Mike,
That's a very serious power supply, more akin to the TGV train in France than a FetZilla.
Have you watched the voltage at the amp rails on a CRO during music passages? I'm a little concerned it might dip precipitously during music peaks, since these represent abrupt current pulses into the amp and will meet with stiff resistance from the chokes.
Lovely setup you have there, and thanks for the comments on the pcbs. I'm kinda pleased with how they turned out. I've just mounted 36 of the ZVP2210Gs onto the boards, then cleaned off the flux and tested them with the DMM. I have only 30 to go!
Cheers,
Hugh
That's a very serious power supply, more akin to the TGV train in France than a FetZilla.
Have you watched the voltage at the amp rails on a CRO during music passages? I'm a little concerned it might dip precipitously during music peaks, since these represent abrupt current pulses into the amp and will meet with stiff resistance from the chokes.
Lovely setup you have there, and thanks for the comments on the pcbs. I'm kinda pleased with how they turned out. I've just mounted 36 of the ZVP2210Gs onto the boards, then cleaned off the flux and tested them with the DMM. I have only 30 to go!
Cheers,
Hugh
Hugh,
As more than one person has queried this and to satisfy my curiosity I did some spice research to find out what the story was.
Subjectively I have no worries in this regard - At the reasonably spirited levels that I play my system, I hear no problems with this supply.
In SPICE I compared different R values with L remaining constant and then tried different L values with R remaining constant. All this with 4 different values of R representing the load.
Just looking here at the smallest R load value of 6.25R ( drawing about 3.5amps )
I found that if I doubled the choke R value from 0.075R to 0.15R per rail I lost 0.22V
However, when I quadrupled the L value from 3mH to 12mH per rail I only lost 0.02V. ( it was hard to see a difference with just doubling L )
So my conclusion is that although increasing L causes a very small loss of voltage under heavy load conditions this represents a very small fraction of the voltage drop caused by the resistive elements of the chokes.
Actually, these chokes really are not working that hard - just a few volts of ripple on a DC flow. I actually checked the primaries of the trannies I am using as chokes to see if lethal voltages were being induced there and I could only measure 5VAC !
However, if chokes are placed between the transformer and the diodes then it is a very different story. In this configuration the chokes are dealing with massive alternating polarity current spikes and this does indeed create a supply with very poor regulation - probably only suitable for single ended O/P class A Amps where the current demand remains constant.
So all in all as long the R values are kept quite low, there should be no substantial difference in regulation between CLRC & CRC. I chose 200VA devices to achieve the combination of reasonable levels of L along with low values of R.
cheers
mike
Last edited:
Thanks Mike,
That really does give empirical proof this concept is more than viable; a CLC works the best it seems, and there is no penalty, in fact quite the contrary, with large values of L.
Group Buy:
Folks, I'm all set to mail out the group buy pcbs and associated semiconductors, T1, T2 and T3 (but not the outputs).
Accordingly, have a look at my thread on this amp at the Aspen forum and if you have not yet paid by paypal, please do so and I will expedite your pcbs ASAP.
I must say it's a great board, very well made, and STRONG, being 2mm FR4, double sided. These boards will take any mistreatment, I should say.
Cheers,
Hugh
That really does give empirical proof this concept is more than viable; a CLC works the best it seems, and there is no penalty, in fact quite the contrary, with large values of L.
Group Buy:
Folks, I'm all set to mail out the group buy pcbs and associated semiconductors, T1, T2 and T3 (but not the outputs).
Accordingly, have a look at my thread on this amp at the Aspen forum and if you have not yet paid by paypal, please do so and I will expedite your pcbs ASAP.
I must say it's a great board, very well made, and STRONG, being 2mm FR4, double sided. These boards will take any mistreatment, I should say.
Cheers,
Hugh
I am away upstairs to try to find out why NFetZilla took out my bass driver yesterday.
It was on, but not playing at the time. I didn't notice until an hour or two later.
The metal driver cone was warm and sooked in like a soor ploom.
Oh bugger, I wanted to try to assemble my new PC tonight. Priorities ! I thought retired folk didn't need to concern themselves with priorities.
It was on, but not playing at the time. I didn't notice until an hour or two later.
The metal driver cone was warm and sooked in like a soor ploom.
Oh bugger, I wanted to try to assemble my new PC tonight. Priorities ! I thought retired folk didn't need to concern themselves with priorities.
Last edited:
I may have an answer.
I had noticed during testing that losing my -ve supply brings the output offset to near +ve rail, (+26V). removing +ve supply has minor effect.
After taking NFetZilla apart I found a broken connection between the removable voltage amp stage and the fixed on heatsink output stage.
The power ground (PG) on volt stage was not connected to PG on the output.
Soldering the connection seems to have returned all to normal.
Removing the Zero Volts connection between PSU and Output stage does not reproduce the effect.
I unsoldered the repaired connection. Output offset goes to -14V.
It must be that broken PG connection that created the -14V offset and cooked the Bass VC.
Keep in mind I am using the inverted version and all my -ve and +ve are swapped.
I had noticed during testing that losing my -ve supply brings the output offset to near +ve rail, (+26V). removing +ve supply has minor effect.
After taking NFetZilla apart I found a broken connection between the removable voltage amp stage and the fixed on heatsink output stage.
The power ground (PG) on volt stage was not connected to PG on the output.
Soldering the connection seems to have returned all to normal.
Removing the Zero Volts connection between PSU and Output stage does not reproduce the effect.
