I am researching a new project, thinking about the best direction for it and what I need to acquire in terms of reading / knowledge at first. My objective is a monobloc power amp with the following parameters:
· High current / low impedance capability as much as high power. The main speakers I own at present are 90db/W but a demanding load at LF – 4R impedance at 30Hz / 45 degrees phase lead rising to 10R at HF. I’ve been down the high efficiency / low power route and it did not work for me so SET designs need not apply!
· Custom OPTs are OK
· Heavy driver requirements are OK – I will look at choke-loaded / small power tube etc drivers later
· Current production tubes for the outputs
· I want to keep voltages reasonable say 600V max: no transmitter tubes
The following things are maybe ‘in the mix’ for me at present - parallel (perhaps more than two pairs) push pull, triodes, cathode feedback with or without ultralinear taps, generally keeping down the turns in the OPT and / or toroids. What do you all think?
I’ve built a big amp before, it works well and this is my attempt to go beyond it – I know what I am getting myself into!
· High current / low impedance capability as much as high power. The main speakers I own at present are 90db/W but a demanding load at LF – 4R impedance at 30Hz / 45 degrees phase lead rising to 10R at HF. I’ve been down the high efficiency / low power route and it did not work for me so SET designs need not apply!
· Custom OPTs are OK
· Heavy driver requirements are OK – I will look at choke-loaded / small power tube etc drivers later
· Current production tubes for the outputs
· I want to keep voltages reasonable say 600V max: no transmitter tubes
The following things are maybe ‘in the mix’ for me at present - parallel (perhaps more than two pairs) push pull, triodes, cathode feedback with or without ultralinear taps, generally keeping down the turns in the OPT and / or toroids. What do you all think?
I’ve built a big amp before, it works well and this is my attempt to go beyond it – I know what I am getting myself into!
Member
Joined 2009
Paid Member
I'm inexperienced with big amps, but I have noticed a lot of good things are being said about the KT120 for big outputs and if I were looking at building a big amp this is where I would start. I would not parallel them if possible. I would look at a strong driver like the ECC99(?) and I would use a very good input transformer/phase splitter. Maybe another gain stage is needed in there somewhere. No substitute for big iron.
I've built a 4 pack KT120 already with Lundahl LL1693 outputs and cathode follower drivers. It delivers 150W output, and handles my difficult speakers well. I'll be posting the optimised design soon, but suffice it to say it is a straight Mullard 5-20 topology with RC coupling to the output stage and nothing too clever going on in circuit terms.
I've hugely enjoyed doing this, and I'm thinking about how I might build something better. I'm going to experiment whilst listening to the amp I've already made and see what I might be able to achieve. If nothing else I will teach myself more about Spice and tube / circuit theory.
I've hugely enjoyed doing this, and I'm thinking about how I might build something better. I'm going to experiment whilst listening to the amp I've already made and see what I might be able to achieve. If nothing else I will teach myself more about Spice and tube / circuit theory.
I am working in the last years with a particular output trafo made custom with these characteristic:
for a push-pull of KT88/KT120: primary 3125 ohm, secondary 5 ohm, ratio 25:1, pentode connection; is a C core laminate 0,1mm
The last amp I have test give me 58 w on 8 ohm and 80 w at 4 ohm ( the FB was 18 dB) with a very low output impedance. Bandwidht 20 to 100 KHz 0,5 dB and very good distortion vs. frequency results (very important test)
Of course with two pairs the ratio will be 12,5:1.
This is the way to have a good current delivery for particular speakers where the impedence is varying specially at low frequency.
Walter
for a push-pull of KT88/KT120: primary 3125 ohm, secondary 5 ohm, ratio 25:1, pentode connection; is a C core laminate 0,1mm
The last amp I have test give me 58 w on 8 ohm and 80 w at 4 ohm ( the FB was 18 dB) with a very low output impedance. Bandwidht 20 to 100 KHz 0,5 dB and very good distortion vs. frequency results (very important test)
Of course with two pairs the ratio will be 12,5:1.
This is the way to have a good current delivery for particular speakers where the impedence is varying specially at low frequency.
Walter
Last edited:
Member
Joined 2009
Paid Member
You could go crazy and use paralleled OPTs, where the advantage is to allow use of higher dc-current in total without very large cores that are difficult to keep cool and which require ever larger diameter wire.
I've read this approach is sometimes used in high power guitar amplifiers - but who's to say it can't be done at the hi-fi level ?
