Category Archives: 1 – KT88 Construction

Fine-tuning the NFB

Having lived with the big KT88 amplifier for the last few months and been mostly happy with it, there were still some niggles that I resolved to get around to. 

Specifically, the amp ran hot, and needed more ventilation holes drilled. 

Also, there wasn’t enough Negative Feedback (NFB). I wasn’t too worried about this until I built the EL84 amp featured elsewhere on this blog, which had more NFB. On hearing the difference, I resolved to correct the situation in the big amplifier, but this would need some equipment I didn’t have.

So, first up was a shopping trip online to get some power metal film resistors for a dummy load, it’s very important to have a non-inductive load for tuning NFB. These were duly mounted to a large heatsink.

Next I needed a pair of old-style variable capacitors, the kind you’ll find in an old valve radio. eBay to the rescue, and these eventually came all the way from Bulgaria.
The method I intended to use for fine tuning the NFB was from Morgan Jones “Building Valve Amplifiers” p.290-291. 

Today, I managed to get the amp back onto the workbench. Pulled out all the valves and attacked it with the drill, to make some new ventilation holes. Problem 1 fixed, and it remains to leave the amp on for several hours to determine its effectiveness.

Problem 2 was also resolved, though this was a good deal more time-consuming. Utilising Morgan Jones’ method, and armed with a healthy stock of film capacitors of various values, I started making the necessary modifications to the circuit, first with potentiometers and variable capacitors, before subbing in fixed components.

First order of business was to reduce the NFB resistor from the (fairly useless) 100K to something lower. After experimenting with the input sensitivity, I dropped this to 33K.
This got the amplifier’s gain to where it needed to be, and eliminated the problem of the very touchy volume control.It did however introduce another very serious problem: high-frequency ringing. The Williamson design is prone to this, and adding NFB in any quantity will exacerbate it.

This was the result (red trace) at the speaker terminals of dropping a 10kHz square wave into the amp, after reducing the NFB resistor from 100K to 33K

Yeah 🙁

Don’t know about you but I don’t want to listen to that amplifier like that. Apart from anything else, it’ll set all the dogs in the neighbourhood howling. And things will be getting mighty hot with all that high frequency energy to dissipate.

So clearly some compensation needed.

So watching the trace on the scope, using Morgan Jones’ method, I arrived at these changes to the circuit:

Added compensating capacitor and resistor parallel to the anode resistor in the first gain stage
Dropped feedback resistor from 100K to 33K
Added compensating capacitor and resistor parallel to feedback resistor

This was the result:

Speaker trace in red

Yeah, I forgot to clip the CH1 probe back on to the input. No matter; it’s the same signal at the same amplitude.

So far so good, this is all textbook from Morgan Jones. However in the course of my experiments I discovered something else that Morgan Jones apparently neglected to mention which I pass on here in the hopes that it may help someone.
Specifically – Jones’ method calls for the feedback resistor to be bypassed by a variable capacitor and resistor, which I did, and I noted that the resultant waveform at the speakers looked pretty much like the above already. 

Thinking I wouldn’t end up needing anything bypassing the anode resistor, I acted on Jones’ recommendation and put a 220nF capacitor across the speaker terminals as a test, and watched the output go absolutely crazy. It looked much worse than the amp with no correction at all, and in fact was on the very edge of falling over into uncontrolled oscillation.

I then decided to bypass the anode resistor in the manner suggested, this resulted in some fine tuning of values as these are all inter-dependent. Eventually I got it to approximately the same level of cleanliness on the output as I’d seen with just the feedback resistor bypassed,

Then I tried the capacitor-across-the-output trick again.

Result: The amplifier barely even noticed the capacitor. A complete fix of the problem 🙂
Conclusion: the anode load resistor bypass doesn’t do much to alter the oscillation into a resistive load, but it makes the amplifier much more stable into a capacitive load.
Morgan Jones did not mention this anywhere I could find.

So for reference this is the circuit diagram of the amplifier now (click to see full size)

The power supply implements the timer circuit which is not shown here for clarity, refer the circuit diagram of the EL84 amplifier on this blog for details on that.

This is the last modification or fine tuning I expect to make to this amplifier.

Morgan Jones “Building Valve Amplifiers”, Newnes press, 4th. Edition. pp. 290-291

Name Plate attached

The courier brought my name plate this morning, I fastened it to the amp and it looks as good as I had hoped:

Looking good with the name plate in place. The image was made by tracing around outlines of photos of the valves in Photoshop. 

Not sure I’m too happy with the placement of the LEDs for the input indicators, but too late to do anything about that now.

