Not every modern DAC can live with the voltage variations at its output such as when using passive IV, which for many (me including) seems to be the best sounding solution. For those DACs, there are basically 2 solutions for IV conversion (useless you want to use output coupling caps as in the Pass D1) – opamp IV, or current conveying into a passive IV (resistor).
I came across Malcolm Hawksford’s discrete IV article when reading his interview with our own Jan. While I have to admit that I am not a fan of error correction, I like the scientific approach of Malcolm Hawksford to audio engineering. I shall leave you to read his article as to why he thinks discrete current conveyer sounds better than opamp IV’s, and the basic principles behind his published circuit.
http://www.essex.ac.uk/csee/researc...Current steering transimpedance amplifier.pdf
One thing that I don’t like about the published circuit is IMHO the unnecessarily large number of active devices. What I do like is the essentially zero input impedance (great for DACs like PCM1704). And you can also get current gain easily, so that you no longer have to parallel 10 DACs to get the impedance of the passive IV down.
As much as I like JFETs, there are no true complementary, low noise devices I can use for the input stage. The 2SK170/2SJ74 pair won’t do, as they have too different transconductance at the same bias. So I stuck to bipolar for those. But for the output device, the cascoded LU1014 in triode mode is the natural choice. One can even use another cascoded LU1014 at the top rail as a constant current source of the output stage.
So this circuit has an essentially zero input impedance, and a current gain of about 12.5. If the DAC will output +/- 1mA, a 220R IV resistor will get you 2Vrms. If it will do +/- 2mA, then you can even go to 110R. The input is at 0V as shown, but you can pretty much hang it at any potential by attaching Vref to the base of Q2. Q8 is just a simple servo to reduce any output DC drift.
It will burn 8W, so a small 3 K/W heat sink should be sufficient.
Patrick
I came across Malcolm Hawksford’s discrete IV article when reading his interview with our own Jan. While I have to admit that I am not a fan of error correction, I like the scientific approach of Malcolm Hawksford to audio engineering. I shall leave you to read his article as to why he thinks discrete current conveyer sounds better than opamp IV’s, and the basic principles behind his published circuit.
http://www.essex.ac.uk/csee/researc...Current steering transimpedance amplifier.pdf
One thing that I don’t like about the published circuit is IMHO the unnecessarily large number of active devices. What I do like is the essentially zero input impedance (great for DACs like PCM1704). And you can also get current gain easily, so that you no longer have to parallel 10 DACs to get the impedance of the passive IV down.
As much as I like JFETs, there are no true complementary, low noise devices I can use for the input stage. The 2SK170/2SJ74 pair won’t do, as they have too different transconductance at the same bias. So I stuck to bipolar for those. But for the output device, the cascoded LU1014 in triode mode is the natural choice. One can even use another cascoded LU1014 at the top rail as a constant current source of the output stage.
So this circuit has an essentially zero input impedance, and a current gain of about 12.5. If the DAC will output +/- 1mA, a 220R IV resistor will get you 2Vrms. If it will do +/- 2mA, then you can even go to 110R. The input is at 0V as shown, but you can pretty much hang it at any potential by attaching Vref to the base of Q2. Q8 is just a simple servo to reduce any output DC drift.
It will burn 8W, so a small 3 K/W heat sink should be sufficient.
Patrick
True, but a sim will let you optimise the circuit and find weak spots, which a listen won't.
jd
I would have to look at it more detailed, but I would start looking at the input Z and the linearity (change with signal current) of that input Z, as well as the flatness over freq.
I'm not sure the pos feedback loop is necessary, possibly biasing with a same-polarity transistor would be just as good, but that's just my gut feeling.
For some inspiration you may want to read this:
http://www.linearaudio.nl/Miscellaneous/smedegaard 4217 aes pp.PDF
jd
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.