As the customer will know better than I because he's familiar with several amplifiers and electric guitars I assumed the amplifier was a bit deaf.
I checked up on typical guitar output levels and typical amplifier inputs. Using RMS or root mean square values the typical sensitivity of an amplifier with a quarter inch jack socket is minus 10dBV or in plain English, a tone input of 0.1 volt should give you a decent volume in the loudspeaker. Next, what's the rating of the specific amplifier? Details on this are confusing, but given a pair of EL34s with 500 volts on their anodes, fixed bias of 36 odd volts and operating in class AB1, I reckon one can expect 50 watts RMS output. The speaker in the case has an 8 ohm coil so, from ohms law, the output voltage across the coil should be the square root of power (=50) x load resistance (=8) or 20 volts RMS.
A brief word about the term "RMS". A waveform having a sinusoidal shape cannot be measured with a DC test meter but you use an oscilloscope to view it. How does one refer to its amplitude though? With a scope the obvious way is to measure peak to peak. On the other hand you may have an AC testmeter. What will this read? Well, it might be anything determined by its designer, but some meters will be calibrated in RMS. Comparing the size of the picture on the screen of the scope with the RMS reading you will see totally different values. RMS is an artificial way of making measurements and is the value of the DC voltage which would produce the same heating effect (say in a resistor) that the AC voltage produces. Its impossible to accurately see the RMS value of our scope waveform but you can calculate it by first measuring the overall height of the sinewave (the peak to peak value), dividing it by two then dividinging the result by root 2 (approx 1.4). My oscilloscope, no doubt like most modern ones indicates the RMS reading alongside the waveform display.
A peak to peak sinewave of 100 volts is said to be about 35 volts RMS. If you had two equal resistors, one with a 100 volt peak to peak sinewave across it and the other with about 35 volts DC across it they would get equally hot. A good quality test meter would give you the same readings when switched to AC at the former and DC to the latter. Note that this applies only to sinewaves. A distorted sinewave or a complex AC waveform doesn't fit in with this analogy.
After trying various frequencies and levels from my audio signal generator the amplifier appeared to give a maximum undistorted output (at this point unmeasured) with around 3.5mVolts input which seemed an awful lot lower than -10dBV. I also noticed there was some intermittency in the output level although why this was exactly was hard to say. The first task was to find some information about the amplifier; ideally its circuit diagram plus rated output power and input sensitivity.
I quickly found a set of circuit diagrams although these were hand drawn and pretty vague when it came to which version of the amplifier they represented. This particular amplifier has a label identifying it as a Type AP, Model SA112 with the serial number 16681. Assuming all the Hiwatt amplifiers use pretty similar circuitry I worked out the most likely power supply, output and pre-amp schematics. In fact the amplifier seemed to have circuitry almost exactly like the set of drawings I'd selected.
|According to the retrieved circuit diagram together with the valve line up in this example of the Hiwatt the pre-amp has four ECC83 valves. One half of the first valve (V1a) and half the second valve (V2a) amplifies the "normal" input signal and the other halves (V1b & V2b) amplify what is called the "brilliant" input signal. The input circuits are identical being a direct grid connection shunted to ground by a 1 Mohm resistor. Each amplifier has its own gain control, a 470kohm log pot. Some measure of balancing feedback is incorporated by combining the cathodes of the V1a and V1b. The "brilliant" amplifier has less gain in the lower registers and higher gain in the upper registers through the simple expedient of a lower value coupling capacitor feeding its gain pot and a 47nF capacitor shunting the self bias cathode resistor in the second stage amplifier.|
|The combined amplified signal from anodes of the second amplifier stage valves is passed through a network of resistors and capacitors and pots providing treble, middle and bass adjustment. Following these tone control circuits there is a further stage of amplification in V3a whose output is governed by the master volume control then the driver stages to the output valves. In more detail; the master volume control connects to half of the third valve V3a which provides amplification. The anode of this stage is then capacitively coupled to the grid of a triode of the fourth valve, V4. The hand drawn schematic shows the second half V4b connected as grounded grid amplifier, such that the cathode voltage will govern the anode current. V3b provides the bias voltage for V4a and V4b.|
The two output valves need to be driven by signals which are in anti-phase. This is why it's called a "push-pull" amplifier. But why use push-pull you might ask? Push-pull circuits are chosen, I suppose, because they can use less expensive valves to develop a given output power.