While I had the amp off the cabinet to screw on the brass plate, I also decided to take a photo of the rear panel, since for some reason I hadn’t done that yet.

The business end. 4-Ohm  and 8-Ohm taps for the output. USB power for the Chromecast. Switched mains output for the separate phono preamp.

Remaining to be done now – just the 3D-printed cover for the power transformer.

The amplifier has been delighting us with its performance and power; despite its tendency to run rather warm, it seems entirely reliable. Its longest runtime so far has been around 6 hours, after which I looked inside carefully to see if there were any signs of thermal distress. Happily, there were none.

Current through the KT88s are stable – I’ve biased it at 28W idle disspiation per valve.
The amp is driving the big 4-Ohm KEF floorstanding speakers with commanding power and finesse at the same time. Noise level (hiss) is negligible, slight hum is audible with ear glued to speaker but 1m away is inaudible. Clarity and separation of the signal is excellent, bass is strong and authoritative and in no way flabby, nothing dominates at the expense of anything else, and when you need the power, there’s ample headroom. We’ve had the walls rattling with this thing already.

Sadly the mains transformer is physically a bit noisy though, hopefully the cover will muffle that a bit. When the music is playing you don’t hear it anyway. 

The transformer slowly gets warmer and warmer as well. Haven’t had it too hot to touch, but definitely on the warm side. 

The remote control and input switch and remote power switching all work well.
I am going to call this project a success.

On to designing the next one….

Uh-oh! First teething problem

So after finishing the build and moving the amplifier to the living room to use, things were going very well. One reliability issue came up fairly early… a low-level but consistent crackling, ticking through the speakers. At first I thought this was due to power supply interference, but it persisted after I installed a mains filter.

Then I started gently tapping the output valves. Bingo! Second KT88 on the left channel, slight wiggle in its seat and the problem immediately stopped.

I remember during initial power-on that two of the valves were drawing no current whatsoever, which I rectified by replacing the contacts in the valve seats from some donor ones.

This particular valve seat didn’t receive donor parts, so it’s going to be worth keeping an eye on it in future.

For now, though, the problem has gone away. But what goes away by itself can come back by itself too. I can see valve seat contact replacement being necessary in the future.

Second problem was a little more dramatic. After the first run, in which I had the amplifier running for a couple of hours, I was happy with the reliability. The next day I ran it for about six hours, and apart from getting rather warm (to use a pleasant euphemism) there were no other issues and the bias voltages remained stable.

Then I turned it off and left it to cool for a couple of hours, before lifting it off the cabinet and opening the bottom, to take a look at whether anything was showing any signs of thermal distress. No evidence to be found, so yet again I was happy.

Then I plugged it back in again and powered it up. After the 30sec delay the HT came on, to be accompanied by a loud hum from the right speaker and the mains transformer, and a very unhealthy blue glow from one of the output valves on the right side.


Having a fairly good idea what caused this, I quickly removed it from service and had the bottom off again, and started probing with the meter, long story short my hunch was correct and the valve in question was for some reason not getting its bias voltage any more.

C’est un Catastrophe!

So a bit of probing with the multimeter quickly revealed the problem. Three days ago, when I replaced the resistors in the bias section on the power supply board, I’d made a poor solder joint. The repeated thermal expansion and contraction had stressed it and it had let go.

So I quickly removed the board, took it to the workbench, re-soldered it and tested, then replaced it – the work of five minutes. Re-assembled and tested and working happily once more.

Hopefully the one and only fireworks incident!

(Almost) the end of the road

With the arrival in the mail today of the Emergency replacement KT88 output valve to replace the one with the internal short, we now have a milestone: a working amplifier
The amp has been sealed up and moved from the workbench to the living room, and connected to the big floorstanding KEF speakers, wired up and biased up, and some listening tests have been successfully conducted.

The moment I’ve been looking forward to for months, since I started this project.
What did the listening tests reveal??

First, this amp has plenty of power. The volume control hasn’t been advanced beyond about 20% and at that level I was anticipating howls of protest from the Teenager upstairs when the parents’ ancient music invaded through her headphones. Such howls were not forthcoming however, but I did keep the high-volume tests to a fairly brief period in the interests of domestic relations.

Second. There’s a bit of hum, as completely expected with valves. If you put your ear up to the speaker you can hear it. If you stand about a metre away you can’t. I am happy enough with that.

Third. The mains transformer makes a bit of physical hum… I had already noticed this when I loaded it up on the test bench. This is by far more more noticeable than the hum from the speakers.

Fourth. The sound quality is exactly what I was hoping for. In an hour of burn-in listening we were delighted with its imaging and separation and clarity.

