Thanks for the feedback. Hopefully you’re right and 50mA is the answer and there’ll be plenty of headroom. I can’t easily modify the pcb to drive L and R separately so need sufficient current for the entire thing. I also just bought a 1622 eval module to try as well which should be similar architecture to the 1612 hopefully so might also work. We’ll see.
If opamp references aren’t your preference then what is? I’m certainly happy to hear divergent views.. The Yamaha uses a Sulzer type regulator I believe based on further reading which was a precursor to the Jung super reg. However not optimized completely in its current config.
Yeah, obviously home theatre receivers/preamps have to be general purpose devices so they aren’t optimized the same way as stereo audiophile products (typically SNR ~70-90db).. There are also not so many options for Atmos enabled preamps that aren’t prohibitively expensive - this is a new to me used device that I paid a fraction of msrp for and am trying best I can to turn into an audiophile caliber source.. within reason. The AVCC regulation here seems to be about the final realistic piece of the puzzle that I can alter so really want to get it right.. system is Yamaha DAC/1656 IV stage —>muses preamp—>neurochrome mod286–> diy elsinores so really have tried to build an audiophile level home theatre setup…
If opamp references aren’t your preference then what is? I’m certainly happy to hear divergent views.. The Yamaha uses a Sulzer type regulator I believe based on further reading which was a precursor to the Jung super reg. However not optimized completely in its current config.
Yeah, obviously home theatre receivers/preamps have to be general purpose devices so they aren’t optimized the same way as stereo audiophile products (typically SNR ~70-90db).. There are also not so many options for Atmos enabled preamps that aren’t prohibitively expensive - this is a new to me used device that I paid a fraction of msrp for and am trying best I can to turn into an audiophile caliber source.. within reason. The AVCC regulation here seems to be about the final realistic piece of the puzzle that I can alter so really want to get it right.. system is Yamaha DAC/1656 IV stage —>muses preamp—>neurochrome mod286–> diy elsinores so really have tried to build an audiophile level home theatre setup…
Last edited:
There are likely multiple pins for the same load to lower impedance between the power supply and the load. Similar thing with VCCA in ES9038Q2M. AVCC current is specified according to AVCC_L and AVCC_R. Pins with the exact same name should be considered as internally connected together.
Regarding AVCC supplies, my suggestion is still to experiment, keep the regulator close as possible to load, ideally on the same ground plane, and see what you think. If LT3042 sounds best to you then fine, you found out what you like.
Regarding OPA1656 for I/V, OPA1612 is one I like better. Actually, MUSES8920 might be worth a try as well. All the aforementioned opamps sound different for I/V use, at least they do IME. You might want to form your own conclusions though. https://www.mouser.com/new/nisshinbo/NJR-MUSES-Amplifiers/
Also, if you would like to take a look at certain discrete approaches to dac output stages, Marantz SA-7 and SA-10 service manuals are available out there on the web. Probably not the best approach for measurement dacs, but they might sound better to most humans.
the architecture of opa1622 and 1612 could not be more different. Its not about hopefully being right mate. i'm right. Ive been working with ESS dacs since the 9008 and i'm fairly familiar with their particular brand of datasheet obfuscation. If only they made an opa1612 in the opa1622 package/pinout. the pinout for opa1622 is a thing of beauty. Nice chip, but has its quirks (especially if used unity gain)
I dont mind opamp buffered references; they have their place. I just dont think this is it. They dont take any feedback from the output, so they are kind of unregulated in a way. they are regulated with regard to the reference voltage, rather than with regard to the load step. you touched on another point; they are often made using chips that dont drive capacitive loads well. the 1656 for example is specified to drive loads of 100pf and you would be driving 47-100uf. I would construct one of jung's opa based shunt regs, or simply lt304X and as Mark says, as close as you can manage to the load. Perhaps the Jung reg isnt a good recommendation though, as although published, has not been picked up as much as some of his other designs and i'm not aware of any PCBs available. Not a design for a beginner in the area either.
