It is, but the logic is bipolar CML, basically ECL. Only the inputs create appreciable noise, if overdriven.
On top, the complement of the output current is send to +5V.
I am still considering this.
The best separation point is DGND.
+5V ties together analogue currents and the positive rail of CML. So we need to return to both -15V and DGND.
DGND on the other hand is apparently purely digital.
In ECL originally the positive rail is "GND" and the negative rail power.
I think 2SA970 (if available) in the highest beta grouping would be best for PNP.
Thor
If you use 250R current to voltage conversion resistor and don't mind the 0V point on the TDA1541 drifting quite a bit, not much.
If on the other hand you want 2V RMS out and perhaps a direct coupled balanced out at 4V RMS, it's quite critical, as there is not much leeway.
Thor
Strong tempco.
A 3.3V low noise zener has a tempco complementary to a Silicon BJT and gives around 2.5V reference voltage.
More noise than a LED, not as thermally stable as TL431 but a lot less noise than TL431.
See here:
https://www.diyaudio.com/community/...s-for-leds-and-zener-diodes.35821/post-417008
Zener diodes around 3.3V are around 1uV noise, LED's are around 0.3...0.4uV but voltage levels are lowers, TL431 are 20uV.
Realistically, with 1uV noise and 2.5V, we get around 2nA noise current in the CCS from the zener noise. The LSB for TDA1541 is 61nA, so I don't think we add undue noise
Even unfiltered 431 is "only" 42nA noise, borderline but not much filtering needed to get enough below the LSB.
Zener vs TL431 have poorer stability with current change and higher impedance.
I switched to these Zenners from LED's in AMR designs. More consistent. Less thermal drift. In no design did we observed more noise.
Thor
Last edited:
Looks to be feeding an active load that holds the operating point reasonably well. Hence doubtful to go much outside the +/-25mV compliance limit spec. I believe that many diyer's are operating the device swinging well into the 100's of mV using a passive resistive network with the TDA1541a. Hence it doesn't seem so important as there are many reports of sonic virtues in doing so.
In direct coupled balanced it seems likely this would be helped by some DC servo network anyway. Seems the TDA1541a operating point would drift some as well.
From noise perspective an LED based CCS seems likely overkill, though if the voltage supply is restricted to 5 Volts it seems a reasonable choice. Of course one could use a diode string... but green LED's are environmentally cool looking... and the red ones are too dangerous looking.
Hi,
I don't want a lengthy debate here, do whatever you like.
HOWEVER, for the circuit I illustrated in this post:
https://www.diyaudio.com/community/...ac-using-tda1541a.79452/page-449#post-7759538
The CCS must be temperature invariant. The reason that a later version replaced the 3.3V Zener diode version shown there with TL431 was that in the simulator the temperature drift was unacceptable.
As long as all BJT circuits are correctly thermally coupled they compensate any thermal drift sufficiently if a stable reference is present. This is a deliberate design feature in order to avoid the requirement for DC servo's etc.
So if anyone elects to use a different type of CCS they need to validate that the thermal performance does not cause problems.
Thor
I don't want a lengthy debate here, do whatever you like.
HOWEVER, for the circuit I illustrated in this post:
https://www.diyaudio.com/community/...ac-using-tda1541a.79452/page-449#post-7759538
The CCS must be temperature invariant. The reason that a later version replaced the 3.3V Zener diode version shown there with TL431 was that in the simulator the temperature drift was unacceptable.
As long as all BJT circuits are correctly thermally coupled they compensate any thermal drift sufficiently if a stable reference is present. This is a deliberate design feature in order to avoid the requirement for DC servo's etc.
So if anyone elects to use a different type of CCS they need to validate that the thermal performance does not cause problems.
Thor
A Temperature stable, low noise CCS can be made in several ways.
1) Start with a temperature stable voltage reference, which invariably is bandgap and thus noisy. Compensate the temperature dependence of the actual CCS (Baxandall Super Pair, Differential Amplifier etc), filter the noise from the bandgap. This approach is highly reliable und unlikely to be affected by process/parameter drift but most complex.
2) Start with a low enough noise reference source that has the complementary temperature characteristics to the CCS element and make a CCS. This just needs 3 parts (BJT, Resistor & Reference). The thermal behaviour of Zener diodes is well defined as is the reference voltage and as the technology is mature process/parameter drift is unlikely (but cannot be reliably excluded). Low voltage Zener's are fairly low noise and adding a big value parallel capacitor can lower noise further due to their relative high impedance.
