The biasing of Q1 and Q2 is terrible, the base current is known, then, the collector current depends of the transistor current gains which is notoriously non accurate.
Q2 as a dynamic load is obscure and about unpredictable.
Much of what you say is true. That's why I argued from the start that while I don't like the particular setup of the Ultra Simple Dienoiser, it deserves to be tested equally. What can be lost, if a controlled test is done?
There is also a lot of empirical in the compensation that the other configurations need to function properly. What I'm trying to do is find a setting where trial and error is reduced, or at the very least, trial and error fits any parameter other than stability.
Best regards
I prefer your original dienoiser version: it is certainly more deterministic, and I think that in practice, some compensation would be required to make your newer incarnation acceptably stable. It needs to be tested.
It may be so, Elvee.
Also, although under an imprecise configuration, it is not complex to test.
Best regards
To see the flexibility of the Ultra Simple Dienoiser, I will vary R3 in a decade.
It only remains to know if it requires any compensation, which I do not think is necessary.
Best regards
It only remains to know if it requires any compensation, which I do not think is necessary.
Best regards
Attachments
Last edited:
True.
I haven't made detailed measurements on the dienoiser, but the nonoiser and dienoiser look relatively similar from a performance perspective, and the total wideband noise (10Hz to 10kHz) of a ZTX-fitted nonoiser has been measured at 90nV. I don't think that a naked, modern reg can better that (but you can show me wrong)
You are missing one important aspect of these modern low ultra low noise regulators. They have a very small footprint allowing to place them close to target as local regulators. Show me how to achieve datasheet performance of modern DAC chips (e.g. AKM 449x) with nonoiser or dienoiser.
Show me how to achieve datasheet performance of modern DAC chips (e.g. AKM 449x) with nonoiser or dienoiser.
There's sot223 LM317. For DAC purposes the voltages are low so you can get away with small caps. I've added a dienoiser to Vref

I think the title of this thread needs to be taken into account again. This is mainly for retrofitting existing LM317 type reg gear. You could integrate the xnoisers into new designs but the appeal of these circuits is that you can improve existing gear (most times).
You are missing one important aspect of these modern low ultra low noise regulators. They have a very small footprint allowing to place them close to target as local regulators. Show me how to achieve datasheet performance of modern DAC chips (e.g. AKM 449x) with nonoiser or dienoiser.
Exactly the point I also tried to make clear before things got ugly. With a new design one adapts to the modern "local regulator" system instead of the old "central PSU" idea. Bonus is that every section of a device gets its own regulators as close as possible to the IC's. Way less risk on stray in too.
I do use LM723 or LT108x sometimes for the main PSU and modern ultra low noise LDO regs locally which means carrying regulated DC over the PCB and the ultra low drop makes for very little heat generation in general which electrolytic caps really like. Ferrite beads, common mode coils and/or coils (CLC) make it complete. In todays environment one has to take into account that HF/RF (cell phone signal, wireless, bluetooth, LED lighting) is literally everywhere and you don't want that in your audio devices. Any excess wire might pick up signals certainly in devices that have XO's and/or high frequencies in themselves like DACs. SMD is not just OK it is a necessity in such mixed signal environments. Even the smallest xxnoiser circuit is already as large (and higher) as 4 SMD regs grouped together....
For this very reason a wooden casing was nice 40 years ago but you can't get away with that now.
Last edited:
Diego I've done a test with the simple dienoiser. The good news is that it's stable without compensation, the bad news is that it didn't perform as well as the original dienoiser. I used 100k and the transistors still got hot. It also took about 5 minutes to reach the best performance, where the original dienoiser needs about a minute max. There's around 15-20dB difference.
I will try to play with it again as I didn't have much time to poke at it.
Red trace is normal dienoiser and yellow trace is simple dienoiser.
I will try to play with it again as I didn't have much time to poke at it.
Red trace is normal dienoiser and yellow trace is simple dienoiser.



