• WARNING: Tube/Valve amplifiers use potentially LETHAL HIGH VOLTAGES.
    Building, troubleshooting and testing of these amplifiers should only be
    performed by someone who is thoroughly familiar with
    the safety precautions around high voltages.

KT88 Parallel Push Pull

Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.
I am thinking of building (starting with breadboarding) the attached design as monoblocs.

I'm interested in this because:

1. I need a lot of power (for big three-way monitors)
2. I am not a tube newbie (I have scratch built a ST70 using the Welborne Mod) but equally I am not a guru and want a circuit I can understand.

I am going to use a Sowter UO71 for the output, and a custom built power transformer with 360V centre-tapped at 500mA for 450V b+ with solid state rectification, 8A 6.3V for the KT88 heaters, a heater winding for the driver, another heater winding for the elevated heaters on the cathode followers and 60V for the bias.

Any advice, suggestions for improving the schematic etc? I was thinking of solid state current sources for the input stage plate load, and the LTP 'tail'.

It's going to be a big one!

HK13 Design.jpg
 
Welcome to the serious power bunch.

I am currently completing a 2x225W monster using 8 x EH KT90's for a commercial application; two power amps are on the cards...a programmable 300/600 B+ equates to 50W and 225W o/p power resp. The second amp is an even higher power of 250W although the differences are minor. The weight is not minor....60Kg as 15Hz is required.

The application is wide b/w hifi studio amp with programmable MI power droop when required. There is more to this but for the moment;

Circuit diagram currently being modified but the audio technology is simple. A concertina front end with a slam Williamson wide B/W low Z driver with a sep power supply; but the pfc power supply is the "black end" has to supply 1kW. 8x EH KT90's sim driven in bridge tied mode will suck 180mA each at the 225W level requiring a kW headroom. This will test the psu.
For those "critics" wanting to bash the concertina phasesplitter, there is technically absolutely nothing wrong with it.
The pfc power design has to be properly tested for every worstcase scenario....notice the large double dummyload resistors in the black box on bench centre left....a relic from the breaking circuit from a previous generation alpine railway locomotive.

You are wise staying with 500V electrolytics......600V is an awkward voltage requiring "series up", as the same goes for semiconductors, 1000V is the next commercial segment.

Enclosed is a graph of expected conditions of another of my amps, sim well as per your conditions. The 6550's JSC are better than the TungSol new ed versions. Soon after 2000 hrs running, the TungSols "greyed" badly and soon after one failed. Parallel p-p has awesome power throughput capabilities, a watchdog bias fail-safe circuit is vital to protect the circuitry as is sufficent fusing. Use wirewound 10R cathode resistors for the o/p stage....ony others, m/F or m/o are useless with transients and will creep high.

Heed the mechanical layout of the tubes regarding heat and proximity to electrolytic caps which for every 10°C rise, their life halves. Also consider the heater wiring, the power tubes furthest away will see a drop, depending on the CSA of the heater wiring......8A is a bit low considering other tubes are also down-stream.

As the power goes up, so does the project complexity as in may case as a commercial project requiring a large amount of documentation with graphs and test results. It will test ones knowledge.

Expect a superlative performance.

richy
 

Attachments

  • EH KT90& JSC 6550 winged C thd comp.JPG
    EH KT90& JSC 6550 winged C thd comp.JPG
    34.6 KB · Views: 2,696
  • lab view with new amp.jpg
    lab view with new amp.jpg
    766.2 KB · Views: 2,896
I am thinking of building (starting with breadboarding) the attached design as monoblocs.


Some comments:-
When using LTP as a driver which has a higher output impedance than the Williamson driver, be careful to keep o/p grid/ component wiring reasonably symmetrical as the output stage frequency response has been improved by the lower Ra A-A impedance hence better bandwidth. This still implies the unscreened first stage be kept at a fair distance from the large electromag emitting o/p stage anodes.
In very brief; an amplifer that responds to a 15kHz signal beit harmonics implies the 3rd harmonic must not droop by 3dB at 45Khz. So the amplifier response bandwidth is vitally important to fulfill the slewrate rise time criteria. There may not be much to be heard at 15Khz, but sound imaging and definition are openly claimed benefits.

