Okay.... explication mode ON
Dear Batteryman, whatever LiPoFe, lead or whatever...
Are you more qualified, have you tried all the rest as well as I did, do you know what all I heard already at home or elswhere ?
Sorry but here as far I know it is your opinion too, you have not the last word of what works best, many disagree with just a traffo here for the task,, as for Sowter, but sorry there are many better made if you ask me ! In my journey I appologize too, you don't seem more qualified more (or less) than I am, if words have sense ! It seems people do not understand anymore the concepts behinds the words they use ! Words have sense, it's not just blah ! Of course the yours arguments of few brands that used a traffo has nothing to see with a formal demonstration nore a scientifical one ! And does I say it is bad ?
Has for what you say to tubes, I don't care more than others solutions ? Common, what is difficult with HV, have you tried a simple 160H Sowter with an ECC88 (very linear between 65V to 90/95 V plate), or a CCS or a Mu follower, or etc... What you know more than others ? If we could make it quieter please ! It is my opinion too my "b....s" are bigger than others 😉 !
Think it twice ! Imho ! You don't talk to the dude that tried the TDA1541A 10 y ago only ! Like many here. And I hope it is clear too for the guy that make WWII plane modelism and that pop ups the screens with measurements took on so-so layouts perhaps and matos! And if it is only about THD, go buy an AKM or an ESS ! Ca va bien 5 minutes !
Be cool guys, it is the end of the world, but the few guys here than have good ideas like Thorsten Loesch, Pedja Rogic or John Brown, I see noone that can be a rooster ! 😉
Explication Mode OFF (this is not an attack, just to unflat the baloon !) ! If we can make it quiter for the good guys to writzze without take a METH or coco to survive the agressiv feelinh it could be better ! Thinks Olympics, not Olympiggs !
regards,best, ymmv, etc. ANd I only write because I respect you, if I just thougth it, it would have mean worse for what I can thinnk about you ! So please don't see it as an attack, it is a little a defense mode but not so !
Bob Marley mode ON !
🙂
Edit : @ThorstenL : please tell us more,on how today you will mix maybe all that I/V tools relative to that IC (if anyome have the real output impedance and output capacitance, I am glad to know for my book). You talked not anymore of tubes but CCII few posts above.... ?! Look forward to know more !
Dear Batteryman, whatever LiPoFe, lead or whatever...
Are you more qualified, have you tried all the rest as well as I did, do you know what all I heard already at home or elswhere ?
Sorry but here as far I know it is your opinion too, you have not the last word of what works best, many disagree with just a traffo here for the task,, as for Sowter, but sorry there are many better made if you ask me ! In my journey I appologize too, you don't seem more qualified more (or less) than I am, if words have sense ! It seems people do not understand anymore the concepts behinds the words they use ! Words have sense, it's not just blah ! Of course the yours arguments of few brands that used a traffo has nothing to see with a formal demonstration nore a scientifical one ! And does I say it is bad ?
Has for what you say to tubes, I don't care more than others solutions ? Common, what is difficult with HV, have you tried a simple 160H Sowter with an ECC88 (very linear between 65V to 90/95 V plate), or a CCS or a Mu follower, or etc... What you know more than others ? If we could make it quieter please ! It is my opinion too my "b....s" are bigger than others 😉 !
Think it twice ! Imho ! You don't talk to the dude that tried the TDA1541A 10 y ago only ! Like many here. And I hope it is clear too for the guy that make WWII plane modelism and that pop ups the screens with measurements took on so-so layouts perhaps and matos! And if it is only about THD, go buy an AKM or an ESS ! Ca va bien 5 minutes !
Be cool guys, it is the end of the world, but the few guys here than have good ideas like Thorsten Loesch, Pedja Rogic or John Brown, I see noone that can be a rooster ! 😉
Explication Mode OFF (this is not an attack, just to unflat the baloon !) ! If we can make it quiter for the good guys to writzze without take a METH or coco to survive the agressiv feelinh it could be better ! Thinks Olympics, not Olympiggs !
regards,best, ymmv, etc. ANd I only write because I respect you, if I just thougth it, it would have mean worse for what I can thinnk about you ! So please don't see it as an attack, it is a little a defense mode but not so !
Bob Marley mode ON !
