I think it would be ideal if there was some kind of feedback that automatically adjusted the current source to compensate for the DC offset of the DAC. The reason is because there isn't a certainty that each PCM1794 DAC will have exactly -6.2mA of DC offset current. I think the offset can vary from one sample to the next. I have not investigated this myself however and it may not be that important after all.
Well, this is what I've been working on for a while now. The Lundahl transformer is a special unit with extremely low winding resistance. There is a LP RLC filter in each of the collector circuits. It's basically similar to Dr. Leach's except that instead of coupling capacitors I put in a follower stage. Gain is set by the two resistors R1 and R2, which also control the currents in the common base transistors. The 12V supply is floating.
Since the DAC DC current offset appears on both output pins, the transformer with a center tap cancels it out, as well as providing balanced to single ended conversion. The transformer also cancels some of the distortion from the DAC. In theory, the noise should be low because there is only one I/V converter and no other opamps or stages. I've been working on a circuit board as well. I'm just waiting on money.
The performance in simulation is good in my opinion, but not stellar. Noise is very low, with a S/N of over 130dB (@ 1kHz) in simulation.
Anyway, my 2 cents.
Since the DAC DC current offset appears on both output pins, the transformer with a center tap cancels it out, as well as providing balanced to single ended conversion. The transformer also cancels some of the distortion from the DAC. In theory, the noise should be low because there is only one I/V converter and no other opamps or stages. I've been working on a circuit board as well. I'm just waiting on money.
The performance in simulation is good in my opinion, but not stellar. Noise is very low, with a S/N of over 130dB (@ 1kHz) in simulation.
Anyway, my 2 cents.
Attachments
Last edited:
http://www.diyaudio.com/forums/analog-line-level/216557-xcen-balanced-single-ended-converter.html
The one above is one of the early versions.
Patrick
The one above is one of the early versions.
Patrick
It costs about $70 here in the USA. There is no gap as far as I know. It is an output transformer so it's much bigger than a typical input type, so it takes a lot more current imbalance to cause problems. It's good to +24dBu before saturation.
Data sheet is here:
http://www.lundahl.se/pdfs/datash/1517.pdf
If the primary windings are in parallel, DCR is about 4.5 ohms. Likewise, secondary windings in parallel produce very low DCR of about 4.5 ohms. If I could find another transformer with center tapped primary and lower DCR that would be super, but 4.5 ohms is pretty darn low. Lundahl is making noises about making silver wire available on all of their transformers, which will lower DCR a little bit more.
With 7.8mA p-p input, the voltage on the secondary is only about 5mVAC, so the transformer is really just moving current and not producing much voltage. It's not really acting as an I/V converter in other words.
today i was trying to find a solution for that, but unfortunately i have not find a topology elegant, the circuit in post71 dont work like the way i was thinking, its not possible to put ccs I1 in that position , such a stupid mistake. Using a current mirror works, but not that well. so i will leave this road for now.
tomorrow i would like to discuss the transistors we can use, if that's alright with you.
dirkwright
the distortion caused by a transformer increase as the ratio between the inductance and dcr in parallel with the impedance of the amplifier, if you have a transformer with zero ohms of dcr, the distortion will be also zero, silver will bring down dcr but will increase the costs. 140 dollars its a lot of money for that, but is your call.
Personaly i use auto transformers in the output of my amplifier , they serve to convert my low impedance speakers (4ohms) to (68, 34 or 16ohms), the highest the load, the less crossover distortion, and they give me peace of mind , as i am always doing mods, and some times i make mistakes, they are the ultimate speaker protection. they ware made by me and they are big, the dcr is very low and as the amplifier has very low output impedance the distortion they produce is also very low.
the distortion caused by a transformer increase as the ratio between the inductance and dcr in parallel with the impedance of the amplifier, if you have a transformer with zero ohms of dcr, the distortion will be also zero, silver will bring down dcr but will increase the costs. 140 dollars its a lot of money for that, but is your call.
Personaly i use auto transformers in the output of my amplifier , they serve to convert my low impedance speakers (4ohms) to (68, 34 or 16ohms), the highest the load, the less crossover distortion, and they give me peace of mind , as i am always doing mods, and some times i make mistakes, they are the ultimate speaker protection. they ware made by me and they are big, the dcr is very low and as the amplifier has very low output impedance the distortion they produce is also very low.
Joachim, are the base current in class A line level or phono circuits really a problem which affects sound negatively, or rather an academic one?
I've mulled over the use of a transformer in my circuit for a while now. I think it's justified for this particular DAC. All of my effort went into dealing with the weirdness of the PCM1794a DAC. If I had a voltage output DAC, or some other DAC with a lot less interface issues, then I would not use a transformer.
Zinsula, in my Paradise and Nobrainer phono stages i shunt the base current to ground with a big electrolytic. The distortion i get from optimized open loop class a circuits is around -100dB. With the idea presented here the distortion can get even lower. If that is audible i can not say. I can personally not hear better then 0.1% low order.
On the other hand if have experienced that improvements in a circuit can sound better, even when it is theoretically under the audible threshold. That may have to do with dynamically changing distortion like energy storage ( settling time ).
On the other hand if have experienced that improvements in a circuit can sound better, even when it is theoretically under the audible threshold. That may have to do with dynamically changing distortion like energy storage ( settling time ).
Joachim
I will use the bc327 (a base resistance of 30 ohms is low enough) in the prototype.
In circuit of post 49, ,i'm not yet convinced that the capacitor in the bases of the input transistors is necessary for lower noise, this bjt are already driven by a low impedance source, the r18 and r19 , the output impedance of the driver transistors (q5, Q2) is very low. what do you think ?
I will use the bc327 (a base resistance of 30 ohms is low enough) in the prototype.
In circuit of post 49, ,i'm not yet convinced that the capacitor in the bases of the input transistors is necessary for lower noise, this bjt are already driven by a low impedance source, the r18 and r19 , the output impedance of the driver transistors (q5, Q2) is very low. what do you think ?
Last edited:
For noise you have to watch through what components the noise path goes until it reaches ground. In your latest schematic it enters R1 and R2 then goes through the emitter-base junction of Q1, Q4 and then through the emitter-base junction of Q2, Q5. Then there is the bias generator for Q2, Q5. Say we use Leds for bias and shunt them with a big electrolytic to ground we can ignore the noise of the bias generator. We are still left with the emitter-base junctions of Q2, Q5. Here i am only talking about the positive input. In the case of the non bypassed R9 and ignoring the effect of the idle ( it is quite high at around 10mA ) we can do the following noise calculation : 4.7 Ohm plus Rbee`of Q1 plus Rbee´of Q2 ignoring the bias generator / 2 = 32.35 Ohm. Shunting the R9 resistor gives 4.7 Ohm plus 30 Ohm / 2 = 17.35 Ohm, an improvement of a bit less then 3dB plus more leeway to design the bias generator. Looking at that circuit balanced the noise will go up 3dB because the noise sources appear in series. Balanced is always noisier then single ended provided you use the same amount of input devices. This can of cause be solved by doubling the devices.
Why do you not like the mirror version ?
Why do you not like the mirror version ?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.