O2 amp CRC, diode, cap, and heatsink mods

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
@agdr, if my numbers are correct your LME48720 mod lets you run your K550's up to a very loud 112dB before it runs out of current at 25mA. If you keep 'em paralleled like the O2 runs the NJM4556, then you can get to a whopping 118dB. It only takes 1.58V to get there.

I agree, that all sounds right. I've left them paralleled due to the only hiccup I see on the data sheet, a (only) 100pF direct-drive (no output resistor) capability per each chip half, probably not enough for a typical headphone cable. Leaving them in parallel should double that to around 200pF, then leaving the 1R balancing resistors in is the old trick (series resistance) to increase stability in the face of load capacitance. So in short, just leave the O2 output section the way it is (paralleled) and swap the NJM4556s for LME49720s. :) Perfect for the AKG K550s.

One interesting note on the DC offset issue. I heard back from AKG and their answer is essentially the same as what RocketScientist said, above. A DC offset voltage just moves the diaphram slighly one way or the other, meaning that at maximum excursion it would run into the mechanical limit of travel in that particular direction a bit soon. But normal listening levels are no where near maximum excursion, so small DC offsets won't matter AKG says. I specifically asked about 1 - 5mV and they said that is fine.

Putting numbers to that, I measure about 80mV(rms) to each channel of the phones as being about as loud as I can take it for normal listening. In the case of the O2, with that 3mV typical DC offset on the output, with an 80mV swing the transducer would wind up with 83mV in one mechanical direction and 77mV in the other. But the "hearing damage 94dB" level on the box is 135mV, and as you've noted above max excursion is probably somewhere around 158mV. So that extra 3mV of DC offset is still well within the mechanical travel bounds.
 
I don't think a measly 1 ohm series resistor will do much in terms of capacitive loading isolation. I'd expect something in the order of several ten ohms to make things bulletproof.

Definitely something to try (higher output resistance) if someone starts experiencing oscillations with LME49720s in for the NJM4556s powering the AKG K550s. I've scoped it and no HF oscillation that I can see so far. The LME49720s themselves are most likely stable with the amount of headphone (cable) capacitance involved, at least in this parallel configuration.
 
Bass boost mod won't work for gain = 1 (no R17 & R21)

I've been meaning to post a correction for awhile to the bass boost mod I posted earlier in this thread. Thanks go out to OPTiK for running into this issue and letting me know.

The bass boost mod won't work if the O2 amplifier is set up for a gain of 1, ie with R17 and R21 clipped out of the circuit on the low gain switch position (or if R19 and R23 are removed in the high gain position).

The reason is that the bass boost modification increases the voltage gain of the O2 first stage for low frequencies, which means there must be voltage gain there to start with. A gain of 1 means no voltage gain by the non inverting op amp formula [1 + (R16/R17)] = [1 + 0] = 1, essentially a unity gain buffer, so altering the value of R16 and R22 with the boost circuit will produce no effect.

My apologies to anyone else who has run into this. I should have made the effect clearer in the writeup!
 
O2 amplifier increased battery run time modification

This modification allows the O2 amplifier batteries to run bit longer before the power management circuit cuts them off. RocketScientist’s original values for R25 and R9 did a similar thing. This mod puts the values roughly in the middle between his existing O2 values for R25 & R9 and his original values.

The modification drops the trigger threshold for the O2 amp's power management circuit to turn “on” by about 0.75V per battery, from around 8.33Vdc to around 7.59Vdc. The threshold to turn “off” is dropped by about 0.35Vdc, from 7.07Vdc to 6.72Vdc. RocketScientist’s original values of R9 = 40.2k and R25 = 2.74M put the “on” threshold at about 6.95Vdc and the “off” threshold at about 6.33Vdc per battery.

This mod was inspired by a recent post in the main O2 there here:

http://www.diyaudio.com/forums/head...headphone-amp-diy-project-57.html#post3193411

The change involves increasing the value of R9 from 33k to 36.5k and the value of R25 from 1.5M to 2.1M. Suitable resistors from Mouser are part 270-36.5K-RC and 270-2.1M-RC:

270-36.5K-RC Xicon | Mouser

270-2.1M-RC Xicon | Mouser


Details:

In RocketScientist's existing O2 power management circuit, when the MOSFETs are "off", U2 pin 7 is near the negative supply rail. That puts the 1.5M hysteresis resistor R25 essentially in parallel with the 33K R9, to give a parallel resistor value of around 32,290 ohms. This voltage goes to one input of the U2 comparator, pin 2, to be compared to the LED voltage on U2 pin 3.

The photo below shows the measured voltage across the O2's LED at about 1.78Vdc. The photo is taken with the O2 on AC power, but I also tested it on batteries and the LED voltage is nearly the same, as it should be. RocketScientist is using the LED essentially as a voltage reference.

