MPP

Hi Joachim,
I was quite intrigued with your starless circuit and decided to give it a go. So in real life, good matching of input cascade jfet's gives offset in uV range, so nothing to worry about. With current mirrors it was another case. Even with matched bjt's, 0.1% low ppm resistors and good thermal tracking, output wonders -+100mv, so yep, no chance of going DC coupled even with substantial efforts without involving servos. Did some quick fft to see if everything is as expected and thd (with 1Vpp output) was as close to sim'ed as it gets. Only 2d dominant, and that 3d is my ancient SB artifact.
Now the fun part. I suspected that operating my DL103 with it's high'ish impedance (45Ohm) to basically virtual ground, would be problematic, and I would need all the gain I could get, so I omitted that parallel resistor and opted with different riaa values (based on mirrors output impedance) that gave me couple extra dB. The sad part is that I'm stuck there (at some 0.7 Vpp output). Cant think of anything else to bump gain more or increase input impedance without hurting performance too much, maybe I'm missing something ?
Other than that, it performs excellent. I don't want to get carried with my subjective listening experience, but level of micro details is just staggering. It makes me wonder how DL103 can pull it off and why I didn't tried transimpedance input earlier :)
 

Attachments

  • fft.JPG
    fft.JPG
    127 KB · Views: 323
  • Proto.jpg
    Proto.jpg
    112.8 KB · Views: 334
Many build my Nobrainer that has a transimpedance circuit too. Also in this case the DL103 works very well. I even tried a 20 year old Audio Technica High Output MC and even that worked satisfactory. The only thing that can happen is that you get some treble loss when the inductance of the cartridge is too high. Could you please post your actual schematic and then i can help you.
 
It's dead quiet Joachim, only broadband noise (2mV/div in foto). Yes, you are right, in my case resulting gain is ~52dB. With 6Ohm cart. that would be way much more. I attached circuit as build and some sims, any tricks and tips would be appreciated. I really like it, I'm even considering upgrading my preamp for more gain if that's the last resort.
I see you started new thread for nobrainer. That looks elegant/interesting too, I may give it a try, just were did I put those low noise 2sc/2sa...
 

Attachments

  • Noise.jpg
    Noise.jpg
    68 KB · Views: 310
  • starless.pdf
    116.2 KB · Views: 153
Ok, you left out the parallel resistor and let the mirror do it´s work. Clever simplification.
Noise looks really low and does not have a particular spectrum. That is very good.
First you could add the helper transistors. That makes the matching of the mirror less critical and reduces distortion. Do you remember that twist or should i look it up for you ?
You could increase the emitter resistors in the mirror. When you double them output impedance doubles too. You can then raise the value of R8 by two and lessen the value of C1, C2 by half. That gives you 6dB more gain. The new emitter resistors drop some voltage. You could compensate that by raising the PSU voltage. When you take the PSU as is you loose some overload margin but it still will be fine. All other modification are more drastic. You could substitute the Fet output buffer with an Opamp with gain plus servo. That gives you DC coupling. I do that too because i found that ones the voltage is high enough the Opamp does no harm to the sound any more. That is the reason i prefer discrete inputs but have no objection to use Opamps in the following chain. I do that even in my 30.000,-€ Opus Magnum. It could be something like the second stage in the Nobrainer Servo but without the bass EQ components. The best Opamps i found are
OPA1611, LME49990 and OPA1641. An NE5534A will do for experimenting. I use OPA134 for servo most of the time. Another drastic measure would be to build another mirror with better bias stability. I have shown here the mirror with cascode that has excellent stablity. Again it was published here earlier. If you can not find it i can search it for you.
That may allow DC coupling. You could use my "strange" servo with the circuit you have.
Also published before. Anyway, i do not think that a 10nF cap does much harm or you would not experience the excellent result other then doing a soft subsonic filter. You could raise the value of the 10nF a bit. Maybe you like that better. I am very happy that it works so well.
 
