Cool. One thing that might help is to have in mind that at RF (where the nominal oscillation is occurring) that's a lot of high impedance wiring a long way from ground. Potentially makes what the radio folks call a Hartley oscillator.
Also, 100 Ohms is way too small for a 2A3 grid stop. Maybe start with 1 K or several K Ohms to keep it damped.
All good fortune,
Chris
Also, 100 Ohms is way too small for a 2A3 grid stop. Maybe start with 1 K or several K Ohms to keep it damped.
All good fortune,
Chris
I recall when Tubelab was tutoring me through the design of my 6L6 amp, he was working on a 'universal driver' circuit that had a differential pair with 10M45 constant current devices for the anode loads and for the tail of the differential pair. If I recall correctly, he had difficulty getting this arrangement to work as it oscillated. Replacing the anode loads with resistors cured the problem.
I also note on pages 142-144 of Morgan Jones 'Valve Amplifiers' 4th edition (yes the new one that there are no copies of...) under the topic 'Using Transistors as Active Loads for Valves' the he describes a differential pair with active loads for the anodes and a constant current sink on the tail. He states:
It would appear that the anode constant current sources in your circuit are superior to the simple LM317 CCS of the tail of the pair. I would suggest this is the root of your troubles.
Regards,
Chris
I also note on pages 142-144 of Morgan Jones 'Valve Amplifiers' 4th edition (yes the new one that there are no copies of...) under the topic 'Using Transistors as Active Loads for Valves' the he describes a differential pair with active loads for the anodes and a constant current sink on the tail. He states:
It would appear that the anode constant current sources in your circuit are superior to the simple LM317 CCS of the tail of the pair. I would suggest this is the root of your troubles.
Regards,
Chris
Rod, the MKPs I find at my supplier are for high voltage, big and expensive for an experiment. How about stacked ceramic. I'm thinking X7R.
Thanks
Ian, The Arcotronics 82D 5mm-pitch MKT are widely available, if MKP is out of reach. You could try the ceramics, and they will probably work in the circuit - but X7R has fairly poor stability and may affect the sound somewhat. But you can always refine the parts after getting to the point of a working circuit....
Thanks Rod. Most of the attraction was that the X7R's are 1/4 the price plus small and easy to solder directly to the 547C leads. Will try the circuit tomorrow.
I did try biasing them up before I took them out. White noise still there but with buzz added to the mix. Also tried pulling all the cathode coupling caps out and got no audio at all. Then followed Rod's suggestion to pull the CCS and try cathode bias 150R. I did that and then added bypass cap. No white noise. Sounds good though I don't think it's quite as good as with the 317 set below oscillation point.
I'll throw in here that I'm also wondering about my SVCS board layout. I say that because of the slightly erratic movement of voltages as I set it. It still happens with the cathode R's in. I have confidence in Wavebourn's SVCS circuit per se (I've built a few versions so far) , so will probably have to try a new layout.
Does this help?
Posted by John Swenson (A) on November 3, 2008 at 16:55:49
In Reply to: ccs PLATE loads on long-tail-pair posted by nikongod on November 2, 2008 at 18:10:15:
What I think you want to do is called folding. I don't have a schematic editor handy right now to draw the picture, that will have to wait till I get home.
The verbal description is:
LTP with CCS on tail as shown. Another CCS on each plate directly connected to the plate, no resistor. The resistor goes from the plate to ground. The plate CCSes are set to a current greater than half the LTP CCS. The difference in current flows through the load resister which determines the DC set point.
Here is a concrete example. Lets say the tail is set for 10mA. Thats 5 mA through each each side of the pair. Then set each plate CCS for say 7mA. 7 - 5 = 2 mA. Thus there will be 2 mA flowing through each load resistor. If you have 50K load resistors that puts 100V across the resistor, which sets the ground to plate voltage at 100V. If you use an adjustable CCS for the plate CCS you can easilly trim the plate DC voltage without changing the B+.
There are some big advantages to this approach:
The load resistor is now referenced to ground rather than B+, radically increasing the PSRR.
Its a piece of cake to dial in just the right voltage on the plate, which is great for DC coupled stages.
You can decouple tube current, load resistor value and B+ requirements. For example, lets say your amplifier works best with 10mA through the tube and you want a 100K load resistance. If you did that with just a simple plate resistor you would need a B+ of over a thousand volts! With this arrangement you can set the plate CCS to be 1 mA greater than the tube current and which sets the plate at 100V and you only need 120V or so B+.