So with the power management circuit just barely "off", the voltage at U2 pin 2 from the R5/(R9 || R25) voltage divider must be right at 1.78Vdc to match that reference voltage from the LED on pin 3. That gives a current through the R9 || R25 (parallel combo) and into R5 of 1.78Vdc / 32290 ohms = 55uA. Multiplying that current by the value of R5 gives a voltage drop across R5 of (55uA)(270K) = 14.88Vdc.

Now adding that voltage back to the 1.78 volts across R9 gives the rail-to-rail voltage at the point where the power management circuit activates and turns "on": 14.88Vdc + 1.78Vdc = 16.66Vdc. Assuming the two batteries are approximately equal, dividing this by two will give the voltage of either battery at the point where the PM circuit turns on as the batteries "recharge" once their load is removed (PM circuit had turned off): 16.66Vdc / 2 = 8.33Vdc.

That voltage matches up with the 8.30 volts measured in the linked post above in the main thread. The batteries in that case were charging back up to 8.30Vdc, just a hair below the 8.33Vdc level need to turn the PM circuit on again, and have it oscillate on and off to signal the batteries are in need of charging.

To find the other trip point to turn the PM circuit "off", with the power management circuit already "on", just work through all the math above but this time with the hysteresis resistor R25 in parallel with R5, the 270K resistor, instead of R9 since the comparator output U2 pin 7 will be near the positive rail when the PM circuit is on. 1.78Vdc across the 33K R9 gives 54uA through R9 and into the 228.8K parallel combo of R5 and R25, producing a (228.8K)(54uA) = 12.34Vdc. Adding 1.78Vdc and dividing by 2, for the two batteries, gives the "turn off" voltage trip point of around 7.07Vdc per battery.

Going through similar math with RocketScientist’s original values of R9 = 40.2K and R25 = 2.74M result in a power management turn “on” voltage of around 6.95Vdc per battery and a turn “off” voltage of around 6.33Vdc each.

8.4V NiMH batteries have a useful charged voltage range of around 8.5Vdc - 8.7Vdc from here:

http://support.radioshack.com/support_tutorials/batteries/Images/nimh-9v-lodis.gif

NiMH 9-Volt Battery Engineering Data Sheet

after which the voltage across the battery drops off rapidly. But as those curves show the discharge rate has a lot to do with it, and various brands of batteries may have slight differences in their per-cell voltage (8.4Vdc is a 7 cell battery x 1.2V per cell). So a different "8.4Vdc" battery may have its useful charged voltage range a bit lower.

This modification changes R25 to 2.1M and R9 to 36.5K to produce a turn “on” voltage of 7.59Vdc per battery and a turn “off” voltage of 6.72. Putting all of this in a table gives, per battery:


R25 = 1.5M, R9 = 33K (current O2 values): turn on = 8.33Vdc, turn off = 7.07Vdc

R25 = 2.1M, R9 = 36.5K (this modification): turn on = 7.59Vdc, turn off = 6.72Vdc

R25 = 2.74M, R9 = 40.2K (original O2 values): turn on = 6.95Vdc, turn off = 6.33Vdc

The 6.72Vdc per battery trip point to turn the power management circuit "off" in this modification matches up with the bottom of the discharge curve for the "8.4Vdc" NiMH batteries in the link above. That works out to be around 6.72Vdc / 7 cells = 0.96Vdc per cell. RocketScientist's initial value of 6.33Vdc to turn off works out to be 0.90Vdc per cell, which is about as low as a NiMH cell should probably be discharged without affecting its lifespan. Note that the discharge graphs for the battery in the link don't even go that low - probably a hint from the manufacturer not to go there. :)
 

Attachments

  • IMG_1631.JPG
    IMG_1631.JPG
    220.8 KB · Views: 647
Last edited:
O2 amp longer runtime #2, low power op amps from BOM

RocketScientist's writeup on the O2 amp includes an option of using a set of low power op amps he specified for increased runtime. It seems the option has been largely forgotten or unused, since as RocketScientist wrote that the distortion and noise floor can be a bit higher, depending upon what one is doing.

However, RocketScientist did not include the effect of headphone sensitivity in his writeup about the low power chipset. This modification shows that for very sensitive headphones, like the AKG K550s at 114dB/V, the distortion levels are well into the "great" range from the dScope measurements he has posted with the low power chips. He shows "very low" distortion up to about 0.7Vrms output swing at 15R output impedance. The K550s only require a maximum of abour 80mV (rms) swing at 32R, well into the low distortion part of his dScope curves.

The writeup is here on his blog,

NwAvGuy: O2 Details

and search for the heading "low power option" and the heading "low power distortion & output" for his dScope distortion graphs with the chipset. The chips involved are listed on any of his BOM's for the O2 amp, under the gain resistor section. The output NJM4556 chips are replaced with TLE2062CPs and the NJM2068 gain op amp replaced with the OPA2277PA.