LED's are often used, but sadly I found them subject to process/parameter drift and so unless the specific LED model from a specific maker with stable processes is listed and applied in circuit, the behaviour of the circuit is not reliable. So "Red LED" is not a sufficient specification any longer, it was in the 1970's.
3) Use a J-Fet or Depletion Mosfet selected to have it's zero tempco current at the desired current.
Here we see that J111 has zero-tempco near 5mA, so it looks good for a CCS at ~5mA.
Add a trimmer to compensate the J-Fet Idss variation. The wide variation of J-Fet's means that probably devices must be selected for correct zero tempco current.
While 1 & 2 can be made reliable without adjustment baring process/parameter drift, 3 must be adjusted manually.
Above a comparison of the CCS current stability with Temperature for "Bandgap (TL431)", "Red LED", 3.3V Zener (BZX55C series) and J-Fet with correct zero tempco current.
Both Bandgap and suitable FET have near ideal thermal behaviour. The Zener diode CCS is a little worse, but may be acceptable under most conditions. LED? Maybe. I think the current variation is huge.
Impedance of each CCS (same colours as before):
Noise of each CCS in pA|/Hz (same colours as before):
Multiply with 141 for a 20khz bandwidth noise current or 316 for 100kHz) - for example the noisiest option, LED with 11pA|/Hz gives 3.4nA for 100khz bandwidth.
It is questionable if all noise sources are adequately modelled, so especially noise is probably not accurate.
Thor
Nice work Thor.
Are you aware of any contributing factors specific to the temperature stability and set-point variation of the 2mA output current of the TDA1541a's? If the set point is critical this may require adjustment in any event.
One of my primary concerns would relate to tight control of the operating point from the perspective that the network shows a coupling capacitor (seemingly likely an electrolytic) at the output. IMO these have always taken time to "settle" from a psycho-acoustic perspective, if not days sometimes weeks. Meaning that having the potential across these devices drifting around may limit the extent they can sonically settle over time.
I am not a fan of any dielectric materials really, hence any storage elements are always suspect.
Are you aware of any contributing factors specific to the temperature stability and set-point variation of the 2mA output current of the TDA1541a's? If the set point is critical this may require adjustment in any event.
One of my primary concerns would relate to tight control of the operating point from the perspective that the network shows a coupling capacitor (seemingly likely an electrolytic) at the output. IMO these have always taken time to "settle" from a psycho-acoustic perspective, if not days sometimes weeks. Meaning that having the potential across these devices drifting around may limit the extent they can sonically settle over time.
I am not a fan of any dielectric materials really, hence any storage elements are always suspect.
And Again - Decoding TDA1541 Power and Ground Pin's, the definitive edition
So, this has been quite amusing. Trying to decode the various pin's on the TDA1541 and getting their actual function made me realise JUST HOW MISLEADING the labels are!
Please note, the following is not for anyone with fixed believes, who feels that the Philips Engineers knew everything, that one specific individual has all the truth in the universe. Stop reading now. You do NOT want to know this, it will only offend. Additionally it will not give you information you can use.
As additional prefix, many folks, even electronic engineers view things like "ground" as a real "thing". It is not. Ground is a convenient fiction that originated in early Radio's that had a terminal that needed to be connected to "ground" (literally the dirt under the house, aka "earth") power terminals (for batteries, +A, +B and sometimes -C) and the Antennae.
Ever since ground has stuck around as term in electronics and as a convenient fiction that simplifies drawing and understanding schematics. In truth however, just like in "The Matrix" where the truth is "there is no spoon":
And in electronics the truth is "there is no ground".
Ok, so much for mysticism, back to something concrete, the TDA1541.
Ok, TDA1541 has several groups of pins:
Digital Inputs (1/2/3/4) - these have non-standard inputs and reference/return to the Pin "DGND"
Format Pin (27) - actual reference unclear, probably -5V.
Output Pins (6/25) - these are current out and their return is the true analogue reference, which is not AGND.
DEM Filter Capacitor Pin's (13-7/18-24 MSB first) - these connect the capacitors perform averaging and act as "hold" capacitors for the current sources for all pin's that are fed by the DEM matrix. Their return is the true analogue reference, which is not AGND. They also "deglitch" the extremely high frequency glitches of the DEM switching matrix.