Half of V3 and half of V4 provide the normal signal to one EL34 and the second half of V4 provides an identical signal but with a shift of 180 degrees. The latter is obtained by connecting together the cathodes of V4 and operating the second half as a grounded grid amplifier. A measure of balancing is obtained by having slightly different value anode load resistors. A "presence" feature is incorporated. This leaks some phase imbalance via a 100kohm linear pot. V4 is supplied with the full HT but has a 24.2kohm cathode resistance (22kohm + 2.2kohm) to reduce the anode-cathode potential to a figure within the valve's voltage tolerance and to provide a distortionless waveform for the output stage.
The push-pull circuit uses a pair of EL34 valves supplied with 490 volts on their anodes and capable of producing 50 watts RMS output. Each valve is supplied with 36 volts of negative bias and its screen voltage supplied via a dropper resistor which together set its operating point and therefore its anode current. The valve control grids are AC coupled to V4a and V4b.
Output valves can be operated in various ways. Practically speaking the more efficient are the valves the least good the output quality. Least efficient is called Class A operation. Anode current is basically constant and valves will get pretty hot. The permitted anode dissipation governs the power output. Class B is fairly efficient but not as linear as Class A. Linearity describes how the output waveform follows the input waveform. Class B operation maintains the no-signal anode current at a low value. As drive to the output valves increases so their anode current rises and this rise must reflect very accurately the input change to get distortionless output. It's easier to choose a compromise, hence Class AB which is sub-dived into AB1 and AB2. By increasing the standing or no-signal anode current to a nominal value its much easier to get decent linearity. This nominal value will reflect the power handling capacity of the valves. The standing anode current is usually determined by applying a fixed negative DC bias voltage to the valve control grid, in the case of this Hiwatt, minus 36 volts has been chosen.
|Finally the power supply. This is simplified by generating only the highest voltage required. Lower voltages for the pre-amp are then derived via dropper resistors combined with capacitor balalancing resistors. Rectification is by a pair of BYX94 diodes in a full wave configuration. This allows for a simple bias supply to be provided by a single BYX94 fed by a potentiometer (110kohm plus 27kohm) which divides the voltage so that minus 36 volts is available.|
Armed with circuit prints I started to remove the chassis from the box and immediately noticed I wasn't the first to do this as a lump of wood was jamming the rear of the chassis tight against the speaker baffle and the wood securing fillet at the rear was split. The chassis came out OK and I rested it, valves downwards, on the transformers, with the output transformer on a wooden block to ensure the EL34s were clear of the bench. The wooden chassis cover was held in place with an odd set of bolts, clearly not original, and inside the chassis were more tell-tale clues of previous visitations. One of the pots had been changed, being mounted 180 degrees out and connected to the circuit board with short lengths of wire. There were re-attached bits and pieces around one of the valveholders and at this point I suspected someone had been chasing the same fault as I was.
The first step was to sort out test equipment. My signal generator has a minimum output of 3.5mV which is too high for test purposes so I made a simple attenuator providing initially tenfold attenuation (or 20dB) to reduce the generator output to 0.35mV. The signal generator was marked "600 ohms" at the output socket so I decided to make a matched attenuator from a 560 ohm and 56 ohm resistors. This had the effect of making the generator display read double the output voltage ie. a generator reading of 2 volts RMS provides a voltage at the 600 ohm load of 1 volt RMS. If I'd chosen say 56kohm and 5.6 kohm the output voltage would have been just about that on the display, say 2V RMS, and the voltage at the junction of the two resistors 20dB less or 200mV.
As tests progressed I found I needed to be able to monitor the audio input on my scope. As this has a maximum sensitivity of 20mV per cm I decided to use around something of the order of 200mV from the signal generator for monitoring purposes but some 40dB attenuation between the generator and the amplifier. A minimum voltage of 100th of half the displayed voltage. In other words, with 200mV on the display I would get 100mV into the attenuator and 1mV out.
I decided the ultimate aim was to see exactly what input level was needed to produce 50 watts output at maximum volume control settings, and to confirm that this was consistent with the output from a guitar.
When I'd finished the next tests I was feeding in 14mV indicated on the generator display, through my 40dB 600 ohm attenuator, which represented 70uV input to the amplifier for maximum undistorted output. Ideally, for levels in this range, using my signal generator, a 60dB attenuator would be best.
To avoid over-dissipation in the EL34s I connected a 10ohm dummy load (I had such a 50 watt resistor to hand) across the speaker output connections (in place of the 8 ohm speaker). With the scope across the 10 ohm load (in this amplifier one side of the output transformer is connected to the amplifier chassis) I increased the amplifier gain settings to maximum and cranked up the signal generator to the point at which distortion became apparent. As the input is increased above this point the output initially distorts then with extra input ends up as a square wave with maximum and minimum values not much greater than the undistorted peak-to-peak value.
