I was wondering if anyone knows better ways to supply screen grid voltage in small signal pentode , the usual off the anode series resistor is one way of doing it, a potential divider another , I have very low hum on the plate supply so does that mean deriving my voltage from there is a good idea or could there be a benefit from making a seperate rc chain from further back in the psu specifically to feed the screen? Its all class A so no major current differences between stand still and full throttle ,probably about 2ma of screen current is required .at maybe 150 or 180 volts and supply to the anode load is 320Volts. Could a string of Zeners be of use here or maybe a gas discharge tube Oa2/3 etc or will I only increase noise by doing that?
stablised screen grid supply
- Thread starter Tubetec
- Start date
Help Support GroupDIY Audio Forum:
ruffrecords
Well-known member
Rupert Neve's first tube console used an EF86 pentode stage direct coupled to an ECC83 (I think) cathode follower, The cathode of the CF was connected to the decoupled screen grid of the EF86 via a highish value resistor. This provided a dc stabilised screen grid voltage by NFB which was decoupled at audio frequencies by the screen grid decoupling cap. I think I may have a schematic of it somewhere.
Cheers
Ian
Cheers
Ian
Thats an interesting idea Ian ,the triode shares some voltage with the screen and allows feedback back to earlier in the chain at the sametime.
In this case ,its just a single c3g pentode ,with a baxandall tone control across it , so plenty of feedback already from plate to grid.
Just the one valve in this box and its own psu in a seperate enclosure .Ive used screened chokes to make the ht rail very quiet , Ive now moved the second choke from the psu to the preamp ,that made a huge difference to my hum levels .I made the Bax circuit with 1meg and 470k pots , a bit high really considering the 10k anode load, I'll probably switch to 100k pots to try and lower noise a bit. I measure about 50pf on the inputs which goes directly to the treble pot ,would it be a correct assumtion to say a plastic case pot would have less stray capacitance to ground than a mini metal alps style?
Just changing the output coupling cap from a boxed style foil to an axial and lifting it up and away from chassis I lost 11pf across the output jacks as measured by my lcr meter, is it valid to measure the grids capacitance in this way ? Likewise when I measure across the output jacks of a rc coupled preamp with an lcr meter and I get a certain reading in the pF range, is this a valid reading for stray and other capacitance on the output stage ?
I have plenty of HT current available in the psu and an extra wire in the multicore up to the preamp ,so I might try a potential divider arrangement ,and dump enough current that the 2ma screen draw doesnt make much difference.
I see in valve power amplifiers the choke was the traditional way of decoupling the screens ,resistor/large cap in the modern age.
Something is lost in the sound without the choke though ,all the best guitar amps I can think of use one . I could use some kind of transistor regulation of either the ht line or just the screen supply but I'd rather not
In this case ,its just a single c3g pentode ,with a baxandall tone control across it , so plenty of feedback already from plate to grid.
Just the one valve in this box and its own psu in a seperate enclosure .Ive used screened chokes to make the ht rail very quiet , Ive now moved the second choke from the psu to the preamp ,that made a huge difference to my hum levels .I made the Bax circuit with 1meg and 470k pots , a bit high really considering the 10k anode load, I'll probably switch to 100k pots to try and lower noise a bit. I measure about 50pf on the inputs which goes directly to the treble pot ,would it be a correct assumtion to say a plastic case pot would have less stray capacitance to ground than a mini metal alps style?
Just changing the output coupling cap from a boxed style foil to an axial and lifting it up and away from chassis I lost 11pf across the output jacks as measured by my lcr meter, is it valid to measure the grids capacitance in this way ? Likewise when I measure across the output jacks of a rc coupled preamp with an lcr meter and I get a certain reading in the pF range, is this a valid reading for stray and other capacitance on the output stage ?
I have plenty of HT current available in the psu and an extra wire in the multicore up to the preamp ,so I might try a potential divider arrangement ,and dump enough current that the 2ma screen draw doesnt make much difference.