🙂
Edit : @ThorstenL : please tell us more,on how today you will mix maybe all that I/V tools relative to that IC (if anyome have the real output impedance and output capacitance, I am glad to know for my book). You talked not anymore of tubes but CCII few posts above.... ?! Look forward to know more !
Last edited:
That is true , but what the man is qualified for, " word salad " maybe 🤣
ps : that plane is a real one , I almost flew it 😎
.
Sorry guy, I find your words are poors, we know all the two cents method of ping pong arguments you use because of no better to say or to find to say!
Words Salad is what said Thorsten to me on just a question among many you even had not the idea to ask, so you perpetual attacks are ridiculous as you only show what you try to critic ! IMHO. You used three times or more thaat words to attack me in this thread poping up to what a guy (Thorsten) said without agresssivity. Mods already advert you, and put me in the same bag than you !
You almost flew ?! SOmetimes people are just almost.....clever ! Hey you start, ,no? As usual btw, not the first time you attack! Do you dream of me too ? We are all finish sim binned because of your agressiv behavior; that also makes guys in love mode like Thorsten to flee to write ! Just because you believe you'tre a plane ! Too bad 🙂
... And as said Schrödinger, there are many ways to skin a cat... and for us, more gentle, to make it singing !
Words Salad is what said Thorsten to me on just a question among many you even had not the idea to ask, so you perpetual attacks are ridiculous as you only show what you try to critic ! IMHO. You used three times or more thaat words to attack me in this thread poping up to what a guy (Thorsten) said without agresssivity. Mods already advert you, and put me in the same bag than you !
You almost flew ?! SOmetimes people are just almost.....clever ! Hey you start, ,no? As usual btw, not the first time you attack! Do you dream of me too ? We are all finish sim binned because of your agressiv behavior; that also makes guys in love mode like Thorsten to flee to write ! Just because you believe you'tre a plane ! Too bad 🙂
... And as said Schrödinger, there are many ways to skin a cat... and for us, more gentle, to make it singing !
Last edited:
Cool man nobody attacks you , i am just kidding , remember the last two months you got holydays , it is time to take some more , it was so cool 😎
.
.
Why do that? Image planes in part work by coupling.
On the DEEM system connected between AGND & -15V we have in principle constant current. Everything is either differential Current Mode Logic (CML) or current sources.
In practice, the limited logic speed and the fact that we eventually transition from CML to single ended saturated switches means each time the DEM oscillator switches we have a significant current glitch on both the AGND to -15V Loop (the glitches that propagate into the AGND pin from the switching circuitry) on the DEM Filter Capacitor to -15V Loop and the -15V to Aox/+5V loop.
These glitches are very narrow, as said, estimated 10....20nS (or less, I'm eyeballing) and not directly related to the DEM clock frequency (Fdem). They are always ~10nS wide, Fdem determines how often they happen.
Glitch currents appearing on AGND and CDEM Pin's should return to -15V (which is the actual "reference" for the entire "DEM Current Sink" circuit) via low impedance at 50...100MHz and allowing for harmonics.
The stacked planes under the TDA1541 form a distributed capacitor of around 220pF (based on 2 mil Kapton and an area of 30 X 12mm for the AGND plane) with a first resonance in the GHz range, that is it forms a low ESR (copper plates) decoupling capacitor.
Additionally, the top -15V plane forms a low impedance return path for glitch currents that pass through either AGND decoupling pin (note that I consider AGND a Power/Signal pin, not "ground", the closest to a "ground" (that is a common place all currents in the loop return to in DEM is -15V).
The decoupling I would (will) choose stacks 1206 Capacitors, first 100nF C0G Ceramic (these are second only on linearity etc to Teflon) and 680nF ECPU Film. So 780nF per DEM Filter Pin. This lowers the edge rate of currents at the DEM pin's substantially.
So I fail to see why anyone would choose anything except stacked planes under the TDA1541 for RF decoupling.
Once we have taken care of RF and removed it as much as possible from the wider circuit, we can consider the DEM current ripple caused by the tolerance of the BJT current divider. If we do Fdem => 176.4khz we are already done. Extra, larger value capacitors in a longer loop on the outside are of questionable value in the chosen setup.
If we do 50/60Hz or even 25/30Hz DEM clocking e need large value external capacitors of temns to 100's uF low leakage.