These two low power chips also have much better input bias current numbers (pA vs. nA) then the NJM chips, resulting in lower DC offset at the output of the O2.

The first photo below shows the new chips. The TLE2062CPs are at the top and the OPA2277PA at the bottom. The next two photos show the quiescent (idle) current of the O2 with the original NJM chips still installed. 22.2mA on one channel and 22.1mA on the other, which matches up perfectly with RocketScientist's mention of around 22mA idle for the NJMs in his writeup. The batteries are freshly charged for all this, both running at about 9.1Vdc.

The next two photos show the new idle current with RocketScientist's low power chipset installed. 4.8mA on both channels, representing a 78% reduction in idle current. RocketScientist says "below 8mA" in his writeup. I'm at a loss as to where he came up with the extra 3.2mA of current usage. The idle current is significantly more than the current used by the headphones, as RS notes, so this idle current reduction represents a huge decrease in current draw. Assuming around a 30mV normal (rms) listening level on the AKG K550s, that is just (30mV / 32R)(2 channels) = 1.9mA, a drop in the bucket next to that 22mA - 4.8mA = 17mA idle current reduction!

Assuming Tenergy is lying about the 250mAhr rating on the 8.4Vdc cells (I've been working with high power LED flashlights too long, lol) and they are really 220mAhr or less, a 17mA drop in current per battery represents 220mAhr/4.8mA - 220mAhr/22mA = 35 hour increase in run time! That means charging your O2 once every 4 days rather than every day.

The next two photos show the reduction in output DC offset into the headphones. 380uV and 1.03mV vs. about 3mV per channel with the NJM4556 output chipset. That is better than a 3:1 reduction.

I have left the output resistors at 1R, rather than reduce them to 6R as RocketScientist mentions in his low power write, since the very low voltage swing needed for the AKG K550s would still result in low chip to chip balancing currents even at 1R each. For headphones requiring larger voltage swings, closer to that 0.7Vrms "knee" in his dScope distortion curves, using the 6R output resistors would probably reduce distortion slightly.

All measurements aside, the final photo is some actual listening time. :) Sounds great! In this case the O2 amp is being fed with the HiFimeDIY ODAC clone here, into the K550 headphones, O2 running on batteries.

I should re-iterate that this mod is assuming very sensitive headphones or IEMs, like these 114dB/V AKGs. The mod would probably work very well down to a sensitivity of 110dB/V or so. Lower than that you will run into all the issues that RocketScientist originally wrote about for this low power chipset, namely significantly increased distortion due to the higher voltage swing requirements and higher noise. This low power chipset would not work well at all with my 100dB/mW Shure SRH940s, for example. Those need closer to 1.8V rms max. From RocketScientist's dScope curve for the low power chipset you can set the distortion starts rising exponentially (at 15R) around 0.7Vrms output.
 

Attachments

  • IMG_1655.JPG
    IMG_1655.JPG
    203.3 KB · Views: 96
  • IMG_1651.JPG
    IMG_1651.JPG
    213.5 KB · Views: 78
  • IMG_1650.JPG
    IMG_1650.JPG
    214.7 KB · Views: 69
  • IMG_1649.JPG
    IMG_1649.JPG
    232 KB · Views: 85
  • IMG_1648.JPG
    IMG_1648.JPG
    231 KB · Views: 578
  • IMG_1647.JPG
    IMG_1647.JPG
    229.3 KB · Views: 586
  • IMG_1646.JPG
    IMG_1646.JPG
    231.6 KB · Views: 610
  • IMG_1644.JPG
    IMG_1644.JPG
    203.9 KB · Views: 617
Last edited:
O2 amp power mgmt latch modification - no oscillate

This O2 headphone amplifier modification adds a latching circuit that keeps the power management circuit "off" permanently - until the amp power switch is turned off and back on to reset - regardless of any re-charge behavior by the batteries.

The battery re-charge issue is described in the posts above about hysteresis. Once the load is removed from the batteries when they get low, when the power management circuit turns off, the battery voltages tend to rise on their own which can turn the circuit, and hence the whole O2 amp, back on again. The O2 then just oscillates on and off like that making a noise in the headphones.

This modification adds a couple of mosfets across the 33K R9 resistor in RocketScientist's schematic. When the power management circuit turns off, the gate of mosfet Q1 goes high (to the positive rail, whatever voltage that is at the moment). That signal now goes to the gate of a new mosfet in this modification that in turn "shorts out" R9, bringing pin 2 of the comparator U2 to the negative power rail. Lowering that pin voltage forces the gate of Q1 high, which is already is, thereby creating a latching feedback loop. If the battery voltage subsequently rises it will have no effect, since that would ordinarily increase the voltage at U2 pin 2 by the action of the R9 / R5 voltage divider. With U2 pin 2 now latched at the negative rail voltage the power management circuit will stay off permanently, regardless of battery voltage, until the O2 power switch is turned off and back on.

