Building the ultimate NOS DAC using TDA1541A

[ecdesigns]

Hi phixphi,

Isn't it the case with the balanced operation too? I.e. the I/V R is half the value for a single chip?

For -same- output voltage using 1:8 step-up transformer:

Single, 68R I/V, 2.18Vpp, output impedance 4352 Ohms.
Paralleled, 2 x 68R I/V, 2.18Vpp, output impedance 2176 Ohms.
Balanced, 2 x 34 Ohms, 2.18Vpp, output impedance 4352 Ohms.

For maximum output voltage, single and paralleled stay the same, for balanced we get 2 x 68 Ohms, 4.35Vpp, output impedance 8704 Ohms.

I personally would suggest to go for lowest output impedance when using a step-up transformer and try to eliminate all active circuits.

The popular common cathode tube amplifier introduces very high distortion for a DAC that has to maintain distortion below 0.0015%.

Typical common cathode triode circuit produces approx. 1% ... 8% distortion, that's roughly 700 ... 5000 times too high!
Depending on tube characteristics (plate resistance) and gain, the output impedance will be higher than the output impedance at the transformer and you need a very, very good coupling cap or you get even more distortion.

In an old DI8 project I used a matched ECC83 as differential amplifier (connected to a balanced TDA1541A configuration of 4 paralleled + 4 paralleled TDA1541A chips). Differential amplifiers offer lower distortion. Then I used paralleled ECC82 in common anode configuration to get reasonably low output impedance.

So it makes much more sense to maximise the output signal at the transformer while maintaining low output impedance.

The limit for I/V resistor (with 2mA bias!) equals approx. 120 Ohms.

With two TDA1541A chips in parallel (RIV=60 Ohms) you get 60 Ohms output impedance and 480mVpp output.

1:2 step-up, 960mVpp, 240 Ohms output impedance.

1:3 step-up, 1.44Vpp, 540 Ohms output impedance.

1:4 step-up, 1.92Vpp, 960 Ohms output impedance.


With four TDA1541A chips in parallel (RIV=30 Ohms) you get 30 Ohms output impedance and 480mVpp output.

1:2 step-up, 960mVpp, 120 Ohms output impedance.

1:3 step-up, 1.44Vpp, 270 Ohms output impedance.

1:4 step-up, 1.92Vpp, 480 Ohms output impedance.

1:5 step-up, 2.4Vpp, 750 Ohms output impedance.

1:6 step-up, 2.88Vpp, 1080 Ohms output impedance.


With eight TDA1541A chips in parallel (RIV=15 Ohms) you get 15 Ohms output impedance and 480mVpp output.

1:2 step-up, 960mVpp, 60 Ohms output impedance.

1:3 step-up, 1.44Vpp, 135 Ohms output impedance.

1:4 step-up, 1.92Vpp, 240 Ohms output impedance.

1:5 step-up, 2.4Vpp, 375 Ohms output impedance.

1:6 step-up, 2.88Vpp, 540 Ohms output impedance.

1:7 step-up, 3.36Vpp, 735 Ohms output impedance.

1:8 step-up, 3.84Vpp, 960 Ohms output impedance.


With balanced operation you have equal negative DC voltage on both outputs, so 0V DC across the step-up transformer primary. However, there will be considerable negative DC voltage on each DAC output, limiting signal swing and significantly increasing distortion. So it is essential to use the bias current injection, even when using a balanced configuration.

This is a very cool idea. Not sure if I'm getting it correctly, but what will then close the loop for the 2mA DC? I attach a drawing of how I understood it. As far as I could get how TAD1541 works, the output current loop is between the output pins and the +5V?

The bias current runs from the positive supply voltage (+5V ... +200V) through the bias resistor into the TDA1541A output where it then flows to the -15V supply.

-When connecting the ground of the tube supply to the ground of the DAC power supply, the bias current can flow. But when the anode supply is isolated from the DAC (step-up transformer isolates both circuits) this won't work. Then you need a separate supply for bias that is galvanically coupled to the DAC circuit.

As for the CCS, the simplest in nature is a choke Have you thought in that direction? Instead of a resistor to use a choke, with its DCR tailored to balance the 2mA and the reactance blocking any HF residuals?

It is essential to maintain -equal- impedance over a very large bandwidth and that can be achieved with a suitable resistor that has low inductance and low stray capacitance.

Sounds more complicated than the previous solution, esp. with a cap and a new corner added. Is there any sonic advantage of this solution wrt the previous?

This setup is very suitable when only low supply voltage (+5V) is available. The capacitor filters out power supply noise. It can be removed but then you have more noise. Think of this cap as a decoupling cap. This solution offers better results compared to a relatively low value bias resistor of 2K5 for +2mA, 1K25 for 4mA and so on.

 

[phixphi]

If the two grounds are not connected, you are right. No sinking anywhere.
You can use the +5V (and a 1,25 Kohm resistor ) or any other positive supply with ground common to the chip (and of course a properly scaled resistor)

Exactly my plan now - to use +5V An interesting perspective was put bu John - what about using a choke with a suitable DCR? It will imitate CCS


Edit: Just noted John's reply, we need a constant impedance so choke is a no go.