Power Pins (5/14/28/26/15) - I include pin's labelled "GND" here because the labels are not strictly functional. Actually, both pin's labelled something "GND" are POWER SUPPLY inputs while the pin's labelled as power pin's are the actual FUNCTIONAL "Grounds". More below.
There are two main pin pairs. One is:
+5V/DGND - these pin's run multiple duties, like all power/gnd pins
Internal logic in the TDA1541 is bipolar, differential so-called "Current Mode Logic" (CML) which is actually essentially identical to ECL. Here a classic ECL latch, most of the logic circuits inside the TDA1541 are latches:
The convention for ECL Logic is to have the positive connection of the logic acting as logic ground and a negative power supply applied to the negative pin's. So classic ECL actually has DGND and -5.2V power supply. This is because any capacitance imbalance or any single ended connection causes noise on the positive rail, while the CCS tails to the negative rail mean -5V is very quiet.
The +5V Pin also acts as dumping ground for the "waste" current of the DAC output, this is the inverse of the actual output, 0-4mA P-P per channel. Here and above the Bit switch(es) of the TDA1541:
Inherently current that is not directed to the output, is directed to the +5V rail, which actually is the "common" or "ground" of CM Logic.
Thus:
+5V = Digital "Ground" for the CM/EC Logic
DGND = -5V "Supply" for the CCS tails of CM/EC Logic
DGND = reference for the TTL -> ECL level conversion of the digital inputs
+5V = Analogue "Ground" for the complement of the output current both channels mixed to mono maximum 8mA P-P
In other words, +5V is VERY noisy due to a number of noise sources.
It needs to be decoupled directly to the ECL supply Rail labelled "DGND" which is in no real way a "ground" type pin but actually the power supply for the digital section AND to the true analogue reference, which is not AGND.
Decoupling must be effective not only for audio signals and the actual sample rate send to the TDA1541 and decoupling to the audio current return and to the digital power supply (aka "DGND").
The digital side, between the pin's labelled +5V (actually DGND) and DGND (actually -5V ECL Power) needs decoupling for frequencies from the word clock upwards. CM/EC logic produced strong higher harmonics, so mainly we need to consider ~32kHz to > 36MHz (if operating with 12MHz bit clock).
Analogue decoupling requires pretty wideband low impedance, from near DC all the way to the switching glitch current of the DEM matrix, which has narrow (< 20nS) glitches every time the DEM circuit switches, regardless of the DEM frequency. So that's 20Hz - 50MHz+!
Optimum RF decoupling of DEM pin's can "bypass" these glitches directly back to the to the true analogue reference, which is not AGND. This is needed even when operating a low DEM switching frequency (e.g. 50Hz).
If RF bypassing of the DEM pin's is not practiced these glitches end up in the +5V Pin and the audio current output. So dealing with the DEM Filter pin's correctly helps making +5V to analogue common decoupling less demanding.
DGND has no analogue function and is not required to be linked directly hardwired to AGND. The datasheet permits as much 0.3V (bipolar) DC voltage difference between DGND and AGND pins.
The analogue and digital functions are "hardwired and shot-gunned" together into pin 28, there is just no way to separate them. there Thus DGND and AGND Pin's much be separated to allow resolving digital and analogue decoupling separately. Hence directly linking AGND and DGND which we see frequently is a questionable practice.
To avoid noise in DGND and via coupling on the IC to DEM Pin's, +5V and Audio Outputs; the digital input's MUST not be driven with TTL or CMOS levels. Doing so causes glitches that appear on all relevant analogue pin's including the output:
This is 55mV P-P of BCK and additional noise from other switching! At MHz frequencies.
MVAudiolabs has an excellent page covering this subject, please peruse this:
TDA1541(A) Digital Input Attenuation
With correct conditioning of input signals, we end up with 2.2mV HF noise, a 29dB improvement.
Yes, this is a heck of a lot goings on on +5V & DGND.
But it seems we are (mostly) finished. Our key question remaining is": "where do we decouple +5V to?", where is that frequently dropped "true analogue reference, which is not AGND"?
Will you be really surprised if I tell you that the true analogue reference or true analogue "Ground" is -15V AND that the "AGND" Pin is actually the power supply pin to DEM system and current generation? You still are? Well, there it is.
