I've been wanting to make another Circlotron for a while; I built my first one in the early 80's.
I have 60 unmatched LU1014, rather than going through the tedious process of matching up a quad set, I might give this a go as I have an untouched matched quad set of THF51S from Watanabe sitting in a draw.
OK, the H3 probably won't be as low, but it can have plenty of grunt to drive my inefficient speakers - 86dB 3R.
Apologies for the off topic.
Schematic removed by moderation at poster's request.
I have 60 unmatched LU1014, rather than going through the tedious process of matching up a quad set, I might give this a go as I have an untouched matched quad set of THF51S from Watanabe sitting in a draw.
OK, the H3 probably won't be as low, but it can have plenty of grunt to drive my inefficient speakers - 86dB 3R.
Apologies for the off topic.
Schematic removed by moderation at poster's request.
Last edited by a moderator:
I don't believe I conducted this specific experiment, at least not with the optimized final circuit mounted on the larger heatsinks. Most of the measurements I performed were on a test rig with an undersized heatsink, but they were primarily done with the compensation circuit in place.
However, I did carry out an experiment that effectively demonstrates why it is important to use temperature compensation in this amplifier circuit and adjust it based on the specific circuit details/implementation. In these measurements, I monitored the bias current and heatsink temperature for amplifier circuits initially biased at around 1.6A at ambient temperature. I suspected that the required compensation depended on the value of the LU1014 drain resistor (Rd), so I recorded measurements for two different Rd values (0R17 and 0R66).
As an afterthought during the second measurement (0R17), I removed the temperature sensing transistor from the heatsink and allowed it to cool down to ambient temperature. By doing this, I effectively restored the initial bias and eliminated temperature compensation. Subsequently, I turned the circuit back on and continued recording the bias current and temperature. Without compensation bias current rises rapidly (approximately 40 mA/ºC).
Assuming that the coefficient of 40 mA/ºC does not vary significantly with Rd (although this may not be a safe assumption), the bias current of the constructed amplifier would increase by approximately 1.3A due to the 33 ºC temperature rise.
The exact duration of this increase and where it would stabilize would need to be determined through further experimental investigation on the actual amp. Maybe I can find the time to do the test one of these days.
However, I did carry out an experiment that effectively demonstrates why it is important to use temperature compensation in this amplifier circuit and adjust it based on the specific circuit details/implementation. In these measurements, I monitored the bias current and heatsink temperature for amplifier circuits initially biased at around 1.6A at ambient temperature. I suspected that the required compensation depended on the value of the LU1014 drain resistor (Rd), so I recorded measurements for two different Rd values (0R17 and 0R66).
As an afterthought during the second measurement (0R17), I removed the temperature sensing transistor from the heatsink and allowed it to cool down to ambient temperature. By doing this, I effectively restored the initial bias and eliminated temperature compensation. Subsequently, I turned the circuit back on and continued recording the bias current and temperature. Without compensation bias current rises rapidly (approximately 40 mA/ºC).
Assuming that the coefficient of 40 mA/ºC does not vary significantly with Rd (although this may not be a safe assumption), the bias current of the constructed amplifier would increase by approximately 1.3A due to the 33 ºC temperature rise.
The exact duration of this increase and where it would stabilize would need to be determined through further experimental investigation on the actual amp. Maybe I can find the time to do the test one of these days.
Attachments
@NicMac - Thank you for posting your results! 
I don't want to take the discussion too far off track. As is clear, I am a total novice.
Do you have any insight as to why at the same starting bias current of 1A6 the heatsink temps began at ~57C (for the experiment w/o the PNP on the heatsink)? I read that you allowed the heatsinks to cool to ambient between experiments, correct? Am I misinterpreting?
Edited for clarity.
I don't want to take the discussion too far off track. As is clear, I am a total novice.
Do you have any insight as to why at the same starting bias current of 1A6 the heatsink temps began at ~57C (for the experiment w/o the PNP on the heatsink)? I read that you allowed the heatsinks to cool to ambient between experiments, correct? Am I misinterpreting?
Edited for clarity.
great idea!
Is this amp a concept or something that will have some PCBs available?
What matching values are you looking for for best performance?
Did you evaluate the amp at lower rails lower/lower power (say ~+/-20 or 25VDC)?
thanks
I allowed the temperature sensing transistor to cool down. Not the heatsink - to lazy to wait for that to cool down properly.
I/we will likely make PCBs and properly matched actives available at some point, but not anytime soon. The basic circuit is shown in Patrick's paper and I hope impatient builders will try to make their own implementation - possibly better than mine
Mine are possibly matched for both Vgs and Yfs to 0.1%, at temperature and across the whole current range. If you don't care for best H2 cancellation you can properly get away with less well matched devices.
The amp sounds and measures just fine with lower rail voltages, but remember that the cascoded triode cell eat up a fair share.
The amp sounds and measures better at higher currents, but look out for the heat dissipation limits of the LU1014
I had good results in creating a low value variable resistance using a nichrome wire adjusted using a modified connector block as a wiper.
https://www.diyaudio.com/community/...lower-and-amplifier-build.349081/post-6078261
Because those values were the subject of all the experiments that Nic went through.
So you'll have to wait for the whole story.
As a starter, you can also take Nelson's triode cell in Zen V9.
Which was also our starting point.
Vbias will vary anyway with the JFETs you use and the bias you choose.
Patrick