ctechdx,
I don't know the reasoning behind MooCow's suggestion of OPA1688. I can tell you that the original designer tested many different opamps in the O2 and standard choices are known to work well in the design. NwAvGuy was very clear that the specific opamp is less important than implementation. Maybe start with the O2 as designed and then try different opamps after you know what it sounds like?
NwAvGuy: Op Amps: Myths & Facts
As most IEM's are fairly sensitive, I would certainly suggest you take a look at the gain settings. The O2 has 2 gain settings, and I think in your case one of them should likely be 1X. This is achieved by leaving R17 and R21 on the board blank.
As for the O2 being "analytical" or "cold" it was designed to be transparent - that is that the amplifier should not be affecting the sound. The idea is to take the amp out of the equation and hear the music as it was recorded along with whatever coloration your headphones provide. Since coloration can't be avoided with headphones my goal is always a transparent amp with headphones that provide the sound signature I like.
I don't know the reasoning behind MooCow's suggestion of OPA1688. I can tell you that the original designer tested many different opamps in the O2 and standard choices are known to work well in the design. NwAvGuy was very clear that the specific opamp is less important than implementation. Maybe start with the O2 as designed and then try different opamps after you know what it sounds like?
NwAvGuy: Op Amps: Myths & Facts
As most IEM's are fairly sensitive, I would certainly suggest you take a look at the gain settings. The O2 has 2 gain settings, and I think in your case one of them should likely be 1X. This is achieved by leaving R17 and R21 on the board blank.
As for the O2 being "analytical" or "cold" it was designed to be transparent - that is that the amplifier should not be affecting the sound. The idea is to take the amp out of the equation and hear the music as it was recorded along with whatever coloration your headphones provide. Since coloration can't be avoided with headphones my goal is always a transparent amp with headphones that provide the sound signature I like.
Part of this reputation might also stem from people accidentally sending the gain stage into clipping - after all, it cannot support full CD player level at 6.5X at all, and at 2.5X it generally still requires mains power.
6.5X also is a bit hard on the stock NJM2068 in general, which is an oldschool low-noise opamp without an awful lot of output current capability that's breaking a bit of a sweat driving little more than 1.5 kOhms at high levels. In NwAvGuy's own opamp testing at a gain of 7X you can see how the more powerful NE5532 and OPA2134 managed to wring out about half a volt higher clipping level and also showed noticeably less high-frequency (crossover) distortion at 2 Vrms out - while the notoriously wimpy TL072 is in epic fail territory altogether. When sticking with the '2068 or the low-power option (OPA2277) at high gain, I'd definitely advise going easy on feedback network resistor values - maybe a factor of 2 higher or so. If you absolutely have to use a TL072 or 4558, it's more like a factor of 4, with some matching of input resistor value - and stack two of 'em.
Optimizing the input stage for minimum noise was a bit of a questionable move in the original design. Try finding an output that's anywhere near that level. A bias slightly more towards level handling would have been better.
6.5X also is a bit hard on the stock NJM2068 in general, which is an oldschool low-noise opamp without an awful lot of output current capability that's breaking a bit of a sweat driving little more than 1.5 kOhms at high levels. In NwAvGuy's own opamp testing at a gain of 7X you can see how the more powerful NE5532 and OPA2134 managed to wring out about half a volt higher clipping level and also showed noticeably less high-frequency (crossover) distortion at 2 Vrms out - while the notoriously wimpy TL072 is in epic fail territory altogether. When sticking with the '2068 or the low-power option (OPA2277) at high gain, I'd definitely advise going easy on feedback network resistor values - maybe a factor of 2 higher or so. If you absolutely have to use a TL072 or 4558, it's more like a factor of 4, with some matching of input resistor value - and stack two of 'em.
Optimizing the input stage for minimum noise was a bit of a questionable move in the original design. Try finding an output that's anywhere near that level. A bias slightly more towards level handling would have been better.
Last edited:
Interesting points!
