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

[Audioguru]

Why does each input use two opamps and many parts to have a fairly low input impedance, a gain of only 1.32 times and pass earthquake frequencies down to 0.17Hz?
I simplified it but it still has a fairly low input impedance, a gain of only 1.32 times but it cuts frequencies below 14Hz.
I would probably simply use a piece of wire. Here it is simplified: 

[paulca]

On headphone output levels.

The headphone spec says, the max power is 1.6W at 1Khz.  Is 1Khz likely to be the worst case?

Anyway directly driving the 38 Ohm headphones 7Vrms gives me 1.3W output with 184mA.  Sounds awfully high and requires 20Vpp through 38Ohms.

I found this test report for the headphones:
http://reference-audio-analyzer.pro/en/report/hp/audio-technica-ath-m50x.php

Which gives their sensitivity as 114.6db / V and their Vrms needs for 96db as 0.09Vrms.  ~2.6Vpp

I had a read through this as well, on headphone outputs:
http://www.apexhifi.com/specs.html

I'd like have at least 110db, concert level, but 130db is the maximum.

Using the spreadsheet in the last link, for a sensitivity of 114.6db / Vrms, into 38 Ohms, aiming for 130db, it gives me 912mW, 5.89Vrms and 155mA around 16 Vpp!

Trying something a little lower gives:
110db gives 0.59Vrms - less than 2Vpp
120db gives 5.9Vrms (yes I see the pattern) - ~17Vpp

So it seems due to the log10 nature of things small increases above 110db result in large changes in voltage. So if I go too low I have quiet headphones, if I go to far I have a dangerous (to other equipment) potential voltage on my headphone out.

As BrianHG said, the line level 2.5-2.7Vpp sounds reasonable.... but do I want reasonable?  5Vpp sounds better, giving close to 120db.

Thus, if I keep things around 2.5Vpp from input gain, summer, the headphone amp needs a gain of 2 and the ability to drive at least 50mA.  That would be 5 5532s in parallel to be audiophile, 3 would probably work.  Or more likely one of those OPAs.

[paulca]

Why does each input use two opamps and many parts to have a fairly low input impedance, a gain of only 1.32 times and pass earthquake frequencies down to 0.17Hz?

Because nearly every circuit you find with a search on "Active volume control" will have a buffer amp.  To answer that further I would have to go back and read the whole chapter on volume controls again.  The answer is most likely because someone far more knowledgeable about these things than I proposed it.

Also, on the gain, remember that R2 and R7 are the volume control!  They will go to 0R and 10k, or 10K and 0R.

Reading this however:
http://objectivesounds.co.uk/articles/active-volume-controls/

Suggests the circuit was designed to have high input impedance was due to accomodate valves.  I'm not sure of this, the link is an only source with 1 amp for an active volume control.

Their simplification to 1 amp does not resemble yours however.

[BrianHG]

On headphone output levels.

As BrianHG said, the line level 2.5-2.7Vpp sounds reasonable.... but do I want reasonable?  5Vpp sounds better, giving close to 120db.

This was for line level.  Using this level out of my PC sound card, my headphones burn my ears, though, 5vp-p would sound better.  Again, if you have a series 32ohm resistor in the line, then. double this to 10vp-p with no load, or at the op-amp output.  Note that the damping factor at low frequencies in the headphones improves without that series resistance.  If you need to overshoot a club/dance scene, like a DJ listening preping up the next track to mix, double this voltage again giving you the +10db out.  This can potentially burn out some consumer grade headphones and make you go death.

Anyways, stick with OPA551 for the premo-grade quality headphone amp output.  5x NE5532s in parallel will just consume a ton of extra current on you power supply and be lower quality sound due to slight timing differences and routing between each parallel IC.  In fact, if you want to drive a killer line level out, add another 2 there.  Use 100 ohm in series from each spare auxiliary cloned parallel output to isolate each one instead of using multiple ones.

[paulca]

The series out resistors, what are the pros and cons? 

