Dynaco ST-70 Project
- Thread starter SwiftEnthusiast
- Start date
Which resistor is the bleeder?
What's your cathode bias voltage on the EF86? It's possible that you can't shove through enough voltage to the grid to get decent swing at the plate. Unfortunately upping the bias voltage with a larger value cathode resistor will decrease plate current, and you are already undoubtedly running very, very little plate current, so some rework on that triode input stage is worth investigation.
SwiftEnthusiast
Site Supporter
The bleeder resistor is the 330 kOhm that connects between the input stage V+ and the EF86 cathode.
The EF86 cathode bias is 1.614 Volts on the left channel and very similar on the right (away from computer at the moment).
Thank you for your insight. Thinking about that.
The EF86 cathode bias is 1.614 Volts on the left channel and very similar on the right (away from computer at the moment).
Thank you for your insight. Thinking about that.
SwiftEnthusiast
Site Supporter
I tried some further load line permutations, and I was only able to slightly improve on my prior examples (higher V+ combined with that same moderate reduction in the anode resistor). But I think the results were still essentially as you said - right on the edge of being able to get away with it. So, I also started exploring variations with more significant reductions in anode resistors. I have been thinking about the configuration below…
Increased V+. Anode resistor reduced to 100 kOhm. Significant increase in operating voltages and currents. The calculator estimates similar THD at 14 Vrms and 28 Vrms (compared to prior examples). Based on the Mullard charts and some basic valve math (with a lot of help from a great book I am reading: "Valve Amplifiers" by Morgan Jones), it seems the gain reduction would not be too significant. Note that the headroom shown still represents a 25% increase beyond where I am running into problems now.
Does this seem like a good operating point? Or at least a reasonable one to trial?
Note that I am not trying to achieve any specific objective THD measurement. It's just that I’ve only heard this with the operating point that resulted from the original component values, so the estimated THD at that operating point serves as a relative point of reference for me as I consider others. I am willing to stray away from that though - it's actually something else I would like to explore at some point.
Some thoughts about the bleeder resistors...
If I were to scale up the bleeder resistor currents to maintain the original proportion of bleeder current to cathode current, the power dissipated in them would increase significantly (roughly 1/2 Watt -> 4 Watts, total for both channels). I believe the current ratio was roughly 2:1 in the original design. Then triode strapping with the original components resulted in a current ratio of roughly 1:1. If I were to leave the bleeder resistor as is while making other changes that further increase the cathode current, the ratio would fall further and its effects further diminished. Restoring the original ratio would increase the bleeder resistor current roughly 5x.
For scenarios like this one where there is both more margin and power dissipation, I think it might make sense to leave the bleeder resistors at their original values (with the lower ratio and reduced effect) or just remove them altogether. If it turns out more bleed current is needed or desired, it is relatively straightforward to revise that later by altering one or two resistors per channel (may also need to change a cathode resistor, depending on the details).
SwiftEnthusiast
Site Supporter
Decided to trial that most recent operating point. I didn’t have 100 kOhm resistors on hand with sufficient power rating, so I used two 47 kOhm in series. Other than that, it is per my prior post. Bleeder resistors removed for this trial.
Quick initial observations: Gain is down as anticipated, but not by as much (only down about 1 dB in a quick 1 kHz test). A quick output power test (117 Volt / 8 Ohm scenario) showed it is up to about the level originally anticipated in that scenario. Sounds really good with a few songs I sampled so far. More to come once I have more time to listen and measure.
Still very open to suggestions and feedback.
Edit: I have realized that while this new first stage operating point has improved headroom in that stage, it has reduced it in the following phase splitter. The result seems to be a net improvement in headroom through those first two stages as amp voltage / power output has increased. I believe the phase splitter is now the limiting factor up to that point. Need to investigate and think about this further, but seems like it might be a good idea to back off the first stage anode operating voltage to provide a better compromise of the operating points for both stages. Back to the drawing board.
Quick initial observations: Gain is down as anticipated, but not by as much (only down about 1 dB in a quick 1 kHz test). A quick output power test (117 Volt / 8 Ohm scenario) showed it is up to about the level originally anticipated in that scenario. Sounds really good with a few songs I sampled so far. More to come once I have more time to listen and measure.