This works very well, is very stable and sounds great. The problem is complexity, you need lots of CCSes.
I use this approach a lot for hi gain pentode stages. The pentode behaves pretty much like an ideal transconductance amplifier. But the ones I like take high currents to get very low noise and very low distortion. To get high gain you need high value load resistors which need very high B+ if just using plate resistors. With folding its very easy to get a lot of gain with one stage.
I hope this makes sense.
John S.
Posted by John Swenson (A) on November 3, 2008 at 16:55:49
In Reply to: ccs PLATE loads on long-tail-pair posted by nikongod on November 2, 2008 at 18:10:15:
What I think you want to do is called folding. I don't have a schematic editor handy right now to draw the picture, that will have to wait till I get home.
The verbal description is:
LTP with CCS on tail as shown. Another CCS on each plate directly connected to the plate, no resistor. The resistor goes from the plate to ground. The plate CCSes are set to a current greater than half the LTP CCS. The difference in current flows through the load resister which determines the DC set point.
Here is a concrete example. Lets say the tail is set for 10mA. Thats 5 mA through each each side of the pair. Then set each plate CCS for say 7mA. 7 - 5 = 2 mA. Thus there will be 2 mA flowing through each load resistor. If you have 50K load resistors that puts 100V across the resistor, which sets the ground to plate voltage at 100V. If you use an adjustable CCS for the plate CCS you can easilly trim the plate DC voltage without changing the B+.
There are some big advantages to this approach:
The load resistor is now referenced to ground rather than B+, radically increasing the PSRR.
Its a piece of cake to dial in just the right voltage on the plate, which is great for DC coupled stages.
You can decouple tube current, load resistor value and B+ requirements. For example, lets say your amplifier works best with 10mA through the tube and you want a 100K load resistance. If you did that with just a simple plate resistor you would need a B+ of over a thousand volts! With this arrangement you can set the plate CCS to be 1 mA greater than the tube current and which sets the plate at 100V and you only need 120V or so B+.
This works very well, is very stable and sounds great. The problem is complexity, you need lots of CCSes.
I use this approach a lot for hi gain pentode stages. The pentode behaves pretty much like an ideal transconductance amplifier. But the ones I like take high currents to get very low noise and very low distortion. To get high gain you need high value load resistors which need very high B+ if just using plate resistors. With folding its very easy to get a lot of gain with one stage.
I hope this makes sense.
John S.
Sorry Shoog, Quick add. My language above wasn't right.
When, with the 317's biased up, I pulled all the intercathode coupling caps and played, the noise level was so high I have no idea whether any audio signal was getting through or not. The thought I had at the time was that perhaps the coupled cathodes had been allowing the two tubes to cancel a large portion of the noise.
Thanks for that. Interesting to see that others do it. I did try it as posted earlier but it didn't seem to get rid of the problem in this particular case.
More after trying a new tail CCS tomorrow or Tues.
If the stage is stable with cathode resistors, it should be easy to get it working with the BC547C CCS.
The LM317 has a high gain error-amplifier and noisy reference diode that cause nothing but trouble in your application. And the error-amplifier cannot easily be frequency-compensated externally.
The twin-BC547C CCS has no error amplifier beyond the action of the single NPN per channel - and compensation caps C2 & C3 can take any value that brings the stability required. Since the CCS is bypassed by big caps, there is no requirement for the CCS to respond across a wide bandwidth, and it can therefore be rolled off at lower than usual frequency. Maybe try a few very different C2-C3 values, and listen out for their impact on the sound.... maybe there will be very little.
The CCS self-noise will mostly be the BC547's. At 30mA this will be very low, probably 100x lower than the 317's horrible bandgap diode.
The LM317 has a high gain error-amplifier and noisy reference diode that cause nothing but trouble in your application. And the error-amplifier cannot easily be frequency-compensated externally.
The twin-BC547C CCS has no error amplifier beyond the action of the single NPN per channel - and compensation caps C2 & C3 can take any value that brings the stability required. Since the CCS is bypassed by big caps, there is no requirement for the CCS to respond across a wide bandwidth, and it can therefore be rolled off at lower than usual frequency. Maybe try a few very different C2-C3 values, and listen out for their impact on the sound.... maybe there will be very little.
The CCS self-noise will mostly be the BC547's. At 30mA this will be very low, probably 100x lower than the 317's horrible bandgap diode.