I looked up the data sheet for the 2068 and you're right, the maximum output voltage drops off a cliff below 1k or 1.5k Ohm load. Clearly NwAvGuy pushed it to the limit in the interest of minimizing noise. What effect does the volume pot setting have on the total load that the gain stage opamp sees? For his tests he rounded things out to 1.3k Ohms. Is there any other consequence other than a bit higher noise to increase the values of the feedback resistors?
I suppose if NwAvGuy was going to prove his point he wanted the measurements as impressive as possible. To your point though, even he admitted that the output stage dominates the noise in the circuit.
I looked up the data sheet for the 2068 and you're right, the maximum output voltage drops off a cliff below 1k or 1.5k Ohm load. Clearly NwAvGuy pushed it to the limit in the interest of minimizing noise. What effect does the volume pot setting have on the total load that the gain stage opamp sees? For his tests he rounded things out to 1.3k Ohms. Is there any other consequence other than a bit higher noise to increase the values of the feedback resistors?
I suppose if NwAvGuy was going to prove his point he wanted the measurements as impressive as possible. To your point though, even he admitted that the output stage dominates the noise in the circuit.
There is another question I have been wanting to ask about the O2/ODAC, let me know if this is the wrong forum for it.
I'm using both of the outputs on the ODAC circuit board. I wired one of them to the pins on the underside of J2, and the other outputs to panel mounted RCA jacks so that I can use the DAC in my home stereo system.
When using the RCA outs with the stereo I can leave the O2 off, but the ODAC outputs are still wired to J2. Is this arrangement reducing the impedance that ODAC sees? For example, if both my stereo and my O2 have a 10k Ohm input impedance, is the ODAC seeing 5k Ohm when the RCA plugs are connected? Does this even matter? It does seem that I get better sound through the stereo when I turn my iTunes volume down a bit. The previous discussion has me thinking maybe the ODAC is clipping because the load is too difficult.
If so, I could plug a dummy 1/4 inch to mini plug adaptor into J2 to short the inputs when using RCA outs.
I'm using both of the outputs on the ODAC circuit board. I wired one of them to the pins on the underside of J2, and the other outputs to panel mounted RCA jacks so that I can use the DAC in my home stereo system.
When using the RCA outs with the stereo I can leave the O2 off, but the ODAC outputs are still wired to J2. Is this arrangement reducing the impedance that ODAC sees? For example, if both my stereo and my O2 have a 10k Ohm input impedance, is the ODAC seeing 5k Ohm when the RCA plugs are connected? Does this even matter? It does seem that I get better sound through the stereo when I turn my iTunes volume down a bit. The previous discussion has me thinking maybe the ODAC is clipping because the load is too difficult.
If so, I could plug a dummy 1/4 inch to mini plug adaptor into J2 to short the inputs when using RCA outs.
Yes he did! In fact I believe he wrote it was the limiting factor on both noise and distortion. So the way around that is to replace it (them, the NJM4556A chips).
Either with the OPA1688 chips or my O2 Booster Board creation (OPA827 + LME49600 looped). That comment of his was one of my big motivations in making the Booster Board, although I haven't been able to do a dScope test to verify the THD really is any lower than the NJM4556As, just going by datasheet numbers and toy (QA401) THD testers.Yep. I asked that question to George Boudreaux (co-creator of the ODAC) once, was curious myself, and if memory serves he said that the 10K input of the O2 was the minimum really that the ODAC should be driving. I forget exactly what he said the consequences were now of going below 10K, but I think it may have been a small increase in THD. Ideally you should add a current buffer between the ODAC and the RCA jacks if the O2 input remains attached. That would also isolate the RCA cable capacitance.
Last edited:
The limit also is somewhat asymmetrical. Look at the Maximum Output Voltage vs. Temperature graph, it says about +13.7...+13.8 V and -12.6...-12.7 V for a 2 kOhm load on +/-15 V. On Maximum Output Voltage Swing vs. Io, you can see that this levels out around the 40 mA mark, at which point the positive side proceeds to drop like a rock - which is where that steep drop around the 100 ohm mark is coming from.