Fifth. These KT88s get HOT! They’re biased up to 28W idle disspiation and they will readily scorch unsuspecting fingers. The rest of the chassis gets warm, but nowhere near too hot to touch.

The build process is not quite complete however. The brass nameplate is the next item on the agenda, that will be in the centre under the row of indicator lights. Also I might make a few aesthetic changes yet, and the 3D-printed cover for the butt-ugly mains transformer still needs to be designed and made.

But for now, we have an extremely powerful yet easy-on-the-ear home designed and hand-assembled amp to listen to. Win!

A few photos of the usable (I won’t say finished yet, until the aesthetics are completed) product:

Milestone: Working Amplifier (kind of)

Today I have finished the build phase of the amplifier, and it’s working… kind of.

There’s a problem however – one of my KT88 output valves is no good. It has a short-circuit between the control grid and the cathode, which only manifests when the valve is heated.

So this means I have three functioning output valves. So until I can source a replacement (and these things should really be bought as a set) then I can not use this amplifier other than one-sided.

This wasn’t the only problem I hit during the construction of the output stages either. It turns out my valve sockets for the output valves were a bit less than ideal… when powering up, two of them were drawing no current at all. Some of the pins were not connecting properly. This necessitated procuring and installing replacements, with all the attendant re-soldering and swearing that entailed.

However, each channel now works perfectly with a functioning pair of valves. Watch as I power the amp up, feed some signal through it and listen to it through a tiny little monitor speaker. Then stick around for a tour of the internals

After this successful test I re-connected the dummy load and connected the output to the oscilloscope and ran the signal generator into the input. Then I increased the volume until THD reached 1%

The result is that this amplifier will deliver 80 watts into a 4 ohm dummy load at 1% THD. I am very happy with this result.

A few photos of the completed internals

Worth mentioning one item was left to experimentation – the negative feedback. By luck I got the phase correct the first time around, and I used a potentiometer to determine the optimum level of NFB, while watching on the oscilloscope. The correct point became very clear in short order: too low resistance and the feedback was visible. Too high and noise waveforms were visible. Just right and neither were present. That Goldilocks point came at 100K, so that’s the value of the NFB resistors I used.

There’s still some work to complete before the amp is completely finished… specifically, the input selection LEDs and remote sensor need to be mounted properly, the way they’re set up at the moment is a bit rude.

Also, there’s going to be a brass nameplate on the front, that needs to be designed first. Once these two parts are done, the amp can go the right way up and migrate to the living room

But first, before any of these steps, replacement output valves are needed!

Construction, Pt. 3

A good amount of progress has been made since the last update:

  • The output transformers have been mounted
  • Output transformers secondaries wired up to speaker terminals
  • The volume / input selector / remote power-on switch has been wired up to the standby transformer
  • The USB power supply has been wired up to the standby transformer
  • The Bias pots have been mounted to the chassis
  • The driver valves (2 x 12AU7) have been wired up and tested

The books I’ve been reading (mainly by Morgan Jones) advocate a big-bang approach to first switch-on – complete with an unintentionally hilarious guide to the smells you may encounter if something goes wrong:

Contrary to this advice, I’ve been powering up the circuit in stages, as each is built, I’ll add dummy loads to sub out the following unbuilt stages, then apply power and check voltages and signal levels on the oscilloscope.

So far there has been no drama – for which I credit the prep work I did in building the test rig. When I powered up the circuit today to check the 12AU7s, everything just worked… voltages, signal levels – all exactly as per my test rig.

The only real difference, other than there being twice as many valves (two channels… this little thing called STEREO…) is that I’m trying – as best I can – to make the wiring tidier than the haphazard mess that was the test rig.

So – today’s photos:

Preamp section is complete, the coupling capacitors can be easily seen
Note component leads bent at right-angles wherever possible, and heatshrink to avoid accidental shorting

Next step will be to build the output stages – wire up the bias supply and the KT88 sockets, then power it up and see what sort of current we’re pulling through them. That will be an exciting day because – unless the amp goes into crazy oscillation (always a risk with the Williamson topology) that will be the day I could theoretically put a signal through it and connect it to a pair of speakers.

I’ve decided I’ll get a set of cheap disposable speakers from the auction site for the first actual test, I don’t want to risk this thing going unstable and pummelling my beloved KEFs with the best efforts of an angry pair of KT88s driven at 560V plate voltage.

Stay tuned… 

Construction, Pt. 2

Owing to other commitments that can’t be avoided, at the moment I can only snatch 15 minutes here and there to work on the amplifier. Frustrating but that’s life. Each 15-minute stretch is divided up into two halves. The first half is spend pondering where to put things and how to make everything fit and look as elegant as possible.

