Post 75,76,77 use of a servo
The weak point is the extreme tetchiness of output offset control. It depends hugely on the pinpoint accuracy of the current source which supplies Q1, the JFET. This would be best solved with a servo, I suspect.
Thanks to Michael for an excellent design.
Cheers,
Hugh
I would suggest using the concept of servo of the Kevin Gilmore.
The weak point is the extreme tetchiness of output offset control. It depends hugely on the pinpoint accuracy of the current source which supplies Q1, the JFET. This would be best solved with a servo, I suspect.
Thanks to Michael for an excellent design.
Cheers,
Hugh
I would suggest using the concept of servo of the Kevin Gilmore.
Attachments
GEirin said:
GEirin,
I don't see how that servo can work with voltage reference (LED) placed right at the base of the current source BJT. I think the voltage reference needs to be placed at the positive input of the opamp, and a cap in parallel with the base emitter of the current source BJT, where the LED is now to filter noise.
some ideas . . .
it may be oscillating, do you have a scope ?
You may need power supply decoupling caps from rails to ground close to the o/p devices. 2 x 0.1 - 1.0uF caps
Also that long beefed up track to the FB resistor may be adding inductance - not good on very fast amps.
Also you seem to be adding you PSU & signal earth down the same single track - is that right ? If so not a good idea - there are current pulses coming down PSU earth which will interfere with the signal. Should have different earth tracks coming from a single star earth connection point on the board - at least power ground and signal grounds should be routed separately. the decoupling caps (above) should be connected to the power earth
hope some of this helps
mike
it may be oscillating, do you have a scope ?
You may need power supply decoupling caps from rails to ground close to the o/p devices. 2 x 0.1 - 1.0uF caps
Also that long beefed up track to the FB resistor may be adding inductance - not good on very fast amps.
Also you seem to be adding you PSU & signal earth down the same single track - is that right ? If so not a good idea - there are current pulses coming down PSU earth which will interfere with the signal. Should have different earth tracks coming from a single star earth connection point on the board - at least power ground and signal grounds should be routed separately. the decoupling caps (above) should be connected to the power earth
hope some of this helps
mike
Last edited:
Also power tracks and signal tracks should not run parallel to each other - this can cause inductive coupling which will interfere with the signal.
On this thread we never once discussed PSU design or PCB layout and these - along with component choice are critical for a normal functioning & good sounding amp.
I hope all these details are not too daunting, I know that an amp that doesn't work can be so frustrating. Good Luck !
mike
On this thread we never once discussed PSU design or PCB layout and these - along with component choice are critical for a normal functioning & good sounding amp.
I hope all these details are not too daunting, I know that an amp that doesn't work can be so frustrating. Good Luck !
mike
Hugh is away this week so I wouldn't expect a reply from him until the weekend.
I was wondering about oscillation too but was nervous to suggest it. If the the circuit is wired up correctly and the feedback cap is not shorted there must be some serious AC finding its way through R7 to make it smoke. The only way this could happen is oscillation.
Despite the fact that single ended inputs are meant to be very stable, I have found that this circuit can be made to oscillate very easily in simulation. A farily large miller cap is needed to slow it all down. Also I would be dubious about using any fets without a gate stopper.
Maybe try increasing the miller cap and adding a gate stopper resistor to the VAS fet??
I was wondering about oscillation too but was nervous to suggest it. If the the circuit is wired up correctly and the feedback cap is not shorted there must be some serious AC finding its way through R7 to make it smoke. The only way this could happen is oscillation.
Despite the fact that single ended inputs are meant to be very stable, I have found that this circuit can be made to oscillate very easily in simulation. A farily large miller cap is needed to slow it all down. Also I would be dubious about using any fets without a gate stopper.
Maybe try increasing the miller cap and adding a gate stopper resistor to the VAS fet??
Last edited:
It lives!
