I`m not exactly sure that an quite big OS-CON (16v) is the best idea to use it to decouple the DAC chip on AVCC. I`m not exactly sure that is a good idea to add this 1000µ directly on the output of the power supply... But you have anyway now an idea what about...
I will suggest you to use many 100µ/6v ceramic or tantalum SMD (no terminals, lowest ESR) soldered together in a pack (avoid wires to connect those capacitors), directly on the AVCC pins of the ES9018, or the closest as possible to these pins. Not on the main power supply, which is anyway quite far from the DAC chip. In some cases precautions have to be taken to prevent damages because large start up currents, which those large capacitors can cause.
One have to keep in mind using in this case capacitors that have lowest as possible both ESR and parasitic inductance. The best to use here are ceramics or tantalum SMD components.
It is a clue to use 2x1000µ on both sides of the ESS9018 chip (AVCC L/R) and dedicated power supplies/regulators for those pins), It is not by chance that the designer have specified those pins as AVCC L&R...
If you will use the same principle with large capacitors direct on the power pins of the I/V - output stage, specially if you use that TPA I/V stage with op-amps, you will have even more pleasant surprises...
Last edited:
In case of large decoupling capacities on power pins, I would like to use low noise dedicated voltage regulators which have current limiter protection... I can not see any problem to use after a shunt regulator, another voltage regulator with current limiter and large decoupling capacities... I can agree with the statement that is not a good idea to use large capacities on shunt regulators... But this can well be another discussion...
Last edited:
Yes, you right! One have to take in consideration the rest of the system when want to experiment something. I personally adapted the rest of the system I use for the moment, to my needs, and to use the way I wanted. It is not the right way to do it without previously thinking what about.
I feel the need to clarify here in conjunction with the last posts/comments from TPA:
- This thread is not intend to destroy or denigrate the TPA products. My self I appreciate and respect the professionals who designed Buffalo. In the same time, I can easily register a kind of opposition from TPA to everything that is not in theirs way to do it. This is OK enough, but only I want to mention that this thread is not a TPA commercial one. Right?
- I do not make recommendations about how to do it, or how one have to use Buffalo series of products. As the title of the thread show, this is about ESS9018 experiments. I come here only with some suggestions and discussions for peoples who can be interest in experiments with this DAC chip, for they who want to try something else than TPA have previously approved, for they who can do it, and who are enough thinking human been to know what they do.
- At least, experimenting is a risky activity. One can destroy something, and it take his own responsibility in this. This is a way to learn... I`ve destroyed components many times, and this can happen to everyone. This is very normal in this field of activity. I do think that this forum, and they who participate to this thread are enough mature thinking human been, who can take own decisions, without special guidance and close care form TPA specialists...
BTW, if one destroy his Buffalo, than can buy another one. So it still be a good business for TPA...
It could be nice if you are willing to explain little bit (on short but better understandable) what/why could be the problems you refer to. It will be very usefully for everybody to understand why TPA AVCC module (shunt regulator) do not "like to see" large capacity on output... So will be (in my opinion) a real contribution to this discussion...
I have in a circuit an TL431 regulator which have 800µ (SMD ceramic) on output. Everything function very well, the regulated device it feels in a very good condition, and "nobody" died yet... But is not about an ES9018...
Last edited:
Coris,
The reliability problem is in the fact the caps will be storing energy at turn off that the regulator has to dump after the error amp has lost it's power. So the error amp is no longer in control, and the shunting transistor could see far too much current at turn off. I am not saying it won't work, it probably would be fine for a while. But anyone who kills their AVCC module like that will not be getting a replacement.
You should also read up on shunt regulation. Using a big cap at the output of a shunt regulator actually makes it harder for the error amp to do its job. A small amount of capacitance right at the device pins (which Buff 3 has) helps to decouple very high freq, but capacitance that is too large at the output actually becomes counter productive for a shunt regulator. At the input you should use as large as practical.
Now if you want to use large capacitance at the AVCC pins the best idea would be to use a linear regulator there which is probably the case on the DAC you have.
Cheers!
Russ
The reliability problem is in the fact the caps will be storing energy at turn off that the regulator has to dump after the error amp has lost it's power. So the error amp is no longer in control, and the shunting transistor could see far too much current at turn off. I am not saying it won't work, it probably would be fine for a while. But anyone who kills their AVCC module like that will not be getting a replacement.
