Strange distortion on Music Angel Amp.

[Robert Gribnau]

Strange that the grid voltage of the lower triode in the phase splitter is only 108 V. Shouldn't that voltage be the same as the grid voltage of the upper triode?

With the voltages indicated, the current through one of the 42K anode resistors is 1.7 mA. So the current through the cathode resistor of 39K would than be 2 x 1.7 mA. But if that would be true, the cathode voltage has to be 133.7 V. But the measurement says 117 V.

If the voltages around the anode resistor of the ECC85 are correct, the anode current is 3.15 mA. That would give a cathode voltage of only 1.02 V. Could grid current at larger input voltages be (partly) responsible for the distortion?

[John_BS]

If you've not made a mistake with the voltages, then either there's some HF instability at play, or there's a serious issue with the 6SL7. The maximum cathode-heater spec for that valve is 90V, so perhaps that might be an issue, or is the filiament supply to that valve floating? You should be seeing grid voltages around 2V below the cathodes.

John

[yestertech]

Doh ! Clearly a senior moment as of course, as pointed out above, the heater wiring is different between the two …ho hum
There are 6.3 volts at pins 4/5 so clearly ECC85 is correct.
I am tempted to try 6SN7 as the phase splitter as it seems to me that with the voltages
Measured there is insufficient bias for the 6SL7.

Next step is to analyse the output of the ECC85 stage.

A.

PS. Just seen the 2 updates above which for some reason only appeared after I posted. Will recheck all voltages given.

[PJL]

Originally Posted by Robert Gribnau

Strange that the grid voltage of the lower triode in the phase splitter is only 108 V. Shouldn't that voltage be the same as the grid voltage of the upper triode?

That will be the meter loading.

Originally Posted by Robert Gribnau

If the voltages around the anode resistor of the ECC85 are correct, the anode current is 3.15 mA. That would give a cathode voltage of only 1.02 V. Could grid current at larger input voltages be (partly) responsible for the distortion?

This is a great suggestion and might explain why the ECC82 performed better as I would have expected it to be worse based on it's lower gain. Needs to be confirmed with the scope on the grid and can be fixed by increasing the anode/cathode resistor values. Unfortunately, that will also make the stage more sensitive to the miller capacitance loading from the 6SL7.

This is a cheap and cheerful chinese amp and that the circuit says ECC82 but the production unit has an ECC85 (DC coupled) does leave a major question mark on it's original performance

I suggest you do all tests without feedback at the moment.

[Robert Gribnau]

Such (abnormal) low cathode voltages in the first stage is something you can see in more Chinese amplifier kits.

In the attached schematic the cathode voltage is only 0.42 V. But it probably is not a typo because the cathode current can't be much higher than 2.1 mA (= 0.42 / 200) because if it would be higher, the voltage drop over the anode resistor becomes too large to fit with the indicated B+ of 278 V.

[PJL]

Based on some rough estimates, the ECC82 is going to raise the cathode voltage to maybe 1.4V but the anode will now be much lower at maybe 30-40V and that will impact the dynamic range of the phase splitter. However, as power is proportional to V^2 the uplift from 1 to 1.4 would almost double the power output before distortion.

Might be useful to put this through LTSpice so the component values can be optimised. Ideally we would need to know the output transformer characteristics, ratios and winding resistance at a minimum.

[yestertech]

I've now established that the signal from the ECC85 anode is 'clean' . Distortion is occuring in the phase splitter, but why is not clear. I'm convinced the only "fault" in this amplifier is its design, as I can't explain away a problem like this affecting both channels which does not appear related to lack of HT somewhere.
(all the valves have been checked by substitution )
Experiments today with lowering the cathode resistor on the ECC85 stage showed marked improvement on the 'scope. The cathode resistor for the ECC85 is 330 ohms in series with a 22 ohm to ground ( feedback insert ). At a value of 3-400 ohms in parallel (Bridging this with a substitution box ) the waveform at the output gets progressively cleaner and the output power increases. The feedback loop is not operating.
At this point the voltage on the 6SL7 cathode/ECC85 Anode is around 88 volts.
Despite the scope traces looking fairly clean, the measured distortion with this mod. is still of the order of 2% at only around 5-6 watts and nearly 4% at 10 watts.
This suggests there could also be a problem in the output stage.
The cathode currents in the EL34's seem rather low, when compared to the Mullard data.
I'm wondering whether the bias has been backed off for some reason.(all valves read roughly the same - see post 1 )

A.

