I will be recapping and upgrading the opamps on my recently purchased DEQ2496.
I noticed that the digital board is v2.5. Based on the rigorous work of stef1777 of diyaudio and awdiy.com, I noticed that my board layout was different from the photos and descriptions he provided.
As such, I believe this may be the first DEQ2496 mod featuring the v2.5 board.
The AKM AK5393 ADC has been replaced with a Cirrus Logic CS5381-KZZ and the capacitors around this ADC now include a few 1uF 50V radials which were absent on the previous layout.
See attached photo of the digital board. I have annotated some of my plans, including how I plan to follow stef1777's suggestions to piggyback 10uF 10V OSCONS on some of the ceramic caps. It seems as if there are actually several more candidates for similar piggybacking or outright replacement which I have left without annotation.
I am excited to increase my electrical engineering knowledge as I am an economist by trade so I do have one initial question for the gurus of diyaudio - looking at frequency coefficients, it seems that 50V capacitors have higher (worse?) frequency coefficients versus their 25V cousins (usually on a datasheet you can see both variants side-by-side). I find it odd that there would be 50V caps on this board, especially when it seems this is the first time 50V caps have ever been used on the DEQ2496 (at least on the digital board). Can I use 25V instead? I know this is not a simple question, but perhaps my detailed photo and the experience of this community can help me answer this question.
Thanks!
(I have mostly settled on Elna Silmic II caps for the analog board's 47uF 25V caps, the (ridiculously) more expensive Panasonic OSCONS for the 10uF 25v caps on the digital board, and PPS Film caps for the 1n5, 2n2, and 390p ceramic smds. If I'm feeling sadistic, I may replace some of the 100n and 100p smds as well. There are a few others, for example, stev1777 suggested replacement of C1 10uF 25V on digital board with a 22uF 25V and 2.2uF 50V that I will need to choose as well - <strike>assuming that v2.5 C1 cap is the same as v2!</strike> *there are differences because the ADC is different, see post #12)
I have also included stev1777's awdiy.com picture of the v2 board so you can see the differences (at least some of them).
I'll leave the analog board for a separate post in this thread.
I noticed that the digital board is v2.5. Based on the rigorous work of stef1777 of diyaudio and awdiy.com, I noticed that my board layout was different from the photos and descriptions he provided.
As such, I believe this may be the first DEQ2496 mod featuring the v2.5 board.
The AKM AK5393 ADC has been replaced with a Cirrus Logic CS5381-KZZ and the capacitors around this ADC now include a few 1uF 50V radials which were absent on the previous layout.
See attached photo of the digital board. I have annotated some of my plans, including how I plan to follow stef1777's suggestions to piggyback 10uF 10V OSCONS on some of the ceramic caps. It seems as if there are actually several more candidates for similar piggybacking or outright replacement which I have left without annotation.
I am excited to increase my electrical engineering knowledge as I am an economist by trade so I do have one initial question for the gurus of diyaudio - looking at frequency coefficients, it seems that 50V capacitors have higher (worse?) frequency coefficients versus their 25V cousins (usually on a datasheet you can see both variants side-by-side). I find it odd that there would be 50V caps on this board, especially when it seems this is the first time 50V caps have ever been used on the DEQ2496 (at least on the digital board). Can I use 25V instead? I know this is not a simple question, but perhaps my detailed photo and the experience of this community can help me answer this question.
Thanks!
(I have mostly settled on Elna Silmic II caps for the analog board's 47uF 25V caps, the (ridiculously) more expensive Panasonic OSCONS for the 10uF 25v caps on the digital board, and PPS Film caps for the 1n5, 2n2, and 390p ceramic smds. If I'm feeling sadistic, I may replace some of the 100n and 100p smds as well. There are a few others, for example, stev1777 suggested replacement of C1 10uF 25V on digital board with a 22uF 25V and 2.2uF 50V that I will need to choose as well - <strike>assuming that v2.5 C1 cap is the same as v2!</strike> *there are differences because the ADC is different, see post #12)
I have also included stev1777's awdiy.com picture of the v2 board so you can see the differences (at least some of them).
I'll leave the analog board for a separate post in this thread.