You should also read up on shunt regulation. Using a big cap at the output of a shunt regulator actually makes it harder for the error amp to do its job. A small amount of capacitance right at the device pins (which Buff 3 has) helps to decouple very high freq, but capacitance that is too large at the output actually becomes counter productive for a shunt regulator. At the input you should use as large as practical.
Now if you want to use large capacitance at the AVCC pins the best idea would be to use a linear regulator there which is probably the case on the DAC you have.
Cheers!
Russ
why?
Coris/Bunpei:
My understanding is that well designed shunt regulators provide low noise and very low output impedance-lower output impedance than any capacitor. Besides a subjective point of view of the "different" sound, why, technically, would one place large capacitance at the point of load, which raises the impedance the load sees (vs a local shunt regulator). Technically speaking, would not placing large capacitance here slow down the power supply?
Coris/Bunpei:
My understanding is that well designed shunt regulators provide low noise and very low output impedance-lower output impedance than any capacitor. Besides a subjective point of view of the "different" sound, why, technically, would one place large capacitance at the point of load, which raises the impedance the load sees (vs a local shunt regulator). Technically speaking, would not placing large capacitance here slow down the power supply?
I really don't want to sound like I am raining on your parade. I don't want to come across as discouraging experimentation and DIYing. I think it's great that you want to experiment and try things out, and I fully support your efforts.
I just wanted to be sure folks don't damage their gear while experimenting.
It would be easy to try a different regulator with added capacitance (etc)... that's why they are not built into the main board.
I just wanted to be sure folks don't damage their gear while experimenting.
It would be easy to try a different regulator with added capacitance (etc)... that's why they are not built into the main board.
Coris,
The reliability problem is in the fact the caps will be storing energy at turn off that the regulator has to dump after the error amp has lost it's power. So the error amp is no longer in control, and the shunting transistor could see far too much current at turn off. I am not saying it won't work, it probably would be fine for a while. But anyone who kills their AVCC module like that will not be getting a replacement.
You should also read up on shunt regulation. Using a big cap at the output of a shunt regulator actually makes it harder for the error amp to do its job. A small amount of capacitance right at the device pins (which Buff 3 has) helps to decouple very high freq, but capacitance that is too large at the output actually becomes counter productive for a shunt regulator. At the input you should use as large as practical.
Now if you want to use large capacitance at the AVCC pins the best idea would be to use a linear regulator there which is probably the case on the DAC you have.
Cheers!
Russ
Thanks for the explanations. Now they who want to try something else, have a little bit more infos to take in to consideration before proceed...
In my case, yes I do use linear regulation (voltage). But was an situation that i used an 431 regulator with large capacities on the device was powered with. As I précised earlier these regulators have a current limiter function, and that because nothing happen with large capacity on the DAC power pins. But anyway, one have to be carefully...
Large capacities have a clue if one analyse the circuit from AC point of view. I have to say again that here is about that capacitors with lowest as possible ESR and parasitic inductance. It is the case of these ceramics and special made tantalum SMD type.
Having enough energy on the power supply points of some circuits, is a benefit for their slew rate, for their capabilities to good/quick reaction when need it. I mean in the lower range of frequency spectre, and in the higher too. Dynamics increase is more than evident in such cases. The Bunpei post can easily demonstrate that improvements occur at once when using large capacities. It is of course about the rest of the system that one have to think about too. This is very true! But if the rest of the system permit, then large capacities leads to improvements.
I will suggest a simple experiment: take an appropriate op amp with a high slew rate. Let`s say 3000v/µs. Use that as final stage. Decouple it with some 10µ. Hear the result. Change the decoupling with an 1000µ. Hear the result. I let you have own conclusions about the final result. Do the same with an 20v/µs, and compare the results.
This is one thing about large decoupling capacities. They are many others (positive). Not to forget the filter capabilities in such decoupling... I mean DECOUPLING (power in - load) and not filtering the output of a power supply...
Personally I did not have ever a bad experience when using such decoupling in the right place (on the right system).