Noise looks really low and does not have a particular spectrum. That is very good.
Yes, but I must add for future reference (if someone else would decide to ticker it) that output buffer input fet gate stopper is really mandatory. Some 100Ohm close to gate is fine, without it I get nasty oscillation in ~60Mhz range.
First you could add the helper transistors. That makes the matching of the mirror less critical and reduces distortion. Do you remember that twist or should i look it up for you ?
No, thank you. I know what you are talking about. In theory it should even reduce output offset swing and help gain. Good point.
You could increase the emitter resistors in the mirror. When you double them output impedance doubles too. You can then raise the value of R8 by two and lessen the value of C1, C2 by half. That gives you 6dB more gain.
True, I thought about that. The thing is it gives non standard cap sizes for riaa so paralleling is needed, also the poles would go off a bit, but I will check it in multisim. This could be enough, as I only need some 8db more overal gain.
All other modification are more drastic. You could substitute the Fet output buffer with an Opamp with gain plus servo. That gives you DC coupling. I do that too because i found that ones the voltage is high enough the Opamp does no harm to the sound any more.
I'am not afraid of op-amps for second stage either, but it has to be right implementation as lot of things can go wrong. I agree, if done right it can be as good as discrete. JC and others have proven that over time too.
Another drastic measure would be to build another mirror with better bias stability. I have shown here the mirror with cascode that has excellent stablity. Again it was published here earlier. If you can not find it i can search it for you.
Do you mean the one from naked truth bjt version ?
http://www.diyaudio.com/forums/attachments/analogue-source/241859d1317052225-mpp-naked-truth-mc-high-impedance-mirror-bjt.tsc-tina.pdf
I will try it as time allows.
That may allow DC coupling. You could use my "strange" servo with the circuit you have. Also published before. Anyway, i do not think that a 10nF cap does much harm or you would not experience the excellent result other then doing a soft subsonic filter. You could raise the value of the 10nF a bit. Maybe you like that better. I am very happy that it works so well.
No, I don't think that 10nf are any harm either. In parallel with 1M it acts like subsonic filter and that is as intended. Not all my records are flat and I really don't like watching speaker cones moving too much.
Thank you for your input, I will get back when I have some results.
 
Thanks for recommending the gate stopper. I do that too. I found that the helper transistors improve sound. To my ears it is more dynamic. Yes, the naked truth circuit has the mirror with cascode. I found that circuit is rather stable in offset because the cascodes force a constant voltage over the input transistors. When you use an opamp you could use it inverted. That would avoid common mode distortion. I assume that the output impedance at 1kHz of your circuit is 3,24kOhm. You need 8dB of additional gain. A ca. 10kOhm resistor over the opamp would do that. You can then use the positive pin for servo. i can draw you something. Another idea is to double up the 2SK170Bl at the input. That gives you double Gm and 6dB of gain. The J310 could run out of steam then but you can double them too. Thanks for building my circuit
 
diyAudio Chief Moderator
Joined 2002
Paid Member
Hi Joachim,
I was quite intrigued with your starless circuit and decided to give it a go. So in real life, good matching of input cascade jfet's gives offset in uV range, so nothing to worry about. With current mirrors it was another case. Even with matched bjt's, 0.1% low ppm resistors and good thermal tracking, output wonders -+100mv, so yep, no chance of going DC coupled even with substantial efforts without involving servos. Did some quick fft to see if everything is as expected and thd (with 1Vpp output) was as close to sim'ed as it gets. Only 2d dominant, and that 3d is my ancient SB artifact.
Now the fun part. I suspected that operating my DL103 with it's high'ish impedance (45Ohm) to basically virtual ground, would be problematic, and I would need all the gain I could get, so I omitted that parallel resistor and opted with different riaa values (based on mirrors output impedance) that gave me couple extra dB. The sad part is that I'm stuck there (at some 0.7 Vpp output). Cant think of anything else to bump gain more or increase input impedance without hurting performance too much, maybe I'm missing something ?
Other than that, it performs excellent. I don't want to get carried with my subjective listening experience, but level of micro details is just staggering. It makes me wonder how DL103 can pull it off and why I didn't tried transimpedance input earlier :)

What is the signal to noise floor distance at 20Hz?
 
Thank you for your efforts Joachim, I have some OP275 at hand, I will try them.

What is the signal to noise floor distance at 20Hz?
Hi Salas,
I deliberately hided low end in that fft :rolleyes: because it is flooded with 50Hz rubbish from ground loops. I really should get my hands on laptop with nice USB sound card and make some T-pads before posting any measurements. It was intended just for checking basic operation correctness. On scope it looks like no more then 4mV on 10ms/Div, so I assume ~49dB would be somewhere in the ball park. I should note that currently circuit is running on a pack of 8 4.3V Li-Ion batteries. Nice shunt PSU (as yours Salas) could be improvement.