Yes, the multiple pins are there either to reduce impedance, provide layout options and also because I guess, since there are a number of dacs in parallel on the die; it may have been a matter of being easier to bring it out to another pin
I dont mind opamp buffered references; they have their place. I just dont think this is it. They dont take any feedback from the output, so they are kind of unregulated in a way. they are regulated with regard to the reference voltage, rather than with regard to the load step. you touched on another point; they are often made using chips that dont drive capacitive loads well. the 1656 for example is specified to drive loads of 100pf and you would be driving 47-100uf. I would construct one of jung's opa based shunt regs, or simply lt304X and as Mark says, as close as you can manage to the load. Perhaps the Jung reg isnt a good recommendation though, as although published, has not been picked up as much as some of his other designs and i'm not aware of any PCBs available. Not a design for a beginner in the area either.
Yes, the multiple pins are there either to reduce impedance, provide layout options and also because I guess, since there are a number of dacs in parallel on the die; it may have been a matter of being easier to bring it out to another pin
Excellent - well again my lack of familiarity with things like the datasheet specs is exactly what I was hoping to get feedback on! I very much appreciate the time you took to respond.
ESS found that 10uf at the output of an opamp buffer was enough to guarantee stability with most opamps tried (only one required 47uf). The other thing is that if using big electrolytic caps, they don't look much like caps up above 100kHz or so. Eventually they tend to get lossy at HF. Their effect on stability at RF frequencies may be negligible. One of those things that is not fully specified so you just have to try it and see (as ESS apparently did).
EDIT: IIRC one thing recently mentioned by KSTR in another thread was that ESS dacs tend to draw pulses of AVCC current at half the clock frequency. If bypass caps are at anti-resonance at that frequency then measured distortion can go up.
EDIT: IIRC one thing recently mentioned by KSTR in another thread was that ESS dacs tend to draw pulses of AVCC current at half the clock frequency. If bypass caps are at anti-resonance at that frequency then measured distortion can go up.
I used them successfully for years. I guess I just moved on; kind of boring to construct the same old regulator for years and years. I tried all manner of regulators, batteries etc. Yes, ESS moved on too, not just because they required more current (fairly easy to sort that out by adding a buffer, transistor, IC or otherwise) and also because they now offer their own dual regulator (which is actually really nice, within its rather limited operational envelope) but I gather also because it allowed higher performance. I just prefer a regulated voltage there and tend to favour MUCH smaller caps in that position. Horses for courses I guess.
Indeed. Yes, regarding AVCC, I wondered about results of a synchronous regulator with tuning to match the pulses, but in general that is exactly the sot of reason I prefer a more directly regulated voltage there.
Indeed. Yes, regarding AVCC, I wondered about results of a synchronous regulator with tuning to match the pulses, but in general that is exactly the sot of reason I prefer a more directly regulated voltage there.
Quick couple of questions I hope..
Thoughts on output cap values on the opamp circuit?
I see post 3003 says either 22uf film or 47uf polarized - I’m assuming electrolytic.. I also see a number of pages back that the ESS reference board that uses the opa1612 used 1uf, and above you’ve mentioned 10uf makes most amps stable per ESS.
So - if I was to try the opa1656 I already have on hand my on hand capacitor selection would include;
Kemet MMK 10uf film caps
22uf oscons
47uf electrolytic
A variety of ceramics however you’ve already voice your opinion on these so let’s forget them..
Thoughts on the 10uf option? If 10uf wasn’t enough for the film I guess I could parallel two but that risks resonances?
Thanks again.
Thoughts on output cap values on the opamp circuit?
I see post 3003 says either 22uf film or 47uf polarized - I’m assuming electrolytic.. I also see a number of pages back that the ESS reference board that uses the opa1612 used 1uf, and above you’ve mentioned 10uf makes most amps stable per ESS.
So - if I was to try the opa1656 I already have on hand my on hand capacitor selection would include;
Kemet MMK 10uf film caps
22uf oscons
47uf electrolytic
A variety of ceramics however you’ve already voice your opinion on these so let’s forget them..
Thoughts on the 10uf option? If 10uf wasn’t enough for the film I guess I could parallel two but that risks resonances?
Thanks again.
The ESS evaluation boards or datasheet schematics do not have any output cap after AVCC opamp buffer. There is a 1uF cap for the Vref voltage divider. But since according to some the circuit in post 3003 reaches the pinnacle of ES9038Q2M sound (with the exception to discrete regulators) who am I to argue.