Design.
Understanding all the following stuff...There is the all important Fourier analysis correlation between open loop gain,phaseshift, global feedback and stability which all interact. The worst is HF instability cause by reckless birdsnesting when one is looking for the faults elsewhere and have to damp the circuit to get it to behave.
One has gone to the expense & trouble of 4 power tubes per amp, so one might as well put them to the best use for a particular o/p transformer.
So layout plays a big part and the phaseshift plot (should be smooth) as in pic provides the cue... all these are essential for a successful new design.
In the 2x225W power amp, the response is -3dB at 55Khz,; -6dB 90Khz and although the front end ECF80 is unscreened, some very slight interaction is detectable with the output stage which is roughly 6 inches/150mm away.
Referring to Pic. The parallel push-pull o/p stage.The 4 black grid coupling capacitors with short 5K6 stoppers are symmetrically located nearest to the valve holder quandrants, the orientation keeping the anode and screens furthest aside; The two thick purple wires go to the phasplitter outputs with 33K/39K resistors near the balance pot. The grid bias adj pcb for one half o/p pair can just be spotted uppermost near the rectangular tinned copper bus at braid. This copper bus is 4mm thick and earthed only at the amp input. It is thick for a reason.
These days alot of expensive components have only a couple of layers of insulating tape to protect the internals......Purchase some Nomex tape and a couple of turns on plastic components to avoid the likelyhood of soldering iron scorches.

I mentioned previously about the o/p stage cathode resistors being wirewound. I have been asked why. A square wave of each half (biphase) o/p stage anode illustrates the problem as in pic; the o/p transformer leakage inductance spike has alot of short-term energy which m/o and other resistors types cannot handle. Designing high power amps with large output transformers invariably leads to higher parasitic losses and the components have to handle them.
The Radiotron handbook covers most of the full theoretics.

Happy building !

richy

--------------------------------------------------------
Thinking is the hardest work there is, which is probably the reason why so few engage in it.HENRY FORD
 

Attachments

  • EH KT90 anode waveform.JPG
    EH KT90 anode waveform.JPG
    35.2 KB · Views: 2,389
  • output stage comp sym.JPG
    output stage comp sym.JPG
    119.2 KB · Views: 2,337
  • phaseshift graph.JPG
    phaseshift graph.JPG
    56.7 KB · Views: 2,294
  • open loop gain response.JPG
    open loop gain response.JPG
    55.8 KB · Views: 346
Thanks for your reply Rich:

I wish I had the time / skills to build the type of amp you're putting together - maybe if I keep learning for another 20 years...

Some of the things you mention were factored in my thinking, others not..


1. I could (should?) get very seriously math'd up, but I'd have to relearn things I forgot 20 years ago and my brain has a day job ..! I’m a conservative engineer from another discipline so by instinct my plan for getting to a final design is one stage at a time carefully, measuring and understanding what’s going on.


2. I have a ‘hankering’ to try a ‘Swenson’ regulator with foldback limiting. However initially I will build the PSU with the simple choke-based design per the schematic. I have decided to stay below 500V B+ with a safety margin for startup spikes etc for two main reasons -

Firstly I don’t like the idea of stacked capacitors. As a student I had a job as an amp tech (or to be precise a ‘gofer’ for a proper one) and a classic failure mode was one stacked cap going bad and then taking out the other.


Secondly as you say above 500V cost moves to another level for most things.

3. I intend to stick to fuses and an inrush current limiter in the main PSU - measure the inrush with a digital scope and fit the right PTC. When breadboarding the amp my bias watchdog will be the variac and something that says Fluke on the side! In the longer run - what’s a good simple scheme for a watchdog in the finished amp? I think it should be take out the HT on the AC side of the bridge (to avoid trying to break the DC B+) in some ‘latched’ type scheme requiring a hard reset if the bias voltage does not come up. I don’t feel comfortable with a microcontroller type supervisor or think it’s necessary – I was thinking about a comparator and relay probably powered from a separate small transformer.


4. In general I intend to breadboard the amp from the output stage back. I have a plan to use a small spare PP output transformer driven from a low impedance function generator to drive the grids of the output quad to explore phase / frequency / distortion response of that stage. I have KT88’s to play with initially – but I’m very interested to see your results on the 6550. I have a long term loan of a Neutrik XL2 so I can do decent distortion measurements. My basic strategy is to, subject to keeping the circuit reasonably simple use good quality components, keep the distortion down and minimise the NFB.

The Sowter UO71 is very serious if you are not familiar with it – spec’d for 200W at 20Hz with flat response out to 50Khz. So hopefully there won’t be a requirement for seriously slugging the driver etc and I will be able to listen for the things you can't hear any more at 50 years old.


