I haven't studied the jung regs in fine detail but I believe they're fairly layout sensitive.
What're you trying to achieve that other existing super-reg/jung reg pcb designs can't offer already? Or going for the learning experience? Either way is cool just might help with people giving advice
What're you trying to achieve that other existing super-reg/jung reg pcb designs can't offer already? Or going for the learning experience? Either way is cool just might help with people giving advice
The haves and the have not
@hochopeper Do is going for lowest noise on a single pcb. Ticknpop is drawing a 3pin tracking preregulator mit bellz but w/o whistlez. Wolfsin awaits r-cores so Kubota can be slotted into a harness that will allow listening to JC-80 with any of the above. As near as Wolfsin can guess, around 2020 (year not vision stat) he can begin to listen. We all survived ticknpop's avalanche of ideas for easyButQuiet but living to see the dust settle may be difficult.
ticknpop got Wolfsin thinking. Paraphrasing balanced.com faq, power train-line stage interaction is greatly reduced in balanced circuits. There are always two line stage currents. When a signal appears at the input, these two currents will change in unison but in opposite directions. The net line stage current change can be made very small. In the ideal case, the power supply will see no current change at all.
Thus once we can provide 30vDC rails at around a quarter amp, closely matched line stage circuit components offset each other symmetrically but whatever tiny noise remains gets amplified by line stage gain. In the default case that is 14dB. Fortunately it is common noise. If there is a 'next' balanced stage the bulk of the power chain noise will be cancelled there. If the output is applied to headphones, the noise 'nets out' there.
It strikes Wolfsin that Do is onto something. The r-cores, rectifiers, filters and chokes of the remote (placed remotely to weaken magnetic fields) is about all we can do to halt mains noise. If Do can get noise to less than 5uv, compared to twenty times that for Kubota, his regulator will measure better. ticknpop's may also. It is important to note that those guyz are already listening whereas Wolfsin, immanentdomain, and kannan hear the sound of silence.
@hochopeper Do is going for lowest noise on a single pcb. Ticknpop is drawing a 3pin tracking preregulator mit bellz but w/o whistlez. Wolfsin awaits r-cores so Kubota can be slotted into a harness that will allow listening to JC-80 with any of the above. As near as Wolfsin can guess, around 2020 (year not vision stat) he can begin to listen. We all survived ticknpop's avalanche of ideas for easyButQuiet but living to see the dust settle may be difficult.
ticknpop got Wolfsin thinking. Paraphrasing balanced.com faq, power train-line stage interaction is greatly reduced in balanced circuits. There are always two line stage currents. When a signal appears at the input, these two currents will change in unison but in opposite directions. The net line stage current change can be made very small. In the ideal case, the power supply will see no current change at all.
Thus once we can provide 30vDC rails at around a quarter amp, closely matched line stage circuit components offset each other symmetrically but whatever tiny noise remains gets amplified by line stage gain. In the default case that is 14dB. Fortunately it is common noise. If there is a 'next' balanced stage the bulk of the power chain noise will be cancelled there. If the output is applied to headphones, the noise 'nets out' there.
It strikes Wolfsin that Do is onto something. The r-cores, rectifiers, filters and chokes of the remote (placed remotely to weaken magnetic fields) is about all we can do to halt mains noise. If Do can get noise to less than 5uv, compared to twenty times that for Kubota, his regulator will measure better. ticknpop's may also. It is important to note that those guyz are already listening whereas Wolfsin, immanentdomain, and kannan hear the sound of silence.
For me it is more of a learning experience as I have never played with a Jung reg but my friend which is designing the PCB is an audio engineer. To him it is a walk in the park sort of speak... Also, sourcing those kind of super regs like the Super Teddy, are not only sky rocket expensive (euros) but I also need one per rail... So for two JC-80, you guessed it, it is 45 euros X4 = two much money for what it is... But I know they're really good!
My friend just fired me an email this afternoon and recommended to keep the rectifiers and filtering on the same board. He says if cables between pre-reg and rectifiers are long we re-introduce rectifier noise so we need to have more filtering caps on the pre-reg plus those cables act like antennas, picking up noise...
So if we really need two boards, the recommended setup would be rectifiers, filtering and LM pre-reg on one PCB and the Jung reg on the second one. AC noise will never make it pass the Jung reg., especially with the pre-reg in front of it.
