Thanks, CARLSOR for that re-assurance. The little BELLESONs are very easy to install when the pins are removed - the larger ones, even with the pins removed would upset my tight (and messy) installation.
I wish I understood what AVRO ARROW is saying - I feel sure HE knows what he is talking about and look forward to hearing more of his ideas about how best to use PIN 20 on the DAC.
I was just repeating what I had read in application notes for IC CCS (yes that is a more proper designation - the constant part is the most important aspect) that those things were noisier than resistors. I can still hope that the absurd but wonderful, TX2575s will make a difference.
One would think that if a CCS or current sink would work best here that BB would have incorporated one into the chip or at least recommend their use for highest quality applications. (more wishful thinking, perhaps?).
So, if the voltage at the pin is constant it would seem the resistor would be sufficient. I guess this voltage could be nominally constant and could use additional stabilizing?
Nigel, I used batteries for a long time and no question they have some good qualities. Got tired of the charging rituals. Never have tried these new-fangled LiPos. With that vanishing low ESR they very well might have overcome my biggest complaint with batteries - they were not as dynamic as a good AC based supply.
I had K&K ship me one of those boards but was not getting much cooperation whether the thing would work with WaveIO. So, I told Kevin Carter it would be best I refused the shipment. Was very surprised that no one has used the board with WaveIO. Dave Davenport was concerned that the I2S inputs of his DAC are "balanced", they are marked + and - whereas the WaveIO instead of "-" is marked ground. Dave was not sure this would be compatible and I did not want to take a chance on a board that would not allow use of WaveIO. I remain curious to hear what it does but without the use of WaveIO making an assessment of the value of the current source/sink impossible.
Dave is a very clever engineer but I think he is a bit of a victim of the "not invented here" syndrome so famously attributed to GENERAL MOTORS. He was not the least bit curious to know if WaveIO would work with this board. He suggested I use, instead, the board K&K has on closeout which is very old technology. Very disappointing. We all loose with this.
From what one can see from the website Dave is using IC CCSs - all of those 8 fingered things you see are IC CCS/sinks - why there are eleven of them is a bit of a mystery - cannot imagine how they are being used. One part of me is wondering if he is using these same devices in conjunction with the three pin regulators on the board ... Who knows?
Last edited:
The pin 20 question
Everyone seems to have the same question I do.
What is pin 20 really for?
For every pin on an IC, there is a reason, usually
a very good reason, that the engineer brought it out.
Now, one of the big problems here is that the engineer/s
that designed the PCM1794a are no longer employed by TI/BB.
They might not have even been around when the data sheet was written.
TI has admitted there is no one there who knows the inner
workings of the chip.
So, that leaves us to wonder, why did the engineer who
designed the chip chose to bring out the part of the chip
attached to pin 20? Surely it was not just to add an external
resistor. A much more accurate resistor could have been implemented
on die. Same goes for a simple CCS like a JFET.
Everyone seems to have the same question I do.
What is pin 20 really for?
For every pin on an IC, there is a reason, usually
a very good reason, that the engineer brought it out.
Now, one of the big problems here is that the engineer/s
that designed the PCM1794a are no longer employed by TI/BB.
They might not have even been around when the data sheet was written.
TI has admitted there is no one there who knows the inner
workings of the chip.
So, that leaves us to wonder, why did the engineer who
designed the chip chose to bring out the part of the chip
attached to pin 20? Surely it was not just to add an external
resistor. A much more accurate resistor could have been implemented
on die. Same goes for a simple CCS like a JFET.
Now that is really something. Sounds like one of those conspiracy shows on cable!
Agreed that a more accurate (trimmed) resistor could have been included in the die but the quality of die resistors can be questionable - on the analog side that might have been a consideration.
Well, we have the literal black box here, it seems. Hope you are able to divine what is in there. Will continue to look forward to hearing what you think.
Well a few of us is going to try something with pin 20
If it makes a difference at all
There's room for improvement even if what we did was bad
Lifepo4 solved the bass limiting problem
And no need to charge as it's on a floating voltage from regulated linear supplies with no adverse effect that I'm aware of
If it makes a difference at all
There's room for improvement even if what we did was bad
Lifepo4 solved the bass limiting problem
And no need to charge as it's on a floating voltage from regulated linear supplies with no adverse effect that I'm aware of
More Belleson 3.3V reg info
Yesterday I sent the following question to Belleson:
"I have 3.3V SPM Belleson regs supplying only 5ma to a digital circuit. Is there a minimum load that would give the best performance?"
Answer:"Good question! For best transient response we recommend about 15mA, so if you add a 330 Ohm fixed resistor that should give a total ~15mA. You can go higher if you wish but you'll reach a point of diminishing returns at about 20mA"....Dave.
I am going to put 330 ohm resistors across the solder pads of C28 and C30 on the bottom of the DAC board. Any location after the Belleson is ok. An SMD resistor such as the Vishay TNPW1206330RBEEN would be a nice fit between the pads.