This shows a re-creation of the TDA1541 DEM/Analogue section to individual transistor level based on the various published articles covering TDA1540 & TDA1541. Just looking at it makes clear, everything is referenced to -15V. The -15V Pin in reality acts as reference for the current sources and as the origin for all currents related to DEM and analogue outputs. So all related currents must return there,
Now you ask: "Why didn't Philips label -15V as "AGND" and and "AGND" as +15V? And why not label +5V as "DGND" and "DGND" as "-5V" and -5V as -10V, if as you claim this is what they are?".
For one, if you connected these "functional grounds" together, which is what a lot of people do with pin's labelled "gnd" we have a dead TDA1541. So the labelling here is NOT actually functional, BUT physical design related. If you connect AGND to DGND and then to the power supply and supply +5V, -5V and -15V you have a functioning system.
So the pin labels tell you NOT the actual function, but how to wire up the chip!
Understanding this we now see that by knowing what is really hiding under the pin labels we can draw some conclusions.
AGND = +15V "Supply" for the DEM logic, voltage and current references that ultimately produce output current.
-15V = Analogue "Ground" for the Current Sink and DEM Logic
The two most important Pin's are Pin 15 (-15V) and Pin 28 (+5V). After lengthy research, -5V pin is commonly treated very cavalier and appears to relate neither directly to the digital or analogue side.
Anything analogue must be star-grounded (returned) to -15V, NOT AGND.
This includes the DEM Filter capacitors and the "+5V" pin.
So DEM Filter capacitors go to -15V, not AGND. The decoupling from +5V goes to -15V, not AGND. The decoupling from "AGND" (aka +15V supply) goes to -15V. Yup, -15V is where it's at. For RF a plane is needed. RF loops from +5V and DEM Filter pins should be very short.
Seeing that +5V pin to -15V pin is the most critical current loop, we really need optimised shunt regulators so that we have a low impedance DC & Audio connection with 20V DC between those pin's. They need 20V DC and 0V AC from DC to 100MHz+!!!!
The -5V supply is not so consequential, but if we tap it off the 20V circuit (so two 5V shunt regulators and one 10V shunt regulator) anyway, it is made to the same standards. And yes, -5V decouples for RF to -15V, not AGND.
So that's it pretty much.
For my personal build, that triggered all this research, I will use this power supply design, which illustrates a practical approach to all discussed previously:
There is one 20V shunt regulator + supercapacitor chain to deliver 5V + 5V + 10V giving all voltages needed.
There are also 2 X 20V/30mA for a linedriver stage
AGND and DGND are separated by an inductor. The Schottky diodes shown across it protect the TDA1541 C and belong under the IC. I will test a bit before a final inductor is selected, the 1.5R/1.5mH one shown is just a placeholder for simulation
The digital 5V supply has an extra 3F/5.4V supercapacitor bank and other decoupling, connected to DGND. So it forms a low impedance separate PSU, with appx. 60 mOhm above 1Hz, falling to 6mOhm at 100kHz and remaining in this ballpark to beyond 20MHz.
Decoupling +5V to -15V is 100nF SMD C0G ceramic capacitors and two 47uF stacked film types, for a total of 94uF film and then the supercapacitor bank of > 9,000,000uF with 40mOhm impedance above ~0.4Hz. No other chemical capacitors in the +5V/-15V loop.
The rest is pretty generic and not worth debating.
I understand, some will hate TL431 Shunt regulators, even though their noise is reduced by ~60dB at 10Hz as result of the LC filter with the supercapacitor bank.
Others will be outraged at the thought that 24pcs 25F supercapacitors are used to power a DAC and that the power supply takes ~300ma 30+V (10VA) continuous despite this and wish to throw soup at the DAC and then superglue themselves to the DAC.
But hey, it's DIY. If we put in the time, why not do something different?
Thor
So, this has been quite amusing. Trying to decode the various pin's on the TDA1541 and getting their actual function made me realise JUST HOW MISLEADING the labels are!
Please note, the following is not for anyone with fixed believes, who feels that the Philips Engineers knew everything, that one specific individual has all the truth in the universe. Stop reading now. You do NOT want to know this, it will only offend. Additionally it will not give you information you can use.