Attachments
Last edited:
I would leave the ceramic decoupling caps alone, infact unless you know the power delivery requirements I would leave ALL the digital decoupling alone. The SMD caps especially near device power pins should not be changed, you are not going to improve things and if you add bigger packages you are going to change the nature of the power delivery system, adding different packages (bigger) is a no no as are different dielectrics, X7Rs are excellent for local decoupling leave them alone. The larger reservoir caps you could play about with but you will not get any gains and again may bugger things up.
Just trying to follow previous mods (that had measurements to back them up) and updating the process for the v2.5 board.
Here is our own AVP1 from http://www.diyaudio.com/forums/digital-line-level/282346-another-ultimate-behringer-mod-what-can-50-do-deq2496-part-1-a.html
"I again probed circuit and found low distortion level at output of two opamps which feed ADC chip. But there was much higher distortion at input pins of ADC chip itself. I traced circuit on analog board and found that though there is nothing between opamps on analog board and board to board cable, digital processing board has some additional analog circuit before ADC chip. This circuit can be found in a datasheet of ADC chip AK5395. This is a circuit that includes three resistors and one capacitor between opamps that feed ADC and ADC input pins (see picture). In Behringer this circuit is located on digital processing board. Here we again see relatively high value ceramic capacitor. I didn't have 1500pF SMT capacitors on hand and didn't want to delay my project again by ordering new ones (and paying another shipping charge for just a pair of them) and decided to use available polypropylene type designed for through hole assembly. I soldered them to pads after removing ceramic chips.
With that done., I tested ADC again and finally found reasonably low distortion."
Now I will note that he is referring to the chip that got replaced on the v2.5 board by Cirrus Logic CS5381-KZZ, so I should compare the layouts on these chips, but the ceramic caps are similarly placed and if their role is the same, then according to AVP1 they cause harmonic distortion (yes probably inaudible but measurable!)
I already have a Motu 8A for my main setup so this...is for SCIENCE! Plus, the Motu 8A is a bit noisy so if this mod ends up being the high quality I am hoping for I will run my mains through the DEQ2496 instead.
So if anyone with DEQ2496-experience can comment on the 1uF 50V and the 2.2uF 50V radials and whether I can go with 25V instead I'd be much obliged!
What are 'frequency coefficients' ?
And those distortions that were found, are there any graphs for that?
Jan
And those distortions that were found, are there any graphs for that?
Jan
Check out this discussion on Frequency Coefficients:
Capacitor Ripple Current - Frequency coefficient - Electrical Engineering Stack Exchange
As I understand it, capacitors, despite their namesake, still exhibit resistance and this resistance is frequency dependent. The higher the coefficient, the hotter the cap gets at those frequencies. Lots of caps that are suggested in audiophile threads still have coefficients as high as 2 at 10khz which introduces some uncertainty about the importance of this characteristic, but one thing that is common is that OSCONS are frequently suggested for digital boards and they for sure have super low coefficients below 20khz. Another anecdotal observation is that Elna Silmic II caps have been praised for their subjective musicality and they also have significantly lower coefficients than similar competitors.
If you go to www.mouser.com and look at the data sheets for a Nichicon 47uF 10V radial cap, and then check out an Elna Silmic II 47uF 10V, you'll see the frequency coefficient charts. Pansonic OSCONS also have this chart. Unfortunately there is some inconsistency in what the manufacturers choose to publish in this regard.
As for graphs - check out AVP1's thread that I linked to earlier. Pretty impressive work, and he includes the graphs.
I have modified the deq2496 extensively even changing it to BNC input into the ak4113 and tap its i2s into a tda1543 dac. I have move on since then.
But if I would to mod it again now, I believe the biggest improvement would be go passive for the output stage and use an audio transformer (10k:10K) and directly tapping the ak4393 +/- output pins into the primary of the transformer and connect its output directly to RCA out. You can add additional filter caps across the transformer output pin to taste but not required to get great sound. I would use a short cable to connect to a decent preamp.