I see in valve power amplifiers the choke was the traditional way of decoupling the screens ,resistor/large cap in the modern age.
Something is lost in the sound without the choke though ,all the best guitar amps I can think of use one . I could use some kind of transistor regulation of either the ht line or just the screen supply but I'd rather not
In low level stages, screen grids are usually decoupled with a set of rather high value cap and resistor, with a time-constant typically about 1 second, which makes it stable for the lowest audio frequency, so there's no gain modulation. Are you worried about gain variations vs. voltage fluctuations?Tubetec said:
I can see no justification for the added complexity.
Interesting!ruffrecords said:
My understanding of this particular arrangement is that it stabilizes the gain vs. ageing and voltage variations without implying NFB at audio frequencies. That makes sense in a no-NFB operation. I would think it to be irrelevant to a Baxendall circuit.
Thanks guys,
I have 6.8 uF foil off the screen to cathode and the basic series resistor feed (150K), I have a preset wirewound in the cathode of the c3g, as I adjust bias ,so my screen volts vary too ,of course . I was thinking at high signal levels and clipping especially wouldnt the interaction between screen and plate volts cause more intermodulation ? A settable screen voltage which remains constant regardless of how much anode current flows seems like a good plan , at normal opperating levels its probably doesnt make much difference ,but Im worried about recovery time from transient overload might be longer if theres an interaction between plate and screen volts .For me smoothness of overload characteristic and fast recovery is one of my chief design considerations , I dont need the Bax tone stage to drive huge levels on the output but spikes and transients are a fact of life in the real world .
I have 6.8 uF foil off the screen to cathode and the basic series resistor feed (150K), I have a preset wirewound in the cathode of the c3g, as I adjust bias ,so my screen volts vary too ,of course . I was thinking at high signal levels and clipping especially wouldnt the interaction between screen and plate volts cause more intermodulation ? A settable screen voltage which remains constant regardless of how much anode current flows seems like a good plan , at normal opperating levels its probably doesnt make much difference ,but Im worried about recovery time from transient overload might be longer if theres an interaction between plate and screen volts .For me smoothness of overload characteristic and fast recovery is one of my chief design considerations , I dont need the Bax tone stage to drive huge levels on the output but spikes and transients are a fact of life in the real world .
What interaction? Capacitive interaction is non significant, since the screen grid is decoupled. Interaction due to power rail sag is in fact self-limiting, since dropping the screen grid voltage tends to decrease current and thus increases the anode voltage.Tubetec said:
In my experience recovery time is an issue in systems that are submitted to global feedback, when a transient pushes one or several of the circuit parts out of their continuous operation. The conditions for this to happen in a Baxendall circuit are easily predictable and avoidable. The Baxendall network can be hard to drive; that's why some use a cath follower in addition to the gain tube. Thorough analysis of the impedance and drive requirement is de rigueur.
ruffrecords
Well-known member
Hmmm... Looks like a way of stabilizing the plate voltage for optimum voltage swing.ruffrecords said:
Do you know the data for the input xfmr? The impedance at the grid of the pentode is about 4kohms, due to current feedback. Accounting for the 3.3k resistor at the bottom of the secondary, it would see about 7k.
ruffrecords
Well-known member
The transformer secondary is connected from the junction of the 470K and 3K3 to the grid so the transformer secondary still sees the open circuit grid as its load with the 100K in parallel. The transformer secondary voltage in series with the NFB voltage is what the grid sees.abbey road d enfer said:
You can use a similar technique to obtain gain from a cathode follower.
Cheers
Ian
I do not have details of the transformer but I suspect is is 1:10
Wasnt it 'Gardeners' transformers Neve used pre Marinair, with an octal plug , I know a vintage Neve tube console resides at Castle Leslie ,Co Monaghan , along with some other bits of Desmond Leslies studio . I wonder do they have original documentation . I think Rup' had trouble getting paid from Des Les, and swore he'd never build another console without money up front .I guess 30/50 ohm transformers for mic input was more common back in the day.