The one trick to the circuit is that RocketScientist has the power management circuit turn off briefly when the O2 is first turned on to prevent a turn-on pop in the headphones. C1 C16, and C21 are part of that function. This action would trip the lock circuit and lock out the O2 from ever turning on in the first place without the addition of an initial time delay, which is formed from the second mosfet in series with the first across R9. The time delay RC circuit feeding the gate of that mosfet produces an initial power-on delay in the latch circuit of 1 -2 seconds (1 sec = 24Vdc rails on AC, 2 sec = 14Vdc with nearly dead batteries). The two mosfets in series logically "AND" the power-on time delay and latching signals as a condition to short U2 pin 2 to the negative power rail.

The first circuit below shows how the new parts are wired into the O2 power management circuit. The second circuit is just an LTSpice test circuit for function.

Suitable parts from Mouser are:

Two n-channel mosfets: these MUST be +/-30Vdc gate mosfets. A 2N7000 with a +/-20Vdc gate will not work.
FQN1N60CTA Fairchild Semiconductor | Mouser

Resistor: 4.7M 1/8W 1%
RN55D4704FRE6 Vishay/Dale | Mouser

Capacitor: don't use an electrolytic for this. The leakage is too great and will prevent the circuit from working correctly with the tiny currents involved. Use a multilayer ceramic capacitor (MLCC), 2.2uF, X7R, 10%, 50V
FK20X7R1H225K TDK | Mouser

The bad news is that I don't see any way to fit this modification into the standard B2-080 O2 chassis, at least using these through-hole parts. The slightly taller B3-080 case would have to be used. All 4 parts in the circuit have surface mount versions though. It may be possible to fit a SMD layout for the the circuit underneath the O2 PCB in the B2-080 case, and wire the board into the bottom of the O2 PCB, although I haven't tried it. :)

For a power managment circuit "on" indicator, wire a small 3mm led in series with a resistor to run at 1mA or so, across C8. Then just drill a small hole in the O2 front panel and glue the LED into the hole. When the PM circuit is on the LED will be on, and when off and latched the LED will be off. The same thing could be done with a second LED across C9 (careful to observe polarities!) to show the same for the negative rail and Q2.
 

Attachments

  • O2 amp pwr mgmt latch tst ckt.png
    O2 amp pwr mgmt latch tst ckt.png
    37 KB · Views: 151
  • O2 amp pwr mgmt latch circuit 1.png
    O2 amp pwr mgmt latch circuit 1.png
    55.5 KB · Views: 183
Last edited:
O2 power mgmt latch mod does fit under the PCB

It turns out that the RocketScientist O2 amp power management circuit latching modification that I posted above does fit under the O2 PCB in the standard B2-080 case just fine - if you are really good with a soldering iron adn some DIY work. :)

The purpose of the latch circuit, again, is to prevent the natural voltage rise that happens with the batteries when the power management circuit cuts the load off from reaching a level where the pwr mgmt circuit turns back on, hence oscillating

I added the circuit to the bottom of the 8-cell AAA NiMH O2 modification that I posted in another thread to test it. Works great, with one thing that I'll call a feature rather than a bug. Once the latch circuit trips, the O2 amp must be powered off for a full 2 minutes to reset it. It takes that long for the RC time delay lockout circuit in the latch circuit to bleed down. If the O2 is turned back on in less than two minutes the mosfets will just remain off.

All of this should be done on a antistatic mat with grounded wrist strap, grounded tools, grounded soldering iron tip, etc. to keep from blowing the mosfet gates.

The first two photos below shows the resistor and axial capacitor that form the RC time delay circuit soldered across the outer pins of R5 and R9 under the O2 PCB. When R9 was soldered in I left the leads long so I could solder the mosfets to them.

The mosfets source and drain leads are bent at right angles, then the drain of one soldered to the source of the other. Then the mosfet pair are soldered onto the free R9 leads and the leads trimmed.

Finally one mosfet gate lead is soldered to the connection point of the RC delay circuit, while the other is bent around and soldered to the nearby gate connection of Q1, as shown in the 3rd photo.

The whole thing clears the bottom of the B2-080 case just fine, as the last photo shows.
 

Attachments

  • IMG_1732.JPG
    IMG_1732.JPG
    200.7 KB · Views: 121
  • IMG_1733.JPG
    IMG_1733.JPG
    197 KB · Views: 122
  • IMG_1740.JPG
    IMG_1740.JPG
    190.3 KB · Views: 167
  • IMG_1752.JPG
    IMG_1752.JPG
    154.3 KB · Views: 181
Last edited:
I have 2 headphones, an Audio Technical CKS77 and Sen HD650. Using the O2 to power both, I found that it is a bit soft and muddy with HD650. I have been waiting for the desktop version of O2, I think it will take more time... so I done a little test by supplying the O2 with Salas SSLV shunt regulator. I don't have tools to measure but I tell it is cleaner, louder and bass is better.

