I haven't heard from Andrew anymore and tried to decrypt and reverse engineer the component values myself.
Attached you find a schematic of the amp.
Note the components don't have reference designators because they are messed up completely. I think Anthony is a good amp designer but documentation is definitely not one of his strengths.
I hope mine is correct then.
Is there anyone here interested in discussing this amp?
My ideas to the changes in Version 2 and 3:
In Version 3 he changed the input DC blocking caps from 47uF to 22 uF and increased the input impedance from 4k7 to 15k.
He also decreased the low pass caps from 2n2 to 1nF.
I think the modifications don't affect the amp very much. Basically just the input impedance is increased. Right?
In Version 2 he changed the input stage emitter resistors from 400R to a value thats shape looks different and changed them in Version 3 to 150R.
Simulation showed that lowering the resistor values decreases distortion considerably. First I ran simulation with 100R I took from the AV800 amp and noticed increased distortion with higher values. Simulation is only of very limited value because I don't have models for all transistors and the ones I have might be off. Anyway, I think lower values are better here.
In version 2 he changed the transitor of the constant current source from BD681 to MJE340, increased the resistor at its base from 10k to 22k and the cap from 47 uF to 100uF.
I think changing the transitor has no big effect. The most important difference is collector-emitter breakdown voltage.
Increasing the cap seems to be a good idea while increasing the resistor increased distortion a tiny bit not worth mentioning.
He also changed the resistor for the reference voltage in the cascode stage from 10k to 22k.
Same thing here.
VAS:
Now he uses the 2SA1837 and 2SC4793 instead of the weaker 2SA1306 and 2SC3289. The old ones might have been B types that widthstand a bit higher voltage. I feel much more comfortable with the new transistors.
Maybe because of the new transistors their frequency compensation caps changed from 10pF to 22pF where one is now recommended a Mica cap.
In version 3 he changed the common Emitter resistor from 100R to 150R what makes my simulation stop working. Maybe this is caused by my transistor models or wrong bias.
BIAS:
Lowering the value of the bias bypass cap from 1uF to 220nF improves things.
The BJT, 2SC4793, that replaces the IRF610 in Version 2 makes biasing more difficult in simulation. Maybe this gives improved thermal tracking and better compensated bias.
I appreciate the BJT in the buffer stage but in simulation they don't perform well.
To Andrew: Do you know more about this? You recommended to keep the IRF drivers...
Decreasing the source resistor of the buffer stage from 330R to 100R increased distortion in simulation but still might be a good idea. I don't know.
All gate stopper resistors are changed from 470R to 100R. In simulation this decreased distortion a tiny bit. I hope this is still stable since I'm afraid of parasitic oscillation.
I almost forgot the split ground. This is a good idea but it would be much better to have a separate decoupling ground to the star point!
Additional thoughts:
I think that it is pointless to run the output stage at the same voltage like the driver stage because this does not allow the output stage to be fully driven to the rails. In my design I use 90V for output stage and 100V for the driver stage. I'm just not yet sure whether to use an extra transformer or to add some windings to the big ones. With the extra transformer the voltage stays stiff all the time and with the added windings the driver voltage may decrease with the power rails voltage. I think the second solution is not only cheaper but also more save.
I wonder he recommends Tantalum caps for input DC blocking. This might be common but I think film caps are way better.
I wonder he recommends MKT and ceramic caps. I don't think stability relies on dielectric loss and will use MKP and Mica ones.
I wonder about the uncommon values for the Zobel network. I think the standard 10R an 100nF will be better. Maybe he changed that in later versions but I cannot read them.
I wonder there's no hint to place the common inductor with parallel resistor in series with the power output to stabilize the amp with difficult loads. I wouldn't like to see such a powerful amp loosing stability.
now
Thanks
Lee
Attached you find a schematic of the amp.
Note the components don't have reference designators because they are messed up completely. I think Anthony is a good amp designer but documentation is definitely not one of his strengths.
