TonyTecson,
Thanks for the information. I was considering using the 329-B or 832-A in the future. They do have a certain visual appeal, might still try that. But . . . I already have a pair of 5894 tubes. And they are good looking too. Hmmm . . .
A couple of 10 Ohm current ‘sense’ resistors in the plates would make it easy(*) to check the plate current balance in the output transformer primary halves.
Well, not easy(*) . . . you have to be careful of the high voltage ‘test points’.
Thanks for the information. I was considering using the 329-B or 832-A in the future. They do have a certain visual appeal, might still try that. But . . . I already have a pair of 5894 tubes. And they are good looking too. Hmmm . . .
A couple of 10 Ohm current ‘sense’ resistors in the plates would make it easy(*) to check the plate current balance in the output transformer primary halves.
Well, not easy(*) . . . you have to be careful of the high voltage ‘test points’.
yes, and Pete Millet did something about plate current monitoring...
i was never able to make the 829b in parallel SET work good, but in push-pull they are indeed awesome....
i made a 5894 set years ago and the owner in fact liked it better than the 2A3 set that i also made for him....
i was never able to make the 829b in parallel SET work good, but in push-pull they are indeed awesome....
i made a 5894 set years ago and the owner in fact liked it better than the 2A3 set that i also made for him....
Yes - that's what I am using, though I don't show them on my schematic. I just measure the voltage across the 10R resistors and use the potentiometer to adjust the gains to make the currents balance.
I just tried this. It works well - when the current imbalance is 5mA the voltage difference across the plates is 140mV. I can achieve a very good balance by bringing the difference down to a a few mV.
Couldn't agree more! I try to be very careful, never touching meter leads etc even if I believe they are low voltage. When I put this amp in a chassis I am contemplating adding an arduino with voltage/current sensors and a nice display. I'll have it display balance information for both channels as well as a few other thing such as Bias, Screen, B+ and total current draw.
Output transformers have the same number of turns in the 2 halves of the primary.
But they usually do not have exactly the same DCR in those 2 halves.
Therefore, just checking that the plate to plate voltage is 0V, will not test the current balance in a push pull output stage.
Example, DCR end #1 to center tap is 100 Ohms,
and DCR end #2 to center tap is 115 Ohms.
Suppose there is 115mA in plate #1, and 100mA in plate #2.
115mA x 100 Ohms = 11.5V
100mA x 115 Ohms = 11.5V
Plate to plate voltage is 0V, but the current is unbalanced by 15mA.
Instead, use a precision 10 Ohm resistor in each plate lead, and measure the voltages across each resistor.
If you use 1% resistors, and the voltage drops are exactly equal, then the maximum error of the reading is 2%. For 100mA plate current, that is 2mA error in what appears to be an exact match. That is a lot better than 15mA error.
Use 0.1% resistors, and there may be as much error in the repeatability of your meter making two measurements, versus the resistor matching.
But they usually do not have exactly the same DCR in those 2 halves.
Therefore, just checking that the plate to plate voltage is 0V, will not test the current balance in a push pull output stage.
Example, DCR end #1 to center tap is 100 Ohms,
and DCR end #2 to center tap is 115 Ohms.
Suppose there is 115mA in plate #1, and 100mA in plate #2.
115mA x 100 Ohms = 11.5V
100mA x 115 Ohms = 11.5V
Plate to plate voltage is 0V, but the current is unbalanced by 15mA.
Instead, use a precision 10 Ohm resistor in each plate lead, and measure the voltages across each resistor.
If you use 1% resistors, and the voltage drops are exactly equal, then the maximum error of the reading is 2%. For 100mA plate current, that is 2mA error in what appears to be an exact match. That is a lot better than 15mA error.
Use 0.1% resistors, and there may be as much error in the repeatability of your meter making two measurements, versus the resistor matching.
i can make opt's that have halves that are within an ohm of each other, this forum can tell you how it is done, just do your search, it is here...
TonyTecson,
It is great to have really good output transformers. Matched DCR is great.
I only wrote post # 49 for three reasons:
1. Some will use commercial transformers, and some of those have DCRs that are not very well matched, most often some of the less expensive ones that is all a few builders can afford. Not everybody is going to wind their own.
2. A simple method to do reasonably accurate current measurements when the DCRs do not match.
3. A reminder that ultra precision relies not only on precision resistors, but also on the repeatability of the measurement device. Some think that if you measure a quantity, that it is in fact accurate, when it may not be nearly as accurate as they think.