The CCS versions I have always used are the LED biased Pinkmouse ones. These are simple and high impedance at all frequencies.
http://www.google.com/url?sa=t&rct=...sg=AFQjCNG2OnP0s_JVj2R4LB5hr2UVbYbOKA&cad=rja
The cathode to cathode caps are the only means by which the signal can be generated by this configuration - take them out and you get the only result possible - no signal.
I am a little surprised that biasing the cathodes higher caused more noise as there is absolutely no reason why it should do and it has never done so for any of my implementations which have used a similar approach. It actually works to confirm the cathode CCS as the problem as giving them more voltage to work properly has caused them to generate more noise.
At this stage I really think that a good stiff CCS will resolve all the issues, and it should be possible to build the Pinkmouse one from pulled parts in your spares pile.
Shoog
http://www.google.com/url?sa=t&rct=...sg=AFQjCNG2OnP0s_JVj2R4LB5hr2UVbYbOKA&cad=rja
The cathode to cathode caps are the only means by which the signal can be generated by this configuration - take them out and you get the only result possible - no signal.
I am a little surprised that biasing the cathodes higher caused more noise as there is absolutely no reason why it should do and it has never done so for any of my implementations which have used a similar approach. It actually works to confirm the cathode CCS as the problem as giving them more voltage to work properly has caused them to generate more noise.
At this stage I really think that a good stiff CCS will resolve all the issues, and it should be possible to build the Pinkmouse one from pulled parts in your spares pile.
Shoog
Last edited:
Just curious Rod, would this be a good place to use a variable (trimmer) cap?
Like I said before. I can't guarantee that some of the problems that arise aren't from poor implementation on my part.
Thanks for the link Shoog. I actually have a bunch of boards already made up with that circuit (known as C4S) from about 10 years ago - first done by a fellow from Texas named John Camille. There's an article by him in an old Sound Practices on an 811 amp. . . . . I might try resurrecting them with different BJT's than in the original sometime but I have found that they add something to the sound that bothers me over time. That's why I'm interested in trying alternatives.
Ian, No need for a trimmer.
Values like 1n, 10n, 100n will give an idea of what works best, but within each range, the actual value won't be very critical.
We're just calming the bandwidth of the CCS so that the high gm of the 8233s does not create oscillation with the CCS. Above a certain value of C2/3, all should be under control. We can then simply double the value, for the final circuit or triple it - and know that we have a safety margin.
But the BC547C CCS will be much more stable than the LM317 anyway, and it is possible that even without a capacitor it will be stable.
Values like 1n, 10n, 100n will give an idea of what works best, but within each range, the actual value won't be very critical.
We're just calming the bandwidth of the CCS so that the high gm of the 8233s does not create oscillation with the CCS. Above a certain value of C2/3, all should be under control. We can then simply double the value, for the final circuit or triple it - and know that we have a safety margin.
But the BC547C CCS will be much more stable than the LM317 anyway, and it is possible that even without a capacitor it will be stable.
One more question before you turn things off for the night and I start putting the circuit together. Anything wrong with putting it on the already built 16-17V tap under the output tail CCS?
Thanks
Hi Ian, do you want some rolloff on the output tail CCS? This one is not bypassed with big caps, like the input, so it may have an influence on sound. But you could try 22n up to around 68nF from drain to gate of the upstairs FET. I suspect it will not be likely to be useful - but no harm in trying.
BTW, if you have some C4S boards, they will easily support the low-voltage compatible BC547C CCS. Put 1N4148s in place of the LEDs, and mount the upstairs NPN only. Take the emitter of this NPN via a 15R to ground. Please avoid the series-connected trimmer in the emitter circuit - 30mA through one of these is a noisy business.
BTW, if you have some C4S boards, they will easily support the low-voltage compatible BC547C CCS. Put 1N4148s in place of the LEDs, and mount the upstairs NPN only. Take the emitter of this NPN via a 15R to ground. Please avoid the series-connected trimmer in the emitter circuit - 30mA through one of these is a noisy business.
No that wasn'at my intent. It's simply that I had put resistors under the output tail CCS to obtain a relatively steady ~ 18V for biasing the input transformer in order to get higher voltage across the LM317s. When I pulled out the 317s I left the resistors there in case I wanted the voltage for something later on and as a side benefit lower the power dissipation on the CCSs. I thought that if 18V wasn't too high I could use it for your circuit as well.
Parts are here. I'll try it and see.
I'm putting this note here so there's some record of what's been done to date.