The second half is spent trying to make that happen. A Master Craftsman I am not, so it often happens that something is 2mm too long or too short, necessitating rework.
However, despite the time constraints, there’s some progress. A couple of phone pics because I didn’t have time to drag the SLR out and do it properly, but since the phone has a Leica camera on it, I think phone pics are sufficient

Trying to be as tidy with the wiring as possible. Using solid (single) core since this stuff is bendable and stays where you put it
Trying to be as tidy as possible with the component leads. My needle-nose pliers make the right-angle bends in the resistor leads easy

As construction progresses, it will become apparent whether my chassis design is too cramped, or well thought out. So far I haven’t hit any major issues, putting all the power supply components onto a PCB has certainly been convenient. I am really glad I went through the learning curve of how to do that.

There’s a good reason I want this amp to be as tidy as possible: apart from being somewhat of a perfectionist, this has a practical aspect: It’s likely this will not be the last amp I build. In fast I already have another that I’ve been commissioned to build after this one, so this one has to sell my amp-building abilities. 

Construction, Pt. 1

No intelligent thoughts or ramblings or decisions in this post, just a few photos of the construction process.

Internals. Heaters all wired up, taking care to ensure the same phase of each is connected to the same terminals on the output valves, thus cancelling any heater-induced hum. Tip from Morgan Jones. Power Supply board is my own creation, see earlier post. Small mains transformer is for the standby circuit (remote control) plus it’ll power the USB supply for the Chromecast on the back panel 
Cathode earthing for the output valves. These will be running Fixed bias, so the 10Ohm resistor is for measuring anode current
The top of the chassis. Featuring the world’s ugliest mains transformer (eagerly awaiting a decorative 3D-printed cover). Note bias test points behind the Octal sockets
Looking pretty with all the glassware installed and glowing
These KT88s have a fairly discreet glow, no great exposed heater piping here
Easily the most beautiful electronic thing I have ever constructed
My horrible small messy cramped workbench. The computer doubles as my oscilloscope. Note amplifier is doing a headstand, the mains transformer is so heavy that it’s quite stable sitting like this.

Home Stretch

Today’s entry marks the beginning of the home stretch: the chassis is finished and the assembly work has begun.

I estimate I’m going to need two to three full days to complete the assembly, likely including a few trips to the electronics store to collect various fasteners and such like, before the job is complete and testing can begin.

Just to satisfy myself that I’d done something, I began assembling a few components onto the chassis, pleased to say everything seems to fit; the holes all line up and are the right size (so far!)

This amplifier will contain TWO mains transformers: the large main transformer, and a smaller one to provide 9v AC which will be the standby power. This transformer will power the input selector, remote volume control and standby control, and the chromecast USB power supply.

Having two transformers allows the provision of the full remote control, including power switch.

So the top plate of the chassis will need to support the mains transformers and the output transformers. It’s made of 4mm aluminium and it seems equal to the task without bending, based on some benchtop tests. Time will tell if it starts to sag in use

More pictures and progress reports as assembly continues. Some quick phone snapshots for now.

Power Supply, USB supply, rear terminals, volume control and rotary encoder in place
The back panel. Note switched power output – this will run to the separate (valve) phono preamp.
Outputs will be 4 and 8 Ohm.
All these separate panels are gonna need some heavy-duty earthing!

Lots of holes

Other than the occasional messing around with bias resistors and measuring signal levels on the test rig, not much progress has been made at the workbench. This is because I’ve instead been designing the chassis, which needs to be completed before any further construction can proceed.

So – after a bit of research I decided on LibreCAD as my design software of choice. It has one (and only one) redeeming feature – its price: free. (As in speech).

Apart from that, this software is a nightmare. Everything works to about 60% and then you have to work around its shortcomings. For a complete CAD virgin such as myself, this proved somewhat challenging. A few times I came close to abandoning the process and instead cutting out bits of paper and re-arranging them on a sheet to simulate the layout.
However, I persevered, and eventually I had some artwork that I was able to send away to be laser cut. This was an exercise in spatial thinking since I was effectively working on the top and bottom side of the top panel simultaneously. A jigsaw puzzle which took some consideration before everything worked.

After the artwork was sent to the laser cutter, a prototype chassis was cut into MDF. Today I have collected this:

Already I have around 10 changes I want to make before the final chassis is cut into 4mm aluminium. This prototype has admirably served its purpose.

Next steps will be actual construction, when the final chassis is completed.