Still very open to suggestions and feedback.
Edit: I have realized that while this new first stage operating point has improved headroom in that stage, it has reduced it in the following phase splitter. The result seems to be a net improvement in headroom through those first two stages as amp voltage / power output has increased. I believe the phase splitter is now the limiting factor up to that point. Need to investigate and think about this further, but seems like it might be a good idea to back off the first stage anode operating voltage to provide a better compromise of the operating points for both stages. Back to the drawing board.
Last edited:
SwiftEnthusiast
Site Supporter
I have travelled a little farther down the path since my last post...
The load line and operating point in my last two posts reduced headroom in the phase splitter stage. When I trialled it, I saw that there was significant peak distortion at the phase splitter outputs (only on one side of the waveform) when the amp was driven to the onset of clipping. I thought that might be due to pushing the phase splitter near a limit, especially since I had just moved its operating point closer to one limit.
Edited this paragraph to distil and correct after further digesting my observations and notes: I believe some of the load line / operating point combinations I have trialled are indeed pushing a phase splitter limit (0 Volt grid line). I believe that was not causing any problems I was aware of in the output signals because it impacts a region of the waveform where the power tube being driven is inactive. When the amp is driven further, into clipping at the output, the opposite peaks at the phase splitter outputs also start clipping.
Edit followup: As I continue through the learning curve, I am getting a better understanding of the limit that was being pushed in the phase splitter and at least one potential implication. I believe pushing the grid voltage near or beyond the 0 Volt line at the signal peaks amounts to operating the phase splitter in Class A2. I recently read that small signal tubes are typically not designed to dissipate the increased grid heat that can result and it can shorten their life. So, I will err on the safe side at this point and avoid operating the phase splitter near or beyond that 0 Volt grid line. That will place further constraints on the viable phase splitter load line / operating point combinations and some of those I had explored are no longer in the running.
So, that input stage load line and operating point were actually usable and I have been exploring minor variations from there. Also explored some alternate load lines for the phase splitter. Currently back on the input stage load line from my prior two posts, with ~102K anode resistors and an operating point of135 V 130 V. The phase splitter is on a new load line with ~18K anode resistors.
This is the frequency response with the current load lines and operating point, with an 8 Ohm resistive load on the 8 Ohm tap. The relative reduction in gain is minimal. Input stage gain is down about 2 dB, but gain is up about 1 dB in the phase splitter stage.
Output power has increased to about what I was originally anticipating:
(1 kHz, 8 Ohm resistive load)
In this current configuration, power output is still limited by clipping on the positive side of the waveform, but now that happens only slightly ahead of the negative side of the waveform.
To facilitate trialling minor variations, I added trimmer pots that allow increasing the input stage cathode resistors up to about 20%. And I can use paralleled resistors to reduce them. Also, the input stage anode resistors have been moved to the top side of the board to make it easier to adjust their value with paralleled resistors. These were installed using some holes and traces that are now free after some of the prior changes (removing feedback components, etc). So, things look a little experimental right now. I plan to keep it this way for a while. Once I settle on those values, I’ll replace all that with one resistor per role.
Thanks @paul_b for your help in getting to this point!
The load line and operating point in my last two posts reduced headroom in the phase splitter stage. When I trialled it, I saw that there was significant peak distortion at the phase splitter outputs (only on one side of the waveform) when the amp was driven to the onset of clipping. I thought that might be due to pushing the phase splitter near a limit, especially since I had just moved its operating point closer to one limit.
Edited this paragraph to distil and correct after further digesting my observations and notes: I believe some of the load line / operating point combinations I have trialled are indeed pushing a phase splitter limit (0 Volt grid line). I believe that was not causing any problems I was aware of in the output signals because it impacts a region of the waveform where the power tube being driven is inactive. When the amp is driven further, into clipping at the output, the opposite peaks at the phase splitter outputs also start clipping.
