The world's best oscillator....
Well done danspy. Very glad you got it going. Sounds nice huh?
Oh dear, it looks like I jumped the gun with my comments about stability...
I went home today and wired things up exactly as they were before with the only difference being that I added a 1uf cap across the bias resistor.
Turned it on, decided to scope the output before adding a load. Terrible rail to rail oscillations somewhere in the MHz range. This time no amount of poking and prodding the components would get rid of them. I don't think could make an oscillator work this well if I tried!
Hmm, added a 150pf miller cap....no good.
Hooked up a 6R resistor to output. Oscillations gone...hmm maybe an output filter would help?
Added a rc filter to the output with a 2W 10 ohm resistor and 100nf cap. Removed 6R dummy load and turned it on. 2W filter resistor started to smoke...sheesh, there's some serious rf going through that 100nf cap.
So what do you think? When I'm lucky the amp won't oscillate. Sometimes it turns on fine, other times I get rail to rail oscillations. If a low z load is attached it seems fine. Seems ok with a speaker attached.
How disappointing!
Guess I need to drop the olg even more and think seriously about some compensation...
Boy does it have some bass though!
Well done danspy. Very glad you got it going. Sounds nice huh?
Oh dear, it looks like I jumped the gun with my comments about stability...
I went home today and wired things up exactly as they were before with the only difference being that I added a 1uf cap across the bias resistor.
Turned it on, decided to scope the output before adding a load. Terrible rail to rail oscillations somewhere in the MHz range. This time no amount of poking and prodding the components would get rid of them. I don't think could make an oscillator work this well if I tried!
Hmm, added a 150pf miller cap....no good.
Hooked up a 6R resistor to output. Oscillations gone...hmm maybe an output filter would help?
Added a rc filter to the output with a 2W 10 ohm resistor and 100nf cap. Removed 6R dummy load and turned it on. 2W filter resistor started to smoke...sheesh, there's some serious rf going through that 100nf cap.
So what do you think? When I'm lucky the amp won't oscillate. Sometimes it turns on fine, other times I get rail to rail oscillations. If a low z load is attached it seems fine. Seems ok with a speaker attached.
How disappointing!
Guess I need to drop the olg even more and think seriously about some compensation...
Boy does it have some bass though!
Last edited:
Fast amps do need extra attention regarding stablity considering cct design & layout but in my opinion it is well worth the effort.
Personally I would not just add more pF to the VAS Miller position - for me that would be a last option and perhaps just a very small value.
Other options available are 1) what I did in my spice cct above and 2) taking HF feedback from the VAS stage ( I will be playing with this in spice some time soon ) and also loosing OLG with more VAS degeneration.
sorry I can't join in with the fun just now
mike
Personally I would not just add more pF to the VAS Miller position - for me that would be a last option and perhaps just a very small value.
Other options available are 1) what I did in my spice cct above and 2) taking HF feedback from the VAS stage ( I will be playing with this in spice some time soon ) and also loosing OLG with more VAS degeneration.
sorry I can't join in with the fun just now
mike
Last edited:
swordfishy - if a 6 ohm load fixed it - then perhaps 0.1uF + 6ohms would be better ? ? ? or even 4ohms - might be worth a try.
When I was experimenting in spice I found that with capacitive loads the miller cap actually made things worse - which is why I removed it.
When I was experimenting in spice I found that with capacitive loads the miller cap actually made things worse - which is why I removed it.
No i have no scope, and if it Oscillate,s and i don`t hear it and the Output,s don`t get hot it don`t upset me.
I wold say it Sound,s Groovy an very musical.
Yeah, musical is the word...if I had to criticise one thing it would be a lack of treble. However I haven't had a chance to audition it with decent speakers yet, just some bottom line JAMO bookshelves. I also have nothing in the way of supply rail bypassing which can't help.
Anyway, mikelm I managed to move the board around a bit and get it to behave for a few minutes. Here is the 20kHz square wave performance at nearly 40v p-p across a 7R load. What do you think?
By the way, when it's not oscillating the DC offset is stable to within a few mV.
By the way no 2....the rise and fall times aren't quite as impressive as they were yesterday as I have upped the gate stoppers to 500R in an effort to stop oscillation. However, the ringing has entirely disappeared which is very good!
Anyway, mikelm I managed to move the board around a bit and get it to behave for a few minutes. Here is the 20kHz square wave performance at nearly 40v p-p across a 7R load. What do you think?
By the way, when it's not oscillating the DC offset is stable to within a few mV.
By the way no 2....the rise and fall times aren't quite as impressive as they were yesterday as I have upped the gate stoppers to 500R in an effort to stop oscillation. However, the ringing has entirely disappeared which is very good!