When these parts were designed in the '80s, the understanding probably was that they'd be used in low-level or relatively high-gain circuits. And it's not like they'd have been wimpy at the time - the NJM4565 of similar vintage is quite comparable. Wringing maximum dynamic range out of them would require a buffer.
Not much, it drops load impedance from about 1.49 to 1.45 kOhms at 6.5X.
C19/20 will eventually have to be adapted lest bandwidth be reduced too much, as well as R3/R7 and C11/12 in the interest of +/- impedance matching. (You ideally want (R7 + Zsource)||C11||R14 to match R17/19 || R16 || C19.)
Thanks for the reply agdr!
I will look for your posts regarding the O2 booster board. I'll also look up the data sheets for the OPA1688. A brief look seems promising, but indicates it appears to be surface mount only. I guess that would explain why it wasn't considered in the first place.
It's great to have some insider information about the ODAC output Bear with me though, I didn't know the first thing about electronics 6 weeks ago. You are confirming my suspicions about the O2 contributing to a reduced overall input impedance when using RCA outs?
Interesting that 10k is the limit for ODAC. Another example of NwAvGuy apparently optimizing for noise performance in the O2. What if I just bumped the input impedance up to 47k?
I know that ODAC uses a charge pump to generate the negative rail. I don't know much about those but it seems intuitive that this arrangement wouldn't be happy having to push much current.
I did try listening to the ODAC in my stereo last night with the dummy connector plugged into J2. Although it could be my imagination things certainly seemed to sound better.
I will look for your posts regarding the O2 booster board. I'll also look up the data sheets for the OPA1688. A brief look seems promising, but indicates it appears to be surface mount only. I guess that would explain why it wasn't considered in the first place.
It's great to have some insider information about the ODAC output
Bear with me though, I didn't know the first thing about electronics 6 weeks ago. You are confirming my suspicions about the O2 contributing to a reduced overall input impedance when using RCA outs?Interesting that 10k is the limit for ODAC. Another example of NwAvGuy apparently optimizing for noise performance in the O2. What if I just bumped the input impedance up to 47k?
I know that ODAC uses a charge pump to generate the negative rail. I don't know much about those but it seems intuitive that this arrangement wouldn't be happy having to push much current.
I did try listening to the ODAC in my stereo last night with the dummy connector plugged into J2. Although it could be my imagination things certainly seemed to sound better.
Wow, thanks for the explanation sgrossklass. With your guidance I'm making a little more sense of the data sheet. I wouldn't have recognized that asymmetry.
With the last point, I'm following you in terms of bandwidth but I think the impedance matching is a new idea for me. I'll look into that.
Thank you!
With the last point, I'm following you in terms of bandwidth but I think the impedance matching is a new idea for me. I'll look into that.
Thank you!
Hey excellent analysis! This particular issue about the O2 has been bugging me for a long time. It does seem that NwAvGuy really did over-do optimizing for low noise with the low value feedback resistor set.
I've just put together two alternate O2 headamp resistor sets with all the values and Mouser part numbers, plus the changes for the compensation capacitors, if anyone is interested, over here in an old O2 modification thread:
http://www.diyaudio.com/forums/head...rc-diode-cap-heatsink-mods-3.html#post4897870
The first set is based on a 2.37K feedback resistor and adjusted gain resistors, resulting in a worse-case load on the NJM2068 of 2.1K with the 6x gain case. The second set based on a 3.01K feedback resistor results in a worst-case NJM2068 load of 2.5K at 6x gain.
If anyone spots any errors in the calculations, please post!
Last edited:
That is the OPA1622 - luckily the OPA1688 is in a SOIC-8 package.
He is involved with the OPA1688 too, but check out one of the last posts in his OPA1688 thread (#51);
http://www.diyaudio.com/forums/vendors-bazaar/280113-new-audio-op-amp-opa1688.html
His contribution was realizing the OPA2172 would make a great headphone driver, renaming it the OPA1688 and adding some audio specs to the OPA1688 datasheet the OPA2172 datasheet didn't have.