It is great to have really good output transformers. Matched DCR is great.
I only wrote post # 49 for three reasons:
1. Some will use commercial transformers, and some of those have DCRs that are not very well matched, most often some of the less expensive ones that is all a few builders can afford. Not everybody is going to wind their own.
2. A simple method to do reasonably accurate current measurements when the DCRs do not match.
3. A reminder that ultra precision relies not only on precision resistors, but also on the repeatability of the measurement device. Some think that if you measure a quantity, that it is in fact accurate, when it may not be nearly as accurate as they think.
Last edited by a moderator:
tbh, i never heard any difference in sound with a dc balanced opt versus non balanced opts.....
it is the cathode currents that determine voltage drops in the primary opt halves...there will be slight difference in plate voltage but not so much...
a 10 ohm resistor in series with each plate i believe is the best way to do it, although care full attention to safety is a must...
it is the cathode currents that determine voltage drops in the primary opt halves...there will be slight difference in plate voltage but not so much...
a 10 ohm resistor in series with each plate i believe is the best way to do it, although care full attention to safety is a must...
I found that with the toroidal power transformers used as OPTs a DC imbalance of a few mA made very little difference to the performance. What I saw was a slight increase in THD below 100Hz. I'll run some plots and post the results.
I found that balancing the plate voltages did achieve a current balance as measured across 10R (1%) resistors in the anode supply. As pointed out that probably means the DCR is equal in the two primaries.
In the end, since I have an audio analyzer suite I can easily check the effects of DC imbalance for any design/OPT I have and I can design accordingly.
Thanks for all the feedback.
I found that balancing the plate voltages did achieve a current balance as measured across 10R (1%) resistors in the anode supply. As pointed out that probably means the DCR is equal in the two primaries.
In the end, since I have an audio analyzer suite I can easily check the effects of DC imbalance for any design/OPT I have and I can design accordingly.
Thanks for all the feedback.
With my design I'm not sure that this is true. The cathode currents include screen grid currents coming from the bias supply. These currents have not passed through the OPT. That's the reason I am measuring anode currents and not cathode currents. Am I missing something?
Richard
Looking forward to seeing the results. I was very surprised to find a 25mA misbalance that I couldn't detect with my ears at least at low power anyway.
mvstillwater,
Good point. You are correct.
For non-triode-wired pentodes and beam power tubes, the screen current is not included in the ends of the primary windings.
For Ultra Linear, the screen currents are included . . . but they are at a different point on the secondary, often at 40%. Example:
Push Tube has 45 mA x 100% and 5 mA x 40% = effectively 47mA x 100%
Pull Tube has 47 mA x 100% and 3 mA x 40% = effectively 48.2mA x 100%
50mA in each cathode . . . but 1.2mA difference in Amp x Turns Product (in the laminations).
When I use triode-wired pentodes and triode-wired beam power tubes, I use individual cathode self bias resistors. The individual cathode currents are easy to calculate from the cathode voltage/cathode resistance. And the cathode current is the Total current in the ends of the primary windings. Individual self bias is not possible on common cathode tubes like the 829-B.
Good point. You are correct.
For non-triode-wired pentodes and beam power tubes, the screen current is not included in the ends of the primary windings.
For Ultra Linear, the screen currents are included . . . but they are at a different point on the secondary, often at 40%. Example:
Push Tube has 45 mA x 100% and 5 mA x 40% = effectively 47mA x 100%
Pull Tube has 47 mA x 100% and 3 mA x 40% = effectively 48.2mA x 100%
50mA in each cathode . . . but 1.2mA difference in Amp x Turns Product (in the laminations).
When I use triode-wired pentodes and triode-wired beam power tubes, I use individual cathode self bias resistors. The individual cathode currents are easy to calculate from the cathode voltage/cathode resistance. And the cathode current is the Total current in the ends of the primary windings. Individual self bias is not possible on common cathode tubes like the 829-B.
kodabmx,
I once had a Dyna Stereo 70 with the old 'black beauty' coupling caps from the concertina splitter to the EL34 Outputs. One 'black beauty' was a little leaky, so there was voltage on one EL34 grid. One EL34 had 60mA, and the other EL34 had 40mA, the single potentiometer bias voltage was adjusted until the voltage across the 15.6 Ohms was 1.56V. So, the total current was adjusted to be correct, . . . 100mA.