Back to the ES9018, I will say that "my rule" was confirmed again. A huge increase in dynamics and the wider output frequency specter, more details occur only by using/changing the large decoupling capacities. Just measure or try to hear the possible noise on final output when large decoupling are in place, and when the "normal" one are used. Compare the results and conclude your self...
You do not need to trust me. Just try yourself (with the necessary precautions to avoid catastrophic result). Take in to consideration the important currents that are involved, the start up and shut down sequences when such large capacities are used. And not at last, the type of power supply involved in the system.
All the benefits of this approach disappear if one use usual electrolytic`s and solder the long terminals somewhere in the power rails, and by chance on the GND... It can get worse very easy then...
Last edited:
Coris,
The reliability problem is in the fact the caps will be storing energy at turn off that the regulator has to dump after the error amp has lost it's power. So the error amp is no longer in control, and the shunting transistor could see far too much current at turn off. I am not saying it won't work, it probably would be fine for a while. But anyone who kills their AVCC module like that will not be getting a replacement.
You should also read up on shunt regulation. Using a big cap at the output of a shunt regulator actually makes it harder for the error amp to do its job. A small amount of capacitance right at the device pins (which Buff 3 has) helps to decouple very high freq, but capacitance that is too large at the output actually becomes counter productive for a shunt regulator. At the input you should use as large as practical.
Now if you want to use large capacitance at the AVCC pins the best idea would be to use a linear regulator there which is probably the case on the DAC you have.
Cheers!
Russ
Referring to your explanation about TPA shunt regulator for AVCC, I can see (a little bit delayed...) some weak points. I could not see a schematic of the shunt regulators you have designed for Buffalo, but I take as reference your explanation here.
You say that the shunt transistor has to support /dump to much current by turning off the power.
One have not to ignore that the load chip (ES9018 in this case) is always connected to the large capacity in parallel with the shunt transistor. So the current is divided by the load. The possibility that the regulator output transistor will want to consume more current than the chip it self is very low. In this particular case is about 3,3v and some 1000µ capacity. When the chip use 30-50mA or more from the stored energy in the caps is no chance that the shunt transistor will be forced to dump very much current... The current deceasing rate is very fast anyway in this case. Not to forget that at the input of the shunt regulator (power supply) are placed even more capacity (filter capacitors on power supply) on it, and the discharge rate of this very large capacity is much longer than the discharge rate of the decoupling (large) capacity. So the shunt transistor and its error amp, will have enough time and energy to come very well out of the dangerous zone (by turn off sequence).
It could be true that a large capacity on the output of a shunt regulator can makes harder for the error amp to do its job. But when a large decoupling capacity is in place there is not need so much that the shunt regulation be so accurate. The large capacity take over very well what the regulator loose in accuracy. The benefit of a large decoupling capacity compensate this loss. The low impedance (AC point of view) provided by the shunt regulator become even lower when using a large capacity. Here is an important note to be mentioned: There is a big clue not to use a single large capacity. Such capacity have enough parasitic impedance and enough big ESR. Using many quite small capacities soldered together get lower much more the ESR and the self impedance (these are paralleled many times...) for the total capacity. The total impedance of the cap (pack) is paralleled with that of the shunt regulator. This is a benefit. The stored energy in the capacitor contribute to a good/better slew rate for the circuit. This is another benefit. Decoupling with large capacity lower the range of noise frequency to be filter out. This is a benefit (noise point of view).
Taken some precautions one can safe use large capacity as decoupling.
Last edited:
You are basically arguing that an unregulated supply is better than a shunt-regulated supply?
You are the best Brian! Is nobody like you in this world! Is only you who own the truth of the universe!
Are you satisfy now?
Sorry, but your arrogant attitude start to make me sick...
Wow. Relax man. How am I being arrogant? I am just trying to understand what you are proposing.
By dulling the response of the error amp, and letting the cap take more of a leading role in the regulation, it lessens the need to even have the error amp involved. A more simple linear reg and a large capacitance would do the same thing with fewer parts. That's all I am saying.
By dulling the response of the error amp, and letting the cap take more of a leading role in the regulation, it lessens the need to even have the error amp involved. A more simple linear reg and a large capacitance would do the same thing with fewer parts. That's all I am saying.
This is what I said too. I`ve just wrote in few earlier posts that I use linear voltage regulators for the moment.