I hope mine is correct then.
Is there anyone here interested in discussing this amp?
My ideas to the changes in Version 2 and 3:
In Version 3 he changed the input DC blocking caps from 47uF to 22 uF and increased the input impedance from 4k7 to 15k.
He also decreased the low pass caps from 2n2 to 1nF.
I think the modifications don't affect the amp very much. Basically just the input impedance is increased. Right?
In Version 2 he changed the input stage emitter resistors from 400R to a value thats shape looks different and changed them in Version 3 to 150R.
Simulation showed that lowering the resistor values decreases distortion considerably. First I ran simulation with 100R I took from the AV800 amp and noticed increased distortion with higher values. Simulation is only of very limited value because I don't have models for all transistors and the ones I have might be off. Anyway, I think lower values are better here.
In version 2 he changed the transitor of the constant current source from BD681 to MJE340, increased the resistor at its base from 10k to 22k and the cap from 47 uF to 100uF.
I think changing the transitor has no big effect. The most important difference is collector-emitter breakdown voltage.
Increasing the cap seems to be a good idea while increasing the resistor increased distortion a tiny bit not worth mentioning.
He also changed the resistor for the reference voltage in the cascode stage from 10k to 22k.
Same thing here.
VAS:
Now he uses the 2SA1837 and 2SC4793 instead of the weaker 2SA1306 and 2SC3289. The old ones might have been B types that widthstand a bit higher voltage. I feel much more comfortable with the new transistors.
Maybe because of the new transistors their frequency compensation caps changed from 10pF to 22pF where one is now recommended a Mica cap.
In version 3 he changed the common Emitter resistor from 100R to 150R what makes my simulation stop working. Maybe this is caused by my transistor models or wrong bias.
BIAS:
Lowering the value of the bias bypass cap from 1uF to 220nF improves things.
The BJT, 2SC4793, that replaces the IRF610 in Version 2 makes biasing more difficult in simulation. Maybe this gives improved thermal tracking and better compensated bias.
I appreciate the BJT in the buffer stage but in simulation they don't perform well.
To Andrew: Do you know more about this? You recommended to keep the IRF drivers...
Decreasing the source resistor of the buffer stage from 330R to 100R increased distortion in simulation but still might be a good idea. I don't know.
All gate stopper resistors are changed from 470R to 100R. In simulation this decreased distortion a tiny bit. I hope this is still stable since I'm afraid of parasitic oscillation.
I almost forgot the split ground. This is a good idea but it would be much better to have a separate decoupling ground to the star point!
Additional thoughts:
I think that it is pointless to run the output stage at the same voltage like the driver stage because this does not allow the output stage to be fully driven to the rails. In my design I use 90V for output stage and 100V for the driver stage. I'm just not yet sure whether to use an extra transformer or to add some windings to the big ones. With the extra transformer the voltage stays stiff all the time and with the added windings the driver voltage may decrease with the power rails voltage. I think the second solution is not only cheaper but also more save.
I wonder he recommends Tantalum caps for input DC blocking. This might be common but I think film caps are way better.
I wonder he recommends MKT and ceramic caps. I don't think stability relies on dielectric loss and will use MKP and Mica ones.
I wonder about the uncommon values for the Zobel network. I think the standard 10R an 100nF will be better. Maybe he changed that in later versions but I cannot read them.
I wonder there's no hint to place the common inductor with parallel resistor in series with the power output to stabilize the amp with difficult loads. I wouldn't like to see such a powerful amp loosing stability.
nowThanks
Lee
I wonder this thread was viewed over 1000 times and only a few people left comments.
This this project annoying you? Am I annoying you?
My latest idea is to change the current source for the input stage. I found an interesting thread about current sources here:
http://www.diyaudio.com/forums/solid-state/37662-current-source-how-optimize-9.html#post2173577
The "ANF" seemes to perform best with just a minor modification.
Comparing THD in simulation the two transistor current source outperforms the other one by lowering thd around 20%.
