Very nice to see that this amp and thread is still going...
I have build a pair of these amps just a few years ago and although they work I never got them to perform 100% correctly.
The issue with my build is the Q4 transistor (716 ?) runs hot, well hotter then it should.
I used the 2nd version of the Maplin amp (150W) with higher power supply (42/0/42 transformer) but tested with a 35/0/35 transformer and all original mosfet's and transistors, checked and rechecked all components an all is correct and then I decided that enough was enough so now they are on the shelf in the shed waiting for another chance.
Can anyone shed some "light" about this issue ?
I have build a pair of these amps just a few years ago and although they work I never got them to perform 100% correctly.
The issue with my build is the Q4 transistor (716 ?) runs hot, well hotter then it should.
I used the 2nd version of the Maplin amp (150W) with higher power supply (42/0/42 transformer) but tested with a 35/0/35 transformer and all original mosfet's and transistors, checked and rechecked all components an all is correct and then I decided that enough was enough so now they are on the shelf in the shed waiting for another chance.
Can anyone shed some "light" about this issue ?
Attachments
Last edited:
Hi chermann
Yes I did nice clean sin wave square wave with a little ring on the leading edge.
Also tried different transistors all with the same result (heated Q4).
Yes I did nice clean sin wave square wave with a little ring on the leading edge.
Also tried different transistors all with the same result (heated Q4).
miller cap changed? miller caps change for every transistor. not all values fit for another transistor
and the step before?-- clear at square wave? i mean at the input stage
and the step before?-- clear at square wave? i mean at the input stage
Sorry for the confusion
Sinewave at the output is clean but there is a small ring on the leading edge wile using square wave, again at the output terminal on the board.
All capacitor have the value specified in the article (27pf for TR3 and TR4).
For C4 and C7 (6n8 - 100nf) had to use polyester instead of polylayer due to not having and not finding a supplier.
Sinewave at the output is clean but there is a small ring on the leading edge wile using square wave, again at the output terminal on the board.
All capacitor have the value specified in the article (27pf for TR3 and TR4).
For C4 and C7 (6n8 - 100nf) had to use polyester instead of polylayer due to not having and not finding a supplier.
Have you checked that there's no ringing on the square wave input source?
Maybe try it into a 47k resistor just to confirm it's clean,
I think about 6mA ends up going through TR4, although you may wish to confirm that, at 50V that's 300mW, but TR5 should be about the same I'd think.
TR4 is a 2SD756 or similar, TO92 I think, at 300mW, it's going to get quite warm, the original is a longer TO92 (TO-92MOD) so perhaps 150C/Watt, so a rise of perhaps 0.3 x 150 = 45C above ambient, so I'd expect 65C+ on an IR thermometer.
Globulator ?
The wave form generator is perfect and calibrated every year (work place equipment)
Exactly TR3 and TR4 are 2SD756and they are capable of 750mW of dissipation.
Maybe I'm being to paranoid they get hot but nothing of the other world (I can still put my finger(s) on it)
I had music going trough the amp and the sj50/sk135 nice and cold (300x300mm heatsink) with 100mA of bias current.
Transformer is 40/0/40 not 42/0/42 like I wrote above original article calls for a 39/0/39 that is not a standard voltage and not easy to find.
Again maybe it's just me I'm used to see the outputs get hot way before the driver are "very hot"
Going to give it another go with the bench power supply to see if there is any unusual power drain.
Rergards
The wave form generator is perfect and calibrated every year (work place equipment)
Exactly TR3 and TR4 are 2SD756and they are capable of 750mW of dissipation.
Maybe I'm being to paranoid they get hot but nothing of the other world (I can still put my finger(s) on it)
I had music going trough the amp and the sj50/sk135 nice and cold (300x300mm heatsink) with 100mA of bias current.
Transformer is 40/0/40 not 42/0/42 like I wrote above original article calls for a 39/0/39 that is not a standard voltage and not easy to find.
Again maybe it's just me I'm used to see the outputs get hot way before the driver are "very hot"
Going to give it another go with the bench power supply to see if there is any unusual power drain.
Rergards
RS232
Gate stopper resistors R13, R14 should be increased from 100R, to 220R.
The MOSFETs oscillate and burns out R15, the Zobel resistor.
The oscillation isn't visible because its damped by the Zobel, but it will burn out eventually, it happened to all 4 of my amps, and 2 were Hitachi, 2 were Exicon MOSFETs.
You could chose R9 such that the collector voltages of TR3 and TR4 are roughly equal.
The problem with this simple circuit is that the bias currents depend on the supply voltage, all set by R3.
Gate stopper resistors R13, R14 should be increased from 100R, to 220R.
The MOSFETs oscillate and burns out R15, the Zobel resistor.
The oscillation isn't visible because its damped by the Zobel, but it will burn out eventually, it happened to all 4 of my amps, and 2 were Hitachi, 2 were Exicon MOSFETs.
You could chose R9 such that the collector voltages of TR3 and TR4 are roughly equal.
The problem with this simple circuit is that the bias currents depend on the supply voltage, all set by R3.
Just looked thru an old "Melody and Music Maker" magazine that Maplin used to publish, dated 1983.
Actually the original design that you show above was spec'ed for 32VAC transformer and 44VDC. Only in their "second time around" article that you posted above do they spec the voltages at 55V DC-max. I assume that your DC voltage is now about 56V, so approximately 25% higher voltage across the transistors give will give more heat.
In the 1983 article they specify 2SC2547E (witch is standard TO92, 400mW) with the 2SD756 (i the longer TO92mod, 750mW) package. I can only assume for the purpose of increased power dissipation. I also note that for the 2SD756, Maplin does not specify the gain group. What transistor is installed on your PCB?