My approach about shunt regulator and large capacities was (clearly enough) directed to they who are already using the existing TPA shunt regulators (and do not want to replace it for this experiment). That because I agree with the idea that is not a big clue to have both shunt regulator and large decoupling capacities. But as we`ve seen, somebody want to try it anyway in this configuration. It is not catastrophic to try large capacities in those situations. For experiments is not a big issue if the error amp in a such shunt regulator do not function as it should, because that large capacity. But in a finally build device one have to find the best stable alternative.
I`m not pretending to propose a better way than another. I just want to show that I was going this way and I`ve registered improvements. Is up to the members here to adopt it, confirm it or not, or follow it...
As I can see now, you also understood well what I was trying to say/explain. So that because I took your last intervention/question as enough inappropriate in the discussion context...
Coris,
The TPA AVCC module was specifically designed for a purpose. Saying the *it should* do such and such in a situation it was not designed for is silly at best. It is designed to directly control the output voltage on the AVCC pins, not to simply charge a cap. Our older designs did exactly that. I think Brian was simply trying to spare Buffalo II/III users any grief they may have if they destroy either their DAC or their AVCC module, nothing more. I am not saying you should not experiment, I welcome it.
Earlier I was trying to give you some insight as to why the AVCC module is not really suited for that situation. The shunt element in the AVCC module is a PNP transistor. When the base of that transistor goes to GND if there is any charge left in the cap at the output it will be shunted at a *very* low impedance. And could kill that transistor. That impedance would be far lower than the impedance of the DAC itself as the power consumer. Also don't count the idea that the inputs will run down slower than the output caps. This would depend entirely on the rest of the load on the primary supply. It would also depend on which supply was used. Now if you can guarantee that the input voltage will decay slower than the output, then there will not be an issue other than reduced control by the regulator itself.
Also you talked earlier about slew rate, which is not really applicable here (at least not in the same way). I understand your thinking in this regard but your argument would be better applied to other output types. The analog section of the ES9018 is not an amplifier like active device. It is literally just an array of switched ~50K (Its right around that I forget the exact value) resistors. And the cool thing about it is it's symmetry. Which mean that current draw is balanced. Pretty cool. The point to this is while the supply should be low impedance and as a composite device have a high slew rate, that there is not really any slew rate to discuss on the device under power, the DAC. This is because the output impedance is always fixed.
Now have fun!
Cheers!
Russ
The TPA AVCC module was specifically designed for a purpose. Saying the *it should* do such and such in a situation it was not designed for is silly at best. It is designed to directly control the output voltage on the AVCC pins, not to simply charge a cap. Our older designs did exactly that.
I think Brian was simply trying to spare Buffalo II/III users any grief they may have if they destroy either their DAC or their AVCC module, nothing more. I am not saying you should not experiment, I welcome it. Earlier I was trying to give you some insight as to why the AVCC module is not really suited for that situation. The shunt element in the AVCC module is a PNP transistor. When the base of that transistor goes to GND if there is any charge left in the cap at the output it will be shunted at a *very* low impedance. And could kill that transistor. That impedance would be far lower than the impedance of the DAC itself as the power consumer. Also don't count the idea that the inputs will run down slower than the output caps. This would depend entirely on the rest of the load on the primary supply. It would also depend on which supply was used. Now if you can guarantee that the input voltage will decay slower than the output, then there will not be an issue other than reduced control by the regulator itself.
Also you talked earlier about slew rate, which is not really applicable here (at least not in the same way). I understand your thinking in this regard but your argument would be better applied to other output types. The analog section of the ES9018 is not an amplifier like active device. It is literally just an array of switched ~50K (Its right around that I forget the exact value) resistors. And the cool thing about it is it's symmetry. Which mean that current draw is balanced. Pretty cool. The point to this is while the supply should be low impedance and as a composite device have a high slew rate, that there is not really any slew rate to discuss on the device under power, the DAC. This is because the output impedance is always fixed.
Now have fun!
Cheers!
Russ
Last edited:
About my first post on this thread.
I was enough surprised to see that my picture posted in my first post on this thread, have more than 600 views/downloads. This make me think that enough many members here where interest in this subject. That because I want to precise now a little more about what it is in that picture.