We're currently aiming at two layers. I will push four layers but the board will cost more. I'm thinking SMD all the way, except for regulators. It will be easier to put in heatsink for through hole parts.
Rectifiers will most probably be HFA08TA60 (dual common) or TB60. They're really good and are of ultra fast and ultra soft recovery. Also thinking of BYQ28E.
I'd like to know everyone interested in the PCBs to voice their opinion before we lock the design later on. I will post schematics once available.
Thanks
Do
Also, sourcing those kind of super regs like the Super Teddy, are not only sky rocket expensive (euros) but I also need one per rail... So for two JC-80, you guessed it, it is 45 euros X4 = two much money for what it is... But I know they're really good!My friend just fired me an email this afternoon and recommended to keep the rectifiers and filtering on the same board. He says if cables between pre-reg and rectifiers are long we re-introduce rectifier noise so we need to have more filtering caps on the pre-reg plus those cables act like antennas, picking up noise...
So if we really need two boards, the recommended setup would be rectifiers, filtering and LM pre-reg on one PCB and the Jung reg on the second one. AC noise will never make it pass the Jung reg., especially with the pre-reg in front of it.
We're currently aiming at two layers. I will push four layers but the board will cost more. I'm thinking SMD all the way, except for regulators. It will be easier to put in heatsink for through hole parts.
Rectifiers will most probably be HFA08TA60 (dual common) or TB60. They're really good and are of ultra fast and ultra soft recovery. Also thinking of BYQ28E.
I'd like to know everyone interested in the PCBs to voice their opinion before we lock the design later on. I will post schematics once available.
Thanks
Do
Last edited:
Wolfsin moans softly
I think if the conductors, each @30v are twisted, it would take a magnetic BLAST to introduce noise that would not self cancel. While it may make sense to pre-regulate in the remote enclosure then we are back to the case of a pure Jung near the JC-80. hochopeper's query then becomes relevant.
I think if the conductors, each @30v are twisted, it would take a magnetic BLAST to introduce noise that would not self cancel. While it may make sense to pre-regulate in the remote enclosure then we are back to the case of a pure Jung near the JC-80. hochopeper's query then becomes relevant.
For the board, I'll make 2 PCBs in break tab so you can keep them together or separate them easily.
Rectifier board with pre-reg will be all Through hole and Jung will be smd parts.
Ciao!
Do
hmm, so its not a standard super regulator then? normally you are limited to the voltage the error amp can pass and the lme will only pass +/-18v (well it wont swing all the way to the rails, so more like +/-16v). thats why i asked, there must be some further VAS stage with the feedback/sense voltage somehow attenuated for the error amp? i have to say this isnt making much sense, as that wouldnt work well at all. and yes these regs with the very fast opamps are not happy with long leads, nor driving capacitive loads; are you using the 4 wire kellvin sense function? you will probably want to remove some of the onboard capacitance on the amp/load side or the amp will perform worse than a nice chip reg.
Last edited:
Uh oh, reality rearz 'n barkz! Thanks for bringing us back to earth, qusp. Now, let me re-ask that earlier question. What needs to be done between the rectifier board and the regulator, a distance of four feet, to hold down noise, avoid reflections, etc.
Wolfsin goes off to google 'kellvin wire sense'
Wolfsin goes off to google 'kellvin wire sense'
Hi qusp,
Basically the regulator can give the max tension of the opamp in single supply mode but not in dual supply mode, so up to 36V max (single supply) on the output. Our opamp version is bootstrapped, basically the latest version of the regulator.
Ciao!
Do
Confession time
The pcb referenced above has been rev'ed. The trace no longer needs to be severed and it will accept two legged as well as three legged diode rectifiers.
The above comment(s) are untrue for power train noise, for several reasons. First, supply noise is applied at various points in the circuit, not at the input. The gain will be less than fourteen, and not all supply noise is common since the supplies (in our case at least) have separate positive and negative rail circuits. The noise from, say, a zener on the positive rail circuit, is NOT common with noise on the negative rail and will not cancel out. Hence the importance of not adding that noise in the first place!
@pinnocchio Are you taking orders for that (pre)regulator yet?
The pcb referenced above has been rev'ed. The trace no longer needs to be severed and it will accept two legged as well as three legged diode rectifiers.