5. I am indeed worried about stability at these power levels both on the breadboard and in the finished amp. I am not a bird’s nest kind of person – when I built my ST70 I read, digested and acted upon Morgan-Jones ‘Building Tube Amps’ book, with short screened connections to the grids, stoppers directly mounted on the sockets, etc. I used lots of heatshrink in that to support connections to sockets & switches, P clips to support B+ wiring etc etc– the way I see these things is at these voltages and currents you have several lines of defence against mechanical problems. I shall add Nomex tape to my armoury – thanks for the tip.

Take a look at the attached picture of the ST70 part way through testing on the breadboard – this might reassure you I have some chance of avoiding building a very expensive AM transmitter.

When I breadboard this amp I am going to do it on turret board – take a look at the reusable turrents on www.ampmaker.com – brilliant IMHO. I have a possible eventual plan to design a PCB for the bias power supply and driver stages .. more reading required I think.


6. I found out about the need for wirewound cathode resistors again in my amp tech days. I was told if the output tubes flashed over MF cathode resistors would frequently leave conductive residues inside the amp and things would keep cooking or somebody would get shocked, whereas wirewounds would give a clean fuse break.



Should I buy a fire extinguisher ...?

ST70 Breadboard.jpg
 

6L6

Moderator
Joined 2010
Paid Member
Your high-power schematic looks very intriguing! One thing worth mentioning, concerning Japanese magazine schematics, is to use the specified iron. Which, almost invariably being Tango and/or Tamura, is breathtakingly expensive. However, it greatly simplifies the endeavor.

BTW, I have a Kimmel/Welborne ST70. It is an outstanding sounding amplifier. If there was a way to scale that amp to about 80 watts, I would do it in an instant. :D :D :D
 
xpersephone:

A rapid way to pick up.....alot of oscilloscope manuals have alot of tech tips and clues,
I learn't from the Cold War service REME; which was invaluable source of trade notes for just about everything technical and with it masses of equipment. Books, Radiotron Handbook 4th ed, and more recent Morgan Jones.
We all go through blowups...ear defenders and eye protection for HPower SS stuff is wise.....600V and 1000uF uniting in TO220 & 247 packages, the effect is a small grenade.

costis:

Decoupling Yes there is.....same as O/P stage on the HT.

A short note regarding transformer locations and proximity. The chassis up and under is cramped.
Examine pic, a tape ruler shows the distance between the mains transformer and both output transformers; the mains transformer is at 90 degrees to minimise mag field leakage into the o/p tranny iron and these are large transformers.
The nub is,with this proximity (unworkable for SE) there is some hum pick up on start up; one can hear it low in the speakers, but as the amp warms up the global feedback loop 20dB does a brilliant job in reducing it to nothing.
The benefits of global feedback are collosal in that not only for the audio side but also the power supply ripple rejection aspects (PSRR). A spectrum of the 2x200W amp at both channels driven at the 150W output level shows the astonishingly little amount of 100Hz sideband present.

Working with power is an exercise in doing it with the minimal amount of space. It will train one, to trade off one against the other.

richy
 

Attachments

  • transformer prox.JPG
    transformer prox.JPG
    115.4 KB · Views: 466
  • spectrum150W.JPG
    spectrum150W.JPG
    187.5 KB · Views: 426
6L6 - I am actually going to use Sowter transformers; a British brand with an excellent reputation. Not quite in the Japanese exotica territory but the UO71 is spec'd for -3dB at 60Khz. I am aware in general terms that the change in output xformer will mean:

1. I should produce a Bode plot and do some relatively straighforward calculation set the value of the capacitor in the NFB loop to ensure ir is stable. Rich : I will aim for 10-20dB GNFB.

2. I may also get more local problems causing ringing - I have read a number of good articles on how to suppress that and I'll ask for more advice here if I can't sort it. I had ringing on the ST70 I built at around 10khz and at about 20% of the amplitude of a sqaure wave at 1kHz. I could certainly hear that (as harshness) and squashed it with a Zobel network.

I certainly will start testing this at low power and reduced B+ .. and the safety glasses are always on. I read a great piece of advice on one of these forums that building tube amps was like deep sea diving - there are few trivial accidents. When I first built a ST70 a few years ago I have to admit to smiling at the idea that literally hundreds of thousands of people assembled these on their kitchen tables in the 60's. It must have been a considerable Darwinian influence on the population ...
 
Finally I built something!

Here’s where I have got with my design...

1. Having read a lot of posts on these forums I decided to use 6SN7s / 6SL7s in a Mullard topology.

2. Decided now much gain I would need from the driver (basically 1.5V RMS from a modern D/A converter pre-amp to give about 120V P/P out of the driver to take PP KT88’s at 480-500 V B+ to clipping). I then discovered (via LT Spice sims) that anything with a 6SL7 would have too much gain. So I decided on a 6SN7 triode gain stage to a 6SN7 LTP and 6SN7 cathode followers.