I unsoldered the repaired connection. Output offset goes to -14V.
It must be that broken PG connection that created the -14V offset and cooked the Bass VC.
Keep in mind I am using the inverted version and all my -ve and +ve are swapped.
Last edited:
I'm sorry to hear this, Andrew, it is a fact that prototypes are not so reliable as full on pcbs. In any event, you have Pmos VAS FetZilla arriving quite soon, the real deal, and you will be able to rebuild the inverted beast.
I believe you will find the pcb easy to build and hopefully sound just as good.
Oh, and retired folk DO have priorities. And worries. The whole human condition thang, really......
Cheers,
Hugh
I believe you will find the pcb easy to build and hopefully sound just as good.
Oh, and retired folk DO have priorities. And worries. The whole human condition thang, really......
Cheers,
Hugh
NMOS
I certainly wouldn't want to lose my best speakers to any experimental device. Naturally, you feel that DC protection would have saved the day.
Bonsai's recent Output Relay thread discussed a VMOS DC switch as a more reliable device than a relay. Perhaps this is the time for something basic
but at least bulletproof like this with, err...cheaper speakers on test?
http://www.diyaudio.com/forums/solid-state/191449-output-relays-7.html
I certainly wouldn't want to lose my best speakers to any experimental device. Naturally, you feel that DC protection would have saved the day.
Bonsai's recent Output Relay thread discussed a VMOS DC switch as a more reliable device than a relay. Perhaps this is the time for something basic
but at least bulletproof like this with, err...cheaper speakers on test?
http://www.diyaudio.com/forums/solid-state/191449-output-relays-7.html
My second version had been playing for over a week and it had been thoroughly tested before hooking up.
However, you are right, it is at an experimental stage and part of my work broke.
That's my fault.
I am not blaming anyone else.
Just reminding Fellow Members that DC detection and some form of protection is useful and that I normally have it on any completed amplifier I use.
This may be an occasion that prompts the design and build if a temporary stand alone detection and trip device that can be used with any experimental device.
And yes, I have been following that thread, but I don't plan to put a relay, of any type, in the audio output. My protection will cut the supply rails, faster than a fuse.
However, you are right, it is at an experimental stage and part of my work broke.
That's my fault.
I am not blaming anyone else.
Just reminding Fellow Members that DC detection and some form of protection is useful and that I normally have it on any completed amplifier I use.
This may be an occasion that prompts the design and build if a temporary stand alone detection and trip device that can be used with any experimental device.
And yes, I have been following that thread, but I don't plan to put a relay, of any type, in the audio output. My protection will cut the supply rails, faster than a fuse.
Hello All,
Here is my variation of FETZilla. The trick is with BSS84 on VAS. BSS84 has ultra low input capacitance and satisfying linearity which significantly reduces overall THD. In OS I am using IRF 140/9140 pair. As they have positive temperature coefficient biasing LEDs should be mounted on the heatsink to achieve thermal stability. Of course you're welcome to use any other MOSFETs as output pair. VAS is running 4.5 mA only but it's sound good. If you use small SMD heatsink 10mA won't be a problem.
http://mach-audio.eu/media/1250/prosty_wzmacniacz.jpg
THD is less than 0.09% for 45W@20kHz
I will tray to post higher quality schematic ans some simulation results soon.
Here is my variation of FETZilla. The trick is with BSS84 on VAS. BSS84 has ultra low input capacitance and satisfying linearity which significantly reduces overall THD. In OS I am using IRF 140/9140 pair. As they have positive temperature coefficient biasing LEDs should be mounted on the heatsink to achieve thermal stability. Of course you're welcome to use any other MOSFETs as output pair. VAS is running 4.5 mA only but it's sound good. If you use small SMD heatsink 10mA won't be a problem.
http://mach-audio.eu/media/1250/prosty_wzmacniacz.jpg
THD is less than 0.09% for 45W@20kHz
I will tray to post higher quality schematic ans some simulation results soon.
This is the most perfect amplifier I have spiced.
It is simple but yet very, very good.
- JFET input with minimal offset problems.
- MOSFET VAS for speed and gain
- LATERAL output for be a light load to VAS
When CLass AB operation, 350mA bias, it shows only
THD 0.00018% at 1 Watt
And there is almost no increase of THD at 25 Watt output
Enjoy!!
".......but what's the point of wasting 20 years to end up right where we started ? "
Harmonic Distortion over the Decades
------------------------------
Last edited:
What distortion is that plot purporting to be showing?
D.Self's Blameless from around the mid 90s (the last dates plotted on the graph) was of the order of 0.0006%. That puts the point on approximately the "d" or modern opamps.
Now we have designs that are around the 0.0001%, all from ClassAB and relatively high feedback. This point from year 2010 would be roughly on the "%" of 0.0001% out by (or on) the margin of the plot.
Some modern opamps (as of 2010) are also in this ballpark for ClassAB distortion. I think there are some quoting ~0.00007%
D.Self's Blameless from around the mid 90s (the last dates plotted on the graph) was of the order of 0.0006%. That puts the point on approximately the "d" or modern opamps.
Now we have designs that are around the 0.0001%, all from ClassAB and relatively high feedback. This point from year 2010 would be roughly on the "%" of 0.0001% out by (or on) the margin of the plot.
Some modern opamps (as of 2010) are also in this ballpark for ClassAB distortion. I think there are some quoting ~0.00007%
Last edited:
Equally, what's the point of whittling irrelevant fractions off what already is proven adequate for that same 20 or 30 years ?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.