Attachments
Last edited:
With a 20khz square wave like that I don't think there is any fear of lack of treble !
it looks like it's soaring way past 100khz
it looks like it's soaring way past 100khz
Yeah not bad huh? What loads would you like me to try it with?
I tried reducing the output filter resistor as you suggested to no avail.
My laptop has died tonight but I plan to spend tomorrow simming all the possible modifications we have discussed to see how far I can push the degeneration, etc. I've just been rereading some of your posts. I like the idea of the cascode, will have to try it.
It is possible that when I replace the zvn3310a with the jfet the lower transconductance will also help things.
I can tell that it's close to stable. If I put my hand on the heatsink I can often get the oscillation to stop due to the capacitive loading. A nice pcb layout would help too. Hopefully we can crack this thing over the next few days.
I tried reducing the output filter resistor as you suggested to no avail.
My laptop has died tonight but I plan to spend tomorrow simming all the possible modifications we have discussed to see how far I can push the degeneration, etc. I've just been rereading some of your posts. I like the idea of the cascode, will have to try it.
It is possible that when I replace the zvn3310a with the jfet the lower transconductance will also help things.
I can tell that it's close to stable. If I put my hand on the heatsink I can often get the oscillation to stop due to the capacitive loading. A nice pcb layout would help too. Hopefully we can crack this thing over the next few days.
Last edited:
Thanks Mikelm.
Sure here it is.
I know the miller cap is very large but it is the smallest I have lying around. I also intend to bypass the bias pot with a 1uF cap, but for the moment it seems to make things worse so I have taken it off.
Also my source resistors are actually 0.47R as I didn't have any 0.1Rs handy.
I have tried all of the following both separately, in combination and all together at once:
1) Output filter with 100n cap and a 10R and 5R resistor.
2) With and without the miller cap.
3) Original VAS degen resistor was 50R, now it's 100R.
4) I upped the gate stoppers on the main fets from 221 to 500R.
5) I tried adding 220pf to gate of n channel fet to balance them (rod elliot had problems with this).
Look forward to your comments. Yikes it's 2:20am! Time for bed I think.
I will try dropping the OLG further tomorrow.
Sure here it is.
I know the miller cap is very large but it is the smallest I have lying around. I also intend to bypass the bias pot with a 1uF cap, but for the moment it seems to make things worse so I have taken it off.
Also my source resistors are actually 0.47R as I didn't have any 0.1Rs handy.
I have tried all of the following both separately, in combination and all together at once:
1) Output filter with 100n cap and a 10R and 5R resistor.
2) With and without the miller cap.
3) Original VAS degen resistor was 50R, now it's 100R.
4) I upped the gate stoppers on the main fets from 221 to 500R.
5) I tried adding 220pf to gate of n channel fet to balance them (rod elliot had problems with this).
Look forward to your comments. Yikes it's 2:20am! Time for bed I think.
I will try dropping the OLG further tomorrow.
Attachments
Yes the 500 gate stopper,s are a way to go, or 220 for the N channel and 470 for the P channel Fet,s like Hafler does it with the Hitachi,s.
Sound way,s it it has very strong mid,s but the strong bass leak,s a bit in the last Octave because of the low ohm Feedback-network 680 and 47/1000uF to ground.
The hight,s are just like i like it not overtoned, and very acurately, or Golden instead of the Silvery Sounding won,s that you can Buy of the Shelf in any Discounter.
Sound way,s it it has very strong mid,s but the strong bass leak,s a bit in the last Octave because of the low ohm Feedback-network 680 and 47/1000uF to ground.
The hight,s are just like i like it not overtoned, and very acurately, or Golden instead of the Silvery Sounding won,s that you can Buy of the Shelf in any Discounter.
danspy, if you can, try to increase C6, for example 330~470uF. Let me know what results you get.
Also, let R6 be 1K and R7 be 1~1.5K.
Also, let R6 be 1K and R7 be 1~1.5K.
Last edited:
Åhh fellow inmates,
hope I can be allowed my 2p's worth...
This trend started wonderfully and then it seems like the 'SIM Head' crace took over?
The advantage of a 1st LTP (Long Tailored Pair), is by many other things - DC stability also thermally. Price is that the Pair will by default suppress even order harmonics, which for Audio, is quite undesirable. Therefore it's a fabulously good idea with a current mirror at the 'Tails'. The main bulk of the Amplification is not done in this stage anyway, but mainly in the 2nd VAS stage. The current mirror does of course enhance this, by providing a much higher impedance at this place, than simple resistors would do, but raw amplification is still magnitudes less than at the VAS stage. Current mirror recommended with a suitable db. BPT. Fet's can be used but will almost certainly make DC drift, in a non mono crystalline form!