As additional prefix, many folks, even electronic engineers view things like "ground" as a real "thing". It is not. Ground is a convenient fiction that originated in early Radio's that had a terminal that needed to be connected to "ground" (literally the dirt under the house, aka "earth") power terminals (for batteries, +A, +B and sometimes -C) and the Antennae.
Ever since ground has stuck around as term in electronics and as a convenient fiction that simplifies drawing and understanding schematics. In truth however, just like in "The Matrix" where the truth is "there is no spoon":
And in electronics the truth is "there is no ground".
Ok, so much for mysticism, back to something concrete, the TDA1541.
Ok, TDA1541 has several groups of pins:
Digital Inputs (1/2/3/4) - these have non-standard inputs and reference/return to the Pin "DGND"
Format Pin (27) - actual reference unclear, probably -5V.
Output Pins (6/25) - these are current out and their return is the true analogue reference, which is not AGND.
DEM Filter Capacitor Pin's (13-7/18-24 MSB first) - these connect the capacitors perform averaging and act as "hold" capacitors for the current sources for all pin's that are fed by the DEM matrix. Their return is the true analogue reference, which is not AGND. They also "deglitch" the extremely high frequency glitches of the DEM switching matrix.
Power Pins (5/14/28/26/15) - I include pin's labelled "GND" here because the labels are not strictly functional. Actually, both pin's labelled something "GND" are POWER SUPPLY inputs while the pin's labelled as power pin's are the actual FUNCTIONAL "Grounds". More below.
There are two main pin pairs. One is:
+5V/DGND - these pin's run multiple duties, like all power/gnd pins
Internal logic in the TDA1541 is bipolar, differential so-called "Current Mode Logic" (CML) which is actually essentially identical to ECL. Here a classic ECL latch, most of the logic circuits inside the TDA1541 are latches:
The convention for ECL Logic is to have the positive connection of the logic acting as logic ground and a negative power supply applied to the negative pin's. So classic ECL actually has DGND and -5.2V power supply. This is because any capacitance imbalance or any single ended connection causes noise on the positive rail, while the CCS tails to the negative rail mean -5V is very quiet.
The +5V Pin also acts as dumping ground for the "waste" current of the DAC output, this is the inverse of the actual output, 0-4mA P-P per channel. Here and above the Bit switch(es) of the TDA1541:
Inherently current that is not directed to the output, is directed to the +5V rail, which actually is the "common" or "ground" of CM Logic.
Thus:
+5V = Digital "Ground" for the CM/EC Logic
DGND = -5V "Supply" for the CCS tails of CM/EC Logic
DGND = reference for the TTL -> ECL level conversion of the digital inputs
+5V = Analogue "Ground" for the complement of the output current both channels mixed to mono maximum 8mA P-P
In other words, +5V is VERY noisy due to a number of noise sources.
It needs to be decoupled directly to the ECL supply Rail labelled "DGND" which is in no real way a "ground" type pin but actually the power supply for the digital section AND to the true analogue reference, which is not AGND.
Decoupling must be effective not only for audio signals and the actual sample rate send to the TDA1541 and decoupling to the audio current return and to the digital power supply (aka "DGND").
The digital side, between the pin's labelled +5V (actually DGND) and DGND (actually -5V ECL Power) needs decoupling for frequencies from the word clock upwards. CM/EC logic produced strong higher harmonics, so mainly we need to consider ~32kHz to > 36MHz (if operating with 12MHz bit clock).
Analogue decoupling requires pretty wideband low impedance, from near DC all the way to the switching glitch current of the DEM matrix, which has narrow (< 20nS) glitches every time the DEM circuit switches, regardless of the DEM frequency. So that's 20Hz - 50MHz+!
Optimum RF decoupling of DEM pin's can "bypass" these glitches directly back to the to the true analogue reference, which is not AGND. This is needed even when operating a low DEM switching frequency (e.g. 50Hz).
If RF bypassing of the DEM pin's is not practiced these glitches end up in the +5V Pin and the audio current output. So dealing with the DEM Filter pin's correctly helps making +5V to analogue common decoupling less demanding.
DGND has no analogue function and is not required to be linked directly hardwired to AGND. The datasheet permits as much 0.3V (bipolar) DC voltage difference between DGND and AGND pins.