The above comment(s) are untrue for power train noise, for several reasons. First, supply noise is applied at various points in the circuit, not at the input. The gain will be less than fourteen, and not all supply noise is common since the supplies (in our case at least) have separate positive and negative rail circuits. The noise from, say, a zener on the positive rail circuit, is NOT common with noise on the negative rail and will not cancel out. Hence the importance of not adding that noise in the first place!
@pinnocchio Are you taking orders for that (pre)regulator yet?
Here is a three terminal regulator which gets about the best you can expect from these type of designs. a good shunt or Jung regulator is better and the parts count is getting up there. You need at least 5 volts above V out for your input voltage. The 1N4148 is for input short protection and low capacitance for low RF feedthrough. The unlabelled diode is 1N4007 is for output short protection or the 100uf C adjust will destroy the regulator if the output is shorted. The unlabelled zener is an LM 329. As a preregulator this is overkill, but that's what Diyaudio is all about.
There is a 0.1 uf film cap in series with a 10 ohm resistor at the input of the second regulator to ground that i can't add to the drawing because it exceeds the site upload limits it seems, but it must be included for stability.
Radjust is calculated as (V out divided by .00908 ) minus 908
typical R adjust values
35v - 2946
30v - 2395
24v - 1735
15v - 743
12v - 413
device tolerances and odd resistor values mean I usually use a pot for Radjust
The best is seperate positive and negative bridges each using the 317 circuit as the 317 has better performance than the 337 - though you can use the same circuit with 337 regulators if you reverse all 3 diodes and all the electrolytic capacitors
There is a 0.1 uf film cap in series with a 10 ohm resistor at the input of the second regulator to ground that i can't add to the drawing because it exceeds the site upload limits it seems, but it must be included for stability.
Radjust is calculated as (V out divided by .00908 ) minus 908
typical R adjust values
35v - 2946
30v - 2395
24v - 1735
15v - 743
12v - 413
device tolerances and odd resistor values mean I usually use a pot for Radjust
The best is seperate positive and negative bridges each using the 317 circuit as the 317 has better performance than the 337 - though you can use the same circuit with 337 regulators if you reverse all 3 diodes and all the electrolytic capacitors
Last edited:
The regulator I posted is a compilation of other's published circuits. The led should be green or red - don't use blue they are noisier. The 100uf R adjust bypass cap makes the circuit ramp up over a couple of seconds. You must have a C out of at least 100 uf or the circuit may not regulate ( especially if you use LT 1085 instead of LM 317).
Last edited:
This certainly has an air of originality about it, tnp. Thanks. Was it drawn on the back of a napkin at Wandas?
Given the need for 30v, that would suggest about 0.9mV noise voltage before the tracking pre so maybe 110uV with the tpr given a guesstimate of 8x improvement. That is not bad for such a simple circuit. Wolfsin now realizes this is what you and John were suggesting in the beginning.
I ordered a pair of 317/337 regulator pcbs with lotsa components I can leave out and fill with your circuit. You did proto and test that circuit, right
Given the need for 30v, that would suggest about 0.9mV noise voltage before the tracking pre so maybe 110uV with the tpr given a guesstimate of 8x improvement. That is not bad for such a simple circuit. Wolfsin now realizes this is what you and John were suggesting in the beginning.
I ordered a pair of 317/337 regulator pcbs with lotsa components I can leave out and fill with your circuit. You did proto and test that circuit, right
Wolfsin,
I haven't tested the LED part, but i will in the next week as i am updating my preamp power supply because i'm installing R cores to replace the EI cores. The LED is in the acoustica design so i expect it to work.
The LM 329 appearently lowers the noise by a factor of 6 (whatever that means). And improves regulation, thermal drift etc. everything else works very well in my raw supply for BlowTorch and Borbely line amp for 5 years now without fault.
Yes Wanda's napkin, did you know Help in Rio before they closed it? The Dollhouse In Tampa, the Bank in Bangkok? Angeles City in Manila? I don't just waste money on Vishay bulk foil resistors.
I've spent 90 % of my life's earnings on women, cars, audio, and good scotch - the rest i wasted.
I haven't tested the LED part, but i will in the next week as i am updating my preamp power supply because i'm installing R cores to replace the EI cores. The LED is in the acoustica design so i expect it to work.
The LM 329 appearently lowers the noise by a factor of 6 (whatever that means). And improves regulation, thermal drift etc. everything else works very well in my raw supply for BlowTorch and Borbely line amp for 5 years now without fault.