Maybe simulation is wrong.
Maybe there is in deed a possibility for improvement.
Changing a 100 times proven schematic might be a bad idea.
What do you think?
Is it safe to replace a current source of an amp?
Thanks
Lee
This this project annoying you? Am I annoying you?
My latest idea is to change the current source for the input stage. I found an interesting thread about current sources here:
http://www.diyaudio.com/forums/solid-state/37662-current-source-how-optimize-9.html#post2173577
The "ANF" seemes to perform best with just a minor modification.
Comparing THD in simulation the two transistor current source outperforms the other one by lowering thd around 20%.
Maybe simulation is wrong.
Maybe there is in deed a possibility for improvement.
Changing a 100 times proven schematic might be a bad idea.
What do you think?
Is it safe to replace a current source of an amp?
Thanks
Lee
Further investigation of the current sources made me stick with the one Anthony chose. I think he knows better than me.
Note: The schematic I posted above contains an error: The driver transistors need to be connected to the supply rails instead of behind the rail filter circuit.
Looking closer to the rail filter simulation showed a strange behavior. The diode and the 100R resistor seem to be no good. The voltage after the filter looks completely different than expected and there's even some ringing. I decided to kick the diode what makes the voltage look better. My only concern without diode is that the R-C filter might act as a remote snubber.
I don't ask for ideas this time and just attach the files I'm simulating with for someone who might be interested...
Note: I'm not good at spice. This is the first project I use it for.
Another note: The extra filter for the output stage driver transistor is unnecessary and only required because my board layout doesn't allow the 80uF cap to be close to the transistor and the 1R resistor is to prevent parallel resonance between the two caps. I'm not happy with this solution but seem to have no choice.
Thanks
Lee
Note: The schematic I posted above contains an error: The driver transistors need to be connected to the supply rails instead of behind the rail filter circuit.
Looking closer to the rail filter simulation showed a strange behavior. The diode and the 100R resistor seem to be no good. The voltage after the filter looks completely different than expected and there's even some ringing. I decided to kick the diode what makes the voltage look better. My only concern without diode is that the R-C filter might act as a remote snubber.
I don't ask for ideas this time and just attach the files I'm simulating with for someone who might be interested...
Note: I'm not good at spice. This is the first project I use it for.
Another note: The extra filter for the output stage driver transistor is unnecessary and only required because my board layout doesn't allow the 80uF cap to be close to the transistor and the 1R resistor is to prevent parallel resonance between the two caps. I'm not happy with this solution but seem to have no choice.
Thanks
Lee
Thank you for your replies.
Feels good to see I'm not on my own.
@Federmann: I started reading your website but need some time for it since the google translator makes it difficult for me.
@Piersma: Thank you for your recommendation.
From my very little amp knowledge the transistors look better for VAS than the ones in place now. Their pinout make them swapable with the 2SA1837/2SC4793. I'll check availability to me and maybe I'll even find models.
Thanks
Lee
Feels good to see I'm not on my own.
@Federmann: I started reading your website but need some time for it since the google translator makes it difficult for me.
@Piersma: Thank you for your recommendation.
From my very little amp knowledge the transistors look better for VAS than the ones in place now. Their pinout make them swapable with the 2SA1837/2SC4793. I'll check availability to me and maybe I'll even find models. Thanks
Lee
Hi Piersma
Thank you for the great idea with the CCS for the cascode Zener. This is in deed rock solid.
I also changed the other CCS.
Maybe performance could be better with some tweaking...
By the way: What do you think about having BJTs instead of FETs for bias and output stage drivers?
In simulation this gives worse THD but completely different gain / phase plot (given I got that right and I doubt it).
Thanks
Lee
Thank you for the great idea with the CCS for the cascode Zener.
This is in deed rock solid. I also changed the other CCS.
Maybe performance could be better with some tweaking...
By the way: What do you think about having BJTs instead of FETs for bias and output stage drivers?