Here is the setting:
+12.16 320mA
-12.19 290mA
 

Attachments

  • IMG-20121019-00666.jpg
    IMG-20121019-00666.jpg
    356 KB · Views: 210
coolhead: interesting! The two major things the different power supply would probably affect are output impedance vs. frequency (looking back into the power supply output pins) and noise generation.

The MC/LM7xxx regulators are fairly noisy vs. some other regulators, but both RocketScientist with the O2 and opc with the Wire amp said they found it made no difference on their dScope/AP testers.

The output impedance of the MC/LM78xxx goes up a bit with frequency, but RocketScientist has the supplies bypassed with the 220uF caps.

So I'm not sure what is making the difference. But I'm always of the mindset that measurements are great, and important, but in the end it is how something sounds that matters! :)
 
Any ideas for adding upc1237 for DC protection?

Any ideas for adding upc1237 for DC protection? I'm not sure if +ve thresh hold 0.7V & -ve threshold of -0.2V would be of any use in h/p as its basically a spk. protection IC or will it require some sort of DC instrumentation amp (to amplify few mV DC offset to couple of V for upc1237 to detect).
Ragards,
availlyrics
 
will it require some sort of DC instrumentation amp (to amplify few mV DC offset to couple of V for upc1237 to detect).

The thrshold levels in that IC at 700mA are too high for headphones, but like you say a 10x amp circuit that would bump 70mA DC at the phones up to 700mA DC for the chip would work. That is essentially what AMB's e-12 protection circuit is doing

The ε12 Muting / Protect Circuit

so that 70mA DC input gets amped up 10x to 700mA DC to turn on the B-E junction of the pulldown transistors.

A couple of years ago I whipped up a schematic for a "super e-12" that had separate low pass filters for each channel. The filters were upgraded to 3-pole active for a very sharp cutoff. I think I lowered the corner frequency from 1Hz to 0.1Hz as I recall, and since there were two stage of gain on each channel (for the filters), was able to lower the trigger voltage to 7mV at the phones (100x gain) instead of 70mA.

I'll poke around and see if I find that over the holidays and post, in case someone wants to give it a try. I've never posted it before but have sent AMB a copy. He is trying to keep the e-12 to a certain size to fit in a chassis and this version would wind up substantially larger physically.
 
O2 amp dual-rail voltage doubler and quadrupler for +/-15Vdc rails and beyond

This modification of RocketScientist's O2 amp adds a few diodes and capacitors to the power supply to create a dual-rail voltage doubler and dual-rail voltage quadrupler using the same AC input transformers. The net result is that a 9Vac transformer with 2A secondary will produce +/-18Vdc rails going into the regulator chips (120mA load per channel), allowing for the use of 15V regulators and supply rails (MC7815 and MC7915).

A note on terminology first. RocketScientist's power supply in the O2 is already a voltage doubler if you look at it just right. It all depends on where the ground reference is placed (negative DMM lead for the discussion here). If the ground reference is in the middle, at one end of the transformer secondary as in the O2, you get dual rail 1x, which is 9Vac rms X 1.414 for sine = 12.7Vdc. Subtract a couple of volts ripple for the half wave rectification and another volt for AC line fluctuation you get +/-9.7Vdc into regulators. If however you move the ground reference to the negative rail being produced, the result is a single ended votlage doubler of 9.7V x 2 = 19.4Vdc.

So the "dual rail voltage doubler" circuit presented here is really a "single ended voltage quadrupler" if the ground reference is moved from the middle to the negative rail. Similarly the "dual rail voltage quadrupler" here is a "single ended 8x" with the ground reference moved.

Another note is in order on currents and voltages. Anyone building these circuits should pay close attention to the ripple currents needed in the capacitors at each stage and make sure cap voltage ratings around around 20% higher than the peak voltage, including line voltage fluctuations, at each stage. And finally, a voltage doubler or quadrupler can produce some high DC voltages if higher AC voltages are used on the input. Use caution! 24Vac into a dual rail voltage doubler will produce around +/- 60Vdc.

The first circuit and plot below show the dual rail voltage doubler. As shown a 9Vac rms transformer with a 2A secondary, like this one

9 Volt 9V AC 2000mA Adapter for Line 6 PX 2 US Pod Power Supply Cord Charger PSU | eBay

will produce around +/-18Vdc rails going into the regulators with a 120mA load on each channel. 15V regulator chips can then be used, along with appropriate resistor value changes in the PM circuit and clip on heat sinks on the output chips (as shown in the first couple of posts in this thread).

