I have just completed my first valve amp design. Rather than start from scratch I used a "building block" approach to come up with the design attached.
Specs:
Ultralinear Push Pull 4 x KT88 "cheap"ish Hammond 1650T Output Tranny.
120W at clipping 135W at hard saturation.
Frequency response at 50W Output -3db at 65 kHz top end and couldn't get a -3dB at the low end due to transformer saturation giving severe waveform distortions below 18Hz (was flat and clean down to 20Hz).
The design is Curcio Audio Engineering Board front end, followed by direct coupled cathode follower to drive 2 pairs of KT88s with Hammond 1650T Output Iron.
Note the UNEQUAL Zobels on between the Ultralinear (screen) taps and anode connections of two sides of the output transformer.
The zobels were worked out using info from this site.
http://www.siteswithstyle.com/VoltSecond/Damping_ringing_XFMRS/Damping_ringing_in_xfmrs.html
Hope this of interest to most especially the beginners out there.
DUD120 stands for "Down Under Design" 120W - a bit of tougue in cheek.
Cheers,
Ian
Specs:
Ultralinear Push Pull 4 x KT88 "cheap"ish Hammond 1650T Output Tranny.
120W at clipping 135W at hard saturation.
Frequency response at 50W Output -3db at 65 kHz top end and couldn't get a -3dB at the low end due to transformer saturation giving severe waveform distortions below 18Hz (was flat and clean down to 20Hz).
The design is Curcio Audio Engineering Board front end, followed by direct coupled cathode follower to drive 2 pairs of KT88s with Hammond 1650T Output Iron.
Note the UNEQUAL Zobels on between the Ultralinear (screen) taps and anode connections of two sides of the output transformer.
The zobels were worked out using info from this site.
http://www.siteswithstyle.com/VoltSecond/Damping_ringing_XFMRS/Damping_ringing_in_xfmrs.html
Hope this of interest to most especially the beginners out there.
DUD120 stands for "Down Under Design" 120W - a bit of tougue in cheek.
Cheers,
Ian
Questions about the design
If some of you experts would oblige:
1) Is the power supply decoupling to the cathode followers (R104, C100,C101 etc) I've added necessary or should it be deleted?
2) given that I'm seeing transformer saturation below 20Hz at high power should I reduce C103 thru' C106 to 47nF to give corner frequency at 34Hz instead of 16Hz as it is now?
Any other comments/suggestions greatfully received.
Thanks,
Ian
If some of you experts would oblige:
1) Is the power supply decoupling to the cathode followers (R104, C100,C101 etc) I've added necessary or should it be deleted?
2) given that I'm seeing transformer saturation below 20Hz at high power should I reduce C103 thru' C106 to 47nF to give corner frequency at 34Hz instead of 16Hz as it is now?
Any other comments/suggestions greatfully received.
Thanks,
Ian
> Is the power supply decoupling to the cathode followers... necessary...
Maybe. While the ripple nominally cancels, it won't exactly, and the raw HV is just ripply enough that I would not bet on it being dead-clean at the output grids. It costs more in thinking-time to check than it does to just throw-in the caps, so unless you mass-produce this is moot.
You might ask if both grids on V2 could be tied together, saving one and maybe two caps. If both halves are matched, it should work, and even with a little mismatch, but maybe not in all cases with available tubes.
As for the rest of it: it isn't how I would do it, but I don't see anything that looks wrong. The LM317-TIP50 rig is something I had not seen just this way before, resistant to over-volt accidents. I'm sure the unequal Zobels should be "select on test": Hammond probably can't hold HF un-balance to that degree of precision from lot to lot, year to year. I wonder how stable the Q2 bias is, you seem to be asking for a lot of forward gain under the R9,R10, R19 feedback loop, but without doing math I suspect it could work.
Gain of 48 is higher than many amps, is that what you want? Like 0.7V-0.8VRMS sensitivity?
> given that I'm seeing transformer saturation below 20Hz at high power
Is this a shaker-table amp or a music amp? There are only a few tracks with significant power below 20Hz. If that's your meat, don't shorten the amp, get bigger iron. But for most music (and speakers), power bandwidth to 20Hz or even 32Hz is more than ample.
Yes, low-power frequency response of amps like this is usually an order of magnitude deeper than the full-power response, and should be. If you expect subsonics, put in a proper 2-pole filter upstream, don't try to filter in the amp.
