FULL.

To listen at a modest 96dB the headphone output (or master volume) would need to come back to -18db for a signal at 0dB internal.

But it "looks" like it's all starting to check out.  If it's correct I can have unity gain on the h/phone output amp, making it a voltage follower config.  I will also be able to meter the 4.9V max peak with a 5V rail-to-rail peak detect circuit and a 5V ADC.

[paulca]

So this arrived today.  Only took 6 days for manufacturer and shipping from AllPCB.

The only bemish is my own doing.  It's too easy to get distracted by the zoom functionality and forget how small the board actually is when you plop a graphic on it.

The top black graphic is suppose to say "Protect from sunlight"

A Baxandall active volume control/pre-amp.

[paulca]

And Tada!

It works.  Slightly more an anti-climax this one, compared to the DAC as I can't really put this into proper use on the desk as it still requires the +-15V supply which isn't really practical yet.

However the point was simply to check the circuit works.  By building small prototypes of each major component I get to iron out any anomolies, rather than try to debug them all at once on a large board with all of them on it.  At potentially considerable overall cost, however spread over months.

I still have to scope it and see how it responds to different torture, but it functions from a quick ear test.



The next board already in the post is the PSU with filter.

Following that is the headphone amp prototype, which I home to actually put into a case and make generally useful.

[paulca]

So a little update on progress.  Things are coming together.  Or... rather they aren't yet.  They things are starting to exist.

Linking in a few cross threads on sub projects:

Headphone (power) amp - created as a standalone.
https://www.eevblog.com/forum/beginners/circuit-review/msg1471160/#msg1471160

Ideas and prototypes for meters.
https://www.eevblog.com/forum/beginners/audio-bar-graph-small!/msg1465125/#msg1465125

The DAC:
https://www.eevblog.com/forum/beginners/audio-dac-output-low-pass-filter/msg1450490/#msg1450490

So right now the DAC and headphone amp are in service and working fine, with caveats.

I added gain to the DAC (swapping 12K input resistors for 3.9K) and while it doesn't upset my gear it upsets my brothers amp which clips horribly.  <shrug>  I gained to to see if it would run headphones, it did, but weakly.  However it drives as a preamp absolutely fine, it drove my existing headphone amp to where I no longer needed full volume for loud listening.  In the full circuit I will put the gain back to what it was as it has a premp stage after it.
Issues: The L and R output are swapped, doh!

The headphone amp.  As it's standalone I couldn't build it for my internal chosen levels, so I gave it a whack of gain, 7.666 actually.  Powered it from a voltage divider rail split with two caps.  It performs excellently.  I "can" get it to full throttle with my headphones, but I can't tolerate it for very long.  Face melting.  I've used it all weekend and on a dial that goes from 1-10, 3 is comfortably loud, 4 is kicking loud, 11 is only tolerable for a short period.  But, this achieves my goal, I don't need to run my volume controls to MAX all the time it's back in my control.  The downside of high gain is it amplifies noise, so even touching the bare metal pot make audible noise... and angers the voltage divider noise below.

Issues: The voltage divider virtual ground is highly sensitive to power rail noise.  Also the input biasing to that virtual ground is sensitive to cap matching.  When I was using 220uF input caps to minimise low end roll off, they were unmatched by about 20% and the right hand channel ended up biased about 2V lower than the feedback reference, so... no output.  22uF are easier for the 22K bias resistors to pull, but I'm tempted to lower them and increase the cap size again.  If I lower the 22k bias resistors (biasing the amp side to the virtual ground) it will form an RC filter with the cap and I might put them back up to 220uF.  Alternatively, I could buy higher tolerance caps.

The voltage divider noise, in the short term with the board I have, I might lower the 4k7's to 2k2's or even 1ks to make it a bit more powerful to avoid uA range pick up.  Burning a few mA is worth it. 

I'm thinking of making a revision two of the amp with these improvements and ... maybe a opamp virtual ground buffer and a chain of power rail filtering caps.

What I'm learning is that when you don't build things to a price, but to a spec they sound better.... or maybe that's just confirmation bias and pride.

Still on the list:
Prototype the bluetooth module.  Probably breaking it out with mod wire on a perf board.  It needs a differential to single ended amp stage and it would be nice to combine that with the preamp, but I doubt I have the experience to do that... or it might not work.
Prototype a few bar graphs for the meter module.
Start trying to layout a full board with everything on it.
Start thinking about how I house this thing.  I'm already thinking it's going to end up in a case that's 4" or 5" sqaure with the meter board off the back, angled up, console style.

[paulca]

The thought of EQ is torturing me now.  Bass and treble controls at least, but what I'd probably prefer is a mid-range attenuator, ideally parametric...  I'll explore.

[BrianHG]

You should have just done an all digital mixer.  Included everything in the DSP, volume, vu meter, mixing, auto volume limiting, full graphic eq, built in MP3 player, WiFi, separate headphone and line out cross channels, and a few more capabilities...