Hi All,
I got impatient waiting for my 2SK170 jfets to arrive and lashed up a version of this amplifier using a ZVN3310A mosfet as the input device. Also, after Danspy's potential stability problems I added more degeneration to the input stage and added some to the VAS to reduce the OLG. The final circuit is shown in the figure below. One good thing about the reduced OLG was that my simulations indicated that I could remove the miller cap entirely.
Anyway, I lashed it up using one of my infamous "air sculptures" (figure 2).
I slowly upped the voltage with a variac while adjusting the DC offset and bias. The dc offset is quite unstable, but it was relatively easy to keep the variations within 50mV either side of 0v. They tended to jump up and down by a maximum of 10mV, as an estimate.
I ended up leaving the bias at 0.5A.
Please excuse the blurry photos. Also, for some reason my scope has decided to not use the entire screen and I can't seem to fix it.
Anyway, I removed the variac and plugged it in with the scope, sig gen and dummy load attached...hmm some oscillation (fig 3).
Hmm, what could it be? Well it was pretty easy to figure out. The P channel fet was getting much warmer than the N channel one, and when I touched the gate pin the wave tidied right up (fig 4). I just added an alligator clip and lead to the pin to load it up a bit and the problem was solved. Clearly some gate oscillation going on that a better layout would probably eliminate.
BTW, I forgot to get a scope shot, but the 20kHz square wave performance is near perfect. The best I have seen in an amplifier I have built. Stunning rise and fall times and just a hint of ring which I think would disappear with a better layout.
Anyway I finally plugged in a speaker using 10000uF of series capacitance for safety.
So, how does it sound? Well I need to get another channel built and get in some serious listening times, but it seems to be quite spectacular really. This is the most "tubey" sounding amp I have built. It has loads of warmth and surprisingly good bass. I need to listen a bit more, but my mind is made up, I am definitely going to pursue this topology further. It's late here and I have a lot more I could say, but I think this will do for now.
It sounds great. Build it if you can and help me decide on the optimum circuit parameters!
I will be making PCBs when the design is finalised.
P.S. Danspy, I will look into your problems further tomorrow. I need to go to bed now!
Hi All,
I got impatient waiting for my 2SK170 jfets to arrive and lashed up a version of this amplifier using a ZVN3310A mosfet as the input device. Also, after Danspy's potential stability problems I added more degeneration to the input stage and added some to the VAS to reduce the OLG. The final circuit is shown in the figure below. One good thing about the reduced OLG was that my simulations indicated that I could remove the miller cap entirely.
Anyway, I lashed it up using one of my infamous "air sculptures" (figure 2).
I slowly upped the voltage with a variac while adjusting the DC offset and bias. The dc offset is quite unstable, but it was relatively easy to keep the variations within 50mV either side of 0v. They tended to jump up and down by a maximum of 10mV, as an estimate.
I ended up leaving the bias at 0.5A.
Please excuse the blurry photos. Also, for some reason my scope has decided to not use the entire screen and I can't seem to fix it.
Anyway, I removed the variac and plugged it in with the scope, sig gen and dummy load attached...hmm some oscillation (fig 3).
Hmm, what could it be? Well it was pretty easy to figure out. The P channel fet was getting much warmer than the N channel one, and when I touched the gate pin the wave tidied right up (fig 4). I just added an alligator clip and lead to the pin to load it up a bit and the problem was solved. Clearly some gate oscillation going on that a better layout would probably eliminate.
BTW, I forgot to get a scope shot, but the 20kHz square wave performance is near perfect. The best I have seen in an amplifier I have built. Stunning rise and fall times and just a hint of ring which I think would disappear with a better layout.
Anyway I finally plugged in a speaker using 10000uF of series capacitance for safety.
So, how does it sound? Well I need to get another channel built and get in some serious listening times, but it seems to be quite spectacular really. This is the most "tubey" sounding amp I have built. It has loads of warmth and surprisingly good bass. I need to listen a bit more, but my mind is made up, I am definitely going to pursue this topology further. It's late here and I have a lot more I could say, but I think this will do for now.
It sounds great. Build it if you can and help me decide on the optimum circuit parameters!
