Building the ultimate NOS DAC using TDA1541A

Other is very simple MOSFET based IV. I took mosfet devices Pass used to work with, and Io od 9.15 mA cca. as in the Pass designs. The IRF610 and IRF9610. I am not familiar with MOSFETs... But I persume that other types and other working points can be successfully used just didnt try in the Spice... 🙁
.
Circuits has Inverted Phase ov Voltage at Output, so I checked deliberatley for very low value of RIV to have Tube amplifiaction stage after it. To brig the phase corrected back in, and to have sufficent Vp-p Output voltage.
.
Thes circuits are for 200mVp-p on RIV. But with very LOW THD, good harmonic arrangement and deep PSRR (note tht abstract PS used, probably be some losses with real one, but if it is quality one there will be no significant degradation...)
.
The results are very good. Very low THD, Very low PSRR...
But Zin is somehow higher about 4 Ohm. I am not the expert but I persume that using other types with smaller Rdson can give smaller Zin? Maybe i am wrong...
Zout=Riv
.
Circuits can be even more simple, with just Rs against CCS circuits, but with CCSs results are better. This is s classic IV circuit.
.
Again in negative branch we can have -Ug for direct coupling Tube for additional amplification
.
these mosfet circuits can be used for any type of Iout DAC with slight adjustment of some R values...
.
Since the Gate biasing voltge is higher, maybe additional benefits can be achieved with some low noise Voltage refference circuit on the gate? I dont know, but that will complicate the circuit for sure...
.
Simple MOSFET SCH.jpg


Simple MOSFET FFT.jpg
 
  • Like
Reactions: ZoltanChivay
I make some spice research to empoy IGBT device in classic transimpedance circuit. And it was really good. Better than with mosfet.
.
But i made a discrete version of IGBT from @Lazy Cat concept, with mosfet and BJTs
and got i think really good results. With available and reduced number of parts. 🙂
.
This simple IV has mybe the best chrs. And can be used for module for additional tube amplifiacation.
This is the 0.2V p-p output version with inverted phase and about 100 ohm of Zout.
So for 4Vp-p only 20X tube gain needed. It covers many tubes...
.
Transformer 1:1 is also OK because it is not complicated, not high Henrys need in primary because of low Zout of IV circuit.
.
THD is very low with good harmonic order.
This is the versin for 2mAp-p like AD1862, PCM56 etc. but very easy can be adopted to other Isources with current injection and more Iout. 🙂
.
IGBT_IV_FFT.jpg

IGBT_IV_SCH.jpg
 
  • Like
Reactions: nicoch58
Ok, looking around, I decided to make a few small changes to the design, here the single TDA1541 sensible "Core".

View attachment 1370578



Thor

I got regulators oscillating like crazy. about 20v swing few kHz. Only removing C28 fixes it but i'm still not sure if it can be stable. What can be done to address this issue here? LM317T from ST. I have 3300uF caps close to the inputs.
 
With the LM317, you don't need much capacitance on the input. A capacitor on the adjust pin will help stability. See figure 7 of the ST data sheet. C2 connected to the adjust pin is 10uF. I have seen various RC combinations used. 100R with 1uF. The proper output cap is very important for optimal transient response, if that is needed. 1uF tantalum. 2R7 and 1uF film.
 
According to the datasheet C28 must be in parallel with R32 then, but still when I connect a 10uF cap parallel to R32 it oscillates, not that badly though. I think it has to do with the inductive load. Even connecting an additional cap directly to the output doesn't solve the issue. Without a cap on Adj pin but with the cap on the output it oscillates as well.
 
I got regulators oscillating like crazy. about 20v swing few kHz.Only removing C28 fixes it

Interesting. Just remove it. It only increases the effective output impedance.

LM317 is normally ok with this AC CCS configuration. I used this style of circuit a lot.

The RC turnover is at around 6.6Hz so that's not it.

Thor
 
According to the datasheet C28 must be in parallel with R32

Not really.

but still when I connect a 10uF cap parallel to R32 it oscillates, not that badly though. I think it has to do with the inductive load. Even connecting an additional cap directly to the output doesn't solve the issue.