Two nights ago, seeing that the voltage output of the SVCS does not transition smoothly (eg. turning voltage down from 130V to 125V it runs like : 130, 129, 128, 127676767, 126, 12898989, 123, 12323232 126 , 125) I wondered about my board layout and decided to put a base stopper on Q3. This helped to tighten things up and made them slightly more lineal but not by much and certainly not good enough.
The sound became a little more abrasive or hard on middle and high frequency peaks and as in my experience grid stoppers have a tendency to do the same thing as value increases I took the stoppers out of the 2A3's and returned the 8233 stoppers to their original 100R value. The helped relieve the unpleasant sound but now the white noise was back.
OK, everybody told me I needed the grid stoppers. So I put them back in and removed the base stopper on Q3, returning the circuit to its previously dead silent configuration. However, now the white noise was back.
Next day I removed the driver cathode bias resistors and installed Rod's BC547C CCS. It works well but still the white noise. I believe the sound is improved. More clear and detailed though LF output is a little down.
The problem with setting the SVCS output voltage remains . This led me last night to separating the two output tail CCS's, putting one under each 2A3 and cap coupling the two cathodes as in the driver stage. This sounds really really good and has the curious sonic effect of being in your listening attention while you are doing anything else , ie, reading , eating, washing dishes, working, talking. It's an odd effect and one I've never encountered before. Complete attention on what you're doing, yet the music is being listened to (not just heard) with discrimination yet apparently not with any effort or even intent.
White noise still there and SVCS adjustablty is still not resolved. I now think of Sheldon's description of the SVCS being at its base a standard constant current source. I think that the skittish behaviour is because of a three way conversation between the output CCS, the driver plate/output tube grid governing SVCS, and the output tube/s. It doesn't seem to matter whether the output tubes are in LTP or with separate ccs and cap coupled.
Sometime soon I'll replace the two 2a3 CCS with cathode resistors and see.
Thanks
Two nights ago, seeing that the voltage output of the SVCS does not transition smoothly (eg. turning voltage down from 130V to 125V it runs like : 130, 129, 128, 127676767, 126, 12898989, 123, 12323232 126 , 125) I wondered about my board layout and decided to put a base stopper on Q3. This helped to tighten things up and made them slightly more lineal but not by much and certainly not good enough.
The sound became a little more abrasive or hard on middle and high frequency peaks and as in my experience grid stoppers have a tendency to do the same thing as value increases I took the stoppers out of the 2A3's and returned the 8233 stoppers to their original 100R value. The helped relieve the unpleasant sound but now the white noise was back.
OK, everybody told me I needed the grid stoppers. So I put them back in and removed the base stopper on Q3, returning the circuit to its previously dead silent configuration. However, now the white noise was back.
Next day I removed the driver cathode bias resistors and installed Rod's BC547C CCS. It works well but still the white noise. I believe the sound is improved. More clear and detailed though LF output is a little down.
The problem with setting the SVCS output voltage remains . This led me last night to separating the two output tail CCS's, putting one under each 2A3 and cap coupling the two cathodes as in the driver stage. This sounds really really good and has the curious sonic effect of being in your listening attention while you are doing anything else , ie, reading , eating, washing dishes, working, talking. It's an odd effect and one I've never encountered before. Complete attention on what you're doing, yet the music is being listened to (not just heard) with discrimination yet apparently not with any effort or even intent.
White noise still there and SVCS adjustablty is still not resolved. I now think of Sheldon's description of the SVCS being at its base a standard constant current source. I think that the skittish behaviour is because of a three way conversation between the output CCS, the driver plate/output tube grid governing SVCS, and the output tube/s. It doesn't seem to matter whether the output tubes are in LTP or with separate ccs and cap coupled.
Sometime soon I'll replace the two 2a3 CCS with cathode resistors and see.
Thanks
Replaced 2A3 cathode CCS's with cathode R's. Still low white noise. Vout adjust on SVCS still a little jumpy. I think it might be worth it to try a new layout for the SVCS.
Rod's little CCS circuit is working fine and sounds good.
One other point . The 1M resistor on the neg of the inter-cathode coupling caps shows full cathode voltage at startup. On the 2A3's that puts it up to ~170V. Then it very slowly discharges. Now at 2 hours on it is down to 23VDC.
Rod's little CCS circuit is working fine and sounds good.
One other point . The 1M resistor on the neg of the inter-cathode coupling caps shows full cathode voltage at startup. On the 2A3's that puts it up to ~170V. Then it very slowly discharges. Now at 2 hours on it is down to 23VDC.
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.