Last edited:
You are missing one important aspect of these modern low ultra low noise regulators. They have a very small footprint allowing to place them close to target as local regulators. Show me how to achieve datasheet performance of modern DAC chips (e.g. AKM 449x) with nonoiser or dienoiser.
What a nonsense. I guess sharing a very clever and simplistic idea is starting to become a frustrating experience on behalf of Elvee.
I hope Elvee won't give up for his innovative contributions on diyaudio for being upset for such nonsense.
Nobody proposed about using lm317+xnoiser instead of a low noise modern regulator into new designs. Yet I've seen people arguing this here, while at the same time commending designs using lt3042 type modules added 30cm away from the load, with rounded traces so the electrons don't hurt themselves on sharp corners.
What a nonsense. I guess sharing a very clever and simplistic idea is starting to become a frustrating experience on behalf of Elvee.
I hope Elvee won't give up for his innovative contributions on diyaudio for being upset for such nonsense.
In principle, you will always be able to find a discrete superior to the internal circuits of a VR, but with modern VR's, this is going to be difficult, and perhaps not worth the trouble. With 317 style regulators, it was easy, because the IC has to cope with everything, DC, AC, thermal stability at the same time. With a discrete, you can separate the AC and DC paths, which eases matters considerably. However, IC designs have become so much smarter that bettering them requires a significant effort. It is certainly possible, but any gain is probably going to be marginal, and if you want a real improvement, you need to design everything from scratch, not try to improve on something that is barely improvable
Post #2 in this thread. LM317 is not a religion, you may use other devices and still be able to go to heaven (but only if you don't use blue LED's).
Some people argue that they don't use them in new designs and yet they do 🙂
Last edited:
What a nonsense.
You probably missed the point. Most of the use cases of denoisers shown in this thread benefit from high PSRR but not from low noise. Ultra low noise (i.e. far below LM317 noise) at audio frequencies is seldom needed in audio equipment so the ultra low noise of no-noiser and dienoiser does not have that many use cases. Ultra low noise is required in e.g. modern DAC chips but for those no-noiser or dienoiser is not practical.
And in case you have missed it in this thread I have built and tested several d(i)enoisers. Elvee's idea is brilliant. However there exists lots of confusing information on the stability and applicability of these solutions. It is misleading to state that denoisers are without tradeoffs or that they offer ultra low noise without also stating the shortcomings.
This has been discussed a few times already and I don't understand. How long are people going to come here and mention the shortcomings? Should we schedule this around once a month? Maybe more? We make it a regular thing to once a week remind people of the shortcomings of this circuit? You know, so people "don't burn their house down"?
Why should we keep having these discussions, out of the blue?
Also why mention it if usually for regular gear a denoiser would do the job just fine? Of course in the majority of situations people are not going to be able to tell the difference between a denoiser and dienoiser. But there's some, like me, who would add a dienoiser to a DAC's Vref so it would be nice if these discussions of how the xnoisers are not as good as new regulator would be rare or just not happen. People got it, this has been mentioned and I see it as ill intentioned to keep trying to bring this up once in a while.
Instead of trying to aggregate all useful info into one post or edit the first post, maybe we should edit the first post with references to the posts that mention the shortcomings of the xnoisers so we don't need to keep bringing it up again and again.
Why should we keep having these discussions, out of the blue?
Also why mention it if usually for regular gear a denoiser would do the job just fine? Of course in the majority of situations people are not going to be able to tell the difference between a denoiser and dienoiser. But there's some, like me, who would add a dienoiser to a DAC's Vref so it would be nice if these discussions of how the xnoisers are not as good as new regulator would be rare or just not happen. People got it, this has been mentioned and I see it as ill intentioned to keep trying to bring this up once in a while.
Instead of trying to aggregate all useful info into one post or edit the first post, maybe we should edit the first post with references to the posts that mention the shortcomings of the xnoisers so we don't need to keep bringing it up again and again.
