Some time ago I had an idea to build a discrete phono stage using bipolar power supplies, but did not really know where to start, so I decided to implement it as a discrete opamp.
I started with this image from Wikipedia showing the basic topology of a discrete opamp with complementary LTPs at the input:
This was the first version of my discrete opamp:
To my surprise, the simulation worked at the first try, but it was not particularly impressive. So I started thinking about improving its linearity. Douglas Self in Small Signal Audio Design says that the two things that help the most are:
1) unloading the VAS by implementing an output stage;
b) improving the linearity of the input stage.
My next step was to add an output stage. Dominant pole compensation was moved from the input differential pairs to VAS (screenshot from QUCS):
These changes, together with adding emitter resistors to the LTPs, reduced distortion by a factor of 15, but it was still not good enough. I ended up replacing tail resistors with constant current sources. Also had to tackle some small capacitors onto the LTPs to make the circuit unity gain stable.
This was almost the final result (shown here with the RIAA feedback network):
I showed it to a friend who is an electrical engineer and a passionate audiophile and he told me that it looks "like an improved version of Acurus P10"
I was really disappointed because I had been so proud of myself...
Anyhow, I designed a small one-channel PCB. Here it is shown built as a simple flat gain stage (no RIAA network):
I later configured it as a unity gain buffer by removing R14 and confirmed that it was, in fact, unity gain stable. I measured 0.0006% THD at 1kHz into 2K load, but my audio interface is an ancient EMU 0202USB which is only capable of about 0.0005% THD in loopback. Which means that the performance of this gain stage is probably somewhat better than that, but since I can't measure it, I am not going to make any claims about it. I did not match the transistors in LTPs by Vbe and hFE, instead I simply took 6 transistors in a row from the same ammo pack for both NPNs and PNPs. The worst DC offset at the input that I measured was about 15mV after an hour, which is definitely not fantastic, but also not a huge problem (at least not to the extent that it would interfere with the normal functioning of the circuit).
Two channels configured as a phono preamp, powered from a very audiophile PSU, and with suitably high-end wiring to boot...
I won't show all measurements because I can't get rid of a ground loop somewhere, but this is the frequency response (measured using R. Williamson's reverse RIAA network):
It works well and sounds great!
Later I thought: "Gee, this discrete opamp thingamajig seems to perform rather well, why not try using it as a regular opamp?" The result was this terrible little thing, with more robust output transistors and the optional ability to adjust quiescent current in the output stage...
To be continued...
I started with this image from Wikipedia showing the basic topology of a discrete opamp with complementary LTPs at the input:
This was the first version of my discrete opamp:
To my surprise, the simulation worked at the first try, but it was not particularly impressive. So I started thinking about improving its linearity. Douglas Self in Small Signal Audio Design says that the two things that help the most are:
1) unloading the VAS by implementing an output stage;
b) improving the linearity of the input stage.
My next step was to add an output stage. Dominant pole compensation was moved from the input differential pairs to VAS (screenshot from QUCS):
These changes, together with adding emitter resistors to the LTPs, reduced distortion by a factor of 15, but it was still not good enough. I ended up replacing tail resistors with constant current sources. Also had to tackle some small capacitors onto the LTPs to make the circuit unity gain stable.
This was almost the final result (shown here with the RIAA feedback network):
I showed it to a friend who is an electrical engineer and a passionate audiophile and he told me that it looks "like an improved version of Acurus P10"
I was really disappointed because I had been so proud of myself...Anyhow, I designed a small one-channel PCB. Here it is shown built as a simple flat gain stage (no RIAA network):
I later configured it as a unity gain buffer by removing R14 and confirmed that it was, in fact, unity gain stable. I measured 0.0006% THD at 1kHz into 2K load, but my audio interface is an ancient EMU 0202USB which is only capable of about 0.0005% THD in loopback. Which means that the performance of this gain stage is probably somewhat better than that, but since I can't measure it, I am not going to make any claims about it. I did not match the transistors in LTPs by Vbe and hFE, instead I simply took 6 transistors in a row from the same ammo pack for both NPNs and PNPs. The worst DC offset at the input that I measured was about 15mV after an hour, which is definitely not fantastic, but also not a huge problem (at least not to the extent that it would interfere with the normal functioning of the circuit).
Two channels configured as a phono preamp, powered from a very audiophile PSU, and with suitably high-end wiring to boot...
I won't show all measurements because I can't get rid of a ground loop somewhere, but this is the frequency response (measured using R. Williamson's reverse RIAA network):
It works well and sounds great!
Later I thought: "Gee, this discrete opamp thingamajig seems to perform rather well, why not try using it as a regular opamp?" The result was this terrible little thing, with more robust output transistors and the optional ability to adjust quiescent current in the output stage...
To be continued...
Last edited:
No need to be disappointed. There have been millions of man-years of R&D going on ever since Flemming invented the "Flemming Valve", so there's nothing truly unique. Been there; done that: lots of bright ideas for both hollow state and solid state design. It may take years, but I've always seen the same thing in the wild eventually.I showed it to a friend who is an electrical engineer and a passionate audiophile and he told me that it looks "like an improved version of Acurus P10"
Thank you Miles. I agree that it is difficult to be truly original, but in my case it was an attempt at self-deprecating humour, since I am just a newbie and haven't done any electronic design since the college days...