We know we get current steps at Fdem of around 5% of the Bit current maximum. The edge rate is the same as our switching glitches, but our RF decoupling under TDA1541 has already "taken the edge off" (pun intended) this. We could calculate the new edge rate from the the capacitor values and current, but with 780nF and 5% of 2mA it's going to be slow. That is the second reason why I consider RF decoupling under the TDA1541 mandatory even for slow DEM clocks.
Our Nichicon KL low leakage capacitors have enough ESL and ESR that we need not worry much about impedance. A plane would be a waste. Return each capacitors negative (positive to pin) as a single line to -15V (star "ground"). So 14 wires/lines converge on Pin 14.
But we WANT "stray" Capacitance, a lot of it!
I would stop AGND/-15V Planes at pin 5. They only exist south of pin 4. With a small "bridge" allowing RF decoupling of Aol.
We may remember that MVAL complains that the I/U conversion circuit needs to handle all these super fast (> 50...100Mhz) Logic glitches. This is because he chooses to not provide a "short" path for such currents. I suggest we supply such a path.
AOL/AOR contain significant HF glitches that are the complement to similar glitches on +5V.
So we need to either decouple Aol/Aor to -15V or +5V. Where is debatable.
I tend towards -15V as return point for both the Aol/AoR and +5V glitch currents. Select decoupling such, that with the actual impedance at Aol/Aor the impedance rolls off at appx. 1MHz and without significant peaking.
The +5V Plane under the PCB extends close to the AGND / -15V plane pair. As the voltage difference from +5V to is 20V need to pick capacitors that can handle 20V. As I have very low ESL/ESR Epcos stacked film capacitors in 100V/47uF in the box, these will be used on +5V to -15V decoupling.
For -5V to +5V the 100n C0G + 680n ECPU is usable. Same for actual -15V to AGND decoupling.
For the specific analogue stage I am intending to use (bipolar current conveyor, "deep class A" plus follower wrapped into a sallen key filter) with ~0.45 Ohm input Impedance in the audio band and rising to ~0.6 Ohm above 20kHz the same 0.68u ECPU + 100n C0G will work ok for compensating the input impedance of the current conveyor.
Note that these two are not free from resonance's but have a controlled resonance with ~400mOhm @ 9MHz, which is ok for the TDA1541, where we see limited logic switching noise from a limited speed BCK, Data etc. and mainly very high frequency stuff. So this combo is ok for the way TDA1541 behaves, but it's not universal.
Anyway it makes sure that the lookback impedance of the Aol/Aor pin remains below 0.6 Ohm from pretty much DC to 100MHz and that unwanted noise is returned to -15V and does not enter AGND or indeed the audio circuit (well, it does, just attenuated a lot).
The attached screen from my Simulation shows the impedance (and also the intended TDA1541 & Analogue stage PSU arrangements). And yes, the power supply for the analogue part is 90V DC.
Tubes could be used, but the main current conveyor (T8/T4/T6) needs to be bipolar anyway (easy temperature compensation, no servo needed). No mosfet or J-Fet with sensible current budget (~10mA) gets a low enough impedance to not have error voltages at Aol/Aor to exceed the highly linear range of TDA1541. So might as well stick to it.
Thor
thanks.
"But we WANT "stray" Capacitance, a lot of it"
I do not understand why, especially with output curent pins ? I think I didn't understand about your short answer here !
"But we WANT "stray" Capacitance, a lot of it"
I do not understand why, especially with output curent pins ? I think I didn't understand about your short answer here !
Interesting.
Actually, in pure objective fact (which may or may not relate to subjective perceived sound) lower values reduce compression and higher values "crush" the tops, in sound engineers speak, that is they compress peaks. Higher values also increase HD for higher level signals.
Back in the 80's (actually I still put the modern equivalent into any club sound system) we had a magic box called Aphex (pronouce Eff-Ex). You dialled the knobs right, it made everything louder, more dynamic, hard to be precise in words, it gives the sound a certain je nais se quois.
SE Tube Amplifiers are very much an "Aphex" style device.
