Super common gate valve I/V converter for TDA1543

[chanmix51]

Hello, this is my first post on DIYaudio in the few months I follow the very intersting topics down here. I want to share the successfull experiment we have made with a valve based «super common grid» I/V converter for a TDA1543.

The idea has slowly emerged from a work with Yves07 and Totof both members of the french forum audiyofan.

The following image illustrates a normal common grid circuit:


The I1 current source represents a TDA analog output. The input impedance must NOT exceed 10Ω to stay within the data sheet specifications of noise and distortion. Here, the input impedance is the cathode's current source which has a very high impedance and the anode load. The anode load is in serie with the internal resistance of the tube and is seen divided by µ+1 which is roughly equivalent to 1/Gm. If we choose a very high Gm tube we would have an input impedance between 30 and 100Ω ... 10 times to high.

The idea is to create a feedback to force the valve's cathod to stay there AND NOT MOVE by a µV !



In this schema, as soon as the cathode moves from a µV, the mosfet gate also does. It amplifies and inverts the change and send it to the valve's grid that also amplifies it. The anode load of 2.2kΩ is seen as 1/(Gmtube * GmMosfet).

In the schema above, the voltage at input varies up to 70µVpp with a 2mApp input which represents an input impedance of 0.04Ω from 20Hz to over than 100kHz. Even OpAmp can not achieve this because their open loop gain greatly decreases with the frenquency. As the DAC current still flows through anode load, the output voltage is 4,4Vpp which means: no preamp ! The output impedance is roughly equal to the anode load: 2.2kΩ (which is ok to me).

You can achieve a full tube circuit replacing the DN3545 with a pentode or a high gain / high transconductance triode, the russian 6F12P seems the best candidate for this purpose. Totof has a working version which he seems to prefer over his TDA1541+OpAmp.

A friend and I bought a cheap TDA1543 board with passive I/V conversion we hacked to use our converter, I will not step back



You can see here my noisy power supply as the converter is managing a 1kHz signal. All signal harmonics are below 90dB. The stereo scene is very large and clear, there are lot of details I have never heard before. I will change my power supply soon

Enjoy !

Chanmix

[chanmix51]

Here is a better schematic with the mosfet version and the full valve version:



Note Totof's version is equiped with a notch filter at the output to block the 44.1kHz noise created by the DAC. You can also use it with the mosfet version.

I have written a blog post about this circuit with some hints about the power suppy.

greg

[chanmix51]

Hello,

Here is the 1.1 version of the converter :


changelog :
– R3 changed from 2,2kΩ to 2,7kΩ
– R5 changed from 24kΩ to 27kΩ
– HT changed from 190V to 220V DC
– C3 changed from 100nF to 220nF
– Added C10 1µF MK

The result is beyond my best hopes, better bass response, better contrast between instruments, better presence. That's the version I am going to put in a box. The input impedance as seen by the DAC is 0,1Ω, output : 2,5Vp

I have updated my blog with this schematic and the power supply part.

greg

[oshifis]

I don't understand the role of C4, doesn't it bypass all AC signal to GND?

[chanmix51]

Hello oshifis,

C4 with the input impedance creates a low pass filter. As the input impedance is very low (rougly 0,1Ω) only signals at frequency over more than 1MHz are bypassed to GND. This is what we found to ovoid the whole circuit to oscillate at several MHz.

I tried with a 150nF at first and that wasn't enough. The circuit was resonnating at 7MHz with 30Vpp on the pentode's grid. This capacitor is soldered against the 1kΩ gate resistor of the mosfet.

Greg

[chanmix51]

Hello,

Using a 2,7kΩ resistor for I/V conversion led to too much gain and resulted into a very aggressive sound. Here is the version 1.2.



Changelog:
- R3 changed from 2,7kΩ to 2,5kΩ
- Measures on the previous version were wrong

greg

[bonalux]

Hi Greg, i'm very interested in your iteration of the I/V converter, but wanna try to simulate it with a jfet instead of the tube (obviously adjusting accordingly the supply voltages) and in LTSpice IV i haven't got a model of DN3545 nor i'm able to put it in the model library so, have you got a model with instruction for placing it in the library?
2nd question, i see that at input you got 3 volts, much higher than the recommended voltage in the datasheet for TDA1543, is this a problem? Anyway i need to try your circuit with an AD1865 that actually has a resistor as an I/V converter.

Thank you very much

[chanmix51]

Hello Bonalux,

No problem with sharing our work, you will find the DN3545 LTSpice models and one of the LTSpice schematic I have been working with here.

As far as I know, the analog output pins must be polarized between 2V and Vdd - 1,8V (3,8V with my board) so I found out any value between 2,5V and 3,5V to be nice.

I am interested in your experiments with JFETs, would you please mind letting us know about your work ?

greg

[chanmix51]

Hello,

After hours listening, I tried 2,5kΩ, 2,4kΩ, 2,3kΩ and I switched back the pentode's anode load to 2,2kΩ. It fixed a bass problem on lot of records and btw gave a more consistant stereo image.



Changelog:
- R3 changed from 2,5kΩ to 2,2kΩ
- HT changed from +220V to +210V

greg

[guglielmope]

Hello,
very interesting work.
Have you never tried with AD1865 DAC?
Can you make me any suggestion to try with this dac?
Ciao
Guglielmo
Italy