This old circuit was designed for easy biasing when all most users were lucky to have an analog Volt Ohm Meter (VOM). They would see where the meter needle was when measuring a common C or D cell; then adjust the potentiometer until the needle was at the same point when measuring the voltage across the 15.6 Ohm resistor.
But because the two cathodes had a Common (single) 15.6 Ohm resistor, you could not detect the fact that the two cathode currents were very poorly balanced.
Negative Feedback 'fixed all that', and you would not hear the problem when the amp was operated at low volume. It took a little careful investigation to see what was causing the real problem in loud low bass notes.
I would bet that with no negative feedback, that even at low volumes, you would hear the difference between two EL34 tubes with 50mA and 50mA, versus two EL34 tubes at completely different operating points (60mA and 40mA).
For those who never heard a difference between balanced and unbalanced primary currents, how many of those amps had negative feedback 'fixing' the problem? And with or without negative feedback, how many mA was the current balance?
Does anybody want to take their Dyna Stereo 70 and modify it, with no negative feed back, and Left channel with 50mA 50mA, and Right channel with 60MA 40mA. Just install a second bias pot for one of the channels to get unbalanced currents. This Ultra Linear amp really should have negative feedback; So to make it have a little better damping factor, and to be a little lower distortion at low volumes, just triode-wire the outputs, and remove the negative feedback. Please let us know how it sounds.
I would just love to hear that we need not pay such close attention to these details of circuits and operating points. "I can't hear a difference" will allow us all to be less careful.
I once had a Dyna Stereo 70 with the old 'black beauty' coupling caps from the concertina splitter to the EL34 Outputs. One 'black beauty' was a little leaky, so there was voltage on one EL34 grid. One EL34 had 60mA, and the other EL34 had 40mA, the single potentiometer bias voltage was adjusted until the voltage across the 15.6 Ohms was 1.56V. So, the total current was adjusted to be correct, . . . 100mA.
This old circuit was designed for easy biasing when all most users were lucky to have an analog Volt Ohm Meter (VOM). They would see where the meter needle was when measuring a common C or D cell; then adjust the potentiometer until the needle was at the same point when measuring the voltage across the 15.6 Ohm resistor.
But because the two cathodes had a Common (single) 15.6 Ohm resistor, you could not detect the fact that the two cathode currents were very poorly balanced.
Negative Feedback 'fixed all that', and you would not hear the problem when the amp was operated at low volume. It took a little careful investigation to see what was causing the real problem in loud low bass notes.
I would bet that with no negative feedback, that even at low volumes, you would hear the difference between two EL34 tubes with 50mA and 50mA, versus two EL34 tubes at completely different operating points (60mA and 40mA).
For those who never heard a difference between balanced and unbalanced primary currents, how many of those amps had negative feedback 'fixing' the problem? And with or without negative feedback, how many mA was the current balance?
Does anybody want to take their Dyna Stereo 70 and modify it, with no negative feed back, and Left channel with 50mA 50mA, and Right channel with 60MA 40mA. Just install a second bias pot for one of the channels to get unbalanced currents. This Ultra Linear amp really should have negative feedback; So to make it have a little better damping factor, and to be a little lower distortion at low volumes, just triode-wire the outputs, and remove the negative feedback. Please let us know how it sounds.
I would just love to hear that we need not pay such close attention to these details of circuits and operating points. "I can't hear a difference" will allow us all to be less careful.
Last edited:
Interesting comments on GNFB "fixing everything". When re-run my tests on the effects of current imbalance I'll run some with no feedback. I hope to get to this on the weekend and post the graphs.
no, just that cathode currents is in direct relation to plate currents,
you are correct to measure plate currents instead with the 832/829b tubes...
a school of thought says, gnfb when done right makes for an even better amp.
i am of the belief that gnfb should not be a "be all and end all", it is just a tool...
careful attention to each stage, local feedback, wide open loop gain bandwith before application of gnfb...
make sure the amp is unconditionally stable even before application of gnfb...
IMHO, opt core saturation at the low frequencies is still the thing to reckon with as far as amps with output transformers are concerned...
edit: Garry Pimm's Tabor amp is an all pentode amp that used no gnfb....
With the 6c33 and its low plate resistance, making amps with OPT and no
gnfb is easy...
- Status
- Not open for further replies.