I will refer first at decoupling in ES9018 case and in general (in my opinion). I want to say that the biggest improvement come from using large decoupling capacities in op-amp stages (here in the case of I/V, and final stage).
By decoupling ES9018 with large capacities lead to improvements. This is a fact in my opinion, and maybe enough many another ones has already experienced this. No matter what some, who think they have monopole about everything in ES9018 "world", and own the "truth" about, will want to say...
Strictly about that decoupling, the picture in my first post show, I want to precise that that my choice to place the capacitors in that way, was not quite the best. Is the best as short connections point of view. But placement of that capacitors over the chip have some negative sides. The capacitors get quite warm and this is not very well for that ceramic capacitors. I will want to place they in other way, but the original design just not permit this. Also the GND points I would want to connect it in a better way. But was not me who designed that arrangement of the DAC chip on the board...
I have been also used two voltage regulators (3,58v, ultra low noise ones, feed it from another main serial 9v regulator) on each side (AVCC L/R) of the chip. At the present those decoupling capacities are 2x1000µF/6v.
If one want to go this way and experiment like this, have one to take in consideration some precautions about large capacities which have to be charge it, at power ON and have quite large energy left inside, when power OFF... Once those cases are fix it, then the result (as I noticed my self) is a clear improvement. I used also the recommended in ES9018 datasheet I/V stage (modified feedback). That I/V circuit permit very well to adjust the DC offset in final stage to the lowest possible. In my case 0,6mV. I have now an 4-5vRMS (not scope it yet that output...) swing on RCA (SE) and no any filter between I/V and the final op-amp.
About the clock oscillator one can see in that picture.
I`ve been used that time an 133,3 Mhz oscillator. I personally noticed a huge improvement in the sound quality after changed that clock. Unfortunately, and even though the sound it self was very high in quality (perceptual point of view), I`ve noticed in the same time a kind of sparkling noise in the silence passages of recordings. This was happened only when a sound file was played (CD, SACD, or wav/FLAC, 192khz/24bit). In the pauses between tracks was no any noise. It is very possible that in this last case, was about a mute function which was ON... Anyway, I concluded that this is not acceptable. So I had to go down to 125Mhz clock frequency...
It still be a very good sound quality, not as high as at 133Mhz, but is very well for me. And no any noise at all now as before. I think to go even lower to 122,88Mhz clock frequency. This clock frequence match accurate (by 5 factor) ones of the usual samplings frequencies. I`m just waiting now to get my ordered oscillator and try this way.
Could be nice to hear comments from somebody else who have tried what that my pictured showed.
As I can conclude for now, that more than 600 views of that picture was shown in my first post exhibit a kind of interest. It was also until now an confirmation about improvements by decoupling with large capacities...
Else I could only noticed many comments in this thread that came only from two members here, who clearly enough have own private interest to minimise, discourage and almost blame that ideas which is not in conformity with their thinking/business way to do it...
But anyway...
I was enough surprised to see that my picture posted in my first post on this thread, have more than 600 views/downloads. This make me think that enough many members here where interest in this subject. That because I want to precise now a little more about what it is in that picture.
I will refer first at decoupling in ES9018 case and in general (in my opinion). I want to say that the biggest improvement come from using large decoupling capacities in op-amp stages (here in the case of I/V, and final stage).
By decoupling ES9018 with large capacities lead to improvements. This is a fact in my opinion, and maybe enough many another ones has already experienced this. No matter what some, who think they have monopole about everything in ES9018 "world", and own the "truth" about, will want to say...
Strictly about that decoupling, the picture in my first post show, I want to precise that that my choice to place the capacitors in that way, was not quite the best. Is the best as short connections point of view. But placement of that capacitors over the chip have some negative sides. The capacitors get quite warm and this is not very well for that ceramic capacitors. I will want to place they in other way, but the original design just not permit this. Also the GND points I would want to connect it in a better way. But was not me who designed that arrangement of the DAC chip on the board...
I have been also used two voltage regulators (3,58v, ultra low noise ones, feed it from another main serial 9v regulator) on each side (AVCC L/R) of the chip. At the present those decoupling capacities are 2x1000µF/6v.