Furthermore, increasing the voltages from 44 to 56 volts will also increase voltage across R3 and the total current thru the LTP input pair from stage from just below 0,92mA to 1,18 mA. I believe Nigel use a constant current source instead of R3 in his slightly modified/improved design to allow less variation in circuit performance and less dependence on supply voltage.
Anyway, increased current will result in increased voltages across the 3.9K "legs" from about 1.8 to 2.3 volts. Assuming a Vbe drop of about 0,7 volt in TR3 and TR4 we will have a voltage across R10 increased from 1.1 volt to 1.6. This is equivalent of a current increase from 11 to to 16mA (almost 50%) in the second LTP pair. By my calculations that should be about 440mW dissipation per transistor at idle. Compare this to Globulator's post above and you should be within the design limits.
This increase is good for charging and discharging MosFET gates, but the sum of increased current and voltage will give more heat dissipation in TR3/TR4.
If you want to bring down heat dissipation closer to that of the 1983 article, consider increasing R3 from 47K to 56K, or even better - put a constant current source in the replace R3 tail resistor for a constant current source.
To LJT
Thank you, best explanation given in this thread.
Was looking at the Hitachi app notes and they give a 65Vdc (unloaded) voltage for the 100W version with 2 pairs of 134/49 with some minor value changes in a few resistors (R3/R4/R5.
Also in the Maplin version there is the R12 47R resistor that never shows up in the official Hitachi paper and actually there are no gate load resistors in the 50W Hitachi version.
Going to try a few things during the week and post back.
Thanks to all
Thank you, best explanation given in this thread.
Was looking at the Hitachi app notes and they give a 65Vdc (unloaded) voltage for the 100W version with 2 pairs of 134/49 with some minor value changes in a few resistors (R3/R4/R5.
Also in the Maplin version there is the R12 47R resistor that never shows up in the official Hitachi paper and actually there are no gate load resistors in the 50W Hitachi version.
Going to try a few things during the week and post back.
Thanks to all
RS232
Remove the 4.7 ohm Zobel resistor and 'scope the output; I saw MHz oscillations.
On a thread in DIYaudio I asked advice, tried a few options before settling on the solution I described above.
I modified two of mine with a current source replacing R3, and an additional cascode common-base transistor emitter to TR3, base to 0v, collector to the current mirror.
Performance doesn't appear to have improved, but bias conditions are now stable, and it made me happy to know that.
One day, I hope to make a PCB for the modified layout.
Remove the 4.7 ohm Zobel resistor and 'scope the output; I saw MHz oscillations.
On a thread in DIYaudio I asked advice, tried a few options before settling on the solution I described above.
I modified two of mine with a current source replacing R3, and an additional cascode common-base transistor emitter to TR3, base to 0v, collector to the current mirror.
Performance doesn't appear to have improved, but bias conditions are now stable, and it made me happy to know that.
One day, I hope to make a PCB for the modified layout.
One more question if I may.
Why do some manufactures use different value resistor as gate loaders ?
Was looking at a Musical Fidelity schematic and the P channel uses 150Ohms and the N channel 1.5Kohms.
This is also mentioned in the Hitachi app note I posted before.
Thanks for all the help
Why do some manufactures use different value resistor as gate loaders ?
Was looking at a Musical Fidelity schematic and the P channel uses 150Ohms and the N channel 1.5Kohms.
This is also mentioned in the Hitachi app note I posted before.
Thanks for all the help
The different polarities have different Cgs values - external capacitances can be used to try to even up the pair so they are more symmetrical. However MOS capacitance is non-linear anyway, so its not super effective in reducing open-loop distortion, but can definitely reduce it, and thus improve performance. Ultimately it comes down the 3:1 difference in electron v. hole mobilities in silicon.
The gate stopper resistors, like grid-stoppers in valve ccts, provide a high enough impedance at RF, at MHz, where it can look like a common-gate amplifier to RF, and all the wiring looks like inductors.
The value then depends on both the MOSFET and the PCB layout as to which type of RF oscillator it resembles.
I found something that worked, that stopped the MOSFETs oscillating at RF, and didn't question it further.
I had an amplifier that received the RF of a CB radio through the loudspeaker cable, demodulated it, and loudly presented the inane chatter to me through the speakers.
I have always found radio a bit of a black art. A bit too voodoo.
The value then depends on both the MOSFET and the PCB layout as to which type of RF oscillator it resembles.
I found something that worked, that stopped the MOSFETs oscillating at RF, and didn't question it further.
I had an amplifier that received the RF of a CB radio through the loudspeaker cable, demodulated it, and loudly presented the inane chatter to me through the speakers.
I have always found radio a bit of a black art. A bit too voodoo.
I altered the input LTP sith a current source here:
It lowers distortion too, which is nice 🙂
If you do this, please check the resistor values around your circuit boards as I may have changed some to balance up the amp after adding the current source.
I managed to get the current source in the orginal board.
If you copy this, please use the original Maplin values (0.68-1.0uF for the input cap and the 47uF feedback cap, bigger is not better here as LF stability is impacted.
Do you know for proven schematic with P type lateral MOSFETs as quazi complementary pair output?
Here is another Hitachi app-note design I bought probably in the early '80s, The PA 105,
It served as a PA amp for some years...
🙂 morten
It served as a PA amp for some years...
🙂 morten
Interesting that it uses 2SD666's instead of the 2SD756 - why would you want less gain and twice the output capacitance of the 2SD756? It has slightly more power handling at 900mW compared to 750mW (though the datasheet actually mis-prints this as 0.9mW !)
Could it be for stability?
I note it's also got 220R gate stoppers.
When playing LTspice I did find that if the open loop gain got too big, it could get a bit 'lively'..