I've read this approach is sometimes used in high power guitar amplifiers - but who's to say it can't be done at the hi-fi level ?
Is not true.
Of course it is very strong to build a parallel tube power amp with 2 or more pairs but you can reach a very good results:
http://www.multitask.it/140/apertura1.jpg
This a pair of amps with EL34 with more than 140 w. The out trafo has one winding with 3 ohm and alla the amp is full balanced
The main problem is to get a right trafos.
Walter
Of course it is very strong to build a parallel tube power amp with 2 or more pairs but you can reach a very good results:
http://www.multitask.it/140/apertura1.jpg
This a pair of amps with EL34 with more than 140 w. The out trafo has one winding with 3 ohm and alla the amp is full balanced
The main problem is to get a right trafos.
Walter
Last edited:
Actually, this is easy.
Acquire a Harman/Kardon Citation II and refurbish it ala Jim McShane. Use the 4 Ohm speaker taps and don't look back. A Cit. 2, AKA the "Deuce", in good repair produces over 60 WPC continuously and over 120 WPC instantaneously.
A pair of PL519 / 6P45S give 200 W, so obviously JJEL509 will do the same and it is still in production.
To get 400 W tube amplifier you should have:
- pentode connected output circuit with 4 pcs. JJEL509
- +Ub = 480 V
- regulated Ug2 = 160 V
- 1k OPT
Ordinary LPT can drive these output tubes.
The above is based on my experience with 200 W PL519 amplifier I built some years ago.
Thanks for all your suggestions.
Something I have definitely decided is that when I build these amplifiers I must avoid what I have come to think of as the 'Pandora's box' effect... ie not listening to the amp until it is almost finished past a point where any fundamental change in the design can be made. The PPP KT120s I have built are so large that I have never run a stereo pair through loudspeakers until I have almost finished them. There has been simulation, experimentation to find the phase margin and optimise the feedback etc, I have a good scope and a decent THD meter... and so on... but I never listened to them. Luckily they sound OK! I must find a way of building these amps to allow for listening tests before committing the design.
Thank you for the suggestions so far; please keep them coming! Gingertube's suggestiosn sit where my mind was going ... along with the 'Norman Koren TENA' amplifier, although I might get custom opts wound than use the Plitron's...my plan might be..
Balanced throughout - 2 x LTPs as gain stages, both with CCS and the second one with a choke / gyrator load to allow it to really push the volts. RC coupled to cathode followers (ECC99's or something like them). Outputs 4xKT120 40%UL with 10%CFB (windings ratios needing much study) say around 2K5 load, with 4R / 8R taps switchable.
I have found the Turner Audio website index to also contain a lot of interesting ideaas.
Something I have definitely decided is that when I build these amplifiers I must avoid what I have come to think of as the 'Pandora's box' effect... ie not listening to the amp until it is almost finished past a point where any fundamental change in the design can be made. The PPP KT120s I have built are so large that I have never run a stereo pair through loudspeakers until I have almost finished them. There has been simulation, experimentation to find the phase margin and optimise the feedback etc, I have a good scope and a decent THD meter... and so on... but I never listened to them. Luckily they sound OK! I must find a way of building these amps to allow for listening tests before committing the design.
Thank you for the suggestions so far; please keep them coming! Gingertube's suggestiosn sit where my mind was going ... along with the 'Norman Koren TENA' amplifier, although I might get custom opts wound than use the Plitron's...my plan might be..
Balanced throughout - 2 x LTPs as gain stages, both with CCS and the second one with a choke / gyrator load to allow it to really push the volts. RC coupled to cathode followers (ECC99's or something like them). Outputs 4xKT120 40%UL with 10%CFB (windings ratios needing much study) say around 2K5 load, with 4R / 8R taps switchable.
I have found the Turner Audio website index to also contain a lot of interesting ideaas.
I meant the second LTP RC coupled to CFs direct coupled to the KT120s. I might also think about source followers.
That was your previous project, 4 x KT120
And the new might be... 4 x KT120. Right ?
Ha ha you've half got me there... however the following things would be different. All balanced topology, cathode feedback outputs, maximum use of local feedback.
Or I might go for something completely different. A triode AB2 design is tempting ala 'the Norman Koren TENA'.
Or I might go for something completely different. A triode AB2 design is tempting ala 'the Norman Koren TENA'.