Its a crafty circuit for sure Ian , it is a little tough to work out everything thats going on in there at first glance .What did he use in the line out amps on this series of Neve ?
Its a crafty circuit for sure Ian , it is a little tough to work out everything thats going on in there at first glance .What did he use in the line out amps on this series of Neve ?
> it is a little tough to work out everything thats going on in there
Simple. The tube works as triode for DC, as Pentode for audio.
Pentode DC bias is always a little uncertain. Triode bias usually hits the mark real close.
I'm ignoring the funny stuff with NFB and input transformer.
Simple. The tube works as triode for DC, as Pentode for audio.
Pentode DC bias is always a little uncertain. Triode bias usually hits the mark real close.
I'm ignoring the funny stuff with NFB and input transformer.
OK, I didn't see the junction of the 470k and the 3.3k; standard voltage FB. Then yes, the secondary sees 100k and makes the xfmr likely a 1:10.ruffrecords said:
ruffrecords
Well-known member
abbey road d enfer said:
Yes, the drawing is not easy to read. Someone has had a go at cleaning up the original. Rupert seemed to be fond of the EF86. The output stage of the mixer is a triode wired EF86 with an 8:1 transformer in the anode circuit.
Cheers
Ian
The ef86 works well as a triode ,I modded up a few Quad 22 preamps and strapped the ef86 ,100k load, 1or 1.5k cathode resistor and arranged switchable feedback back to grid ,works ok with an input transformer on dynamic mics ,if higher input levels are required you can feed the tone stage directly ,its got nice mid notch eq with switchable centre frequencies ,works great for taking the honk off lesser quality dynamics , its all a bit lo-fi by pro standards but on sources that dont have much bandwidth its acceptable. Ef86 doesnt really do it for me in the output stage ,but of course Rupert must have been looking for lowest noise
I understand the Bax's input impedence varies wildly with the settings of the controls ,and that its a touch on the low side for most valves to drive properly ,with limited amounts of cut or boost it doesnt sink too low . I have the option of driving it with either a transformer coupled pre-amp , or a step up or 1:1 transformer depending on signal level .Anything transistorised/op amp based drives it fine ,even though I have +/- 20 db range I almost never need extreme settings of the controls .I know in the world of commercial equipment you cant just have a stage that wonks the impedence like that as an input ,but I do sometimes like to be able to vary the loading on transformers depending on signal level, to dial in small degrees of core saturation on the peaks .With transformer coupled input to the bax, the variable loading ,while not technically correct often gives me the sound I want
There was a revised version of the bax came out later ,transistorised with a peak/shelf control on each band ,22k pots all around, way low for direct valve drive , but I wonder can I scale up the pots x5 and reduce the caps accordingly and make it work with the C3g with 10k anode load .There is one Bax version that uses triode pentode 7199 ,I think, to boost the gain around the feedback network to 1000's of times, I think it allows other semi parametric frequency controls too .
My version is hardwired around the pots and no doubt this adds signifigant pF's , a tiny postage stamp sized board with plastic mini pots and small resistors/caps could help there ,I'd imagine the more I cut the stray capacitance the higher up in frequency the boost will go ,and that could be undesirable, comments or suggestions most welcome.
I understand the Bax's input impedence varies wildly with the settings of the controls ,and that its a touch on the low side for most valves to drive properly ,with limited amounts of cut or boost it doesnt sink too low . I have the option of driving it with either a transformer coupled pre-amp , or a step up or 1:1 transformer depending on signal level .Anything transistorised/op amp based drives it fine ,even though I have +/- 20 db range I almost never need extreme settings of the controls .I know in the world of commercial equipment you cant just have a stage that wonks the impedence like that as an input ,but I do sometimes like to be able to vary the loading on transformers depending on signal level, to dial in small degrees of core saturation on the peaks .With transformer coupled input to the bax, the variable loading ,while not technically correct often gives me the sound I want
There was a revised version of the bax came out later ,transistorised with a peak/shelf control on each band ,22k pots all around, way low for direct valve drive , but I wonder can I scale up the pots x5 and reduce the caps accordingly and make it work with the C3g with 10k anode load .There is one Bax version that uses triode pentode 7199 ,I think, to boost the gain around the feedback network to 1000's of times, I think it allows other semi parametric frequency controls too .