In simulation this gives worse THD but completely different gain / phase plot (given I got that right and I doubt it).
Thanks
Lee
There will be PCB for AV1000.
Attached you find a preview of the board that shows Anthony's amp with some minor modifications.
It's not diy style because it uses a two layer board, smt components and connectors
. I use some hard to find fancy components like the input caps an the output stage source resistors. Thanks
Lee
There will be PCB for AV1000.
Attached you find a preview of the board that shows Anthony's amp with some minor modifications.
It's not diy style because it uses a two layer board, smt components and connectors
Thanks
Lee
Thanks, great work there, but I was looking for the original PCB and schematic version 3 from 2005.
Thanks.
I knew you are looking for Anthony's work. For the schematic maybe ask Andrew
but I think you might not find gerber files or diy exposure prints. You could try to add three more pairs of output transistors with a graphic program. Just load the pdf that lm423 posted into gimp and duplicate... Just an idea. Lee
P.S.:
For the case you find the schematic version 3 please leave me a copy, ok?
BJT versus FET
hi Lee,
I would prefer a BJT bias circuit but you could run into troubles with the temperature tracking of the output FETs. The thermal compensation circuit must track the output devices in a correct way, so to say not under- of overcompensated. So when you stick with the FETs you probably do best leaving the FET in the biasing circuit, if you go for a BJT outputstage you defenitly need a BJT for thermal tracking.
If you really want a "fancy" and perfect outputstage driver, use a CFP configuration. (especially when using a lot of outputput devices) Take a look at AAK's DTV SymAsym.
best regards,
Piersma
hi Lee,
I would prefer a BJT bias circuit but you could run into troubles with the temperature tracking of the output FETs. The thermal compensation circuit must track the output devices in a correct way, so to say not under- of overcompensated. So when you stick with the FETs you probably do best leaving the FET in the biasing circuit, if you go for a BJT outputstage you defenitly need a BJT for thermal tracking.
If you really want a "fancy" and perfect outputstage driver, use a CFP configuration. (especially when using a lot of outputput devices) Take a look at AAK's DTV SymAsym.
best regards,
Piersma
Last edited:
Hi Piersma
Thanks for the explanation.
I'd like to keep the FET output stage and so keep the FET for bias as you recommend.
I looked at the SymAsym that has a topology a bit similar to the AV series amps. There are a lot of cascode things in the amp what looks interesting. I will simulate the fancy driver that looks promising at the first glimpse. I'm not afraid of some added complexity and anyway have some board space left. Maybe my amp will become a hybrid between the two amps.
Today I checked availability of the 2SA1478 and 2SC3788.
Farnell, RS, MSC, Endrich, DigiKey, Codico and Mouser don't stock them. MOQ is >10000 pcs. Lead time about 14-18 weeks what is usual at the moment for most parts. I talked to Sanyo and they wanted me to contact a distributor for samples. The distributor told me it is near impossible to receive samples from Sanyo and for the case I succeed it takes half an in infinity and I won't receive more than 5 pcs.
Thanks
Lee
Thanks for the explanation.
I'd like to keep the FET output stage and so keep the FET for bias as you recommend.
I looked at the SymAsym that has a topology a bit similar to the AV series amps. There are a lot of cascode things in the amp what looks interesting. I will simulate the fancy driver that looks promising at the first glimpse. I'm not afraid of some added complexity and anyway have some board space left. Maybe my amp will become a hybrid between the two amps.
Today I checked availability of the 2SA1478 and 2SC3788.
Farnell, RS, MSC, Endrich, DigiKey, Codico and Mouser don't stock them. MOQ is >10000 pcs. Lead time about 14-18 weeks what is usual at the moment for most parts. I talked to Sanyo and they wanted me to contact a distributor for samples. The distributor told me it is near impossible to receive samples from Sanyo and for the case I succeed it takes half an in infinity and I won't receive more than 5 pcs.
Thanks
Lee
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.