I will be making PCBs when the design is finalised.
P.S. Danspy, I will look into your problems further tomorrow. I need to go to bed now!
Attachments
Last edited:
Great to hear your progess swordfishy - it sounds very promising.
I did a lot of "spicing" over the last week to get my head into this cct and see how it might work best for me.
I attach where I got to so far - this arrangement - at least in spice works into any capacative load up to 1uF without the use of o/p choke or miller cap or o/p zobel cct - but as you can see has a feedback based compensation cct. JLH & john curl said / say that amps sound better without without o/p chokes and JLH also preferred not use miller caps
Someone here was asking what a hawksford cascode looks like and it can be see here above my input jfet - maintaining constant power across it
You can see that I went a bit over board with the VAS degeneration but the figures held up and I think I might enjoy the sound with less open loop gain - and of course this helps with stablility.
I still have some more research to do to explore a few more options and a I am traveling at present so actually building is not an option for me at present.
I did a lot of "spicing" over the last week to get my head into this cct and see how it might work best for me.
I attach where I got to so far - this arrangement - at least in spice works into any capacative load up to 1uF without the use of o/p choke or miller cap or o/p zobel cct - but as you can see has a feedback based compensation cct. JLH & john curl said / say that amps sound better without without o/p chokes and JLH also preferred not use miller caps
Someone here was asking what a hawksford cascode looks like and it can be see here above my input jfet - maintaining constant power across it
You can see that I went a bit over board with the VAS degeneration but the figures held up and I think I might enjoy the sound with less open loop gain - and of course this helps with stablility.
I still have some more research to do to explore a few more options and a I am traveling at present so actually building is not an option for me at present.
Attachments
Mikelm,
OK, I will try to get a scope shot tonight.
Sorry I should have provided more detail.
To get rid of the oscillation I had to short the gate resistor of the p channel fet with a clip lead which seemed to work at the time, though I didn't do it for any scientific reason - all I knew was that the fet was oscillating and I had to change the conditions at the gate. Touching the pin with my finger had the same effect.
I think shorting the p channel gate stopper with a clip lead probably had the effect of compensating for the higher capacitance of the p channel device and making the drive currents more symmetrical.
This will need to be ironed out and a suitable real world implementation devised. Maybe we need a capacitor at the gate of the n channel device to balance the gate capacitances (like Like Rod Elliot's design) or asymmetric values for the gate stopper resistors. Maybe just a good PCB layout would do too....???
I didn't try driving any capacitive loads so I can't comment on the stability in that respect - which is what you are probably interested in. All I used for a load was a 6 ohm wirewound resistor. I guess this is not overly helpful but I did it to ascertain whether the VAS is providing enough current to turn the fets on and off quickly - which it clearly is. The ZVN3310A is doing a fine job and will not need to be replaced with a much less desirable TO220 device.
It may turn out that a capacitive load would cause significant issues. I will try this tonight too. What size load do you suggest?
I have not built many amplifiers. Those which I have built are pure class A with active current sources and I have never been able to get such nice rise and fall times so it was a little victory for me. There was some ringing but I am confident that a better layout would fix things.
Please note that I got this amplifier to work as shown with no apparent oscillation with input frequencies from 1-20kHz. This was with the terrible layout you can see in the photo, no PS bypass caps on the protoboard, no miller capacitor, no feedback compensation and no output zobel network.
Your schematic looks good, though you are moving out of the realms of my knowledge and experience so I can't offer much feedback at this stage. I need to go and do some research and simulate and learn a bit more before I comment. Maybe Hugh can provide more insight when he returns.
It would be nice if we could do away with any form of feedback compensation or miller capacitor though, and I think the design I posted certainly comes close to achieving this.
Last edited:
Finally i Got it Working, and it Sound,s Good to.
Only the Power need to grow a little, +-25V with Fet Output that can not be mush more den 20Watt
Only the Power need to grow a little, +-25V with Fet Output that can not be mush more den 20Watt
An externally hosted image should be here but it was not working when we last tested it.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.