Anyway a 100-120Watt amp shouldn't normally be beating its butt off, it should be loafing at a few dozen watts with ample headroom. Power bandwidth may then be 10Hz. If you have speakers good for 10Hz, please don't move to my neighborhood: you will scare my nearly-deaf dog.
You can also kill LF saturation with more current, more tubes. But that's doing it the hard way.
Maybe. While the ripple nominally cancels, it won't exactly, and the raw HV is just ripply enough that I would not bet on it being dead-clean at the output grids. It costs more in thinking-time to check than it does to just throw-in the caps, so unless you mass-produce this is moot.
You might ask if both grids on V2 could be tied together, saving one and maybe two caps. If both halves are matched, it should work, and even with a little mismatch, but maybe not in all cases with available tubes.
As for the rest of it: it isn't how I would do it, but I don't see anything that looks wrong. The LM317-TIP50 rig is something I had not seen just this way before, resistant to over-volt accidents. I'm sure the unequal Zobels should be "select on test": Hammond probably can't hold HF un-balance to that degree of precision from lot to lot, year to year. I wonder how stable the Q2 bias is, you seem to be asking for a lot of forward gain under the R9,R10, R19 feedback loop, but without doing math I suspect it could work.
Gain of 48 is higher than many amps, is that what you want? Like 0.7V-0.8VRMS sensitivity?
> given that I'm seeing transformer saturation below 20Hz at high power
Is this a shaker-table amp or a music amp? There are only a few tracks with significant power below 20Hz. If that's your meat, don't shorten the amp, get bigger iron. But for most music (and speakers), power bandwidth to 20Hz or even 32Hz is more than ample.
Yes, low-power frequency response of amps like this is usually an order of magnitude deeper than the full-power response, and should be. If you expect subsonics, put in a proper 2-pole filter upstream, don't try to filter in the amp.
Anyway a 100-120Watt amp shouldn't normally be beating its butt off, it should be loafing at a few dozen watts with ample headroom. Power bandwidth may then be 10Hz. If you have speakers good for 10Hz, please don't move to my neighborhood: you will scare my nearly-deaf dog.
You can also kill LF saturation with more current, more tubes. But that's doing it the hard way.
Thanks PRR
PRR - Thanks for your reponse/comments.
The Gain of 48 is an "unintended" consequence of set to work tuning. With no feedback the Zo is 3.5 Ohms. I'm driving VAF Research DC-Xs which have nominal imedance of 6 Ohms but do not dip below 5 Ohms at any frequency. They do however have the usual increase in impedance at higher frequencies. With no feedback on the amp they sound bright. I had intended to use 12 to 15dB of feedback which would have given a more sensible gain and sensitivity figure.
With the circuit as shown feedback is about 5dB giving Zo of 1.8 "ish" Ohms. This was the best compromise betwen fast "live" feel and good speaker control (not too much colouration). That is a damping factor of about 3.
The Q2 hfe is 40 min @ 10mA (Vce = 1V) which is adequate (just).
Another Question:
I was always told get the amp as linear as possible before applying feedback. The Output Transformer Zobels were indeed select on test determined using VoltSeconds methods and were done with the amp open loop before feedback was applied.
Do they need re-adjusting once feedback is applied? Feedback should reduce the output tube ra and so the pole would have moved higher.
On the issue of setting to work - one thing I struggled to come to grips with BUT think I mostly now understand is stabilising the amp with feedback. None of the books I have give this more than lip service. Is there a good text anyone can recommend. I feel sure that this is something that all newbies must struggle with.
Cheers,
Ian
PRR - Thanks for your reponse/comments.
The Gain of 48 is an "unintended" consequence of set to work tuning. With no feedback the Zo is 3.5 Ohms. I'm driving VAF Research DC-Xs which have nominal imedance of 6 Ohms but do not dip below 5 Ohms at any frequency. They do however have the usual increase in impedance at higher frequencies. With no feedback on the amp they sound bright. I had intended to use 12 to 15dB of feedback which would have given a more sensible gain and sensitivity figure.
With the circuit as shown feedback is about 5dB giving Zo of 1.8 "ish" Ohms. This was the best compromise betwen fast "live" feel and good speaker control (not too much colouration). That is a damping factor of about 3.
