Phase Splitter Name
- Thread starter Gold_xyz
- Start date
Ray ok (agreement to the worst of all specs)--I'm no fan of this kit stuff as I've had a couple to repair. ( and modded them to autobias)...
Interesting claim to spec, 0.1% thd at 1W o/p... that must surely be a mistake. Anyone else spot this ? Interestingly no thd figure is mentioned near full o/p.
Appro 3rd harmonic thd. Morgan Jones p.188 < mentions B9A ECC82 dual triodes are particulary gastly>. That's a published fact. so why use'm ?
The only way to get sensible performance with parallel pairs with low thd is either to use cathode follower drivers or an intermediate balanced push-pull voltage driver stage as per Williamson. I favour the latter as it cushions the split load.. I fail to understand why the authors of the kit haven't attempted this avenue.
Academically, readers can come to a swift sum-up by browsing Morgan Jones 3rd ed, classic amplifiers p. 412 onwards "the inherent driver defects of the Mullard 5-20". That's only with a conventional p-p pair not paralleled.
No hiding facts here.
richj
Interesting claim to spec, 0.1% thd at 1W o/p... that must surely be a mistake. Anyone else spot this ? Interestingly no thd figure is mentioned near full o/p.
Appro 3rd harmonic thd. Morgan Jones p.188 < mentions B9A ECC82 dual triodes are particulary gastly>. That's a published fact. so why use'm ?
The only way to get sensible performance with parallel pairs with low thd is either to use cathode follower drivers or an intermediate balanced push-pull voltage driver stage as per Williamson. I favour the latter as it cushions the split load.. I fail to understand why the authors of the kit haven't attempted this avenue.
Academically, readers can come to a swift sum-up by browsing Morgan Jones 3rd ed, classic amplifiers p. 412 onwards "the inherent driver defects of the Mullard 5-20". That's only with a conventional p-p pair not paralleled.
No hiding facts here.
richj
jane said:This is one of the few split load inverters I have seen with grid leak bias.
Jan E
AndreasS said:
Hi
It seemed me that there was something strange.
In the scheme (k4040) I see a 0V grid polarization.
Also ECC82 as preamplifier and ECC83 as Splitter
This other scheme seems me correct, ECC81 with R12
http://www.velleman.be/downloads/0/manual_k8010.pdf
also
http://www.velleman.be/downloads/0/manual_k8011.pdf
Oh boy...
The ECC83 split load as given will give a maximum peak signal output of about 40V before it runs out of steam (quickly hooked up) - and that is at quite visible 2nd harmonic distortion on the scope. (A 47K plate load in parallel with only 110K next grid resistors is looking for distortion.) Then grid-leak bias only works well when fed with a low impedance source. A feeding ECC83 is definitely not, apart from the grid leak resistor being too low - at least 10meg should have been there. Or it could have been direct coupled, although hardly successfull with an ECC83. I have said in the past that I do not see the ECC83 as a good choice for any kind of duty in a power amplifier. A pentode input to low mu triode phase-splitter would run circles round the above (ECF80, ECF82, 7199 etc. - and no comment from the "pentode-is-noisier" clan. That is in theory; certainly not audible in this kind of application.
I thought this kind of circuit has already been thrashed out to such an extent over 6 decades that should no longer appear compromised, . My apologies, but if this is the best Velleman can come up with, they have lost me.
Sorry Gold_XYZ, but that is not it.
The ECC83 split load as given will give a maximum peak signal output of about 40V before it runs out of steam (quickly hooked up) - and that is at quite visible 2nd harmonic distortion on the scope. (A 47K plate load in parallel with only 110K next grid resistors is looking for distortion.) Then grid-leak bias only works well when fed with a low impedance source. A feeding ECC83 is definitely not, apart from the grid leak resistor being too low - at least 10meg should have been there. Or it could have been direct coupled, although hardly successfull with an ECC83. I have said in the past that I do not see the ECC83 as a good choice for any kind of duty in a power amplifier. A pentode input to low mu triode phase-splitter would run circles round the above (ECF80, ECF82, 7199 etc. - and no comment from the "pentode-is-noisier" clan. That is in theory; certainly not audible in this kind of application.
I thought this kind of circuit has already been thrashed out to such an extent over 6 decades that should no longer appear compromised, . My apologies, but if this is the best Velleman can come up with, they have lost me.
Sorry Gold_XYZ, but that is not it.
The ECC83 (EF86, triode of the ECL86, ECLL80) are tubes, that can work with grid leak bias.
So this concertina phase inverter with grid leak bias we also meet in other amplifiers: Revox G36, Telewatt VS55, Grundig NF10.
The datasheet for the ECL86 (triode) gives an output voltage of 9 Volt RMS with THD 0,4%.
Regards Andreas
So this concertina phase inverter with grid leak bias we also meet in other amplifiers: Revox G36, Telewatt VS55, Grundig NF10.
The datasheet for the ECL86 (triode) gives an output voltage of 9 Volt RMS with THD 0,4%.
Regards Andreas
It is not just a question of whether grid-leak bias will work (at all), but how well. I have no reliable model for the ECC83 for Spice, but my Philips Tube Data Sheets show the following comparative data. (The ECC83 is quite similar to the triode of an ECL86).
In all cases the H.T. is 300V, Ra=100K, Rg1 (next stage)=330K, and choice is for approximately equal Ia. Output (signal) voltage is given for quality work, i.e. just before onset of grid current (in that sense maximum):
1. With own Rg1=10meg and gen.resistance=330K:
D=5% at 6Vrms out, Ia=1.23mA, A=50
2. With own Rg1=10meg and gen. resistance=100 ohm:
D=2% at 30Vrms out, Ia=1.29mA, A=54
3. With suitable cathode bias (Rc=1,5K bypassed):
D=3,7% at 30Vrms out, Ia=1,11mA, A=57
Other moderately high mu triodes and pentodes show similar figures.