[PJL]

I am not sure why the 6SL7 phase splitter works better with lower current. The anode resistors are lower then the plate resistance which is going to significantly reduce the gain. https://www.aikenamps.com/index.php/...-line-approach suggests that the original circuit values would have been about right.

Although the 15K should have a constant current through it, it might be worth adding an electrolytic here as suggested above.

You might also want to try increasing the ECC85 anode resistor rather than lowering the cathode as this will give extra open-loop gain and that will reduce the distortion when the loop is closed.

[Diabolical Artificer]

The distortion could be caused by the OP valve capacitance,the PS not having enough ooomph to drive the OP stage. As a rule in my amps I use a 6SN7/ECC82/12BH7 configured as a LTP, HT 300-350v - 47k anode resistors no tail R but a CCS, CCS set to 4mA each triode, this drives most stages.

If you can it might be worth lifting the PS tail R off ground and connect a bench PSU to it with a voltage of -80v or even -30v as an experiment, thereby giving a bit more headroom. Before that though with the change you made in post #27, either short out the 15k or 5k6 HT dropper or tack a 4k7 or 1k across it to raise the HT. Increase the anode R's too as PJL suggested.

[Diabolical Artificer]

After typing the above I drew a loadline for your phase splitter, it looks to me like it's incorrectly biased with hardly any Vgk, meaning it's going into grid current, your readings bear this out. This means you need more Vk at least 1v. So I'd increase Ra to 47k, shunt out the 15k R that drops HT to the anodes and try a 330r in series with the 39k tail R.

There are other things I don't like about the amp. The IP signal is reduced by the 100k/100k at the IP,the gain stage is throttled back, then they use a 6SL7 to get more gain, why? Far better to increase the gain stage &use a 6SN7 as P. I'd reduce the OP stage grid leaks and increase the coupling caps as well as getting rid of the10k grid stoppers. Just a thought.

Andy.

[PJL]

I have knocked up an LTSpice simulation but with a 6AQ8 and a good quality output transformer. This confirms the feedback is minimal due to the low open loop gain caused by the anode loads (I can get an extra 10dB gain by doubling the anode loads). It does not show the open-loop distortion though.

Maybe we need to look at the output transformer as a wrong ratio here would cause problems. What do the 6SL7 anodes look like and are they hitting the rail? Are you definitely using the right tap and dummy load? Are you confident you have the wattage calculation correct?

[PJL]

Originally Posted by Diabolical Artificer

After typing the above I drew a loadline for your phase splitter, it looks to me like it's incorrectly biased with hardly any Vgk, meaning it's going into grid current, your readings bear this out.

It is auto-biased off the ECC85 anode voltage so it will sort itself out.

The simulation shows decoupling of the 15K does help once you get to higher frequencies due to the capacitive loading from the EL34 grids.

[PJL]

According to the simulation, there is maybe 7dB feedback leaving the amp ridiculously over sensitive. My revised component value version has 10dB more open loop gain and about 20dB feedback so sensitivity is down 10dB.

The circuit also shows the output is not connected to 0V and if this is the case it has no feedback.

[PJL]

The revised circuit values and it does well up to 20W at 20KHz with very little visible distortion. The actual transformer could change all this though.

[yestertech]

Originally Posted by PJL

I have knocked up an LTSpice simulation but with a 6AQ8 and a good quality output transformer. This confirms the feedback is minimal due to the low open loop gain caused by the anode loads (I can get an extra 10dB gain by doubling the anode loads). It does not show the open-loop distortion though.

Maybe we need to look at the output transformer as a wrong ratio here would cause problems. What do the 6SL7 anodes look like and are they hitting the rail? Are you definitely using the right tap and dummy load? Are you confident you have the wattage calculation correct?

I really appreciate all the help received so far with this amp !
Other things have got in the way of late so further progress has been slow !
I can confirm that :
the output has one side grounded, and that the feedback loop has been disconnected. I
noticed on removal that there seemed to be very little feedback in any case.
The load resistor value has been checked and is connected to the 4 ohm O/P on the amp.
The load resistor unit contains measurement electronics/meters which read out directly in
Watts. This was made by a commercial company ( unnamed ) but has proved fairly accurate
on all other bench tests previously. I will re-check with a 'VTVM'
The output is very distorted at the phase splitter output ( EL34 grid drive )

Andy

[yestertech]

Originally Posted by PJL

The revised circuit values and it does well up to 20W at 20KHz with very little visible distortion. The actual transformer could change all this though.