But if I would to mod it again now, I believe the biggest improvement would be go passive for the output stage and use an audio transformer (10k:10K) and directly tapping the ak4393 +/- output pins into the primary of the transformer and connect its output directly to RCA out. You can add additional filter caps across the transformer output pin to taste but not required to get great sound. I would use a short cable to connect to a decent preamp.
If you decide to use transformers, then next logical step will be to add balanced buffer with minimum gain to decouple any kind of interconnect. Of cause you always can add another 600:600 transformer to electrically isolate load. But this seems to be overkill to me.
From what I read he was on about caps in the signal path... Not decoupling caps, where X7Rs are used almost exclusively in the world of electronics for local pin decoupling duties..
Relevant more to high power switching circuits etc. than audio, but another gremlin to throw around.🙂
Can also have an effect on cap life and MTBF figures when used in higher frequency applications.
http://www.rubycon.co.jp/en/products/alumi/pdf/Life.pdf
Hey AVP1! Thanks for stopping by this thread! Perhaps you can help, as I still haven't received any advice regarding my initial question (btw your mod and great measurements and testing is what inspired me to mod my deq2496 😀).
Are these 1uF 50V and 2.2uF 50V caps replaceable with 25V variations? (see attachments in first post)
Isn't 12V the max output from the power supply in this unit? 50% De-rating based on that would suggest nothing higher than 25V would be needed anywhere in the DEQ2496. Correct? ESR seems much worse on the 50V variations.
Anyway, I know your mod included opamp replacement and ceramic replacements (but perhaps not radials?) and here I am asking about the electrolytics... but you clearly have a firm grasp of the overall circuit.
Thanks (and thanks for your initial thread back in late 2015 - super cool!)
Last edited:
I took a look at the CS5381 ADC data sheet and learned something interesting about the 47uF 50V cap at C7 on the digital board. Apparently Cirrus Logic provides a diagram that shows a higher capacitance cap results in better low freq. distortion. Also, all the circuits for this board are 5v max so 50V caps is silly. 16v-25v will be fine.
The FILT+ and VQ connections from the ADC diagram are C7(47uF), C29 and C28 are the 0.01uF ceramics, and C6 is the 1uF.
Also, if these are not COG ceramics...I'm not sure how to tell...the ADC data sheet specifically says they need to be (so if they are X7R for example that would be bad): "The use of capacitors which have a large voltage coefficient (such as general purpose ceramics) must be avoided since these can degrade signal linearity. C0G capacitors are recommended for this application."
Furthermore, if you look at the capacitor tech comparison graph, film caps are reaffirmed as being even better as they don't have stability issues with VDC.
Importantly, these lessons tell me that C29 and C28 on the v2.5 board are NOT the same as on the v2 board since it is a different ADC chip. C12 and C11 are the caps that you should bypass with 10uF radials, NOT C29 and C28. (see figure 24 and notice the 10uF caps at the start of the journey towards the opamps and eventually the analog inputs).
Learning lots! (correct me if I'm mistaken)
The FILT+ and VQ connections from the ADC diagram are C7(47uF), C29 and C28 are the 0.01uF ceramics, and C6 is the 1uF.
Also, if these are not COG ceramics...I'm not sure how to tell...the ADC data sheet specifically says they need to be (so if they are X7R for example that would be bad): "The use of capacitors which have a large voltage coefficient (such as general purpose ceramics) must be avoided since these can degrade signal linearity. C0G capacitors are recommended for this application."
Furthermore, if you look at the capacitor tech comparison graph, film caps are reaffirmed as being even better as they don't have stability issues with VDC.
Importantly, these lessons tell me that C29 and C28 on the v2.5 board are NOT the same as on the v2 board since it is a different ADC chip. C12 and C11 are the caps that you should bypass with 10uF radials, NOT C29 and C28. (see figure 24 and notice the 10uF caps at the start of the journey towards the opamps and eventually the analog inputs).
Learning lots! (correct me if I'm mistaken)
Attachments
COG (NP0) ceramics (class I) are very stable and dont have stability issues, used a hell of a lot in RF circuits, they should not be confused with class II ceramics (X7R) etc. Again figure 24 shows ceramics within the signal loops, here I would not use class II ceramics but COG are perfect, small size, very stable and good temperature coefficients.