If one want to go this way and experiment like this, have one to take in consideration some precautions about large capacities which have to be charge it, at power ON and have quite large energy left inside, when power OFF... Once those cases are fix it, then the result (as I noticed my self) is a clear improvement. I used also the recommended in ES9018 datasheet I/V stage (modified feedback). That I/V circuit permit very well to adjust the DC offset in final stage to the lowest possible. In my case 0,6mV. I have now an 4-5vRMS (not scope it yet that output...) swing on RCA (SE) and no any filter between I/V and the final op-amp.
About the clock oscillator one can see in that picture.
I`ve been used that time an 133,3 Mhz oscillator. I personally noticed a huge improvement in the sound quality after changed that clock. Unfortunately, and even though the sound it self was very high in quality (perceptual point of view), I`ve noticed in the same time a kind of sparkling noise in the silence passages of recordings. This was happened only when a sound file was played (CD, SACD, or wav/FLAC, 192khz/24bit). In the pauses between tracks was no any noise. It is very possible that in this last case, was about a mute function which was ON... Anyway, I concluded that this is not acceptable. So I had to go down to 125Mhz clock frequency...
It still be a very good sound quality, not as high as at 133Mhz, but is very well for me. And no any noise at all now as before. I think to go even lower to 122,88Mhz clock frequency. This clock frequence match accurate (by 5 factor) ones of the usual samplings frequencies. I`m just waiting now to get my ordered oscillator and try this way.
Could be nice to hear comments from somebody else who have tried what that my pictured showed.
As I can conclude for now, that more than 600 views of that picture was shown in my first post exhibit a kind of interest. It was also until now an confirmation about improvements by decoupling with large capacities...
Else I could only noticed many comments in this thread that came only from two members here, who clearly enough have own private interest to minimise, discourage and almost blame that ideas which is not in conformity with their thinking/business way to do it...
But anyway...
Last edited:
I stopped using TPA standard shunt type AVCC regulator for my AVCC decoupling capacitor experiments on TPA Buffalo III.
Instead, I made a simple AVCC regulator board with LDO type linear regulator chips. The chip is Toshiba TAR5SB50, which is similar to LT1763 that used to be on TPA Buffalo II board. (The price of TAR5SB50 is only 300 JPY / 10 pieces, approximately, 4 USD/ 10 pieces!)
First I applied 100 microF and 1,000 microF Sanyo OS-CON at the output side and recognized that resulting sounds had remarkable differences with those obtained with the original TPA standard shunt regulator. At least, the first result gave me a very positive impression.
I think this kind of experiments are worth doing very much. Evaluation criteria may depend on evaluators and music categories. However, I think we have a certain opportunity for finding the best setting that matches our own taste from these experiments. In other word, I realized that configurations on a AVCC power supply line have a big influence on analog sounds.
In the next step, I will try some film, ceramic or tantalum caps.
Instead, I made a simple AVCC regulator board with LDO type linear regulator chips. The chip is Toshiba TAR5SB50, which is similar to LT1763 that used to be on TPA Buffalo II board. (The price of TAR5SB50 is only 300 JPY / 10 pieces, approximately, 4 USD/ 10 pieces!)
First I applied 100 microF and 1,000 microF Sanyo OS-CON at the output side and recognized that resulting sounds had remarkable differences with those obtained with the original TPA standard shunt regulator. At least, the first result gave me a very positive impression.
I think this kind of experiments are worth doing very much. Evaluation criteria may depend on evaluators and music categories. However, I think we have a certain opportunity for finding the best setting that matches our own taste from these experiments. In other word, I realized that configurations on a AVCC power supply line have a big influence on analog sounds.
In the next step, I will try some film, ceramic or tantalum caps.
Last edited:
Nice to hear about your results. Nice to have an confirmation about my observations too...
But you know, one do not have to write something on this forum which is not positive about TPA, or something that Twisted Pear Audio peoples possible do not like. One can be moderated at once and the post deleted... As it just happened to me right now after I posted some evidences about their full oscillated Placid PSUs... I`ve wrote it also before something about another problems with their Placid, and it was deleted that post too.
I just wander who gave them the moderation right here... It is suppose to be a free expression forum and not a dictatorial act of one or another...
But anyway... I look forward to "hear" about your new experiments with different decoupling ways, caps types, and so on...