I gather if I put a 40R or similar in series with the output I can just do the sums again for 40+38 Ohm load... ?

For the buffered line out, I should aim a bit higher for the series resistor, say a 2.2K since I have used them elsewhere... ?

I have no intention of making myself deaf... or death   I just don't like people telling me how loud my headphones are allowed to go.  I want them to not be at 100% volume.  I want 100% volume to be too much.

[Audioguru]

Paul, your simulation does not say and does not show a volume control but now you say it is TI's "active" volume control circuit.
Medical experts say that sound levels over 85dB cause hearing loss and you are talking about 120dB that is going to destroy your hearing.
If your headphones survive 120dB, their power is 35dB more that the 85dB that begins to cause hearing loss. 35dB is about 4000 times the power. Hit yourself with a hammer that causes a small wound. Then hit yourself 4000 times harder and see what happens.

I protected my hearing with earplugs when in loud sounds but my hearing is normal for my age (72). The high frequencies are reduced to almost nothing so I got hearing aids which create normal hearing for me up to 8kHz. The missing top octave cannot be boosted anymore without destroying the remains of my high frequency hearing and/or there will be acoustical feedback whistling if there is more boost.

[paulca]

So, taking a break from the opamps for a few days, I looked at the USB DAC.

I'm currently thinking of the PCM2906C.  The typical application circuit doesn't look too challenging to implement.  There are a few design decisions to make, but I'm currently thinking of just tying the HID inputs low as I don't want/need buttons.

The issues I am pondering is power.  Obviously TI are making these "Bus powered" as 5V is available off the USB.  There is an extended schematic "For high performance" which appears to use a linear 3.3V regulator, but that's about the only change.  I'm not sure what advantage one gets from running the chip on 3.3V instead of 5V I might do some digging there.

My issue with power is that it creates a route for EMI noise from my PC.  The DAC being digital will not have noise on the audio, but it will have on the power.  Now obviously I could stick a dummy load, like a 100K resistor on the VCC of the USB port and run the DAC off a 5V regulator off the +15V rail of the project (I know there are caveats, unbalanced load on the rails and I have not thought that through yet). 

The problem is then the ground.  I can't leave it off as it is needed for the D+ D- (I believe).  But once connected to the board it will then potentially pollute the rest of the board, including the analogue amplifier grounds with PC PSU and GFX Card noise.  I can still hear it on my current USB DAC, although very quiet, however it IS bus powered.

How do you filter a ground?  After a quick googling, this is as I suspected non-nonsensical.  However I did figure out why the "High performance" circuit runs off 3.3V.  The LDO and associated filtering caps will cut most of the noise out of the USB power. I might take that option. 

This has also made me wonder if I would get rid of the noise on the power rails from the XP Power +-15V PSU if I dropped them to 12V with LDOs after ferrite and cap filters.  I probably don't need the 15V rails.

[BrianHG]

With analog DAC
https://www.ebay.com/itm/CM6631A-digital-interface-USB-to-I2S-SPDIF-coaxial-32-24Bit-192K-sound-card/132272012272?epid=887353946&hash=item1ecc06bbf0:g:wH4AAOSw3ZNZdgsT

https://www.ebay.com/itm/DAC-384KHZ-32bit-Decoding-Module-XMOS-PCM5102-TDA1308-USB-Decoder-Board/232547127833?hash=item3624e39219:g:hbMAAOSw09xZ-sV~