The Q2 hfe is 40 min @ 10mA (Vce = 1V) which is adequate (just).
Another Question:
I was always told get the amp as linear as possible before applying feedback. The Output Transformer Zobels were indeed select on test determined using VoltSeconds methods and were done with the amp open loop before feedback was applied.
Do they need re-adjusting once feedback is applied? Feedback should reduce the output tube ra and so the pole would have moved higher.
On the issue of setting to work - one thing I struggled to come to grips with BUT think I mostly now understand is stabilising the amp with feedback. None of the books I have give this more than lip service. Is there a good text anyone can recommend. I feel sure that this is something that all newbies must struggle with.
Cheers,
Ian
PRR said:
No don't do it. It will ruin the tubes. On core saturation, magnetic permeabilty effectly drops to the winding resistance to take up the power which waveform tries to become square. Result tubes strained and working life reduced.
The only way to get more umph at the low freq end is bigger output trannies. Judging by my design for Hi Fi, each o/p tranny in my stereo lump 150+150W weighs nearly 10kg. THD will rise when pushed at 20Hz as expected around 1% mark. With the mains tranny plonked on top youre looking at a chassis weight -circa 45-50kg.....are you strong enough to lift this lot ???. certainly won't get stolen in a hurry.
If you are looking for reserve 120+120W then B+ will be around 530V not 500V. Are you happy with this ? Todays tubes do require a bit more. The question arises..DO you really want 100W tube amp sound esp with large reflex LS......it's loud. Far better to run at lower B+ with PPP and attempt class A by upping I quies. Quality is brill which for an output tranny overspecified for lower power..
275V B+ at 80mA per tube quies will give 30W output power.
Output stage...as shown...Presented with a strong HF signal i.e a 15Khz signal rammed through the amp at full power. The zobel networks from g2-anode 1K 3W and 4n7 will stew to destruct. C107/108/109 R124 aren't symmetrical values. Why not ? That implies somethings up in the power circuit. They should all be identical. A misprint ?
For g2-A zobel I use 1K5 6W carbon types and 2n2 polyprop but that depends on o/p tranny characteristics. A 4n7 will create more heating power in the resistors. These will influence high audio frequency square wave quality and be quite sure that these components can take the stick.
Generally when using SS in tube amps especially within interstage psu's... make darned sure they are well designed not to fail.....a 100W tube amp can do alot of damage in fault conditions.
richj
120 watt mono block tube amp
looking for info at the schematic level to build a 120 watt per channel tube amp. Actually have a pair of unbuilt Dynaco Mark Vi's and would rather build them as kt-88's, 6550's, 5881's, 6l6's, el4's or something other than 8417's. Not to much detail out there on doing an actual conversion, so I'm thinking using something like the dud120 schematic to guide my thinking is the way to go. I read the pro's and con's about the dud120 design, but the level of detail is there and the iron size on the Dynaco Mark VI is in the right league.
So basically, looking for additional info, and/or recommendations if any to the dud120 by the folks who actually built it.
thanks.
looking for info at the schematic level to build a 120 watt per channel tube amp. Actually have a pair of unbuilt Dynaco Mark Vi's and would rather build them as kt-88's, 6550's, 5881's, 6l6's, el4's or something other than 8417's. Not to much detail out there on doing an actual conversion, so I'm thinking using something like the dud120 schematic to guide my thinking is the way to go. I read the pro's and con's about the dud120 design, but the level of detail is there and the iron size on the Dynaco Mark VI is in the right league.
So basically, looking for additional info, and/or recommendations if any to the dud120 by the folks who actually built it.
thanks.
I have responded to your PM.
For the benefit of others here is what I wrote to "Speakerfritz":
Mike,
This is a very old thread. I never progressed with the prototype, instead I got side tracked into the "Baby Huey" low power design. I still have the prototype intact and will go back to it eventually.
I've learned a few things since I built that prototype.
First I would get rid of those anode to screen zobel networks on the output tubes. You can always play around with a zobel across the output transformer secondary if required.
The (low frequency) distortion at high power levels is definitely due to output transformer core saturation. There is nothing you can do about this short of going to a bigger output transformer.