Diego I've done a test with the simple dienoiser. The good news is that it's stable without compensation, the bad news is that it didn't perform as well as the original dienoiser. I used 100k and the transistors still got hot. It also took about 5 minutes to reach the best performance, where the original dienoiser needs about a minute max. There's around 15-20dB difference.
I will try to play with it again as I didn't have much time to poke at it.
Red trace is normal dienoiser and yellow trace is simple dienoiser.
It is gratifying to confirm that your setup does not require compensation of any kind, even if it does not match the performance of the original dienoiser.
Have you been able to check if it improves the parameters in relation to the typical schemes suggested by the manufacturer (with Cadj and Cout)?
Most likely, the circuit behaves very flexible in terms of ranges of values for R3.
It is very interesting to note that the Ultra Simple Dienoiser has a much less complex spectral content than the original Dienoiser.
Your work and persistence are very admirable. Thank you very much and all my respects to you.
Best regards
Last edited:
I didn't have time to test more but I'll make some comparisons to Cadj and simple LM317 and also with the denoiser.
If the simple dienoiser has better performance than the denoiser it might be worth it in situations where the extra 40-50mA isn't an issue.
I'll update once I test some more. If the simple dienoiser is around in the middle between denoiser and normal dienoiser I might also test LM337 as that would be interesting.
Indeed it's a good idea to try and find something better than the denoiser that doesn't require extra gear to implement successfully. Especially since it's lower part count and could also be made really small. Also need to check the thermals for sot23 transistors.
If the simple dienoiser has better performance than the denoiser it might be worth it in situations where the extra 40-50mA isn't an issue.
I'll update once I test some more. If the simple dienoiser is around in the middle between denoiser and normal dienoiser I might also test LM337 as that would be interesting.
Indeed it's a good idea to try and find something better than the denoiser that doesn't require extra gear to implement successfully. Especially since it's lower part count and could also be made really small. Also need to check the thermals for sot23 transistors.
As you may have noticed in my latest simulations, the Ultra Simple Dienoiser could draw significantly less current than those proposed 40 mA max, without major detriment to performance. Very surely, you can play around with R3 extensively, above the 100K you initially used (until its own thermal noise starts to prevail).
The work that Elvee originated deserves all my respect and deepest thanks. They have served as triggers for all the other ideas and work that many are doing in that direction.
Best regards.
The work that Elvee originated deserves all my respect and deepest thanks. They have served as triggers for all the other ideas and work that many are doing in that direction.
Best regards.
Last edited:
I made a test size comparison between an adm7150 supply and lm317+dienoiser supply. Both have output caps. LM317 is 0.6$ and ADM7150 is 8.66$.
The complete supplies are similar in size. The LM317 can be made single sided and can be even diyed.
If we are talking about retrofitting an existing lm317 implementation, you mean to say that you are clearly better off adding an ADM7150 to the circuit as it is, rather than adding a denoiser or even a dienoiser (if you have LNA)? Would blindly adding an ADM7150 without measuring guarantee it would be better than just adding a denoiser? Would that extra matter at all objectively to the resulting sound? Would the ADM7150 retrofit present the same issues in pick-up of stray noise as a denoiser would in a noisy gear? We're talking adding extra modules.
Now, new design. As you see, the supplies are similar in size. Would you add an ADM7150 to a new design and "sleep comfortably" that it works without measuring? We're talking the target application for the lm317+de/dienoiser here. Like analog supplies, opamps/VACC stuff like that.
Adding low noise regulators to new designs would imply that you'd measure the result so you know you did benefit and have not made any mistakes or random issues that might appear. If you are measuring then you might as well try the dienoiser if you already have the measuring gear.
I'll take concerns seriously when someone presents cases with measurements and pictures showing how retrofitting a denoiser to existing lm317 gear has a worse outcome (that is audible) than replacing the lm317 with some fancy new ultra low noise regulator module. Just because it's expensive and the pcb is black with gold text on it doesn't mean the end result is better sound compared to just adding a denoiser module to that existing lm317 gear.