Hey whatever works! He was right, the chip does a excellent job.So...by extension...if someone wants the quad OPA1689 which isn't shipping yet they should be able to just use the OPA4172 which is shipping.
Last edited:
Yeah, you guys have given me lots of great stuff to chew on.
agdr, I'll bookmark the resistor values you posted if I feel like playing with this thing again. In the meantime I got an order from Mouser yesterday with my preferred Bourns volume pot and a 5532 opamp for the gain stage based on re-reading NwAvGuy's development blog and sgrossklass's comments. I only run on wall power and my understanding is that the 5532 is a drop-in replacement with several advantages at the expense of about 3dB more noise and increased power consumption in the gain stage. Please correct me if I'm missing something.
For those interested, my idle power consumption on batteries was 21.6mA with the 2068 and 26.0mA with the 5532.
While the case was open (and near midnight) I was auditioning the new opamp and got a very loud transient. I thought the O2 was dead but after my ears recovered realized I could hear the music very softly. Being an asshat I cycled the power several times to make sure I wasn't hallucinating and heard the transient each time. Finally I checked the outputs and realized I was getting DC to the rails. Oops. Sorry Sennheisers. I guess that faint music signal was the dynamic range left in my amp and my headphones when more than 11V DC is flowing through them
With the clock rolling well into the small hours I ran through NwAvGuy's entire troubleshooting procedure. After a few minutes of working on the thing everything started measuring appropriately again. The only thing a little off:
U2 pin 2: -8.7 (NwAvGuy says -8.4)
U2 pin 3: -10.2 (NwAvGuy says -10)
R6 opposite U2 pin 3: +11.8 (NwAvGuy says -11.8 which I assume is a typo)
I am assuming the above discrepancies are related to my LED or R6 being a little out of spec.
Anyway, I don't know what happened (maybe a grounding or shorting problem?) but I'm listening to Miles Davis' new box set right now and everything sounds fine!
R6 will have the positive unswitched rail on one end and the negative unswitched rail less the forward volt drop of the LED and so around +12 and -10.3 would be expected given -/+12 volt rails.
It sounds like you have a serious (but probably simple in cause) fault that does need sorting. If you are getting rail voltage on the output, then the first check is to make sure the 'other' rail is still present.
So +11 on the output suggests the negative rail has disappeared for some reason.
It sounds like you have a serious (but probably simple in cause) fault that does need sorting. If you are getting rail voltage on the output, then the first check is to make sure the 'other' rail is still present.
So +11 on the output suggests the negative rail has disappeared for some reason.
Thanks! This was confusing when I first saw it so I'm glad to hear this is normal behavior (or behaviour)
Okay. My DC offset was actually about -11 and when I first started working on it the supply voltages at pins 4 and 8 of U4 were both negative, so that means the positive rail was out.
I took all the ICs out and did all the supply checks and everything checked out. Then when I put everything back together it measured appropriately. Not sure where in the process my problem resolved itself so it is difficult to know what was happening. If it happens again I'll use that info to check it out. I've checked for the obvious things like solder bridges and loose leads banging around.
I've got a panel mounted power supply connected to the rear panel that was flopping around so maybe I accidentally shorted something?
Anyway I've had my O2 working for about a week prior to this incident and the positive rail has always measured a tenth of a volt less in magnitude than the negative rail. I wonder if this is a sign of some kind of borderline problem?
The datasheet for the 7812 specs 11.5 <= Vo<=12.5. Likewise the datasheet for the 7912 specs -11.5 <= Vo <= -12.5. Worst case is a 1V differential between the two, say, -11.5 on one and +12.5 on the other. Within 0.1V is actually pretty good, it's 0.1/12 which is very roughly 1%. If the regulators are just warm, and voltages are within 0.1V, I'd guess something else.
Unbalanced rails are no problem at all. There would be no DC offset present even with an imbalance right down to the minimum working voltage level for the opamps. So +12 and -3 or -3 and +12 volt rails will all still give zero volts DC offset on all the opamp outputs.