Now there is actually Toshiba Fet's avaliable that would do a sufficiently good high Voltage job here: 2SK373. They can be obtained from various sources incl. eBay. it's a 100V device so that should be plenty enough for a Hi-Fi LPT pair.
Now is current source really needed for the pair made with Fet's? Try SIM'ing with just a resistor defining the current here pls.
VAS stage:
I have also tried Hexfet's, small Hitachi's etc. to some less desirables effects...
The VAS tranny is probably the most important component in the the whole amp, since it (again) provides the most bulk of the raw amplification, together with it's load!
A good, well degenerated Vertical Mosfet will do the job here.
The ZVP3310 is actually quite a good choice!
NXP and Siemens also make's something suitable for this part.
It's correct that current will be quickly limited to not much more than 10mA's, but this is actually plenty good enough for driving 2 pair's of LATERAL MosFets.
The HexFet's has much higher input capacitance and therefore will require more 'drive amp's' at HF. More than a single pair, without source follower's is not recommended!
On this note: If You consider the need for more current, then there's the small Hitachi's in TO-220 as potential for source follower's (ala Erno Borbely) - they could/would even provide enough juice for driving a number of BiPolars, if anybody should have that desire...
I did say loading of the VAS - that also means the current source for it. The higher the impedance of this - the higher the potential amplification. A suitable Vertical MosFet, can be utilized here with great advantages... (one of the schematics did outline this...)
Now pole setting's: it is hugely important that the 'Miller cap' is kept as small as possible or preferably omitted all together, in order to ensure good open loop BW and resulting available FB for HF low distortion. Try setting the poles elsewhere in the circuit pls. A small 2n2 cap can be attempted on the current mirror degenerating resistors for instance.
Hope this provides some nice food for further Simulations and experimentations/builds Pls!
all with the best intentions of encouragement
DocO
hope I can be allowed my 2p's worth...
This trend started wonderfully and then it seems like the 'SIM Head' crace took over?
The advantage of a 1st LTP (Long Tailored Pair), is by many other things - DC stability also thermally. Price is that the Pair will by default suppress even order harmonics, which for Audio, is quite undesirable. Therefore it's a fabulously good idea with a current mirror at the 'Tails'. The main bulk of the Amplification is not done in this stage anyway, but mainly in the 2nd VAS stage. The current mirror does of course enhance this, by providing a much higher impedance at this place, than simple resistors would do, but raw amplification is still magnitudes less than at the VAS stage. Current mirror recommended with a suitable db. BPT. Fet's can be used but will almost certainly make DC drift, in a non mono crystalline form!
Now there is actually Toshiba Fet's avaliable that would do a sufficiently good high Voltage job here: 2SK373. They can be obtained from various sources incl. eBay. it's a 100V device so that should be plenty enough for a Hi-Fi LPT pair.
Now is current source really needed for the pair made with Fet's? Try SIM'ing with just a resistor defining the current here pls.
VAS stage:
I have also tried Hexfet's, small Hitachi's etc. to some less desirables effects...
The VAS tranny is probably the most important component in the the whole amp, since it (again) provides the most bulk of the raw amplification, together with it's load!
A good, well degenerated Vertical Mosfet will do the job here.
The ZVP3310 is actually quite a good choice!
NXP and Siemens also make's something suitable for this part.
It's correct that current will be quickly limited to not much more than 10mA's, but this is actually plenty good enough for driving 2 pair's of LATERAL MosFets.
The HexFet's has much higher input capacitance and therefore will require more 'drive amp's' at HF. More than a single pair, without source follower's is not recommended!
On this note: If You consider the need for more current, then there's the small Hitachi's in TO-220 as potential for source follower's (ala Erno Borbely) - they could/would even provide enough juice for driving a number of BiPolars, if anybody should have that desire...
I did say loading of the VAS - that also means the current source for it. The higher the impedance of this - the higher the potential amplification. A suitable Vertical MosFet, can be utilized here with great advantages... (one of the schematics did outline this...)
Now pole setting's: it is hugely important that the 'Miller cap' is kept as small as possible or preferably omitted all together, in order to ensure good open loop BW and resulting available FB for HF low distortion. Try setting the poles elsewhere in the circuit pls. A small 2n2 cap can be attempted on the current mirror degenerating resistors for instance.
Hope this provides some nice food for further Simulations and experimentations/builds Pls!
all with the best intentions of encouragement
DocO
Last edited:
Sorry I don`t Simulate nor Measure with a Scope, I build my AMP,s only for Sonic purposes.
Last edited:
- Status
- Not open for further replies.