Adding ultra low noise modules is exactly as risky as retrofitting a denoiser to the lm317 gear. Until you measure you can't be certain there's any improvement.
So, until anyone presents a case where a lt3042 or adm7150 shows clear audible improvement over just adding a dienoiser to existing lm317 gear these discussions are not really relevant, it's just fearmongering. And since you'd need to measure to make sure everything is as it should then you might as well add a dienoiser.
I now want to see how anyone would defend that it's better and safer to retrofit existing lm317 gear with any of these new ultra low noise regulators WITHOUT measuring the result, than just adding a denoiser without measuring.
Left is lm317+dienoiser (with output cap) and right is the complete adm7150.
edit: I guess the adm7150 also has the advantage of on/off control.
The complete supplies are similar in size. The LM317 can be made single sided and can be even diyed.
If we are talking about retrofitting an existing lm317 implementation, you mean to say that you are clearly better off adding an ADM7150 to the circuit as it is, rather than adding a denoiser or even a dienoiser (if you have LNA)? Would blindly adding an ADM7150 without measuring guarantee it would be better than just adding a denoiser? Would that extra matter at all objectively to the resulting sound? Would the ADM7150 retrofit present the same issues in pick-up of stray noise as a denoiser would in a noisy gear? We're talking adding extra modules.
Now, new design. As you see, the supplies are similar in size. Would you add an ADM7150 to a new design and "sleep comfortably" that it works without measuring? We're talking the target application for the lm317+de/dienoiser here. Like analog supplies, opamps/VACC stuff like that.
Adding low noise regulators to new designs would imply that you'd measure the result so you know you did benefit and have not made any mistakes or random issues that might appear. If you are measuring then you might as well try the dienoiser if you already have the measuring gear.
I'll take concerns seriously when someone presents cases with measurements and pictures showing how retrofitting a denoiser to existing lm317 gear has a worse outcome (that is audible) than replacing the lm317 with some fancy new ultra low noise regulator module. Just because it's expensive and the pcb is black with gold text on it doesn't mean the end result is better sound compared to just adding a denoiser module to that existing lm317 gear.
Adding ultra low noise modules is exactly as risky as retrofitting a denoiser to the lm317 gear. Until you measure you can't be certain there's any improvement.
So, until anyone presents a case where a lt3042 or adm7150 shows clear audible improvement over just adding a dienoiser to existing lm317 gear these discussions are not really relevant, it's just fearmongering. And since you'd need to measure to make sure everything is as it should then you might as well add a dienoiser.
I now want to see how anyone would defend that it's better and safer to retrofit existing lm317 gear with any of these new ultra low noise regulators WITHOUT measuring the result, than just adding a denoiser without measuring.
Left is lm317+dienoiser (with output cap) and right is the complete adm7150.
edit: I guess the adm7150 also has the advantage of on/off control.

Last edited:
As you may have noticed in my latest simulations, the Ultra Simple Dienoiser could draw significantly less current than those proposed 40 mA max, without major detriment to performance. Very surely, you can play around with R3 extensively, above the 100K you initially used (until its own thermal noise starts to prevail).
The work that Elvee originated deserves all my respect and deepest thanks. They have served as triggers for all the other ideas and work that many are doing in that direction.
Best regards.
That looks good. I'll try with higher values as I'm not that comfortable with the heat resulting from using 100k.
I used bc337/bc327 with the -40 gain rating. I presume in this application the higher gain versions would be desired?
It is preferable to use the signal transistor with the highest possible gain for Q1, while for Q2 it is not strictly necessary. This effect can be observed in the simulations. It is not even necessary to match between pairs.
This is why my choice of different transistors for the pair (not complementary).
Anyway, they are observations that I do not want to advance until they are verified in the tests. The idea is to go step by step.
Best regards
This is why my choice of different transistors for the pair (not complementary).
Anyway, they are observations that I do not want to advance until they are verified in the tests. The idea is to go step by step.