ESS published a app note which used to be available on their website which included discussion of output caps for opamp buffers. The caps were apparently added to assure stability. If an opamp doesn't need a cap for stability, then there may not be any reason to use one. Regarding film caps for AVCC buffers, the time I tried that I used 22uf metalized film caps. Those are kind of big and do tend to become inductive at higher frequencies. Also because they are metalized rather than wound foil, ESR is a bit higher. Using too good of an HF cap might cause more problems than help. That said, adding a small resistor in series with a cap might fix some stability or ringing problems. IMHO can try some experimenting or else just use something you know will work and be done with it. One thing that I have consistently found is that adding a resistor from the AVCC supply output to ground can sometimes help improve SQ. Maybe 47-ohms, 1/2watt, if your regulator can supply the extra current. Or maybe 100-ohms could help. Why exactly it helps I have not attempted to measure. In some cases the type or brand of resistor can make some audible difference. Then again, some people seem to hate experimenting and much prefer to sim, build, measure, and call it done. Different courses for different horses, I guess.
Thanks Mark - much appreciated. As you say the films are quite large - while I could deal with but location this is going in it’s going to be tricky. If the cap is more about assuring stability maybe I’ll try the oscon first and see how goes as by comparison its tiny.. I could also consider paralleling a small film as well maybe as I also have those running around too. I also do now have a MKP4 wima 22uf cap coming too so could swap if I feel compelled..
I whipped up a opa1656 buffer on a TI dual sonic opamp eval board last night so I’ll see if I get time this weekend to try it.. As per your suggestion on the same forum page as 3003 I did also add a 1000uf cap at +ve rail entry to the board.. that local decoupling may also help the native circuit in addition to the reduced load from the add-on board powering the main DAC on the board..
The one thing I’m hesitating on is the question of opamp stability - last thing I want is oscillations enough to cause temporary overvolt but that’s seems fairly unlikely I guess given the ESS sheets
I whipped up a opa1656 buffer on a TI dual sonic opamp eval board last night so I’ll see if I get time this weekend to try it.. As per your suggestion on the same forum page as 3003 I did also add a 1000uf cap at +ve rail entry to the board.. that local decoupling may also help the native circuit in addition to the reduced load from the add-on board powering the main DAC on the board..
The one thing I’m hesitating on is the question of opamp stability - last thing I want is oscillations enough to cause temporary overvolt but that’s seems fairly unlikely I guess given the ESS sheets
That application note is quite old (for ES9008). More importantly the output caps are supposedly needed because local decoupling caps (1uF) may cause possible oscillations. However in later datasheets (and evaluation boards) these caps are not used for LM4562. Anyhow these 47uF caps are not meant to replace local AVCC decoupling caps which normally are 100nF to 1uF X7R caps.
Earlier ESS dac chips allowed worse case AVCC to be a bit more than 3.3v, IIRC maybe 4.7v. Seems like Twisted Pear took advantage of that and ran it a bit high for their dacs. For ES9006 looks like ESS changed the absolute maximum AVCC voltage spec to much closer to 3.3v and only slightly higher than DVCC. Maybe using a different process to make it, don't know. Anyway, there is apparently less of a safety margin if an oscillation were to occur. Also maybe harder to protect AVCC with a zener during testing, although it might be doable.
At least they understand and know what they are doing. Some people don't, and don't care they are ignorant.
Thanks a lot for that - helpful to know some of those implementation nuances.
Mark - yeah I had seen that max voltage spec on the datasheet which is why my hesitation - even temporary voltage spikes could be an issue.. I might consider getting one of these boards as a test bed.. In the meantime I swapped the NE5532 for an opa1622 yesterday and added a 270uf oscon to the 5V rail, added a 10uf ceramic in parallel to the existing 1uf on the opamp 12v rail, swapped the 100uf tantalum for a 10v 100uf ceramic, and swapped the electrolytic 10uf for a 10uf ceramic on the 3.3v reference RC filter. Overall I feel as though the sound gained a hair of extra definition - a slight perceptible improvement in background details, etc. overall a good effect so far.