Very interesting thanks ! - I will look at tacking in these mods on one channel to compare against.

A.

[yestertech]

Finally got some time to get this one back on the bench. I've rearranged the circuitry to resemble the simulation set up, except for the ECC85 cathode resistor which I have left at 330R plus 22R without feedback connected a this is almost the same.

Alas, all is still not well.
The measured voltages are not where the simulation suggests they should be
These are as follows :
HT in volts 370
volts post 22K decoupler 301
volts post 33K decoupler 289
volts at ECC85 anode ( also 6SL7 g1 ) 87.4
volts at ECC85 cathode 0.828
volts at 6SL7 a1 153
ditto a2 180
volts at 27K tail resistor 85.8

The maximum undistorted O/P at 1KHz observed on the 'scope at 6SL7 a2 is around 10 volts. There's noticeably less at a1
Beyond 10v the trace once again shows distortion appearing.
I've replaced the valve with a NOS tested 6SL7 but this makes little or no difference.
Something about this topology is determined not to work.
As a sanity check I'm planning to cross connect the output stage to a LEAK chassis with their trademark cathode coupled phase splitter just to make sure that the output stage and transformers at least are working correctly with 'clean' drive. Interestingly, the LEAK circuit is not DC coupled to the input stage.

A.

[GrimJosef]

Originally Posted by yestertech

... As a sanity check I'm planning to cross connect the output stage to a LEAK chassis with their trademark cathode coupled phase splitter just to make sure that the output stage and transformers at least are working correctly with 'clean' drive ...

Good idea. Otherwise you could drive the secondary of a push-pull output transformer from a good quality power amp and use the primary, loaded if necessary with a resistor across each half so the power amp sees 8ohms, and drive the output stage of the Music Angel from that. The only problem might be if this arrangement would be too 'stiff' a voltage source to reveal problems with one half or the other of the output stage drawing significant current.

Cheers,

GJ

[Diabolical Artificer]

Just wrote a post and clicked the wrong button, doh! Will try again.

First off where did you measure that sine wave? At the LTP anodes? At the EL34 grids? Is this with NFB applied?

Second although in theory the 6SL7 LTP should drive the EL34's a 6SN7 would be better, I drew up a loadline and circuit, see attached. One thing missing, 2nd triode grid de-coupling cap = 100n ish. The 10k EL34 grid stoppers concern me. 1k would be better and I'd change the g2 stoppers to 470r as per the datasheet. EL34 Vgk should be -32v needing 45v P-P Vg1,the circuit attached can supply 85v P-P @ 2.5mA and a rough OP Z of 15k, should be enough to drive the EL34's cleanly. See attached. I''ll build and test this circuit later and report back.

Andy.

[yestertech]

Originally Posted by Diabolical Artificer

Just wrote a post and clicked the wrong button, doh! Will try again.

First off where did you measure that sine wave? At the LTP anodes? At the EL34 grids? Is this with NFB applied?

Second although in theory the 6SL7 LTP should drive the EL34's a 6SN7 would be better, I drew up a loadline and circuit, see attached. One thing missing, 2nd triode grid de-coupling cap = 100n ish. The 10k EL34 grid stoppers concern me. 1k would be better and I'd change the g2 stoppers to 470r as per the datasheet. EL34 Vgk should be -32v needing 45v P-P Vg1,the circuit attached can supply 85v P-P @ 2.5mA and a rough OP Z of 15k, should be enough to drive the EL34's cleanly. See attached. I'll build and test this circuit later and report back.

Andy.

Thanks for this
The scope trace and AC voltage measurements are as taken at the LTP anodes and the f/b loop is disconnected. Vgk for the EL34's on this amp is around the -32 mark.
I do suspect that the main issue is caused by the direct coupling used from the previous stage - Leak seemed to prefer inserting a coupling cap.
I still struggle with the poor performance of the original design as sold, given that both channels are the same and there appears to be no actual 'fault' as such. The low output and high distortion does seem to have its root cause in the original design of the phase splitter circuitry. I'm also struggling with why the practical implementation of the spice simulation kindly supplied by PJL fails to give the expected results ! ( 6SL7 subbed but no difference )
Before I eject this as a lost cause, I will try feeding the output stage from the LEAK LTP topology to determine whether the rest of the amp is viable.

Andy