Without Analog DAC, just I2S interface.  (you can opto isolate (use a high speed 8 channel digital galvanic isolator) these and place your DAC IC on the analog side of your preamp, no more connection to your PC's GND)
https://www.ebay.com/itm/CM6631A-USB-2-0-to-SPDIF-Coaxial-Optical-24bit-192kHz-Sound-Card-DAC-Board/322769627550?epid=1339946835&hash=item4b2691a99e:g:omsAAOSwkklZpQ6B

https://www.ebay.com/itm/XMOS-USB-Digital-Interface-I2S-SPDIF-Output-Board-Module-DSD256-32bit-384KHz/272891906642?epid=17007417855&hash=item3f89a00a52:g:mP8AAOSwevdZ5srx

https://www.ebay.com/itm/Italy-Amanero-USB-IIS-Combo384-module-USB-to-IIS-adapter-DSD512-PCM384K/272563718243?_trkparms=aid%3D222007%26algo%3DSIM.MBE%26ao%3D2%26asc%3D50073%26meid%3D61b938851ef9422b8621e9c6035a28b0%26pid%3D100005%26rk%3D1%26rkt%3D6%26sd%3D272563722080%26itm%3D272563718243&_trksid=p2047675.c100005.m1851 (new chipset with full list of OS support)

Both I2S and DAC with optional case:
https://www.aliexpress.com/item/HIFI-CM6631A-Digital-interface-32-24Bit-192K-Sound-Card-USB-to-I2S-SPDIF-Coaxial-Output/32855527872.html?spm=2114.search0104.3.231.6e5f4d70Qt73Ho&ws_ab_test=searchweb0_0%2Csearchweb201602_4_10152_10151_10065_10344_10068_10130_10342_10547_10325_10343_10546_10340_10548_10341_10084_10083_10618_10139_10307_5711211_10313_10059_10534_100031_10103_10627_10626_10624_10623_10622_10621_10620_5711311_5722415%2Csearchweb201603_2%2CppcSwitch_5&algo_expid=c9a89ce6-7582-41c6-b9ec-6810a48f3dd7-36&algo_pvid=c9a89ce6-7582-41c6-b9ec-6810a48f3dd7&priceBeautifyAB=0

[paulca]

Interesting boards, but I was kind of hoping to get everything onto one board.  Some of those board smell of audiophool too.  32bit?

Further, what are the chances of these boards being class compliant USB audio devices that work without specialist drivers/firmware?

But the idea of dividing the DAC into two stages is an interesting one, but don't opto-couplers for USB cost $$$$?  Due to the higher data rates and timing?  Is SPDIF easier to opto couple?  Does I2S not need a ground connection?

The other option is an actual optical, but I haven't checked how much they cost.

[BrianHG]

Interesting boards, but I was kind of hoping to get everything onto one board.  Some of those board smell of audiophool too.  32bit?

Further, what are the chances of these boards being class compliant USB audio devices that work without specialist drivers/firmware?

But the idea of dividing the DAC into two stages is an interesting one, but don't opto-couplers for USB cost $$$$?  Due to the higher data rates and timing?  Is SPDIF easier to opto couple?  Does I2S not need a ground connection?

The other option is an actual optical, but I haven't checked how much they cost.

Look at the SPDIF, it is already coupled.  That black cube isolation transformer beside the SPDIF jack.  SPDIF is difficult to interface compared to I2S.  I2S is on every audio DAC you can buy today.  Example: http://www.ti.com/audio-ic/converters/dac/products.html#p1021=2;2&p84=32;32&p1339=2;2&p348max=384;384

These 32bit 384khz stereo DAC start at 1$.  High end ones at 3$, differential output ones at 4.25$.  The top ones have over 114db SNR.  This is not audiophool.  What if in windows, you turn down the volume to -50db.  With this, you still get you source 16 bits going right through at low volume because of the depth of 32 bits.  As for sample rate, what if you are playing DVD audio at 48Khz, or CD audio at 44.1Khz, or, HD-DVDA/SACD/TrueHD Bluray at 96/192Khz, 24 bit audio.  Keeping your system at 384Khz, or even slowing it down to the standard 192KHz makes for a non-wobbly output like when you tried to play my .wav files which played perfectly clean like a function generator on my old 192Khz 24 bit PC motherboard.  Also, the differential/balanced DAC versions outputs a killer 4.2vrms levels.  No noise here if in your preamp you wire that input channel to receive the differential audio.  Compare that to your 1vrms single-ended old 48khz 16bit usb dac.  Run TI's PCM5242 at 384Khz in differential out, you will get 114db, for 4$.