The analogue and digital functions are "hardwired and shot-gunned" together into pin 28, there is just no way to separate them. there Thus DGND and AGND Pin's much be separated to allow resolving digital and analogue decoupling separately. Hence directly linking AGND and DGND which we see frequently is a questionable practice.
To avoid noise in DGND and via coupling on the IC to DEM Pin's, +5V and Audio Outputs; the digital input's MUST not be driven with TTL or CMOS levels. Doing so causes glitches that appear on all relevant analogue pin's including the output:
This is 55mV P-P of BCK and additional noise from other switching! At MHz frequencies.
MVAudiolabs has an excellent page covering this subject, please peruse this:
TDA1541(A) Digital Input Attenuation
With correct conditioning of input signals, we end up with 2.2mV HF noise, a 29dB improvement.
Yes, this is a heck of a lot goings on on +5V & DGND.
But it seems we are (mostly) finished. Our key question remaining is": "where do we decouple +5V to?", where is that frequently dropped "true analogue reference, which is not AGND"?
Will you be really surprised if I tell you that the true analogue reference or true analogue "Ground" is -15V AND that the "AGND" Pin is actually the power supply pin to DEM system and current generation? You still are? Well, there it is.
This shows a re-creation of the TDA1541 DEM/Analogue section to individual transistor level based on the various published articles covering TDA1540 & TDA1541. Just looking at it makes clear, everything is referenced to -15V. The -15V Pin in reality acts as reference for the current sources and as the origin for all currents related to DEM and analogue outputs. So all related currents must return there,
Now you ask: "Why didn't Philips label -15V as "AGND" and and "AGND" as +15V? And why not label +5V as "DGND" and "DGND" as "-5V" and -5V as -10V, if as you claim this is what they are?".
For one, if you connected these "functional grounds" together, which is what a lot of people do with pin's labelled "gnd" we have a dead TDA1541. So the labelling here is NOT actually functional, BUT physical design related. If you connect AGND to DGND and then to the power supply and supply +5V, -5V and -15V you have a functioning system.
So the pin labels tell you NOT the actual function, but how to wire up the chip!
Understanding this we now see that by knowing what is really hiding under the pin labels we can draw some conclusions.
AGND = +15V "Supply" for the DEM logic, voltage and current references that ultimately produce output current.
-15V = Analogue "Ground" for the Current Sink and DEM Logic
The two most important Pin's are Pin 15 (-15V) and Pin 28 (+5V). After lengthy research, -5V pin is commonly treated very cavalier and appears to relate neither directly to the digital or analogue side.
Anything analogue must be star-grounded (returned) to -15V, NOT AGND.
This includes the DEM Filter capacitors and the "+5V" pin.
So DEM Filter capacitors go to -15V, not AGND. The decoupling from +5V goes to -15V, not AGND. The decoupling from "AGND" (aka +15V supply) goes to -15V. Yup, -15V is where it's at. For RF a plane is needed. RF loops from +5V and DEM Filter pins should be very short.
Seeing that +5V pin to -15V pin is the most critical current loop, we really need optimised shunt regulators so that we have a low impedance DC & Audio connection with 20V DC between those pin's. They need 20V DC and 0V AC from DC to 100MHz+!!!!
The -5V supply is not so consequential, but if we tap it off the 20V circuit (so two 5V shunt regulators and one 10V shunt regulator) anyway, it is made to the same standards. And yes, -5V decouples for RF to -15V, not AGND.
So that's it pretty much.
For my personal build, that triggered all this research, I will use this power supply design, which illustrates a practical approach to all discussed previously:
There is one 20V shunt regulator + supercapacitor chain to deliver 5V + 5V + 10V giving all voltages needed.
There are also 2 X 20V/30mA for a linedriver stage
AGND and DGND are separated by an inductor. The Schottky diodes shown across it protect the TDA1541 C and belong under the IC. I will test a bit before a final inductor is selected, the 1.5R/1.5mH one shown is just a placeholder for simulation
The digital 5V supply has an extra 3F/5.4V supercapacitor bank and other decoupling, connected to DGND. So it forms a low impedance separate PSU, with appx. 60 mOhm above 1Hz, falling to 6mOhm at 100kHz and remaining in this ballpark to beyond 20MHz.
Decoupling +5V to -15V is 100nF SMD C0G ceramic capacitors and two 47uF stacked film types, for a total of 94uF film and then the supercapacitor bank of > 9,000,000uF with 40mOhm impedance above ~0.4Hz. No other chemical capacitors in the +5V/-15V loop.