 

[phixphi]

For -same- output voltage using 1:8 step-up transformer:

Single, 68R I/V, 2.18Vpp, output impedance 4352 Ohms.
Paralleled, 2 x 68R I/V, 2.18Vpp, output impedance 2176 Ohms.
Balanced, 2 x 34 Ohms, 2.18Vpp, output impedance 4352 Ohms.

John, sorry for being a bit slow, but isn't the output current for the balanced the same as for the parallel, i.e. 2x4mApp (One DAC pushing, other pulling)? Then "2x34R" applies to two channels, so 34R per channel (1 chip = 1 channel +/-) = 2176R ?

With balanced operation you have equal negative DC voltage on both outputs, so 0V DC across the step-up transformer primary. However, there will be considerable negative DC voltage on each DAC output, limiting signal swing and significantly increasing distortion. So it is essential to use the bias current injection, even when using a balanced configuration.

Very interesting info, had no idea about it, thanks! Have you sonically compared balanced vs parallel given all other possible factors equal?

The bias current runs from the positive supply voltage (+5V ... +200V) through the bias resistor into the TDA1541A output where it then flows to the -15V supply.

-When connecting the ground of the tube supply to the ground of the DAC power supply, the bias current can flow. But when the anode supply is isolated from the DAC (step-up transformer isolates both circuits) this won't work. Then you need a separate supply for bias that is galvanically coupled to the DAC circuit.

Ok, just as I suspected.

It is essential to maintain -equal- impedance over a very large bandwidth and that can be achieved with a suitable resistor that has low inductance and low stray capacitance.

So the impedance not the current. This kills the choke idea

This setup is very suitable when only low supply voltage (+5V) is available. The capacitor filters out power supply noise. It can be removed but then you have more noise. Think of this cap as a decoupling cap. This solution offers better results compared to a relatively low value bias resistor of 2K5 for +2mA, 1K25 for 4mA and so on.

My both 1541 have separate power supplies, so I can use higher resistance of 2.5kOhm per chip.


Thanks again

 

[john jones]

Old Mk7 design

A question for John at EC Designs and sorry to interrupt this thread, but I want to try out the Pass D1 style I/V circuit used in the MK7 design in post 3826 way back in 2011.

Here R24, feeding the source of the 2SK216 MOSFET is 2M2. Surely that value is incorrect? With a -15V supply this will alow no curerent flow at all through the MOSFET. I have seen this design used in the Red Barron DAC also...with exactly the same values as I think it was used with John's permission directly from this thread. But the value seems way too high and nothing like the values used in the Pass D1 design. I would have thought that using 4mA bias current, and ground reference around 1v, then 3.75k for R24 was a more suitable value. Can you comment?

 

[koldby]

A question for John at EC Designs and sorry to interrupt this thread, but I want to try out the Pass D1 style I/V circuit used in the MK7 design in post 3826 way back in 2011.

Here R24, feeding the source of the 2SK216 MOSFET is 2M2. Surely that value is incorrect? With a -15V supply this will alow no curerent flow at all through the MOSFET. I have seen this design used in the Red Barron DAC also...with exactly the same values as I think it was used with John's permission directly from this thread. But the value seems way too high and nothing like the values used in the Pass D1 design. I would have thought that using 4mA bias current, and ground reference around 1v, then 3.75k for R24 was a more suitable value. Can you comment?

The circuit and the resistor is correct. The return path for the source is through the TDA1541 . The 2M2 is only there to make sure, that the source potential is lower than the gate, when TDA1541 is starting up AFAIK

 

[koldby]

A question for John at EC Designs and sorry to interrupt this thread, but I want to try out the Pass D1 style I/V circuit used in the MK7 design in post 3826 way back in 2011.

Here R24, feeding the source of the 2SK216 MOSFET is 2M2. Surely that value is incorrect? With a -15V supply this will alow no curerent flow at all through the MOSFET. I have seen this design used in the Red Barron DAC also...with exactly the same values as I think it was used with John's permission directly from this thread. But the value seems way too high and nothing like the values used in the Pass D1 design. I would have thought that using 4mA bias current, and ground reference around 1v, then 3.75k for R24 was a more suitable value. Can you comment?

Remember that if there is nothing connected to the TDA1541 output it has -15V on the output. The 250 Kohm trimmer on the gate makes sure that the current through the FET is 2 mA so the output from the TDA1541 is 0V DC. It is a common gate circuit. Input is source output is drain. Gate is at AC ground.

 

[john jones]

Remember that if there is nothing connected to the TDA1541 output it has -15V on the output. The 250 Kohm trimmer on the gate makes sure that the current through the FET is 2 mA so the output from the TDA1541 is 0V DC. It is a common gate circuit. Input is source output is drain. Gate is at AC ground.