What I find curious is why they have used 50V caps, this may be for MTBF figures as even 50% derating you still get very low figures with electrolytic caps, the more they are derated the better the final MTBF figure will be.
What I find curious is why they have used 50V caps, this may be for MTBF figures as even 50% derating you still get very low figures with electrolytic caps, the more they are derated the better the final MTBF figure will be.
I'll pull a cap or two this weekend and heat them up a bit to see if they are C0G class I or not.
Also, I'm going to slow down on my posts so that I can be sure I don't post false information. Looking at the AK5393 ADC circuit diagram and comparing it with the Cirrus Logic CS5381, it looks like C11 and C12 were both now and then connected to the analog input leads on the ADCs. Stev1777 didn't list these as needing replacement or piggybacking. And according to the CS5381 diagram, as long as they are 10uF that should be fine.
Also, C28 and C29 on the CS5381 are on the Filt+ and VQ circuits (reference output voltages) but on the AK chip I'm confused because these same caps are at the bottom of the chip, and that's not where the reference voltages are located. Also, the AK 5393 says that on the reference voltages there should be a 10uF electrolytic and a .1uF ceramic which, on the v2 board points to C2 and C3 for the radials and two of the four caps nearby (I'm guessing) which would be C6, C7, C8, or C9 and these are at the top of the chip as you would expect.
So basically, prior mods can only serve as guidelines and not pointed instructions...
Also, I'm going to slow down on my posts so that I can be sure I don't post false information. Looking at the AK5393 ADC circuit diagram and comparing it with the Cirrus Logic CS5381, it looks like C11 and C12 were both now and then connected to the analog input leads on the ADCs. Stev1777 didn't list these as needing replacement or piggybacking. And according to the CS5381 diagram, as long as they are 10uF that should be fine.
Also, C28 and C29 on the CS5381 are on the Filt+ and VQ circuits (reference output voltages) but on the AK chip I'm confused because these same caps are at the bottom of the chip, and that's not where the reference voltages are located. Also, the AK 5393 says that on the reference voltages there should be a 10uF electrolytic and a .1uF ceramic which, on the v2 board points to C2 and C3 for the radials and two of the four caps nearby (I'm guessing) which would be C6, C7, C8, or C9 and these are at the top of the chip as you would expect.
So basically, prior mods can only serve as guidelines and not pointed instructions...
Attachments
The use of 50V capacitors may be just a question of economics and inventory: remember that Behringer buys millions of caps...
Another thing that I have read very long ago says that a higher voltage of given capacitor value will give you better characteristics (ESR?). I also may be wrong...
Another thing that I have read very long ago says that a higher voltage of given capacitor value will give you better characteristics (ESR?). I also may be wrong...
Just a quick update, I'll be adding a toroidal psu to the project but it's taking forever to get the transformer and the pre-assembled psu itself. Once I get the psu, I can see what resistors need to be modded to get the correct output, and then I'll be able to finalize my list of parts for mouser or digikey.
Hi friends V6 Vivid Dual X 2 C3 and IC4 of deq2496 were used in output buffer
amplifier sony ta fe600r player ibasso dx90
I tested the deq2496 at flat settings when you plug it into the device, the AB class will sound like A class from A amplifier. A remarkable point,
The V6 Vivid Opamps have a very vivid voice. Neutral and detailed sound with a tight bass effect dynamic field perfectow and medium tones are very accurate and have exactly as needed. Bass is quite low and deep
The V6 Vivid has a clean and clear treble response.
Highly dynamic, impressive and detailed musical.
Attachments
Last night I put two 1:1 transformers hooked directly to the balanced outputs of the AKM4393 going straight to the XLR outs. I believe, no capacitor/opamp swaps can come even close for the difference it makes...
Best, Mark Fouxman
Samar Audio Design, LLC
SAMAR Audio & Omni8 Audio
Best, Mark Fouxman
Samar Audio Design, LLC
SAMAR Audio & Omni8 Audio
+1
In this case, JT11EMCF direct connection. Even better, replace with AKM4396. Everything else being the same, it has a higher V out level and "drives" better.