Having outlined the methodology and having described the measuring equipment I'll turn to the testing, or the baffling bit.
Tests were initially a little odd. Sometimes you could hear the output transformer singing and then for no apparent reason it would go silent. At one point I had a decent input level and absolutely nothing beyond the pre-amp. Was there a problem with one of my test cables, a jack socket or was there a bad valve? I removed the four ECC83 double triodes (three with Mullard markings and the other from China) and tested them with my AVO valve tester. Two or three were pretty good with indicated results of between 80 and 90 odd percent. The fatter Chinese version wasn't too bad but not very balanced. One was perfectly usable but not as good as the others. Next, I checked the pair of EL34s. Both showed emissions of around 50% but improving as they warmed up.
I put back the valves, choosing the best ones for the more critical circuit positions and fired up the amplifier once again. Much the same as before, half decent output then it went completely deaf. I'd already tested all the resistors and capacitors so; unless they were breaking down from high voltages they were blameless. I decided to check more methodically using the circuit diagrams as a guide.
I got as far as the main gain control and found its circuitry was slightly different to the hand drawn schematic. A 22kohm resistor had been moved from the gain control input to the wiper connection going to the grid of the phase splitter. The new circuit would result in a slightly higher output but not too important in the scheme of things. With the amplifier turned off I checked resistance of various points to chassis (being a valve circuit one would expect high resistances to chassis except at cathode connections) and found the earthy end of the gain pot was indeed earthed as it should be, but the live end of the pot showed 32 ohms to ground. Very strange. Maybe a short-circuit capacitor, but no all were OK. I then suspected a short within the pot to its mounting bush; however there were two pots connected at this point, the treble pot wiper and the gain pot live terminal. Maybe a short to chassis at the treble pot wiper? I removed the knob and unscrewed the bush and waggled it clear of the chassis. The short disappeared. With the treble pot bush clear of the chassis I powered up the amplifier and, just as before it came to life, then went deaf after a few seconds.
It had to be the gain control pot. Nothing else could explain the short. I measured this and found it had now changed from 32 ohms to around 15 ohms. I detached the gain control knob, unscrewed the bush and waggled it clear of the chassis. The short had gone and I was now seeing about 72kohm....
At this point I decided to revise my test equipment. It would be nice to be able to read the test voltage without having to divide it by two as explained above. I'd also like to use my oscilloscope for making measurements. It can display a signal together with a reading of its RMS value, however it works best with input signals of over say 0.1 volt.
What I need is an attenuator to add to the signal generator output which can add 20, 40 and 60dB extra attenuation. The generator already has a built in attenuator reading 0 to 60dB so the extra will let me use really low level signals. At a reading of 5mV (500uV) I'd be able to push out 50, 5 and 0.5uV. As I'll be using this set-up for high impedance tests the new attenuator resistor values are not that important, although to maintain reading accuracy when the output impedance is 600 ohms I'd like to have an attenuator input resistance of better than 100 times this, or over 60kohm.
Writing a set of equations for a simple potentiometer type of attenuator I derived resistor values of 99kohm plus 9kohm, 900ohms and 100ohms. The whole lot in series measures 109kohm thus satisfying my aim for better than 60kohm. The new attenuator will use a 6-way rotary switch wired to give me 0dB, -20dB, -40dB and ground plus "open".
With the new attenuator built, I retested the amplifier. The audio input can now be set over a range of -40dB in steps of ten at the signal generators and a further -60dB in steps of twenty at the new attenuator. Connecting the scope across the dummy load however produced some very strange effects. There was wailing and shrieking and all sorts of problems, unexplained until I checked the input gain pot. The ground connection was open circuit. This must have happened when I'd removed then refitted the circuit board carrying the row of pots. Presumably this had been done in the past so many times the circuit tracks had been weakened and the one to the wiper given way?
I reflowed the solder connections to the gain pot and connected the earthy ends of the two adjacent gain pots together and this restored stabilty.
I soldered a dummy load of 8.5 ohms across the speaker jack plugs, but also had to insert a jack plug into one of the sockets as these carry a short-circuit if no speaker is plugged in. I can't figure out exactly why this had been done, other than to protect from excessive voltages if the amplifier is operated without a speaker. But why not a dummy load rather than a short-circuit?