3. Having decided to use cascade CCS for the triode input stage and LTP of the Mullard topology I played with DN2540 cascodes. I had problems with getting stable DC performance from the classic circuit posted many times on this forum. Having read around this subject and excluded oscillation as a cause of the problem I discovered that the bottom MOSFET in the cascode should have lower Vgth for the degenerative feedback in the cascode to work properly. All of my DN2540s were from the same batch and all had very similar Vgth of about 2.6V. When I substituted the BSP135s for the lower MOSFET (these have Vgth around 1.4V) my DC stability problems went away.

4. I then breadboarded something very close to the schematic posted, but without the feedback loop or the zobel network. Open loop response was -3dB at 10Hz and 72kHz, but with gain rising again to another broad peak at 120-150kHz. I mention this because although sine wave performance was good (125W RMS 2kHz at clipping, with less than 1% THD at 50W with no feedback) the amp rang like a bell when driven with a fast 5khz square wave. The ringing was 50% of the square wave amplitude at 130khz and not damped out within a half-cycle. Output transformer is a Sowter UO71 and the KT88s are Electro Harmonix.

5. The ringing was almost totally eliminated with the zobel values shown (thanks SY for a post which pointed to putting these from plate to screen). 630V orange drop caps in this network exploded (literally) and 1W resistors smoked at high outputs (yes I admit 50W of 10kHz square wave may not have been a good idea but..). Big 2kV Arcotronics caps and 2W resistors work well. I may still optimize these values vs the HF roll-off caps in the first stage.

6. I am now pretty settled on the schematic shown, but have posted it looking for suggestions for improvement from the gurus. Having applied 18 or so dB of GNFB it gives 125W at clipping from 480V b+ medium power distortion (I just have not gotten round to measuring it at high powers) is good – I measure -60dB THD+noise at 2kHz. Frequency response is -0.8dB at 20khz and -3dB at 60khz. Input sensitivity is 1.5VRMS to fully drive the output, and square waves now show minimal overshoot. I’ve run the amp for long periods at high power with music inputs into a 8R / 0.47u load with no signs of instability.


I am currently running the bread boarded design from a lashed up variac to power transformer to CLC filter power supply.
My next steps are going to be to design a regulated PSU for B+ and bias.
The posted design shows DC heaters for the input tubes – my current plan is to run an output protection circuit from this, I will use comparators to sample the voltage across the KT88 cathode resistors and switch off B+ it exceeds a safe value.
 

Attachments

  • kt88 ppp v5.png
    kt88 ppp v5.png
    38.4 KB · Views: 1,984
Regulating B+ in high power push pull amp?

OK – I will use small wirewound resistors for the 10R cathode resistors.
I can't see a need for grid stoppers on the 6SN7’s – during extensive testing I have not seen the slightest signs of instability without them.
I am planning to build these as monoblocs – separate power supplies might be a step too far domestically. I do plan to keep the PSU very separate in construction terms.
My idea of what the PSU needs is:
· 6.3V @ 1.5A for the triode gain stage and the LTP ground referenced.
· 6.3V @ 0.7A floating for the cathode follower elevated heaters.
· 6.3V @ 8A (to be on the safe side) for the KT88 heaters
· 80V @ a few mA for the bias circuits
· 480-500V @ 0.6A for B+ - I need 230mA quiescent current, and twice that for peak output. Add 30mA for the preamp stages and a margin for headroom.
· At this power level I am very concerned about burn ups and whilst I don’t have the skills / time to build a microcontroller watchdog I intend to implement fail-safes.
· I am planning DC heaters on the preamp stages.
· I have read these forums on B+ delays and the various 'cathode poisoning' controversies and do not intend to implement this.
· I will softstart the PSU overall using a NTC thermistor on the mains.
My implementation plans are:
· The power transformer will be custom wound something like 6.3V 2A, 6.3V 2A, 6.3V 8A, 60V 0.1A, 380-0-380V 600mA
· Watchdog circuits – I plan 4 off comparators looking at the cathode resistors. These will shut down B+ via either a MOSFET driven by a high-side driver or by shutting down the regulator if I do go for completely regulated B+
· I am currently playing with the Gary Pimm / Swenson high voltage regulator. This gives me the option of shutting down the B+ with an optocoupler (look at the Gary Pimm circuit and replace the current-limit transistor with the optocoupler). This requires drop out voltage of 25V minimum. If I double that to take account of ripple / transformer internal resistance at 600mA out this means 30W of dissipation in the regulator – entirely doable.
· I am planning to use LT1185’s to regulate the 6SN7 heaters, and current limit them for soft start of the heaters.
· I think the TL783 looks like a good option to regulate the bias, which I will have to do if I regulate B+.
The $64000 question to me is if it is worth regulating B+?
 