If you want separation, just use the one of the simple I2S boards.  Mount one sound board in your preamp isolated from the rest of everything.  Those sound boards are self powered from the USB power.  Next, get one of these: https://www.digikey.com/product-detail/en/maxim-integrated/MAX14930CAWE/MAX14930CAWE-ND/5396463  Yes, this is basically a 4 input, 4 output 150MHz optocoupler with 2ns rise and fall time.

I2S only needs 4 digital outputs.  On you preamp, one side of the isolator is powered by the 3.3v output on the USB sound board, & gnd & data fed.  On the other side, you power from your preamp the 3.3v & gnd & the 4 I2S data lines go to your 3$ ti DAC.  The output of the TI dac goes to an input channel on your preamp.  Keep the isolator at the edge of you pre-amp PCB and do not GND or VCC fill under it.  Make the connector off the PCB going to the USB sound board as far to the edge as safely possible.

[BrianHG]

Maybe you should use TI's balanced input high fidelity amplifier design in the PCM5242's example schematic in it's data sheet.  It goes to the volumes that you want and is faster than the OPA551 and it is stereo.  However, to get that full quality, you will need to go all balanced audio for that part of your circuit.  3X cheaper, and more power (1 watt for +/-15v supply), higher frequency than the OPA551.  However, to compare the 2, you would need to convert % distortion for the OPA551 to distortion measured in db for the tpa6120a2.  In fact, the tpa6120a2 roasts your NE5532s to such an absurd degree, so, maybe you don't need such a high quality headphone amp on a mixing console where the sound previously goes through multiple NE5532s.  (In audio applications & driving headphones/line level)

[paulca]

Confession of a blonde moment.

I was concerned about the high current draw.  That current being measured by the INPUT from the bench supply to the dual rail PSU.

Well, of course I forgot to factor the voltage difference.

Highest current pull of the PSU was 70mA.  25mA is the stated idle quiescent input current, but I was reading more like 30mA.  Even if we take the 25mA and subtract that from the 70mA seen...

70-25mA = 45mA
45mA * 12V = 0.540W
0.540W / 30V = 18mA

The opamps are rated for +-10mA, there are two ICs on the circuit + a small amount dropped in the various passives, so they are not being exceeded.  They may be higher than expected, which might be the HF noise I'm dealing with in another thread, I hope, but it's all good.

[BrianHG]

This one was made just for you: https://www.ebay.com/itm/Douk-Audio-PCM1794A-DAC-Decoder-Balanced-HiFi-Parallel-Board-192K-Gold-plated/272982323436?epid=28011370612&hash=item3f8f03b0ec:g:ZKIAAOSwBkRaMjTA

Take a look at the op-amps.
This board has a SNR of 132db since it uses each dac in balanced mode out & is converted to unbalanced using your NE5532s.
Though, you are limited to 24 bit, 192Khz sample rate.  But, this is plenty enough.

In mono mode, that DAC outputs 9v RMS.

[paulca]

I thought I was the overkill suspect?

You seem as overkill on the DACs as I am on active volume controls. 

I think a TI PCM290x will do me just fine.  All I need to do is minimize noise pick up from the PC ground.

[BrianHG]

You seem as overkill on the DACs as I am on active volume controls. 

That's because if you checked the data sheet on the 32bit 4$ DACs, you would see they have a built in pre-amp volume control which can be set through I2C.  No sketchy wired pots in my design.

[paulca]

How do these interface on USB?  Obviously the DAC that produces the i2s, but I mean regarding drivers and sound card emulation.

[BrianHG]

HD Audio drivers for amanero & older windows https://amanero.com/

Xmos has native support HDaudio in Win10, but if you want HDaudio in older windows, you need a a driver.

Same for the CM6631a.