Yes Wanda's napkin, did you know Help in Rio before they closed it? The Dollhouse In Tampa, the Bank in Bangkok? Angeles City in Manila? I don't just waste money on Vishay bulk foil resistors.
I've spent 90 % of my life's earnings on women, cars, audio, and good scotch - the rest i wasted.
Last edited:
A dash of unobtanium
@ticknpop Only 10% waste? I pay more than that in taxes and thought Canada was worse
Wolfsin has his unobtanium ready for the r-cores. Once the transformers, rectifier 'n caps pcbs are mounted he'll post pix. It is a UOS (that's used old stock) Panasonic video camera power supply he found abandoned along the roadside. Well, with a free sign (that translates roughly to Please Haul Away) now that Larimer County landfill prices have skyrocketed. Predrilled for ventillation and JUST the right size.
@ticknpop Only 10% waste? I pay more than that in taxes and thought Canada was worse
Wolfsin has his unobtanium ready for the r-cores. Once the transformers, rectifier 'n caps pcbs are mounted he'll post pix. It is a UOS (that's used old stock) Panasonic video camera power supply he found abandoned along the roadside. Well, with a free sign (that translates roughly to Please Haul Away) now that Larimer County landfill prices have skyrocketed. Predrilled for ventillation and JUST the right size.
Wazzzz happening here?
No news for some time. Wolfsin, you sleepin' man
I noticed while listening/comparing back and forth with
- Buffalo II --> Legato 3.1 --> JC-80 + R2R attenuator --> amplifier
- Buffalo II --> Legato 3.1 + volumite --> amplifier
I'm losing some micro details when using the JC-80... Stuff like the subtle sound of lips or breathing... Some instruments blurring and a bit narrower stage...
Maybe it is just a question of impedance matching between the BII's Legato 3.1 I/V/line stage output to the JC-80 that is causing this but I have similar issue with the IVY III Line stage with Opus to the JC-80...
I decided to connect my venerable BII to my tube I/V stage and BAM! everything is back to normal or at least those micro details are back...
I'm using AMB's o1 R2R shunt ladder volume and o2 input selector. It is set at 25K input and the load is set to 1000K which is what the JC-80 should be at... Both phase are set to 1000K on my shunt ladder.
Anything you guys think I could try?
Ciao!
Do
No news for some time. Wolfsin, you sleepin' man
I noticed while listening/comparing back and forth with
- Buffalo II --> Legato 3.1 --> JC-80 + R2R attenuator --> amplifier
- Buffalo II --> Legato 3.1 + volumite --> amplifier
I'm losing some micro details when using the JC-80... Stuff like the subtle sound of lips or breathing... Some instruments blurring and a bit narrower stage...
Maybe it is just a question of impedance matching between the BII's Legato 3.1 I/V/line stage output to the JC-80 that is causing this but I have similar issue with the IVY III Line stage with Opus to the JC-80...
I decided to connect my venerable BII to my tube I/V stage and BAM! everything is back to normal or at least those micro details are back...
I'm using AMB's o1 R2R shunt ladder volume and o2 input selector. It is set at 25K input and the load is set to 1000K which is what the JC-80 should be at... Both phase are set to 1000K on my shunt ladder.
Anything you guys think I could try?
Ciao!
Do
Tempco?
>>the load is set to 1000K which is what the JC-80 should be at... Both phase are set to 1000K on my shunt ladder.
How sensitive is your JC-80 to temperature? Was it equally warm in both tests? Please solve this soon so Wolfsin has one less thang to wurrie over
Wuz supposed to be in Bouder today but heavy snow forecast delayed the trip. I will likely solder up a second rectifier board with soft recovery diodes so I can measure the various voltage drops when the r-cores arrive. I really am anxious to post a BOM so builders can order parts ONCE and expect to be ready to build when they arrive.
>>the load is set to 1000K which is what the JC-80 should be at... Both phase are set to 1000K on my shunt ladder.
How sensitive is your JC-80 to temperature? Was it equally warm in both tests? Please solve this soon so Wolfsin has one less thang to wurrie over
Wuz supposed to be in Bouder today but heavy snow forecast delayed the trip. I will likely solder up a second rectifier board with soft recovery diodes so I can measure the various voltage drops when the r-cores arrive. I really am anxious to post a BOM so builders can order parts ONCE and expect to be ready to build when they arrive.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.