Best regards
Last edited:
I made a test size comparison between an adm7150 supply and lm317+dienoiser supply. Both have output caps. LM317 is 0.6$ and ADM7150 is 8.66$.
The complete supplies are similar in size. The LM317 can be made single sided and can be even diyed.
If we are talking about retrofitting an existing lm317 implementation, you mean to say that you are clearly better off adding an ADM7150 to the circuit as it is, rather than adding a denoiser or even a dienoiser (if you have LNA)? Would blindly adding an ADM7150 without measuring guarantee it would be better than just adding a denoiser? Would that extra matter at all objectively to the resulting sound? Would the ADM7150 retrofit present the same issues in pick-up of stray noise as a denoiser would in a noisy gear? We're talking adding extra modules.
Now, new design. As you see, the supplies are similar in size. Would you add an ADM7150 to a new design and "sleep comfortably" that it works without measuring? We're talking the target application for the lm317+de/dienoiser here. Like analog supplies, opamps/VACC stuff like that.
Adding low noise regulators to new designs would imply that you'd measure the result so you know you did benefit and have not made any mistakes or random issues that might appear. If you are measuring then you might as well try the dienoiser if you already have the measuring gear.
I'll take concerns seriously when someone presents cases with measurements and pictures showing how retrofitting a denoiser to existing lm317 gear has a worse outcome (that is audible) than replacing the lm317 with some fancy new ultra low noise regulator module. Just because it's expensive and the pcb is black with gold text on it doesn't mean the end result is better sound compared to just adding a denoiser module to that existing lm317 gear.
Adding ultra low noise modules is exactly as risky as retrofitting a denoiser to the lm317 gear. Until you measure you can't be certain there's any improvement.
So, until anyone presents a case where a lt3042 or adm7150 shows clear audible improvement over just adding a dienoiser to existing lm317 gear these discussions are not really relevant, it's just fearmongering. And since you'd need to measure to make sure everything is as it should then you might as well add a dienoiser.
I now want to see how anyone would defend that it's better and safer to retrofit existing lm317 gear with any of these new ultra low noise regulators WITHOUT measuring the result, than just adding a denoiser without measuring.
Left is lm317+dienoiser (with output cap) and right is the complete adm7150.
edit: I guess the adm7150 also has the advantage of on/off control.
I have not seen anybody suggesting retrofitting ADM7150 (or any other ultra low noise regulator) in place of LM317 or to anything operating in audio frequencies. And why would anybody put ultra low noise regulators to retrofittable modules in new designs. BTW it is usually good practice to study datasheets. ADM7150 is targeting RF applications and its exceptional strengths are mainly above audio frequencies.
This is really starting to look like a religion or a cult both of which I have always tried to avoid. Over and out.
It's one of these low noise regulators. I understand it's mostly desired for digital but it was an example small low noise regulator. You can substitute it for any of the desired low noise regulators.
This is not a religion or a cult, just a group of people playing with the denoiser designs.
What I really don't understand is why do people feel the need to mention weekly that the denoisers are not as good as other options. It doesn't make much sense for me.
I've not said this is the be-all and end-all solution of anything analog supply related. I would just like to understand what makes people post about how there's other better regulators and how there's a bunch of shortcomings to the denoiser. It's like minding your coffee in a coffee shop when random people keep barging in saying there's better coffee in other places. Ok, most probably...
This is not a religion or a cult, just a group of people playing with the denoiser designs.
What I really don't understand is why do people feel the need to mention weekly that the denoisers are not as good as other options. It doesn't make much sense for me.
I've not said this is the be-all and end-all solution of anything analog supply related. I would just like to understand what makes people post about how there's other better regulators and how there's a bunch of shortcomings to the denoiser. It's like minding your coffee in a coffee shop when random people keep barging in saying there's better coffee in other places. Ok, most probably...
- Home
- Amplifiers
- Power Supplies
- D-Noizator: a magic active noise canceller to retrofit & upgrade any 317-based VReg