All the other OS, ie Mac & Linux natively support USB Class 2.0 audio.

If you don't have drivers for older windows older than Win10, the boards will run in USB Class 1.0 48KHz audio mode.

I'm  thinking of doing a DAC board with amanero and my Win7 PCs.

[paulca]

So here is an LTSpice file for an example of where I'm at for two mono channels.

This is without a baxandall active volume control on the master, just a pot on the headphone amp input.  The only downside to this is that the pot interacts with the "aux out" amp (top right).

"Input Gain" (x2) and "Master Gain" show pots. Both inputs are set for slight attenuation, master is set to MIN in the up load.  Increase R19 and reduce R20 to turn it up.

C2 and C8 can be removed, I know, along with R7 and R14.

[Audioguru]

What is chopping the waveform?

[paulca]

What is chopping the waveform?

Is that just a Spice artefact maybe?  I'm sure it has to have a resolution it simulates at.  Maybe "minimum step size" in the simulation command might change that.

Anyway I ran through some normal cases and some worst cases.  With a quiet 0.3vpp input gained on the pre-amp I can still get 8vpp and 100mA on the headphone output.  Similarly with a loud 1.7Vpp input and slight attenuation on the pre-amp I can get sensible output levels.

Yes, if I max the input gain on a loud signal and put the master volume full I get close to 24Vpp on the output which will not impress things.  However that would be a mistake.

I don't like that the master volume pot attenuates the aux out though.  Putting another buffer before the volume pot might help.

The next thing I think I need to look at are signal and peak lights, set at sensible points roughly related to dB.  Just not sure what those levels will be, 0db and +10db?  Or the best way to do them.  Zener or voltage divider and a known Vf diode.

[paulca]

So I'm closing in on the rightly.

I have a board in manu for a prototype pre-amp.  Think I have the circuit for the mixer amp and output amp sorted. Also making progress on filtering the DCDC converter power supply.

The open items are:

Master Volume
Limiting the mixer amp output to something non-destructive in the event of a mistake in gain.

The THAT4301 Chip mentioned, by Bassman I believe a number of pages back.  I like the idea behind this.  They are hard to find though.  However I want to explore this idea.

Is there such a thing as a voltage controlled volume / compressor / limiter?  Preferably both in one IC.  ie.  A voltage from a pot selects volume output, which will never exceed an RMS voltage limit set by another pot.  This would ideal, but I would settle with putting the volume control separate (before) the limiter.  High bonus points if it provided indication output when it was limiting, so I can light a "SAFETY" LED or similar.

A quick search on Farnell makes it seem finding one of these chips could be quite difficult.

[Audioguru]

Why is the maximum output level so high that a protection compressor circuit is needed to prevent destruction of the headphones and your hearing??

[paulca]

So I can preamp up quiet signals.

[Audioguru]

The voltage gain allows you to increase the loudness of quiet sounds. The supply voltage and load resistance determine the maximum loudness. Your supply voltage is too high for your load resistance, causing the output power to damage the headphones and your hearing.

[paulca]

I read a few guides on limiters and the first thing proposed and very quickly dismissed was limiting the supply voltage.

Changing the supply voltage to limit the output power just sounds wrong.  All that will achieve is clipping the amps.  That is much more damaging to both ears and particularly speakers and headphones.  Stacking enough resistance onto the output to make the maximum achievable voltage by the mix amp, given a mistake, not be too loud will ruin the gain structure and require running the amps at 20Vpp to get full volume.

Having enough headroom for any output and then limiting the ultimate peak voltage to prevent damaging headphones or downstream equipment seems to be to be a much more professional way of handling things.

Besides, outside of nanny state consumer audio headphone amps will typically melt a pair of headphones if not treated right. 

Further, the limiter will be on the output of the mixer amp, which is effectively my master bus.  It has an aux output.  If I have that connected to something, it doesn't matter if I limit the headphone amp supply rail as it will still fry other equipment unless I limit it's voltage too.

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