Got any square wave pics ?

well done.
Don't force the issue with 10Khz square wave at power..remember squarewave harmonics are rich and one is asking the large o/p transformer to punch through the 3rd harmonic at little loss; this depends on the leakage parasitics of the transformer versus the output stage primary impedance. These are much higher than for smaller transformers. The first victims to be punished will be the series screen resistors of the o/p tubes as more energy is pushed through them as o/p stage inefficiency rises vs. frequency.
Expect a squarewave similiar to the pic; although I used 6550's new edit which weren't up to the job; but used a sim output transformer.
Be careful where one places the osc scope probe earth on high power square wave measurements; as multiple equipment earths will display untrue waveforms.

The Majestic output transformer/amp at 150W per ch using EH KT90's is sim except I used a Williamson diff driver... expect sim curves if circuit is optimised.


richy
 

Attachments

  • 10Khz 50W 8659 6550-12BY7-ECF80.jpg
    10Khz 50W 8659 6550-12BY7-ECF80.jpg
    31.9 KB · Views: 966
  • Majestic D5301-02 Modified op trans THD vs Pout.JPG
    Majestic D5301-02 Modified op trans THD vs Pout.JPG
    52.3 KB · Views: 886
  • spectrum 150W 4xKT90.JPG
    spectrum 150W 4xKT90.JPG
    68.1 KB · Views: 848
Here's a square wave

Have you built this amp?..

I certainly have built it to the prototype stage - it took me six months to get this far (alongside my full time job) and a lot of reading of this forum! I've attached a picture of the amp breadboarded to give you some idea of what you are getting yourself in to - what you can't see on the picture is the temporary power supply containing amongst other things a 8kg power transformer, a 4kg choke and some pretty sizeable capacitors on the metal rack to the right of the driver and output stages.

I have just finished designing a PCB for the driver stages, and am currently having that fabricated. This is so I can concentrate on the PSU, and test it together with the amp in a reasonable amount of space - I don't have enough space to breadboard both at the same time! I expect the PSU to be as complex as the amplifier...

Got any square wave pics ?

Attached is a 10kHz square wave output into a 8R + 100nF load - higher powers do not look any different. This is from the exact schematic posted. I'm not going crazy about the square waves other than as a way of deciding when the amp is more or less damped right - I recognise that any big output transformer has it's limitations. a completely clean 10khz square wave would imply the kind of HF response that my reading clearly tells me does not come from such devices.

I was quite proud of the fact that my messy breadboard only had a few mV hum.

I've torn down the amp now as I am breadboarding the PSU, but the distortion products from a 2khz sinewave at 50W output into 8R were:

2nd - -38dB
3rd - 48db
4th << -55db

I'm getting there with your help!!
 

Attachments

  • breadboard.jpg
    breadboard.jpg
    81.2 KB · Views: 885
Your'e getting there fine: the technical learning curve is considerable. I could go further, by doing the open loop curves, plot what's actually happening vs. frequency with the response etc.
The next job is to put all what's loose on the bench in a box, make it look good....I chose aluminium with reinforcements....others choose different. However, it's when the ferrous components get close then proximity & technical problems can begin. The photo section of this tubes forum has many build examples.

Here is my 250+250W Williamson...completely stable and virtually hum free; some top and under pics. This amp is an extremity; it can be used in bridge mode i.e 500W o/p; however there are caveats....the power supply (PFC) design gets complex and the weight of the whole amp doesn't come light.
65kg /143Lbs if one wants to go down to 15hz as this one does. Notice the "tennis court" matrix earth bus...(traditionally many amps in the past were done like this) with tinned 4mm copper wire. The output stages is where the currents flow and the "court matrix" effectively neutralises them from the input sections. Earthing is only at the input sections and where possible a symmetrical layout used for both channels. As the power goes up, so does the protection too.


Keep it up !

rich<y
 

Attachments

  • DSC06721red.jpg
    DSC06721red.jpg
    621 KB · Views: 582
  • DSC06730 red complete.jpg
    DSC06730 red complete.jpg
    329.6 KB · Views: 676
Last edited:
Status
This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.