With the new attenuator in place I checked the amplifier again. Maximum output before distortion was apparent was about 7 volts RMS across 8.5 ohms. This represents only 5 or 6 watts. As the customer had explained this is too low and here was the proof. To be sure of my measurements, I disconnected the output valve HT feed and inserted a 33 ohm resistor. At max output I read about 3.3 volts. This means the pair of EL34s are drawing an anode current of 100mA which at 490 volts HT is 49 watts input, miles short of what's needed to produce 50 watts output.
At this point I went back to the circuit diagram and carefully checked each component. All measured OK. I even disconnected a coupling capacitor in case this was leaking and altering the bias in the pre-amp or driver stages. Oddly I found this capacitor had already been cut and resoldered back in place. Clearly I'm not the first to try and solve the problem.
Of course the capacitor was blameless so I went back to checking signals with my scope. I carefully checked each stage and found that the first two stages amplified the input signal very well, however, not so the phase splitter, which easily produced clipping in both main and 180 degree signals. In fact the clipping was so bad no more than 7 volts RMS was achievable. At this point I stopped to think more clearly. The phase splitter uses two high value resistors of 82kohm and 91kohm in their anodes and a 24kohm in the cathodes. It's easy to think of a transistor in the position of the 12AX7 triode, however it's far from the truth. A transistor is essentially a current amplifier and as such with an 82kohm collector load will produce lots of voltage swing. A triode is a voltage amplifier and works to a defined curve. Given a grid voltage a corresponding anode voltage is formed. Anode current is defined by the valve characteristics and typically, for a 12AX7 will be around a half to one milliamp. With an 82kohm load the voltage will be 41 to 82 volts, however the anode resistance of the triode is around 60kohm so 30 to 60 volts will be lost and any extra anode current is not possible.
Given the circuit values it looks like the phase splitter is doomed to producing no more than a handful of volts. This surely explains why the EL34s are under-driven.
Maybe a triode with a lower anode resistance and able to supply more current is the answer? We don't have to look far. The 12AU7 or a 12AT7 is the answer. The former has an anode resistance of about 7kohms and a typical anode current of 10mA. Push 5mA through our 82kohm resistor and you'll have 400 volts. The next step is to ditch the 12AX7 and try a 12AU7 and then a 12AT7. See which works best? Maybe the circuit diagram is wrong and maybe the amplifier has a wrong valve as V3 ? As there's only the one circuit readily to be found on the Internet, maybe someone found this then swapped V3 for a 12AX7?
Well, some news to report. I fitted a 12AT7 in the phase splitter and things imroved a lot. A second 12AT7 in position V4 topped off the testing. I plugged in both EL34s and repeated the final test, cranking up the input with -40dB at the audio generator resulting in an indicated 34mV output and -20dB on the external attenuator made the dummy load very hot with 19.5 volts RMS appearing across it. My 33 ohm current monitor resistor also ran very hot because this had over 13 volts across it.
Before I carried out final tests and then reassembled the amplifier I removed the 33 ohm monitor resistor and fitted one with a value measuring exactly 0.69 ohms. In the future, anode current can be measured by checking the voltage across this resistor. I also measured the value of the speaker circuit dummy load at 7.9 ohms. Turning both volume controls to full the amp produced exactly 20 volts RMS across the dummy load. This represents 50.6 Watts, at which point the input was 3.6mV RMS at 1KHz. The anode current produced 200mV DC across the monitor resistor so was 290mA, and the HT voltage had dropped from 490 to 439 volts so the DC input was 127 watts. The efficiency of the EL34s neglecting screen and heater figures is therefore 40%. As far as distortion was concerned I used only my eyes to look at the output waveform. As the input is increased you can see that the shape doesn't change much until 19.5 volts output is achieved, then the lower part of the sinewave starts to get slightly fatter, presumably reflecting design shortcomings in the balance at higher powers, then becoming equally distorted as odd harmonics creep in. At exactly the point where 20 volts output was achieved the waveform looked reasonable, but remained at this voltage level and gradually distorting as the input increased. The power output would be rising still, and the EL34 anode current would be rising also, but the output voltage would no longer be RMS so the 20 volt output cannot be simply mathematically converted to power.
What's the explanation to all of this?
Presumably the amp had an intermittent problem. The owner gave it to a repairer. The repairer downloaded the circuit (with the wrong valves indicated).
The repairer changed V3 and V4 and continued to chase the intermittent fault. He changed the bass pot and cut a number of component leads, then resoldered them.
Finally, he handed back the amp. It wasn't very good so was sold on Ebay. It then had not only the intermittent fault plus low output because the wrong valves had been fitted.
Incidentally... what's the knob on the rear panel? The knob is screwed to a potentiometer which isn't connected to anything. Very strange.