If a rails disappearing (and nothing such as the FET or reg is getting hot which would suggest a short) then my first guess would be a crack in the print where the FET's are soldered, perhaps with a pad just lifting away from the PCB.
If the problem is elusive then adding a direct link from positive rail FET output to the farthest opamp supply pin on the board should fix (fiddle) the problem.
If a rails disappearing (and nothing such as the FET or reg is getting hot which would suggest a short) then my first guess would be a crack in the print where the FET's are soldered, perhaps with a pad just lifting away from the PCB.
If the problem is elusive then adding a direct link from positive rail FET output to the farthest opamp supply pin on the board should fix (fiddle
) the problem.Many thanks for the input, guys!
Hmm. So if one of the rails approached the minimum voltage rating you would just clip the output at a lower magnitude on one half of a sine wave?
Uh oh. This has the ring of truth. In a display of total ignorance I soldered 3 of the 4 MOSFETs in backwards the first time around. I had to re-order all 4 of them plus the power supply diodes. This was my first attempt at removing parts, so it was a little less than graceful. Also, with everything out of the case my power line from the rear mounted power plug was running right over the power regulators. That might have been torquing them a bit and exacerbating a crack.
Is this how the circuit is printed in the PCB?
I am learning a lot from you guys. I've never really participated in an online forum before and I really appreciate your willingness to teach an ignorant newcomer a few things.
Hmm. So if one of the rails approached the minimum voltage rating you would just clip the output at a lower magnitude on one half of a sine wave?
Uh oh. This has the ring of truth. In a display of total ignorance I soldered 3 of the 4 MOSFETs in backwards the first time around. I had to re-order all 4 of them plus the power supply diodes. This was my first attempt at removing parts, so it was a little less than graceful. Also, with everything out of the case my power line from the rear mounted power plug was running right over the power regulators. That might have been torquing them a bit and exacerbating a crack.
if the problem is elusive then adding a direct link from positive rail FET output to the farthest opamp supply pin on the board should fix (fiddle
Is this how the circuit is printed in the PCB?
I am learning a lot from you guys. I've never really participated in an online forum before and I really appreciate your willingness to teach an ignorant newcomer a few things.
Yes, it would just clip at the value of the rail.
If there is doubt over the pads where the regulators are soldered then the standard repair technique is to scrape the resist off the PCB print next to the component lead and then solder a small piece of tinned wire from the component leg itself to the fresh print a few mm away. With practice you can do the job so neatly that it almost looks original.
I don't how the PCB is laid out tbh, I've never even seen an O2 for real I just work off the circuit diagram for fault-finding. That said, the rail to the final opamp (that drives the headphones) would be the one to link the supplies to if it came to doing that. Why ? because that stage is AC coupled and so no fault elsewhere could cause a DC offset to appear through for example an incorrect input to the chip.
If there is doubt over the pads where the regulators are soldered then the standard repair technique is to scrape the resist off the PCB print next to the component lead and then solder a small piece of tinned wire from the component leg itself to the fresh print a few mm away. With practice you can do the job so neatly that it almost looks original.
I don't how the PCB is laid out tbh, I've never even seen an O2 for real
I just work off the circuit diagram for fault-finding. That said, the rail to the final opamp (that drives the headphones) would be the one to link the supplies to if it came to doing that. Why ? because that stage is AC coupled and so no fault elsewhere could cause a DC offset to appear through for example an incorrect input to the chip.Thanks Mooly!
I never would have thought of scraping the PCB to expose the track from the damaged pad. I'll keep that in mind.
Excellent point about the output stage being AC coupled. I believe that wiring from the FET output to the most distal opamp power supply would still cause the sound to go out in case of a fault because the gain stage would go to the rail, but it would protect my headphones.
I never would have thought of scraping the PCB to expose the track from the damaged pad. I'll keep that in mind.
Excellent point about the output stage being AC coupled. I believe that wiring from the FET output to the most distal opamp power supply would still cause the sound to go out in case of a fault because the gain stage would go to the rail, but it would protect my headphones.