Does anyone have any suggestions on a good value for the miller feedback capacitor in my VAS stage?
The VAS stage bias class A is as such: Iq=730uA with 24V-RMS
Hfe=250min Ft= 300MHz (ECG 26 type) biased with a current source.
I am using 8pF (from collector to base) of the VAS transistor right now. The frequency responce begins to suffer from capacitor effects at around 12-KHz. This triangle shaping of the sine wave must be caused by this capacitor since this is the only capacitor in my amp circuit (direct coupled to speaker, transformer coupled to input) I have a 2pF capacitor. Would this be to small? I am not sure how to correctly calculate the value. The circuit does occilate around 120KHz when there is no miller feedback.
Any calculation methods would be of great help!
--or is this just a trial and error problem--
If I can work this bug out and some other minor details, I will have created 20VRMS across 4Ohms with only using two output transistors!
BTW, even with transformer coupled input, I seem to have significant responce at 7Hz!
Not that it really matters, full range is the goal.
The VAS stage bias class A is as such: Iq=730uA with 24V-RMS
Hfe=250min Ft= 300MHz (ECG 26 type) biased with a current source.
I am using 8pF (from collector to base) of the VAS transistor right now. The frequency responce begins to suffer from capacitor effects at around 12-KHz. This triangle shaping of the sine wave must be caused by this capacitor since this is the only capacitor in my amp circuit (direct coupled to speaker, transformer coupled to input) I have a 2pF capacitor. Would this be to small? I am not sure how to correctly calculate the value. The circuit does occilate around 120KHz when there is no miller feedback.
Any calculation methods would be of great help!
--or is this just a trial and error problem--
If I can work this bug out and some other minor details, I will have created 20VRMS across 4Ohms with only using two output transistors!
BTW, even with transformer coupled input, I seem to have significant responce at 7Hz!
Not that it really matters, full range is the goal.
The VAS should be run at a suitable current to drive the bases/gates of the following stage.
If drivers, assuming beta 100 and likewise outputs, then around 8mA is fine for two output pairs and 100W. For a single output pair, say around 60W, an Iq of around 6mA is fine.
For these currents, assuming a single ended voltage amplifier and conventional single diff pair input stage, around 68pF would be a good start, working down for best sonics and no oscillation.
Consider a smallish cap, around 22pF, from collector of VAS to base of feedback transistor. This makes the amplifier tolerant of capacitive loads; an inevitable issue with most speakers.
Your VAS might be too fast. I'd not go over 150MHz. Too much speed may actually encourage instability.
Ensure good HF bypassing across the Vbe multiplier, and use base stoppers on both drivers and outputs.
Cheers,
Hugh
If drivers, assuming beta 100 and likewise outputs, then around 8mA is fine for two output pairs and 100W. For a single output pair, say around 60W, an Iq of around 6mA is fine.
For these currents, assuming a single ended voltage amplifier and conventional single diff pair input stage, around 68pF would be a good start, working down for best sonics and no oscillation.
Consider a smallish cap, around 22pF, from collector of VAS to base of feedback transistor. This makes the amplifier tolerant of capacitive loads; an inevitable issue with most speakers.
Your VAS might be too fast. I'd not go over 150MHz. Too much speed may actually encourage instability.
Ensure good HF bypassing across the Vbe multiplier, and use base stoppers on both drivers and outputs.
Cheers,
Hugh
The reason for using such low current in my VAS is because I really like the ECG 26 type device. It is a very clean amplifier, but operates at low current(50mA-max). It is a SP-92 syle device. Pd=0.2W max, breakdown voltage of 120V though Lo-noise. So you have to have an intermediate emitter follower stage to boost this small current. Instead of having 2 transistors in VAS stage(current source) to output enough current to drive the drivers, ~6mA, I have 6 devices to do the same job, except the bias for all of these is around 2mA.
I have exactly 100V DC, Zener regulated from +V to -V (+/-50V) that supplies all circuit componants except collector of output transistors.
My noise ceiling in terms of voltage, white noise at around 40uV.
Thanks for the idea of using a feedback cap from collector of VAS to base of feedback transistor in the diferential. I will try it and see what happens.
As for as DC stability, there is no drift from 0V greater than 10mV one way or the other.
I have exactly 100V DC, Zener regulated from +V to -V (+/-50V) that supplies all circuit componants except collector of output transistors.
My noise ceiling in terms of voltage, white noise at around 40uV.