This is something for you to note -
The MKIV output transformers are designed for a single pair of EL34 in Ultralinear connection to give 40 watts. Power transformers would be similarly designed for that power level. Changing to a DUD120 design would basically mean starting again from scratch with new power and output transformers. Also, converting to KT88 could be done (transformers are suitable) but will give only 50 watt compared to 40 watt for the EL34s. The output tube grid1 resistors (270K) are too high for KT88 and so some front end / driver mods would be required.
I would suggest building the MKIVs "As is" and once up and running you can then try some upgrade options.
Cheers,
Ian
For the benefit of others here is what I wrote to "Speakerfritz":
Mike,
This is a very old thread. I never progressed with the prototype, instead I got side tracked into the "Baby Huey" low power design. I still have the prototype intact and will go back to it eventually.
I've learned a few things since I built that prototype.
First I would get rid of those anode to screen zobel networks on the output tubes. You can always play around with a zobel across the output transformer secondary if required.
The (low frequency) distortion at high power levels is definitely due to output transformer core saturation. There is nothing you can do about this short of going to a bigger output transformer.
This is something for you to note -
The MKIV output transformers are designed for a single pair of EL34 in Ultralinear connection to give 40 watts. Power transformers would be similarly designed for that power level. Changing to a DUD120 design would basically mean starting again from scratch with new power and output transformers. Also, converting to KT88 could be done (transformers are suitable) but will give only 50 watt compared to 40 watt for the EL34s. The output tube grid1 resistors (270K) are too high for KT88 and so some front end / driver mods would be required.
I would suggest building the MKIVs "As is" and once up and running you can then try some upgrade options.
Cheers,
Ian
Mike,
Sorry "dislexic" reading - I read MKIV where you clearly say MKVI.
Staying away from 8417 is a good idea - from availabilty and reliability point of view. The trannies WILL suit a quad of KT88 or 6550 in Ultralinear and this is the obvious way to go.
Plenty of ways to "skin a cat".
My experience with the DUD120 was that I ran out of gain to be able to apply much global feedback. The basic MKVI design will have this problem since the 8417 had twice the gm of KT88 and the like. That means you need to swing twice as much voltage into a KT88 grid.
You could use the DUD120 design and it will work but you may find it does not have enough damping. I would also substitute 4 MOSFET source followers, one each direct coupled to each KT88 grid with individual fixed bias for each tube applied to the MOSFET gates - then 2 cap coupled to each diff amp side (Instead of the 6SN7 cathode followers and AC coupled output tube grids).
A simpler design which may work is an Allen Wright PPC1A cascode diffamp front end variation driving the output tubes without a buffer stage. Run 6H30, drop the anode loads to about 12K and run the currents up around 15mA per side.
If I were balnk sheet designing I'd try to incorporate a "Baby Huey" shunt feedback scheme from the output tube anodes back to the diff amp. That may require an additional gain stage ahead of the diff amp - something like this BUT with MOSFET Source followers and Output Tubes doubled up.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=78712&perpage=10&pagenumber=3
Cheers,
Ian
Sorry "dislexic" reading - I read MKIV where you clearly say MKVI.
Staying away from 8417 is a good idea - from availabilty and reliability point of view. The trannies WILL suit a quad of KT88 or 6550 in Ultralinear and this is the obvious way to go.
Plenty of ways to "skin a cat".
My experience with the DUD120 was that I ran out of gain to be able to apply much global feedback. The basic MKVI design will have this problem since the 8417 had twice the gm of KT88 and the like. That means you need to swing twice as much voltage into a KT88 grid.
You could use the DUD120 design and it will work but you may find it does not have enough damping. I would also substitute 4 MOSFET source followers, one each direct coupled to each KT88 grid with individual fixed bias for each tube applied to the MOSFET gates - then 2 cap coupled to each diff amp side (Instead of the 6SN7 cathode followers and AC coupled output tube grids).
A simpler design which may work is an Allen Wright PPC1A cascode diffamp front end variation driving the output tubes without a buffer stage. Run 6H30, drop the anode loads to about 12K and run the currents up around 15mA per side.
If I were balnk sheet designing I'd try to incorporate a "Baby Huey" shunt feedback scheme from the output tube anodes back to the diff amp. That may require an additional gain stage ahead of the diff amp - something like this BUT with MOSFET Source followers and Output Tubes doubled up.
http://www.diyaudio.com/forums/showthread.php?s=&threadid=78712&perpage=10&pagenumber=3
Cheers,
Ian
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.