Something about OPA1622 is that it has a ground pin. Some DIP adapters for OPA1622 connect the ground pin to somewhere between the power rails. The problem with that is distortion increases when the pin is not properly grounded. Also IIRC OPA1622 has been described as more of a very low distortion, low-ish power, headphone driver. It can be used for other things too of course, but not necessarily the best choice for some things.
Regarding adding caps to rails, swapping caps, etc., its not always best to play around with that. Caps in parallel can have all sorts of impedance swings and resonances that cause more problems than they help. Also, most ceramic caps are nonlinear with voltage and are piezoelectric. They can turn PCB vibrations into voltages, or vice versa. Lower ESR is not always better than moderate ESR, etc.
The other thing I would say is that most dacs, virtually all of them, are designed to a price point. If they are well designed then corners are cut more or less equally everywhere so that there is nothing that is a very much better than it needs to be and nothing much worse worse than it should be. Caps and opamps, among other components, are usually chosen to be of a similar quality as the rest of the circuitry. Although swapping caps and opamps can change the sound and be fun, it may not turn out to be an improvement in the long run. It may take some time to be sure if the sound is really better or not. For instance, after a change the sound may seem more 'clear' but that can be due to a little bit of low level added distortion instead of due to removing mud (that is to say, it may not be not fixing the real problem).
Regarding adding caps to rails, swapping caps, etc., its not always best to play around with that. Caps in parallel can have all sorts of impedance swings and resonances that cause more problems than they help. Also, most ceramic caps are nonlinear with voltage and are piezoelectric. They can turn PCB vibrations into voltages, or vice versa. Lower ESR is not always better than moderate ESR, etc.
The other thing I would say is that most dacs, virtually all of them, are designed to a price point. If they are well designed then corners are cut more or less equally everywhere so that there is nothing that is a very much better than it needs to be and nothing much worse worse than it should be. Caps and opamps, among other components, are usually chosen to be of a similar quality as the rest of the circuitry. Although swapping caps and opamps can change the sound and be fun, it may not turn out to be an improvement in the long run. It may take some time to be sure if the sound is really better or not. For instance, after a change the sound may seem more 'clear' but that can be due to a little bit of low level added distortion instead of due to removing mud (that is to say, it may not be not fixing the real problem).
Thanks - sorry re: above that was a brain fart - was the 1612 I replaced with. No changes in pinout on that one obviously. The original was SOIC so no need for dip adapters. The caliber of circuitry on this thing is overall good but as you say, everything designed to a price point - and 100uf ceramics I don’t think we’re all that common 5-6 years ago but point taken. Also, it uses 5532s literally everywhere so I think that’s a cost decision too. Re: piezoelectric considerations, yeah that’s fair as well plus resonances etc. unfortunately I do not have a scope or means of formally testing SNR at this point.
Okay, cool. Didn't mean to sound discouraging. A lot can be judged by ear, but it helps a whole lot to have a better quality device available to A/B compare with. Without that it makes learning what to listen for a slower and less certain process. Another option is to get two boards, one to mod, and one as a reference. Makes it easier to A/B test to make sure of what is actually changing.
Nope, all good - totally appreciate your help.
Yeah I had actually priced out a second board the other day.. Yamaha Canada wanted 700 bucks for it haha. None I’ve seen on the used market so far but definitely going to keep my eyes open for one. After the mods yesterday the machine stoped latching onto HDMI signals yesterday for a brief period - I think software/hdmi handshake rather than a problem caused by the mods but nearly gave me a heart attack and it’s been working reliably since I reconnected everything so I think it’s okay and none of the mods should’ve affected power rails dealing with hdmi switching that I’m aware of.. But stakes are high so have to be cautious a bit. I may well as mentioned opt to get one of the 9038 boards here to test stuff out on… Other thing I realized is where I had wanted to mount the separate opamp power supply on the bottom of the board may well not be doable due to space.. but if I optimize the existing circuit for the main DAC then use an opamp for the power rail to the second DAC (the inductor is mounted on top side on that one so directly accessible) that also eliminates power supply crosstalk and further reduces noise floor hopefully. Anyways, again thanks for the help.