JMO, the OPs 90 dB. sensitive speakers need circuitry that provides a substantial damping factor. The H/K Cit. 2 I previously mentioned has, for a tube amp, an enormous damping factor. IMO, the other design possibilities being considered will come up short.
Even derating the speakers to 87 dB. sensitive leaves a "Deuce" in good repair providing all the power needed and the crucial factor of good driver voice coil control is accounted for.
AR3s and AR9s are not the easiest of speakers to cope with. The Cit. 2 handles them with complete aplomb. About the only speakers I would not mate to a "Deuce" are big Thiels, with their outrageous impedance drops to 1 ohm.
Even derating the speakers to 87 dB. sensitive leaves a "Deuce" in good repair providing all the power needed and the crucial factor of good driver voice coil control is accounted for.
AR3s and AR9s are not the easiest of speakers to cope with. The Cit. 2 handles them with complete aplomb. About the only speakers I would not mate to a "Deuce" are big Thiels, with their outrageous impedance drops to 1 ohm.
HK Citation II - I have seen a reconditioned pair in the country next door for about $5k ... they looked like they had come out of the factory. The next time I am there I will have a listen. What makes the circuit sound so good - a ton of feedback both local and global, very good transformers and some very careful engineering?
In the answer to how much power I want - too much is enough. I have always liked big speakers with proper bass. This always seems to work better with big amps, a good 'littlun' always seems to get beaten by a good 'bigun' ..
In the answer to how much power I want - too much is enough. I have always liked big speakers with proper bass. This always seems to work better with big amps, a good 'littlun' always seems to get beaten by a good 'bigun' ..
Mississipi, love the site, really do. An absolute treasure, I will study it with great enthusiasm.
Parallel transformers got mentioned earlier ... here's an interesting take I found somewhere, put in a notebook as interesting without writing down where I got it from! I understand that it's a Quad circuit but with two identical OPTs rather than a custom OPT ... has anybody built anything like this? The original article describing the Quad 'Amplifiers and Superlatives; - D. T. N. Williamson and P. J. Walker Wireless World, Sept. 1952 claimed lower distortion than triode, pentode or Ultralinear with power output similar to Ultralinear.
Parallel transformers got mentioned earlier ... here's an interesting take I found somewhere, put in a notebook as interesting without writing down where I got it from! I understand that it's a Quad circuit but with two identical OPTs rather than a custom OPT ... has anybody built anything like this? The original article describing the Quad 'Amplifiers and Superlatives; - D. T. N. Williamson and P. J. Walker Wireless World, Sept. 1952 claimed lower distortion than triode, pentode or Ultralinear with power output similar to Ultralinear.
Attachments
The QUAD Acoustical connection I believe does not ultilize two identical OPTs - just one - the cathode feedback winding is wound on the same core as the rest of the windings.
Yes sorry, it's the same fundamental topology circuit (distributed load) but the Quad is CFB from an auxiliary winding on a single transformer.
There is nothing new under the sun and there's a thread on this forum about the type of design I posted with two transformers- see 'Crowhurst's Twin Coupled Amplifier' and this circuit is in some ways a 'poor man's Macintosh' using two transformers rather than a single exotically wound bifilar / trifilar ... although I suspect it would not end up being cheap enough to build to merit the term 'poor man's' at all! Then there is the driver problem with very significant driver swings (300 V+) required. Still tempted however ...
As I started saying, I have a lot to learn. My mind is going down the path of 'if 90% of the distortion in a typical amp is in the output stage then deal with it there' rather than sitting on it with GNFB. This is clearly a path well-travelled and most of the answers seem to lead to other compromises. I need to read more and experiment ... CFB interests me as a way of tacking the problem at source.
There is nothing new under the sun and there's a thread on this forum about the type of design I posted with two transformers- see 'Crowhurst's Twin Coupled Amplifier' and this circuit is in some ways a 'poor man's Macintosh' using two transformers rather than a single exotically wound bifilar / trifilar ... although I suspect it would not end up being cheap enough to build to merit the term 'poor man's' at all! Then there is the driver problem with very significant driver swings (300 V+) required. Still tempted however ...
As I started saying, I have a lot to learn. My mind is going down the path of 'if 90% of the distortion in a typical amp is in the output stage then deal with it there' rather than sitting on it with GNFB. This is clearly a path well-travelled and most of the answers seem to lead to other compromises. I need to read more and experiment ... CFB interests me as a way of tacking the problem at source.