Yesterday I sent the following question to Belleson:
"I have 3.3V SPM Belleson regs supplying only 5ma to a digital circuit. Is there a minimum load that would give the best performance?"
Answer:"Good question! For best transient response we recommend about 15mA, so if you add a 330 Ohm fixed resistor that should give a total ~15mA. You can go higher if you wish but you'll reach a point of diminishing returns at about 20mA"....Dave.
I am going to put 330 ohm resistors across the solder pads of C28 and C30 on the bottom of the DAC board. Any location after the Belleson is ok. An SMD resistor such as the Vishay TNPW1206330RBEEN would be a nice fit between the pads.
I forgot your idea for the batteries. Very clever and I meant to (if I did not) compliment you on this at TIR NA.
I have a battery charger for a phono stage (ALTMANN) my last battery powered component, and I cannot hear it when I forget to disconnect it so I know your approach is sound (no pun intended). It, too, is a "floating" charger.
Was told something very exciting today from someone very close to this project. There is something brewing so DDDAC1794 is only to get more interesting. Sorry for the teaser but for anyone who may think they are losing interest ...
That is similar to what I have done.
I am using BELLESON for the analog side, also, and have a similar arrangement. I used leaded resistors, parallel 330R DALE with the lowest temp coefficient and they have copper leads. I have the resistors under the board with the leads used to attach the BELLESONS on the top of the board.. In addition to the resistor there is a capacitor in parallel with the resistor - OSCON on the digital side and SILMIC on the analog which continues throughout the board.
Were you using BELLESON on the analog side or the TENT - cannot remember.
lifepo batteries
Sorry for the OT post Doede!
Those trying to use batteries to power different parts of the dac should consider Ian's passive battery manager:
Ian's I2S FIFO Project - diyAudio
Look at section "2.5 Passive Battery Management Adapter"
I use it to power Ian's FIFO clock from a lifepo, works very good for its intended use, which may work for some of the dac parts in doede's dac. I don't know if he has any left, but you can always ask...
Sorry for the OT post Doede!
Those trying to use batteries to power different parts of the dac should consider Ian's passive battery manager:
Ian's I2S FIFO Project - diyAudio
Look at section "2.5 Passive Battery Management Adapter"
I use it to power Ian's FIFO clock from a lifepo, works very good for its intended use, which may work for some of the dac parts in doede's dac. I don't know if he has any left, but you can always ask...
Sorry, guys (and gals) - it's been impossible to keep up with the conversation, due to other stringent duties...
From the (sketchy) information I was able to find with limited time available, it seems to me that pin 20 has an internal voltage reference that provides some ability for limited adjustment of the output current by varying the value of this resistor from pin 20 to ground. Sounds reasonable per ohm's law. If this is correct, a CCS in that position would simply be a very high performance resistor. Basically, beyond differences such as source/sink etc., I see a CCS as a high-impedance dynamic resistor, in that it allows for a certain set current to flow through it independent of all other factors. In doing that, it bests a simple resistor in basically any performance aspect. Think of it this way - if a 10k resistor placed across a potential differential of 10V passes 1mA, a CCS placed in the same position may show an impedance of 1M for the same duty. If this biases a transistor or provides a plate load for a tube, it places that much more buffer from the power supply with all its garbage. It's a magic resistor!
But I digress. I hope I am able to attach options 3 and 4 that I mentioned this far. Options 1 and 2 are a matter of reading the data sheets of LM334 and 1N5288.
A few words on option 3. 2SK170 is a marginal fit at best, because its Vgs(off) is just a tad too high. In that sense, it's on par with J201, but I'm not sure if it's a long gate JFET (desirable), and its capacitances are not great (very important). I believe the 2N4338 is a far better option, and only requires being screened for an Idss well in excess of 400uA. It also has much less capacitances than the 2SK170. If the CCS is ran close the the Vgs(off) of the JFET, performance decreases substantially (see AN103 which someone here quoted). Mind you, Walt Jung's work is essential for understanding this topic especially with audio in mind.
Option 4. The advantage to this is that it requires no trial and error - the values I list should get the current desired. The pitfall is that it requires biasing - I guess the "+" could go to the 8V B+. This injection of juice from the power supply may also be its other disadvantage, but it comes with all three legged CCSs.
In any case, as the information on this is so scarce, whoever tries this should know that it's an experiment and may put their DDDAC at some risk. The fact that the K&K world tried it is reassuring, but there goes the disclaimer.
I hope the attachments worked.
Radu.
From the (sketchy) information I was able to find with limited time available, it seems to me that pin 20 has an internal voltage reference that provides some ability for limited adjustment of the output current by varying the value of this resistor from pin 20 to ground. Sounds reasonable per ohm's law. If this is correct, a CCS in that position would simply be a very high performance resistor. Basically, beyond differences such as source/sink etc., I see a CCS as a high-impedance dynamic resistor, in that it allows for a certain set current to flow through it independent of all other factors. In doing that, it bests a simple resistor in basically any performance aspect. Think of it this way - if a 10k resistor placed across a potential differential of 10V passes 1mA, a CCS placed in the same position may show an impedance of 1M for the same duty. If this biases a transistor or provides a plate load for a tube, it places that much more buffer from the power supply with all its garbage. It's a magic resistor!