Green and blue in the plot are the voltage outputs of the voltage doubler, going into the regulators. Red is the AC input voltage. Volt scale is on the left. The aqua and magenta current plots using the righthand scale are the currents through the input capacitors, C2 and C6. Half the current waveform looks distorted since these caps charge on one half cycle and discharge on the other. The grey plot is the current spike through the input transformer, showing the need for the 2A secondary.

The second circuit is less of an O2 modification and more of a separate DIY project for someone with high impedance (600R) low efficency cans. The plot shows the result of 12Vac into a dual rail voltage quadrupler circuit, producing +/-60-Vdc going into the regulator. Regulator chips are not available at these voltages, so a simple discrete regulator pair as in the next attachment would have to be used. This one is a simple capacitance multiplier fed with a current sourced zener. The high voltage +/-60Vdc rails here can be used with amps such as opc's "wire class A/AB" amp based on the LME49830 chip and mosfets that goes up to +/-90V rails

http://www.diyaudio.com/forums/soli...amplifier-based-lme49830-lateral-mosfets.html

In both the dual rail doubler and dual rail quadrupler the circuit just replaces D3 and D4 on the O2 PCB, as shown. In the case of the voltage doubler the parts cold be built up on an upside down PCB slid into the top slot on the taller B3-080 case, with wires going to the 4 holes where D3 and D4 would have gone, plus the one ground connection on the AC input jack.
 

Attachments

  • O2 volt quad 12Vac rms in 50mA load plot.png
    O2 volt quad 12Vac rms in 50mA load plot.png
    28.2 KB · Views: 385
  • O2 volt quad 12Vac rms in 50mA load circuit.png
    O2 volt quad 12Vac rms in 50mA load circuit.png
    63.2 KB · Views: 389
  • O2 volt doubler 9Vac rms in 120mA load plot.png
    O2 volt doubler 9Vac rms in 120mA load plot.png
    31.2 KB · Views: 391
  • O2 volt doubler 9Vac rms in 120mA load circuit.png
    O2 volt doubler 9Vac rms in 120mA load circuit.png
    64.2 KB · Views: 399
  • high voltage O2 regulator.png
    high voltage O2 regulator.png
    41 KB · Views: 368
Last edited:
O2 amp dual rail power management modification

Here is a dual-rail version of RocketScientist's power management circuit in the O2 headphone amp for anyone into some DIY modding work. :) This circuit essentially takes RocketScientist's design and just applies a copy to each power supply rail individually rather than measuring the two rails summed together.

One benefit is the ability to detect a single supply failure that doesn't add up (both rail voltages added together in the original O2 PM circuit) to less than 14Vdc where the original PM O2 circuit would trip. An example from a recent post in the O2 thread is using a DC power supply instead of AC. One rail is on AC while the other is one battery, allowing the battery to discharge too low to about 3Vdc. The dual rail circuit here would catch that condition - or any unbalanced rail condition like a battery with partially shorted cells inside - and shut down the O2.

The circuit adds reversed biased diodes across each rail to ground to take care of a situation where one rail is left floating and held up to the opposing rail by rail-to-rail parts, like using a DC supply with no batteries installed.

The circuit incorporates the latch modification earlier in the thread to latch the PM circuit off once it trips. This eliminates any chance of the circuit oscillation when the battery runs down, original PM circuit trips, then the battery recharges a small amount and turns the PM circuit back on. I've left off RocketScientist's hysteresis resistor R25 since it isn't needed with the latch circuit. With the latch the O2 will remain off until the power is switched off and left off for about 15 seconds to reset the latch circuit.

I've also left off the capacitors on the mosfet gates to slow turn on, the "turn on thump" preventer. I'm assuming that a headphone relay cutout circuit like posted earlier in this thread would be used instead to lockout turn on thump. But the caps can be added back if needed.

This dual-rail circuit has two power leds, one for each rail, which also act as voltage references for each rail just as with RocketScientist's original design. The LEDs would also help diagnose missing power supply inputs on one rail or the other.

The output uses a mosfet optical solid state relay, an Omron G3VM-62C1 to improve the syncronization and hence quality of the PM switch-on and switch off. RocketScientist's original design tries to keep the gate switching of the two individual mosfets synchonized so both rails switch at exactly the same time. But in actual practice a lot of posts have noted that turn on thumps are solved by replacing one mosfet or the other, meaning manufacturing differences (and possibily installation static damage) are causing switching time differences with the mosfets.