My version is hardwired around the pots and no doubt this adds signifigant pF's , a tiny postage stamp sized board with plastic mini pots and small resistors/caps could help there ,I'd imagine the more I cut the stray capacitance the higher up in frequency the boost will go ,and that could be undesirable, comments or suggestions most welcome.
Attachments
The SS version relies on emitter followers to provide the necessary low Z to drive to both sides of the network. At 1kHz, the impedance is about 10 ohms, varying to about 300 ohms at 20Hz, because of the coupling capacitors.Tubetec said:
I would think you need to scale up about x20 and even then you will not get the same performance, but who cares if you get +/-16 instead of +/-20?
Typical stray capacitance should not be a sensitive issue compared to the nominal values in the network.
So just made a few small changes , including rearranging the final decoupling cap in the ht line ,and moving the choke to directly in series with the anode load resistor . Hum is totally gone from the output even at full bass boost no trace at all.
Will the inductive reactance of the choke in the anode circuit help present a higher anode load to the pentode ?
Will this increase the open loop gain of the stage ,as I think I'm seeing a slightly higher range on the boost controls after the changover.
For now Ive run the screen off the simple series resistor ,tapped off before the choke , what happens if I tap the screen voltage after the choke and before the anode load resistor ,now my screen will be fed a portion of the output signal ,it starts to look a bit like UL connection , if I disconnect the screen decoupling cap can I expect a change in the open loop gain and impedence due to feedback ? How will the LF constants of the coupling and decoupling components play out ?
Anytime I vary the bias point of the stage I see the screen volts trail along behind, in either positive or negative direction following the anode ,also after switch on settling time is minutes ,no major problem in itself, but in the event of transient overload I wonder how does a rock solid reference voltage for the screen play out vs a supply that allows degrees of negative feedback to anode, or could I even have a switching arrangement ,to give three posibilities ,regulator fed , conventional series resistor bypassed to cathode or UL with screen decoupling removed .
I'll sketch out my ideas on paper and post them once I grab a coffee , for claritys sake.
Thanks to Merlin also, I found the valve tracer circuit on his excellent Valvewizard.co.uk site , it has a very simple but effective screen regulator ,just a hand full of components , specifically designed for small signal pentodes , I think I might give it a try .The valve tracer is such a neat machine , I think thats definately going onto the project list as well , the simplicity is remarkable . I can see how it wont make a commercial design as you need to know what your doing with the patch cables or your going to fuse the unit out and probably burn out/damage tubes. Maybe two seperate 9ba sockets ,one for ef86's and one for dual triodes might simplify the patch bay , at the expence of universality of usage by having all electrodes patchable . Seeing as I have an Arb Gen with usb upload of waveform ,can I generate my ramp voltage on that and simplify the build of the unit an order of magnitude simpler ?
Theres a few scenarios in the drawing , I wonder if someone could explain better the implications of the different setups . For now 1Hy will do and it makes calculations easy ,I know its a low value . I did find a way of making a dual bobbin anode load choke by getting two identical small transformers of the usual kind found in old domestic equipment (with the clear plastic that has seperately removable pri/sec half bobbins) . By combining two primary sections which have a dc resistance of a couple of thousand ohms each on the same laminations E/I stacked and gapped, I can get somewhere between 50-100Hy ,but these are very small transformers and probably wont handle much current . Still looks like a very easy and cheap way to create a multi section anode load choke , If you dual wound each half bobbin to get four seperate pi sections you could probably do very well in terms of frequency responce ,swapping out the plain silicon steel lams for something more exotic might get you hundreds of Hy. Its an interesting experiment into Tab/Siemens V style pres and output stages , that costs nothing . Would it be possible to wind an anode choke in such a way that it acts as a centre tapped load to a push pull pair ,then capacitor couple both anodes to a centre tapped mumetal signal transformer?