But I digress. I hope I am able to attach options 3 and 4 that I mentioned this far. Options 1 and 2 are a matter of reading the data sheets of LM334 and 1N5288.
A few words on option 3. 2SK170 is a marginal fit at best, because its Vgs(off) is just a tad too high. In that sense, it's on par with J201, but I'm not sure if it's a long gate JFET (desirable), and its capacitances are not great (very important). I believe the 2N4338 is a far better option, and only requires being screened for an Idss well in excess of 400uA. It also has much less capacitances than the 2SK170. If the CCS is ran close the the Vgs(off) of the JFET, performance decreases substantially (see AN103 which someone here quoted). Mind you, Walt Jung's work is essential for understanding this topic especially with audio in mind.
Option 4. The advantage to this is that it requires no trial and error - the values I list should get the current desired. The pitfall is that it requires biasing - I guess the "+" could go to the 8V B+. This injection of juice from the power supply may also be its other disadvantage, but it comes with all three legged CCSs.
In any case, as the information on this is so scarce, whoever tries this should know that it's an experiment and may put their DDDAC at some risk. The fact that the K&K world tried it is reassuring, but there goes the disclaimer.
I hope the attachments worked.
Radu.
Last edited:
Also - I have not been able to breadboard option 4 (option 3 is well known and I have played with it extensively, but not with a 2N4338 which I don't have in my stock). It needs to be tested. It was presented in WJ's work.
For whoever has never done option 3 - the pot is there because of large variations in built of the JFETs. Basically, one would breadboard this and apply a voltage (I suggest 2.4V...), then adjust the pot until the desired current is obtained (for instance, a 1 kohm resistor placed in series would want to see 0.4V over it). Then, the variable resistor can be replaced with a fixed one of appropriate value. The CCS would at this point always pass 400uA through it, provided no other variables exceed some absolute maximums for the JFET.
For whoever has never done option 3 - the pot is there because of large variations in built of the JFETs. Basically, one would breadboard this and apply a voltage (I suggest 2.4V...), then adjust the pot until the desired current is obtained (for instance, a 1 kohm resistor placed in series would want to see 0.4V over it). Then, the variable resistor can be replaced with a fixed one of appropriate value. The CCS would at this point always pass 400uA through it, provided no other variables exceed some absolute maximums for the JFET.
(high capacitance for the JFET degrades the performance of the CCS considerably. this is why RF duty JFETs are particularly fit for this)
Avro,
Any thoughts on the led suggestion, I know biasing a cathode with an led improves the circuit dynamics so be interesting to see if it works and may be a simple way to check for the worth of a CCS?
Chuz,
Drew.
An LED is a constant voltage device in that the voltage drop across it stays relativity constant with changing current.
I used it in my example as a stand in for the voltage source at pin 20.
I don't think attaching an LED to pin 20 would be a good thing to do...it would
likely destroy the 1794.
I used it in my example as a stand in for the voltage source at pin 20.
I don't think attaching an LED to pin 20 would be a good thing to do...it would
likely destroy the 1794.
...so their CCS concept is JFET-based?... Or where they just suggesting that because the context was around JFETs? Knowing of the secrecy surrounding the RAKK I'm curious.
(I said already what I think about the 2SK170; I don't really see the 2N3819 working there, though it's capacitances are in the right side of the scale.)
They were saying that their mk2 and mk3 rakk dac just used a jfet, whereas their mk4 uses their new secret recipe, which they claim is a further improvement.
I think their "just use a jfet" advice was a way of helping me (and stopping me asking direct questions about their design in a public forum) without giving away their latest secret, which all seemed fair enough to me.
It's a shame that no high res pictures seem to exist on the net for their mk4 board...
Maybe you should ask on that forum to upload hires photos 🙂
I was a little surprised they answered at all
We probably have enough to chew on for now
I have a Mk III board. I will see what FET is in use.
There is no question the CCS or Sink on the new board is not discrete. Look at all those 8 legged things - so many they must be used for other duties, too - I figure knowing Dave Davenport's affinity for CCSs they are being used with the 3 pin regs, also.
Would be good to know how much modulation of the current there is on Pin 20 with the resistor in place. That should not be too hard to figure out. Just measure the voltage across the resistor for those with a working board.
Hope to get my wire before the weekend so I can finish this.
There is no question the CCS or Sink on the new board is not discrete. Look at all those 8 legged things - so many they must be used for other duties, too - I figure knowing Dave Davenport's affinity for CCSs they are being used with the 3 pin regs, also.
Would be good to know how much modulation of the current there is on Pin 20 with the resistor in place. That should not be too hard to figure out. Just measure the voltage across the resistor for those with a working board.
Hope to get my wire before the weekend so I can finish this.