Thanks for the idea of using a feedback cap from collector of VAS to base of feedback transistor in the diferential. I will try it and see what happens.
As for as DC stability, there is no drift from 0V greater than 10mV one way or the other.
AKSA said:
Thanks for the idea AKSA, I didn't think of this one. I removed the miller feedback off of the VAS completely and connected the capacitor from the collector of VAS transistor to the base of the feedback transisor in the diff. and the problem seems to be solved. Since the impeadence of the diff. and VAS stages are high, I was able to use a 1pF capacitor for this purpose. My noise ceiling also went down some, and the higher frequencies have less distortion.
1 pF, eh? Don't sneeze or you'll lose it. 
The acid test is to load the output with various capacitor values and see what happens. It is easy to make something marginally stable into just a zobel load (you've got one of these, right?) or a resistor. Another matter when you have a capacitive or resonant load connected through long aerials in the form of speaker wires. In my experience fine tuning with extremely small capacitances may be giving you a false sense of security. The very smallest capcitor I tend to use in an amp circuit is 33pF, I normally try to avoid going below 100pF. 1pF is getting down to parasitic levels: the order of stray pcb capacitances and inductances, an uncertain place to be fiddling.
A rule of thumb is to keep the speeds sensible and damp where you can. Oscillation is energy going out of control - just like when the shock absorbers fail on a car. Especially in a low feedback design you don't need much speed (hearing stops at 20kHz) so keeping the thing slow mitigates regenerative parasitic energy. Lots of shock absorbers (resistors) dissipate and discourage resonant energy. Inductors and capacitors don't damp - they store for later release - so they don't always solve the problem. The Miller cap method creates a local feedback loop around the VAS transistor that results in a reduction of transimpedance with frequency, which is really useful for slowing the overall voltage resonse. It is a very stable feedback loop because it only encompasses one transistor junction.
The acid test is to load the output with various capacitor values and see what happens. It is easy to make something marginally stable into just a zobel load (you've got one of these, right?) or a resistor. Another matter when you have a capacitive or resonant load connected through long aerials in the form of speaker wires. In my experience fine tuning with extremely small capacitances may be giving you a false sense of security. The very smallest capcitor I tend to use in an amp circuit is 33pF, I normally try to avoid going below 100pF. 1pF is getting down to parasitic levels: the order of stray pcb capacitances and inductances, an uncertain place to be fiddling.
A rule of thumb is to keep the speeds sensible and damp where you can. Oscillation is energy going out of control - just like when the shock absorbers fail on a car. Especially in a low feedback design you don't need much speed (hearing stops at 20kHz) so keeping the thing slow mitigates regenerative parasitic energy. Lots of shock absorbers (resistors) dissipate and discourage resonant energy. Inductors and capacitors don't damp - they store for later release - so they don't always solve the problem. The Miller cap method creates a local feedback loop around the VAS transistor that results in a reduction of transimpedance with frequency, which is really useful for slowing the overall voltage resonse. It is a very stable feedback loop because it only encompasses one transistor junction.
It is all about the impeadence. A higer impeadence circuit uses much less current than a lower impeadence. The 1pF capacitor discussed is a feedback capacitor. The impeadence of this cap is around 6MOhms at 20KHz. The diff. circuit uses no more than 800uA total and +&- input transistors are driven using only a few micro-amps for the base currents. Any more capacitive feedback and the freq. responce suffers. Transistors tend to operate faster at lower currents. Therefore the feedback through 1pF is enough current to affect the operation. If the impeadence of my diff. was lower, say a 10 milli-amps or so, 10pF -100pF would have the same effect. 10 times less current = 10 times larger impeadence, or 10 times smaller capacitor.
With an intermediate EF stage, 730uA Iq for VAS stage is plenty enough. It does work, and to hear any noise or hisssssssssssss, you have to bury you ear right against the tweeters of the speaker and it is barely audiable, and if the house fan is running, you can't even hear the hisssssssss. I have heard more hisssss out of a factory designed and built circuit having half the power of this one.
BTW, it really helps to have a metal enclosure.
With an intermediate EF stage, 730uA Iq for VAS stage is plenty enough. It does work, and to hear any noise or hisssssssssssss, you have to bury you ear right against the tweeters of the speaker and it is barely audiable, and if the house fan is running, you can't even hear the hisssssssss. I have heard more hisssss out of a factory designed and built circuit having half the power of this one.
BTW, it really helps to have a metal enclosure.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.