Sorry about the long and winding post ,any ideas at all that spring to mind graciousy recieved .
Will the inductive reactance of the choke in the anode circuit help present a higher anode load to the pentode ?
Will this increase the open loop gain of the stage ,as I think I'm seeing a slightly higher range on the boost controls after the changover.
For now Ive run the screen off the simple series resistor ,tapped off before the choke , what happens if I tap the screen voltage after the choke and before the anode load resistor ,now my screen will be fed a portion of the output signal ,it starts to look a bit like UL connection , if I disconnect the screen decoupling cap can I expect a change in the open loop gain and impedence due to feedback ? How will the LF constants of the coupling and decoupling components play out ?
Anytime I vary the bias point of the stage I see the screen volts trail along behind, in either positive or negative direction following the anode ,also after switch on settling time is minutes ,no major problem in itself, but in the event of transient overload I wonder how does a rock solid reference voltage for the screen play out vs a supply that allows degrees of negative feedback to anode, or could I even have a switching arrangement ,to give three posibilities ,regulator fed , conventional series resistor bypassed to cathode or UL with screen decoupling removed .
I'll sketch out my ideas on paper and post them once I grab a coffee , for claritys sake.
Thanks to Merlin also, I found the valve tracer circuit on his excellent Valvewizard.co.uk site , it has a very simple but effective screen regulator ,just a hand full of components , specifically designed for small signal pentodes , I think I might give it a try .The valve tracer is such a neat machine , I think thats definately going onto the project list as well , the simplicity is remarkable . I can see how it wont make a commercial design as you need to know what your doing with the patch cables or your going to fuse the unit out and probably burn out/damage tubes. Maybe two seperate 9ba sockets ,one for ef86's and one for dual triodes might simplify the patch bay , at the expence of universality of usage by having all electrodes patchable . Seeing as I have an Arb Gen with usb upload of waveform ,can I generate my ramp voltage on that and simplify the build of the unit an order of magnitude simpler ?
Theres a few scenarios in the drawing , I wonder if someone could explain better the implications of the different setups . For now 1Hy will do and it makes calculations easy ,I know its a low value . I did find a way of making a dual bobbin anode load choke by getting two identical small transformers of the usual kind found in old domestic equipment (with the clear plastic that has seperately removable pri/sec half bobbins) . By combining two primary sections which have a dc resistance of a couple of thousand ohms each on the same laminations E/I stacked and gapped, I can get somewhere between 50-100Hy ,but these are very small transformers and probably wont handle much current . Still looks like a very easy and cheap way to create a multi section anode load choke , If you dual wound each half bobbin to get four seperate pi sections you could probably do very well in terms of frequency responce ,swapping out the plain silicon steel lams for something more exotic might get you hundreds of Hy. Its an interesting experiment into Tab/Siemens V style pres and output stages , that costs nothing . Would it be possible to wind an anode choke in such a way that it acts as a centre tapped load to a push pull pair ,then capacitor couple both anodes to a centre tapped mumetal signal transformer?
Sorry about the long and winding post ,any ideas at all that spring to mind graciousy recieved .
Attachments
Yes it will, in a frequency-dependant manner.Tubetec said:
That's correct.
Correct.
Indeed in order to produce the "UL" effect, the screen should not be decoupled.
That depends very much on the relative values of the plate resistor and the choke. The "UL" effect is deemed beneficial when the percentage of AC screen voltage to the AC plate voltage is between 20% and 50%. Since the ratio depends on the choke/plate res impedance ratio, there will be little effect at low frequencies (vg2=0, or Vg2=constant)=> pentode-ish operation, at high frequencies, the impedance of the choke is very high, then vg2=va => triode-ish operation. In-between, various degrees of "UL" operation.