The question comes up: Why this almost fixation with the ECC83 in this type of circuit in many power amplifiers? If one insists on this, the ECC81 will be considerably better, at the penalty of slightly lower gain. But there still remains the matter of Miller capacitance for the first stage, unless a low input impedance (lower than most volume controls seen there) is used. With due respect for other designers, I rather fear that the operation of the ECC83 is not well understood if used in this application, at least for uncompromised quality. Spectrum analysis (harmonic content) of such instruments is revealing.
This explains my choice of a pentode-triode (moderate gain) here, as said. Yes, if one wants to use the ECL86, making use of the given triodes, than condition 3 above. The grid-leak bias method is to my mind specifically for low output conditions (pre-amplifiers), under the above conditions, because a low enough feed impedance is not always feasible (certainly not another ECC83 triode, anode fed).
Gold_xyz,
I would use an ECF80 or ECF82 for your case, with the triode drawing several mA so that it can use Ra=Rc=22K or so. But as others have said, there are better circuits - either an extra double triode driver, or the long-tail pair (Schmitt) phase inverter. Personally I don't feel the saving of an extra tube in a circuit of this total cost is merited. One is then looking at say an EF86 and ECC81/ECC88 or similar.
Apologies for a rather lengthy comment, but it might be worthwhile to get matters clear before choosing - at least my take.
In all cases the H.T. is 300V, Ra=100K, Rg1 (next stage)=330K, and choice is for approximately equal Ia. Output (signal) voltage is given for quality work, i.e. just before onset of grid current (in that sense maximum):
1. With own Rg1=10meg and gen.resistance=330K:
D=5% at 6Vrms out, Ia=1.23mA, A=50
2. With own Rg1=10meg and gen. resistance=100 ohm:
D=2% at 30Vrms out, Ia=1.29mA, A=54
3. With suitable cathode bias (Rc=1,5K bypassed):
D=3,7% at 30Vrms out, Ia=1,11mA, A=57
Other moderately high mu triodes and pentodes show similar figures.
The question comes up: Why this almost fixation with the ECC83 in this type of circuit in many power amplifiers? If one insists on this, the ECC81 will be considerably better, at the penalty of slightly lower gain. But there still remains the matter of Miller capacitance for the first stage, unless a low input impedance (lower than most volume controls seen there) is used. With due respect for other designers, I rather fear that the operation of the ECC83 is not well understood if used in this application, at least for uncompromised quality. Spectrum analysis (harmonic content) of such instruments is revealing.
This explains my choice of a pentode-triode (moderate gain) here, as said. Yes, if one wants to use the ECL86, making use of the given triodes, than condition 3 above. The grid-leak bias method is to my mind specifically for low output conditions (pre-amplifiers), under the above conditions, because a low enough feed impedance is not always feasible (certainly not another ECC83 triode, anode fed).
Gold_xyz,
I would use an ECF80 or ECF82 for your case, with the triode drawing several mA so that it can use Ra=Rc=22K or so. But as others have said, there are better circuits - either an extra double triode driver, or the long-tail pair (Schmitt) phase inverter. Personally I don't feel the saving of an extra tube in a circuit of this total cost is merited. One is then looking at say an EF86 and ECC81/ECC88 or similar.
Apologies for a rather lengthy comment, but it might be worthwhile to get matters clear before choosing - at least my take.
Hi Andreas,
True. I did not suggest that it was, but just tried to quote similar conditions, showing the relative disadvantage of grid-leak biasing under those conditions. In the original circuit the phase splitter was having an equivalent Ra of about 100K and was fed by roughtly 50K equivalent. That would still give unacceptable results, further aggrivated by having only 820K as a "grid-leak" resistor and not 10meg - not nearly enough to give proper bias without resulting in an unacceptable Ig1.
The results mentioned in my first post (#9) was from a quick experimental hook-up; I can assure you it did not look good (wish I had the easy facility to post a waveform here). I must ask again, why the preference for that mode of operation with the shown resistors, when a single extra resistor (Rc) shows a considerable improvement?
I respect known designers, but in this case it would need explanation when the results are as clearly second-best as is easily determined here. With all respect, one cannot simply discard data provided by tube manufacturers (Philips is also a reliable manufacturer) showing such discrepancies, simply because Studer, Klein and Hummel (and not many others) do something. I am not being arrogant/dismissive, but believe it is reasonable to ask for explanation. (Also, I do not find that there is something wrong with my own oscilloscope ...so?)
Regards
True. I did not suggest that it was, but just tried to quote similar conditions, showing the relative disadvantage of grid-leak biasing under those conditions. In the original circuit the phase splitter was having an equivalent Ra of about 100K and was fed by roughtly 50K equivalent. That would still give unacceptable results, further aggrivated by having only 820K as a "grid-leak" resistor and not 10meg - not nearly enough to give proper bias without resulting in an unacceptable Ig1.
The results mentioned in my first post (#9) was from a quick experimental hook-up; I can assure you it did not look good (wish I had the easy facility to post a waveform here). I must ask again, why the preference for that mode of operation with the shown resistors, when a single extra resistor (Rc) shows a considerable improvement?
I respect known designers, but in this case it would need explanation when the results are as clearly second-best as is easily determined here. With all respect, one cannot simply discard data provided by tube manufacturers (Philips is also a reliable manufacturer) showing such discrepancies, simply because Studer, Klein and Hummel (and not many others) do something. I am not being arrogant/dismissive, but believe it is reasonable to ask for explanation. (Also, I do not find that there is something wrong with my own oscilloscope
...so?)Regards