For want of a better way to put it, if a chain is too clean, the sound is lifeless and sounds "thin", which is the result of having real dynamics without compression. If I was mastering such tracks, I would put on a optical compressor (likely emulated in my Yamaha 01V/95V2VCM) with just a tiny bit of bottom compression (1 : 1.5 maybe) below -40dB, add a tape emulator (clipping only) to crush the tops by around 6dB and add a touch of Aphex. I'd probably also try if the pitch is boring bog standard 440Hz to pitch down to 432Hz (more relaxing) or 448Hz (more aggressive, more energy).
I have no issue with people who seek to emulate something like that by selecting a high value I/U conversion resistor and like the sound. It is however not one of my design goals, as the DAC will be used mainly with Tube Amplifiers and it would produce too much of a good thing in this case.
Plus, I disagree with tuning equipment for a specific sound in a Chain that validated as "neutral". But that is just me.
"Do what thou wilt shall be the whole of the Law" 93 93/93
Thor
Yes. Simply because it is not possible to avoid obvious and audible distortion from TDA1541 if the I/U conversion resistor produces much more than 30mV Peak into one direction.
A 20 ohm I/U conversion resistor was used as compromise, the added distortion was very modest and swamped by the tube tube stage. The secondary loading at 47k lowers this resistor somewhat.
The 80mV P-P after 20R were stepped up to around 0.25V RMS after DCR losses of the transformer.
The E182CC in SRPP with un bypassed lower cathode gives a gain of around 12, so with 0dBfs we end up with around 3V RMS.
The DAC was intended to be used with "audio synthesis passion" style passive controllers.
Without step-up transformers a much less linear and higher gain analogue stage is needed.
If we use a larger value resistor the sound get's more coloured, but not in an objectional (to my ears) way until we get far past 50 Ohm.
It depends upon design goals.
Yes, but I have my own style doing such things.
Thor
At least, this was to know if you add or remove something in Sergio's I/V stage (intended for ESS9018-8 mA output) ; Many find it nice because no complex sorting out of the parts (live with H2) . The THD is cool, well for the onesthat has obsession as an alpha and omega point of view !
Not only Sowter makes good transformers. I had an AD1865 DAC (battery powered) with Sowter I/U conversion transformers to line level. It was procured on the insistence of an AMR Dealer who declared it the best ever at full retail and it was not cheap.
Despite correct output levels it sounded silent, small sound stage and thin/bright in many systems. It was eventually passed on at a loss. The AMR Dealer also fairly rapidly cooled on the DAC and soon dropped it from his shop.
Much of the engineering was interesting and good, just the sound was lacking.
I can understand why with this solution you want to "wind up" I/U conversion resistor values even if extra gain is not needed. I tested the transformers on the AP-2 and remember being unimpressed. Now this over two decades ago, so memory is a bit hazy.
But I remember the distinct sound you described for low value I/U conversion resistor from that box with those transformers.
Thor
ES9018 != TDA1541
IME the ES9018 sounds best in voltage mode, with fast local clock (100MHz and up) and with 1:1/1:2 transformers (10k :1 0k + 10k) that are intentionally bandwidth limited (so no bifilar windings please) and have a high nickel core. The 10k windings with nominal +26dBu handling means the rise in LF distortion is minimal at ES9018 voltage out levels.
Because I can, in my design the output was buffered with a 5687WB set up as "Wright Super Linear Cathode Follower" (read cascoded and CCS tail).
The result still had great SNR and low enough HD, but sounded rather nice. I have a Chinese 9038 Pro based DAC here (decent board and layout) and I'm waiting on some good transformers to replicate this as another "DAC Reference" which will be rather different to TDA1541.
Thor
Actually....
I disagree with tuning equipment for a specific sound in a Chain that IS NOT validated as "neutral".
To enlarge this point. If designing in a singular chain and "by ear" there is a tendency to tune individual elements in a complementary fashion. It can give excellent results, but any conclusions as to what a given technical change does for or against sound lacks any applicability outside this system, unless we can somehow quantify and "dial out" the rest of the chain, which is challenging.
Thor
I didn't say it was the same chip. I just said he has not designed it for TDA1541A. And the H2 (so furthers) is so low, there is no needs to sort out parts. Of course it is a sim. It seemed to me a good candidate with 0.2 something Z input, if we add the 2 mA injection to have +2mA/-2 mA offset. Some others here prefers Jfet inputs, etc. Was more intented to have your 'pinion as you know the TDA1541A.
