I'll write a little about one project I've been working lately.
The idea was dropped after I had problems with overheating of NAD S300, which is a copy of Gryphon Tabu amplifier. The amplifier do not use the NFB through the individual levels and had to be put on higher current to work in linear mode. With a lot of current through signal transistors they are burned, so they themselves desolder of amplifier board and the amp is a frequent guest of service. There's a little more about that who is interested: diyaudio.com.hr • Pogledaj temu - NAD S300 (on croatian language).
In this type of structure where LTP at the input has a large current, the problem is that this arises noise at the output due to high current of base, so the input must see a small resistance of sources output. They (Gryphon/NAD) put signal from source directly to the potentiometer, since the amplifier has a substantial gain. They put on the input opamp that is connected to itself to canceling the input current.
Designed is a circuit where is added cascode to led dissipation to some normal level and there is no DC servos.
Below is a scheme of that version. The problem is still the need of input buffer to hold the base LTP to a low impedance, else there is a high noise. But even with the input buffer, there is some amount of noise at the output (I can hear noise from a distance of ten centimeters of speakers, they are 87dB sensitivity).
I made this version and it plays really nice (almost awesome) without exaggeration. There is something in this topology.
---
The idea for the new version is to put JFET follower directly on the PCB, on both sides of LTP, so avoid coupling capacitor and at the same time reduce the electrolytic cap on the dc feedback. In the first version from VAS was taken DC feedback, for compensation of the offsets. The signal is filtered directly to ground through the electrolytic cap. VAS is connected to ground, and the two resistors 2.2 mega from VAS is connected to JFET, so that the 10uF film capacitor be sufficient. Output follower is not connected to the DC feedback, but offset is stable, few mV +/-.
On the new version offset is still great and stable, and i moved predrivers from heatsink. Now it is OK with 10R on bias emitter.
But there is a strange problem. Open input creates a noise like BJT version, input closed reduce noise. There is no logic for JFET... I tried separate power supply for the JFET followers, from the source 1n ceramics to +15, 15V and GND... i receive the same noise. But with 100uF from source to gnd stop noise!
Here is a few pictures of the new version. I had to improvise heatsinks for pre-drivers
Several videos with oscilloscope (measured at the output):
Closed input:
https://www.youtube.com/watch?v=UJ4xouH4msU
Open input:
https://www.youtube.com/watch?v=fOr864-YxxI
On 1mV scale has no movement when I connect probe on source:
https://www.youtube.com/watch?v=Ouje0H6TeUc
Here's a test on a rectangular signal, oscilloscope is at 5V/cm and a frequency slightly higher than 20kHz. Load is 8R. No signs of instability.
In the simulator an amplifier is rather quickly, over 100k -0.1db, at -3 dB already approaching 1MHz. The PCB is designed to avoid the pitfalls and to possible fluctuations reduce to a minimum.
Here is layout versions with JFET buffer:
Does anyone have an idea why noise occurs at the output and how to remove it? Where could be the problem with this circuit?
Thanks in advance.
ps: Sorry for the bad english. I hope that is problem well described.
The idea was dropped after I had problems with overheating of NAD S300, which is a copy of Gryphon Tabu amplifier. The amplifier do not use the NFB through the individual levels and had to be put on higher current to work in linear mode. With a lot of current through signal transistors they are burned, so they themselves desolder of amplifier board and the amp is a frequent guest of service. There's a little more about that who is interested: diyaudio.com.hr • Pogledaj temu - NAD S300 (on croatian language).
In this type of structure where LTP at the input has a large current, the problem is that this arises noise at the output due to high current of base, so the input must see a small resistance of sources output. They (Gryphon/NAD) put signal from source directly to the potentiometer, since the amplifier has a substantial gain. They put on the input opamp that is connected to itself to canceling the input current.
Designed is a circuit where is added cascode to led dissipation to some normal level and there is no DC servos.
Below is a scheme of that version. The problem is still the need of input buffer to hold the base LTP to a low impedance, else there is a high noise. But even with the input buffer, there is some amount of noise at the output (I can hear noise from a distance of ten centimeters of speakers, they are 87dB sensitivity).
I made this version and it plays really nice (almost awesome) without exaggeration. There is something in this topology.
---
The idea for the new version is to put JFET follower directly on the PCB, on both sides of LTP, so avoid coupling capacitor and at the same time reduce the electrolytic cap on the dc feedback. In the first version from VAS was taken DC feedback, for compensation of the offsets. The signal is filtered directly to ground through the electrolytic cap. VAS is connected to ground, and the two resistors 2.2 mega from VAS is connected to JFET, so that the 10uF film capacitor be sufficient. Output follower is not connected to the DC feedback, but offset is stable, few mV +/-.
On the new version offset is still great and stable, and i moved predrivers from heatsink. Now it is OK with 10R on bias emitter.
But there is a strange problem. Open input creates a noise like BJT version, input closed reduce noise. There is no logic for JFET... I tried separate power supply for the JFET followers, from the source 1n ceramics to +15, 15V and GND... i receive the same noise. But with 100uF from source to gnd stop noise!
Here is a few pictures of the new version. I had to improvise heatsinks for pre-drivers
Several videos with oscilloscope (measured at the output):
Closed input:
https://www.youtube.com/watch?v=UJ4xouH4msU
Open input:
https://www.youtube.com/watch?v=fOr864-YxxI
On 1mV scale has no movement when I connect probe on source:
https://www.youtube.com/watch?v=Ouje0H6TeUc
Here's a test on a rectangular signal, oscilloscope is at 5V/cm and a frequency slightly higher than 20kHz. Load is 8R. No signs of instability.
In the simulator an amplifier is rather quickly, over 100k -0.1db, at -3 dB already approaching 1MHz. The PCB is designed to avoid the pitfalls and to possible fluctuations reduce to a minimum.
Here is layout versions with JFET buffer:
Does anyone have an idea why noise occurs at the output and how to remove it? Where could be the problem with this circuit?
Thanks in advance.
ps: Sorry for the bad english. I hope that is problem well described.
Last edited:
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.