The Omron SSR is an integrated DIP-8 unit so the two mosfets will be highly matched. Easily replaced, too, since the whole thing fits in a standard DIP8 socket. The mosfets are optically triggered by two LEDs in the package that are placed in series between the outputs of the voltage comparators on each rail, fed with two mosfets in a similar fashion to RS's original design. So now the mosfets don't supply the power rails directly, but rather together turn on the SSR control LEDS and simultaneously switch both SSR mosfets. If one comparator chain or the other is slightly off in time it doesn't matter since the SSR LEDs are in series. First one wins. The constant current source maintains the datasheet recommended 7.5mA current through the SSR LEDs through the whole power supply range of 12Vdc to the trigger point of 7Vdc on each rail. The downside is 7.5mA of additional current draw, about the same as one of the op amps, which will run the batteries down slightly sooner. Another external LED is included in series with the SSR LEDs as a PM trigger indicator that comes for "free" in terms of current draw. The PM "on" LED indicator should help with diagnosing problems.

The LEDs in the schematics for the voltage reference are just something that was living in LTSpice. For the actual circuit use the LED in the O2 and adjust the series resistor accordingly for the same current from one rail rather than rail to rail. The voltage divider resistors get adjusted for the 1.7Vdc of the O2 LED.

The first plot below shows the positive battery (green trace) discharging from from +9.5Vdc intial down through the +6.5Vdc trip point of the comparator and down to +6.3Vdc. The orange trace is the current through the SSR LEDs, showing the switch at the trigger point. The battery then recharges back up through the trigger point to +7.5Vdc, but the SSR remains off due to the latching circuit.

The second plot shows the same thing for the negative rail (blue trace). The negative rail circuit has an extra comparator to make an inverted copy of the output signal for the negative rail latch circuit. The third plot shows the effect of a sudden complete disconnection or failure of one supply (loose battery terminal, one battery removed, AC voltage regulator failure). The SSR switches off as it should.
 

Attachments

  • O2 V2.0 PM plot C.png
    O2 V2.0 PM plot C.png
    18.6 KB · Views: 73
  • O2 V2.0 PM circuit C.png
    O2 V2.0 PM circuit C.png
    52.1 KB · Views: 73
  • O2 V2.0 PM plot B.png
    O2 V2.0 PM plot B.png
    21.2 KB · Views: 70
  • O2 V2.0 PM circuit B.png
    O2 V2.0 PM circuit B.png
    53.2 KB · Views: 86
  • O2 V2.0 PM plot A.png
    O2 V2.0 PM plot A.png
    20.1 KB · Views: 100
  • O2 V2.0 PM circuit A.png
    O2 V2.0 PM circuit A.png
    53.4 KB · Views: 115
Last edited:
I'll poke around and see if I find that over the holidays and post, in case someone wants to give it a try. I've never posted it before but have sent AMB a copy. He is trying to keep the e-12 to a certain size to fit in a chassis and this version would wind up substantially larger physically.

It would be very helpful indeed if you can post your design here.
My plan is to build desktop ver. of O2 with stepped attenuator+ DC offset amp ( preferably opamp based over discrete design)+upc1237 with mute relays.
Thanks,
availlyrics.
 
Last edited:
It would be very helpful indeed if you can post your design here.
My plan is to build desktop ver. of O2 with stepped attenuator+ DC offset amp ( preferably opamp based over discrete design)+upc1237 with mute relays..

Here you go, below. :)

I found my old design with the single 3rd order filter but I had done it in a schematic editor rather than LT Spice. While entering it in I had a better idea. Instead of a single piece circuit, here is an "e-12 booster" that could simply be added in front of (electrically, before the e-12 inputs) existing AMB e-12 board(s) stacked vertically with spacers between the screw holes. I'm a big one on not re-inventing the wheel. Most of parts needed are just the existing e-12 (power supply, relay and relay driver, and now-secondary filter on each channel). This modification just adds a front-end on each channel that does several things:


  • Separate filters for each channel all the way to the pull down transistors to prevent the (highly unlikely, granted) chance of the two channel's DC offsets summing to below the trip point. For example, if one channel DC offset is +90mV and the other -30mV the existing e-12 would sum those to +60mV, below the 70mV trip point, and not trip. This modification treats each channel separately and compares each independently against the (new, lower) 7mV trip point.
  • Added an additional cascaded filter stage on each channel to convert the filter from second order to 3rd order. The original e-12 has a first order active low pass followed by a discrete low pass (the 1K and 100uF) to make a cascaded second order LP filter. The higher 3rd order filter increases the roll-off slope from -40dB/decade in the original e-12 to -60dB/decade here for an even more definite cut-off.
  • Filter feedback caps increased from 1uF to 10uF XLR MLCC to drop the corner frequency from about 1Hz in the original e-12 to 0.1Hz. This helps with handling very low frequency inputs like 10Hz. The filter gain and phase plots won't drop smoothly if the corner frequency gets too close to the minimum signal frequency.
  • The added filter stage on each channel has a gain of 10x, increasing the total cascaded gain from 10x to 100x, and thus dropping the DC offset trigger voltage being detected on the headphone amp output from 70mV to 7mV. The O2 typically runs at 3mV DC offset per channel, less than the new 7mV trip voltage, so it would not erroneously trigger the circuit in normal operation. In conjunction with this the op amps specified for the new stages are very low inherent DC offset OPA627 units. The OPA627 is a precision DC op-amp which has about 0.1mV (100uV) of offset vs. 3 - 15mV for the TL082. This becomes especially important both with the small DC offset being detected (7mV) and with cascading gain stages. A 0.1mV input offset voltage become a 10x0.1 = 1mV output offset on the first stage, and 1x10 = 10mV after the second cascaded stage. The larger DC offset of the TL082 makes a smaller difference in the second stage. However, another change, I'm also specifying TL082B chips with have a maximum input offset of 3mV, vs up to 15mV for the base TL082 to help keep the second stage input offset under control. The OPA627 is also extremely expensive, about $30 each for the high grade part (lowest offset) and $20 each for the low grade part, meaning each op amp costs about as much as all the basic O2 amp parts combined. That would probably make RocketScientist bang his head against a wall if he were still around, since his goal was always low cost and affordability. :) There are some (much) lower cost chopper stabilized op amps out there with even 10x better offset numbers, but I wanted to keep yet-another filter out of the filter circuit here and wanted something that didn't chop as was all through hole for easy DIY.
In the schematic below the parts in the outline called "AMB e-12 #1" are one e-12 board fully stuffed with the regular parts except all 10K input resistor left out but one. The second stacked e-12 board labelled "AMB e-12 #2" is a short-stuffed board with the power supply and relay parts left off. The power and ground lines are jumpered over from board 1. Just the filter and open collector transistor are used here, with the open collectors tied to those on the first board to form a wire-OR circuit.

All of this can just be DIYed too, of course. The new parts with the OPA627s could probably be laid out on a board sized to stack vertically with the other two e-12 boards. If someone were laying out a PCB with the two new OPA627 filter stages it might be easiest just to add the parts shown for the second e-12 PCB, the TL082B and open collector transistors, reducing things to just the one new board and the one existing e-12 board.

And a link to AMB's e-12: The ε12 Muting / Protect Circuit
 

Attachments

  • e-12 booster for AMB's e-12 protection circuit.png
    e-12 booster for AMB's e-12 protection circuit.png
    65.9 KB · Views: 93
Last edited:
e-12 booster relay protect - part 2, applied to O2 amp

Here is some more info on the "e-12 booster" above. If used with the O2 amp the O2's power supplies can be used, as shown in the circuit below. Just don't populate the power supply and virtual ground parts on e-12 PCB #1. Wire the +12Vdc, -12Vdc, and ground over from the power switch on the O2.

Also a typo warning on the circuit I attached above - it has the negative rail attached to ground in the top half when I did a cut and paste. The negative rail should attach as shown here.

I've added the device models for the OPA627 and TL084 in the LT Spice sim below. The first plot is the AC response from 0.01Hz to 200kHz, showing the 0.1Hz corner frequency and 20dB/decade, 40dB/decade, and 60dB/decade rolloffs for the output of the 3 cascaded filter sections. Green, blue, and red respectively. Labels are at the top of the plots.

The second plot shows the transient response when a fault of 20mV occurs at the 1 second point (20mV of DC shows up on one headphone output). The circuit takes about 2.3 seconds to switch the relay, which is the transient delay through all the filters. The the input signal is in green (2.5khz so it looks nearly solid here) and relay current is in purple. The output of filter section 3 (1K in series with the 100uF), the aqua plot, clamps at 0.7Vdc since it is feeding a transistor B-E junction, of course. Voltage scale on the left and relay current scale on the right.

The final two transient plots show the more "difficult" cases for the filters of lower signal frequencies at 150Hz and 20Hz. The filter delay at 150Hz is still about 2.3 seconds until the relay switches off, while at 20Hz it shrinks down to about 1.5 seconds.
 

Attachments

  • e-12 booster with O2 pwr circuit.png
    e-12 booster with O2 pwr circuit.png
    62.8 KB · Views: 105
  • e-12 booster with O2 pwr AC plot.png
    e-12 booster with O2 pwr AC plot.png
    29.5 KB · Views: 71
  • e-12 booster with O2 pwr transient plot 20mV at 1sec.png
    e-12 booster with O2 pwr transient plot 20mV at 1sec.png
    19.8 KB · Views: 47
  • e-12 booster with O2 pwr transient plot 20mV at 1sec 150Hz.png
    e-12 booster with O2 pwr transient plot 20mV at 1sec 150Hz.png
    29.7 KB · Views: 59
  • e-12 booster with O2 pwr transient plot 20mV at 1sec 20Hz.png
    e-12 booster with O2 pwr transient plot 20mV at 1sec 20Hz.png
    38.3 KB · Views: 56
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.