The rest is pretty generic and not worth debating.
I understand, some will hate TL431 Shunt regulators, even though their noise is reduced by ~60dB at 10Hz as result of the LC filter with the supercapacitor bank.
Others will be outraged at the thought that 24pcs 25F supercapacitors are used to power a DAC and that the power supply takes ~300ma 30+V (10VA) continuous despite this and wish to throw soup at the DAC and then superglue themselves to the DAC.
But hey, it's DIY. If we put in the time, why not do something different?
Thor
Last edited:
Should be ok at 0.02% per K.
We never trimmed individual TDA1541 for gain or offset at AMR in mass production.
Bipolar electrolytic. These are lot cleaner and have less variation, plus I will use 100V rated devices at 100uF. This places the -3dB point at ~2.7Hz for a 600 Ohm load, not that I intend to use such low loads.
Everything is a compromise.
Usually direct coupling replaces a visible coupling capacitor that we take great care of with a PSU cap or a servo, where peep's take less care. The result is them often worse than just using a coupling cap.
Again, for me there is a trade-off between circuit simplicity, performance and the need for DC blocking capacitors.
Thor
Not really, quite the opposite, in my current design, i use smd 431 shunt reference, and low V super caps. Regs should handle that much virtual short. Just with a concept you had before. I'll rework it for this.
But it will have to wait, vacation time is upon me, no laptop just phone, salting my bottom at seaside.
In the meantime, if you can have a look at this. Pdf full doc in attachment.
In some CD players I see -6 V instead of -5 V, and -13.5 V instead of -15 V. Is it the result of further optimization by the designers?
No, it was by the old/first datasheet as a max value. It turns out it burns chips after a while. Some probably had better tolerance, some were not so lucky. It was remedied in latest datasheet to 5.5V.
Quite noisy, stability depends on reference, noise too, up to the high self noise. high cost and parts count. Not really that useful for audio. What do you want to with it?
A single half of 5532 + a PNP transistor can do better on noise with sufficient filtering of the reference and otherwise as good and is a better choice for audio.
Thor
I checked the models (about BC847B to BC567B it seems that It was about models - I use first generic model than Philips model. With philips was good...
.
In real world probably a slight changes for biasing resistors will occure. But that coul be solved. More tunning will be with Rgs For JFETs but there is not a big problem....
.
In real world probably a slight changes for biasing resistors will occure. But that coul be solved. More tunning will be with Rgs For JFETs but there is not a big problem....
Nothing big is changed with J111 and 2N4393. ith +5V as current sources. Booth has the around 5mA Thermal stability. Except that the J111 will opperate in linear region a bit less than Vds=<1V, and 2N4393 will worki in saturation region Vds=>4.4V
(I put the 100 ohm Riv for o.4Vp-p, that can be amplified with the direct coule tube stage...)
FFT was time=21ms, start=1ms, step=100ns, thd for 7 harmonics... Will put 9 or 11 next time but there will be no significant changes.
.
(I put the 100 ohm Riv for o.4Vp-p, that can be amplified with the direct coule tube stage...)
FFT was time=21ms, start=1ms, step=100ns, thd for 7 harmonics... Will put 9 or 11 next time but there will be no significant changes.
.
Hi Maybe someone will go for Germanium version ? 😡
This is about the max level reached until 5th harmonic goes higher... But 3.2Vp-p is probably better to use with preamplifier?
Harmonics order look a like the triode stage. 😎
.
(The issues was significant dips at the HF cutt-off. I tried to supress them with Rbase and some C coupling. Tried a different connections , in the same time using a lower C values. Not sure that I was did it the best way? The minimum phase shift I try to obtain is about -17dg @ 20KHz and with -0.3db, that will add to NOS mode rollof of about -2.7db).
.
This is about the max level reached until 5th harmonic goes higher... But 3.2Vp-p is probably better to use with preamplifier?
Harmonics order look a like the triode stage. 😎
.
(The issues was significant dips at the HF cutt-off. I tried to supress them with Rbase and some C coupling. Tried a different connections , in the same time using a lower C values. Not sure that I was did it the best way? The minimum phase shift I try to obtain is about -17dg @ 20KHz and with -0.3db, that will add to NOS mode rollof of about -2.7db).
.