But I can understand it is not that simple and here a curent injection migth not work in that circuitry (so not good aat the end for the TDA1541A... or will it works with - 2 mA zero centered : flip the polarity of the loudspeakers ??? (I claim to salad word myself but at least I ask ! Not intented to you T.)
But I can understand it is not that simple and here a curent injection migth not work in that circuitry (so not good aat the end for the TDA1541A... or will it works with - 2 mA zero centered : flip the polarity of the loudspeakers ??? (I claim to salad word myself but at least I ask ! Not intented to you T.)
Last edited:
Whaou ,frightening words from the mad pagan Crowley 😳
" To enlarge this point. If designing in a singular chain and "by ear" there is a tendency to tune individual elements in a complementary fashion"
yes correct , this is why measurements are usefull , and why we should have more than one equipement , say amplifier , speakers , etc...
.
Last edited:
H2 cancellation in Simulators is always perfect, in reality. Depends.
If you look at my take, which id strictly for TDA1541, correctly set up, it should keep the TDA1541 output at +/- 1mV with less than +/- 1mV thermal drift across any sensible range.
One of the big issues using BJT's in current conveyors is the emitter current dependent base current, I like to call that "current escape".
The best solution is to use very high beta transistors. Using a SS9014 or MPSA18Nas NPN and a BC560C as PNP for a folded Darlington compensates the thermal drift and gives a compounded beta of > 250,000 for AC.
That means a +/- 2mA current swing in the emitter will have only have current swing in the BC550C base of +/-8nA while LSB current of the TDA1541 is around 61nA.
Adding another BC550C to the SS9014/MPSA18 as inverted Darlington will further reduce base current modulation or as I like to call it "signal current escape" from our normally well defined current loops.
It also produces an input impedance around 0.3 Ohm, though not all that linear.
Switching to FET's doesn't really improved on the circuit I show for the extended audio range, unless we up the current a lot.
Now, if we do not have any "current escape" from our current conveyor the current entering the compound emitter must exit from the compound collector, minus any AC current escaping via the Base of the compound transistor.
The nonlinear input impedance doesn't really matter, as it is very low and keeps AC and DC to a small fraction of limits. So the current on the TDA1541 output (sources from a MOhm Impedance) remains unaffected.
If the collector current is substantially identical to the current at the TDA1541 Aol/Aor and we keep that current linear, we can add a resistor to convert the current to voltage that is limited not by the compliance of the TDA1541 output, but by the current conveyor output compliance.
For a number of reasons the design target for this DAC is to be adjustable from 0dBu to +20dBu. This means we have need of a very high compliance and a high power supply voltage.
Now +20dBu = 22V P-P needing a 5.5k I/U conversion resistor. We need extra overhead of at least 3 X the 2mA offset current, so at a minimum we need 8mA DC flowing in our circuit.
Hence the second cascode as the lower current conveyor is made with low voltage BJT's (which tend to be faster and higher beta). To avoid "current escape" in this circuit, a J-Fet is used at the input, as some thermal drift is inconsequential. This circuit is naturally made with high voltage transistors.
So we now in effect still have the exact current that entered the original compound transistor emitter from the TDA1541 exiting the second cascode compound transistors collector towards a +75V rail, via a resistor adjustable from 0.5k to 5.5k.
This resistor converts the TDA1541 current to voltage completely passive, there is no active current to voltage conversion.
All that is active are current conveyor cell's that have minimal or zero current escape and simply shift the reference level and increase output compliance for voltage swing.
The technology is actually not that different from current conveyors inside the TDA1541 and is indeed inspired by looking in detail inside the TDA1541.
As 5.5k is a very high output impedance, a buffer is needed.
The high voltage current conveyor cell also makes a nice follower, so it is used stacked on a CCS, as it performs better than available MOSFETs, in the sim at least.
The 5.5k current to voltage conversion resistor also form part of the 2nd order sallen key lowpass (first resistor) which is wrapped around the buffer.
This filter is switchable between a critically damped 75kHz lowpass (for sample rates > 48kHz) or a peaking (+3.7dB @ 22.05kHz nominal) lowpass for 44.1kHz sample rate which compensates the SINC rolloff.
The same filter remains for 48kHz (pretty rare anyway) though it will result in a very mild treble boost.
If absolute phase worries you and you think doing it in the digital domain is a "bad idea", yup.
Thor