OK, then I understand. But this limits the bias current through the mosfet to 2mA in the MK7 circuit as shown...which is pretty low for such a beefy device. The Pass D1 uses IRF610 and 10mA (and many say that bias current is too low). If you want to increase the bias current in the mosfet you need to reduce R24 to a much lower value...to allow, say, a 4mA bias current to flow and thus current through the IV resistor varies between 4mA and 8 mA

 

[koldby]

OK, then I understand. But this limits the bias current through the mosfet to 2mA in the MK7 circuit as shown...which is pretty low for such a beefy device. The Pass D1 uses IRF610 and 10mA (and many say that bias current is too low). If you want to increase the bias current in the mosfet you need to reduce R24 to a much lower value...to allow, say, a 4mA bias current to flow and thus current through the IV resistor varies between 4mA and 8 mA

The 2SK216 and the IRF610 are two very different MOSFets. 2SK216 is actually pretty linear at low currents, whereas IRF610 ar more non linear in this region.
It is the difference between s-type and d-type mosfets.

 

[john jones]

The 2SK216 and the IRF610 are two very different MOSFets. 2SK216 is actually pretty linear at low currents, whereas IRF610 ar more non linear in this region.
It is the difference between s-type and d-type mosfets.

Yes, I need to hunt for a suitable MOSFET, ideally with high transonductance to lower input impedance (2sk216 are virtually impossible to find.. and transconductance is pretty low)

But if I want 2.5v ground reference then I need 5mA flowing through the 500R i/v resistor at digital silence. This will give 3mA through R24 (because 2mA of the 5mA is biasing the TDA1541A) and hence R24 should be 5K. Then I get 2.5V ground reference voltage. 2.5V across the I/V resistor at digital silence, and voltage across I/V resistor varying from 1.5-3.5V.

 

[koldby]

Yes, I need to hunt for a suitable MOSFET, ideally with high transonductance to lower input impedance (2sk216 are virtually impossible to find.. and transconductance is pretty low)

But if I want 2.5v ground reference then I need 5mA flowing through the 500R i/v resistor at digital silence. This will give 3mA through R24 (because 2mA of the 5mA is biasing the TDA1541A) and hence R24 should be 5K. Then I get 2.5V ground reference voltage. 2.5V across the I/V resistor at digital silence, and voltage across I/V resistor varying from 1.5-3.5V.

I would say no you don´t want high transconductance, you need linearity.

 

[john jones]

I would say no you don´t want high transconductance, you need linearity.

Agree that linearity is important, but so is transconductance as this directly affects imput impedance, and ideally we need TDA1541 to feed a virtual ground.

I have a few Toshiba 2SK2013 lying around and they were designed for audio use and have much higher transconductance than the Renesas lateral mosfets. I'll give them a go and see how it sounds.

 

[diyiggy]

thanks for that, looking forward to...

 

[978HK]

I'm looking forward to sanity.
Voltage compliance 1541A is like my old hat: done to death already.
HK

 

[rfbrw]

I'm looking forward to sanity.

And you came here ?

 

[diyiggy]

I'm looking forward to sanity.
Voltage compliance 1541A is like my old hat: done to death already.
HK

@ rfrwb,


We are mutants here, we transcript music from readings and the tempo comes from the datasheet reading. Been a long time we are deaf... evolved with a big *** for sofa, no legs, big iron solder hands...

In the kingdom of the deafs, the one-eared is king

 

[diyiggy]

as anyone tried the old 2SK30A (genuine I mean) for the voltage compliance of the TDA1541, better than moderner BFA245 for instance ?


@HK (tubes is too much expensive and risky for my skills)


This thread is so long than I don't remember what John used to select in all the 18 versions of the TDA i/V, nore guys lile JockoHomo... True it's an old chip and so many old designs see their BOM not made anymore !

 

[sumotan]

as anyone tried the old 2SK30A (genuine I mean) for the voltage compliance of the TDA1541, better than moderner BFA245 for instance ?


@HK (tubes is too much expensive and risky for my skills)


This thread is so long than I don't remember what John used to select in all the 18 versions of the TDA i/V, nore guys lile JockoHomo... True it's an old chip and so many old designs see their BOM not made anymore !

Oldie but a real Goodie chip just like the resurgance of vinyl wil never die. lol
Isn't it strange we're going back to the beginning Se Amps, High Effiicency speakers, field coil etc just to attain great sound

 

[diyiggy]

Hey Kaneda proof

same same but diffferent... bf245A not the same than J113, sk170 not the same as newer LSK equivalent and so on... but I dunno, so... not sure the spec of the sk30a are less good than moderner... good equilibrium said kaneda that has study it to death...though didn t know for the todays chips'of course....

 

[Hierfi]

This thread is so long than I don't remember what John used to select in all the 18 versions of the TDA i/V, nore guys lile JockoHomo... True it's an old chip and so many old designs see their BOM not made anymore !

Has anyone ever made a list of the posts that show the 18 versions having evolved?

 

[koldby]

Has anyone ever made a list of the posts that show the 18 versions having evolved?

I don´t think so, but if I remember correctly (and I think I have been with this thread since it started) and read John correctly, all the active I/V solutions has been abandoned and transformator solutions preffered. That has also been my conclusion after trying all sorts of active I/V stages.