That doesn't sound like the good old LM317.

If an output cap and a cap on adj makes it stable, go with that.

Thor
 
I make some spice research to empoy IGBT device in classic transimpedance circuit. And it was really good. Better than with mosfet.
.
But i made a discrete version of IGBT from @Lazy Cat concept, with mosfet and BJTs
and got i think really good results. With available and reduced number of parts. 🙂

My experience is that noise snd thermal stability are suboptimal.

Going the opposite way, with a Baxandall superpair and MOSFET Sziklai creates a thermally stable circuit, with minimal and trimable offset to the reference voltage, which is thermally stable across a wide range of temperature.

One problem, most MOSFET have a high enough threshold voltage to loose enough voltage to limit levels and HD on 5V rails.

It is usable for a negative rail of > 5V.

Say MPSA18 NPN up front, 2N4403 second stage and to complete, 2N7000 or similar MOSFET.

Thor
 
Hi
Thanks I will check it.
.
I finished SMD version, jfet instead of mosfet, simplifying "IGBT" circuit for just needed elements,
minimising number of different parts and chose jfets and BJTs that are available in the stores.
IMHO the results on the spice tests are very good.
Only the input impedance is say, high. Around 6.25 Ohma with the specific JFET in "IGBT" module. But it is uniformly constant over large BW. Output inpedance of the module is RIV in this case 750 ohm which is more than enough, but with the buffer it dramatically lowering to 18 Ohms. These 18 Ohms of source allowing usage of non-complicated line transformers with just 4Hy Lprimary for very low LF range.
THD is very small even for complementary designs, this is Single ended design. Thad is typical little less than 0.003% (more than 50% less than all discrete versions before, that was about 0.008%) I add a buffer to safely drive optionally transformer for return the phase...
Noise floor is huge, around -205db
Note that Output voltage is 3Vp-p.
.
Anyway, the circuit can be used without the buffer for users with single stage tube preamps, because they also shifting the phase for 180deg. So the amp and speakers will see the righta phase...
.
SMD_IGBT_SCH.jpg

In IGBT module Currents troug BJTs should be adjused to be the same or very close. That could be done with trimming Req1 or Req2. Values of these R determine current for JFET too.
.
SMD_IGBT_FFT.jpg

Buffer add 2nd harmonic only from -134db to -110.8db (+23.2db)

SMD_IGBT_Phase.jpg


SMD_IGBT_Zout.jpg

Voltage Y scale is 1[V] = 1 [Ohm]

SMD_IGBT_Zin.jpg

Voltage Y scale is 1[V] = 1 [Ohm]
 
Last edited:
@Zoran ,in practice, two jfets with small Yfs cannot work as CCS, I haven't tried it with 2n5458 but I know that two 2sk208(2sk246) don't work, the lower one must have a larger Yfs(2sk209) and then it's OK. That's one of the reasons I prefer depletion mosfets for CCS.
I used it many times. But the "upper" one has to be from higher Idss tna "lower" one. I just put 2N5458 because I think that is available and that is in to-02 package to match device packages.
In that position more JFETS can be used because of low Io that contols...
Othet current source can be used but on the sims that with 2xJfet performing the best.
.
But feel free to suggest other type of CSS?
.
The key device is 2N4391 smd version and standard, for all of these IV circuits. With other CCS results are significantly worse.
.
Can You recommend some to92 mofets other than these on the list i checked? For lower Zin further
with BS170 is the lower from those from the list. 3.9 ohms that is high... 🙁
 
Last edited:
Yes I can post.
I just have to go back to windows form OSX, do some work and get back in OSX... Pain in the ***... Didnt install Parallels 🙁
,

It is basically 5 times (or more taps) in parallel RLC serial notch. With one R infront. Taps are nxFs
So it is the issue to solve to control the relays for changing the value of C for 44.1/48 KHz base.
But i dont see the reason to be maual, even most of the digital interfacess have indication of Fs tha can be used...
.