Edit followup: As I continue through the learning curve, I am getting a better understanding of the limit that was being pushed in the phase splitter and at least one potential implication. I believe pushing the grid voltage near or beyond the 0 Volt line at the signal peaks amounts to operating the phase splitter in Class A2. I recently read that small signal tubes are typically not designed to dissipate the increased grid heat that can result and it can shorten their life. So, I will err on the safe side at this point and avoid operating the phase splitter near or beyond that 0 Volt grid line. That will place further constraints on the viable phase splitter load line / operating point combinations and some of those I had explored are no longer in the running.
So, that input stage load line and operating point were actually usable and I have been exploring minor variations from there. Also explored some alternate load lines for the phase splitter. Currently back on the input stage load line from my prior two posts, with ~102K anode resistors and an operating point of
This is the frequency response with the current load lines and operating point, with an 8 Ohm resistive load on the 8 Ohm tap. The relative reduction in gain is minimal. Input stage gain is down about 2 dB, but gain is up about 1 dB in the phase splitter stage.
Output power has increased to about what I was originally anticipating:
(1 kHz, 8 Ohm resistive load)
8 Ohm output tap, 117 Volt supply voltage
17.2 Watts/ch, one channel driven
16.0 Watts/ch, both channels driven
4 Ohm output tap, 117 Volt supply voltage
14.1 Watts/ch, one channel driven
13.6 Watts/ch, both channels driven
8 Ohm output tap, 124 Volt supply voltage
20.3 Watts/ch, one channel driven
18.7 Watts/ch, both channels driven
In this current configuration, power output is still limited by clipping on the positive side of the waveform, but now that happens only slightly ahead of the negative side of the waveform.
To facilitate trialling minor variations, I added trimmer pots that allow increasing the input stage cathode resistors up to about 20%. And I can use paralleled resistors to reduce them. Also, the input stage anode resistors have been moved to the top side of the board to make it easier to adjust their value with paralleled resistors. These were installed using some holes and traces that are now free after some of the prior changes (removing feedback components, etc). So, things look a little experimental right now. I plan to keep it this way for a while. Once I settle on those values, I’ll replace all that with one resistor per role.
Thanks @paul_b for your help in getting to this point!
Last edited:
SwiftEnthusiast
Site Supporter
Since my last post, I have trialed various permutations of the first two stages - load lines, operating points, power supply dropping resistor values, bleeder resistor presence/values, and grid resistor presence/values. I learned that I could stagger the operating points of the first two stages while keeping them DC-coupled using a voltage divider or large grid resistor. That allowed exploring some higher input stage operating anode voltages, although I ultimately ended up settling on a lower operating voltage closer to where I had started.
This schematic reflects the configuration I have settled on for now.
It was interesting to see how the distortion generated in different stages can interact to produce a wide range of results. The black line in this graph is THD (1 KHz, first eight harmonics) for this current configuration, as reported by the Pico scope. The blue and red lines are based on my reading of the second and third harmonic levels in the Pico spectrum analyzer view. I don't know how accurate the absolute values are, but I think the relative values are meaningful regardless (which was my main interest in collecting this data). I believe the dominant second harmonic at higher distortion levels is due to the input stage approaching a limit before the other stages. The resulting asymmetry in the signal prevents symmetrical clipping downstream. If anyone is interested, I have similar distortion graphs for some other configurations I tried.
Frequency and phase response (8 Ohm resistive load, 8 Ohm tap).
Power into 8 Ohm resistive load, 8 Ohm tap, ~125 Volt line voltage, both channels driven:
~16 Watts/ch @ 1% THD
~22 Watts/ch @ 5% THD
Hum: 84/85 dB (L/R) below 22 Watts/ch
I switched to a set of new JJ output tubes (vs used JJ tubes that came with the amp), which increased output power a tad. Also, I decided to take the variac out of my system and return it to test bench duty.
The board and sockets are mounted on a stainless steel plate with an anodized aluminum cover plate (SendCutSend).
The perforated steel piece was something I just made from scrap to make the assembly more rigid (reduce flexing when inserting tubes). Those holes to either side of the oval vents are for probe jacks that double as the bias measurement points and mounts for the output stage cathode resistors. The original bias measurement terminals are now used for the driver power supply stages.
There is an additional tube now because I opted to use a separate tube for the phase splitter of each channel instead of sharing one. This allowed for a cleaner layout and restored the original left/right heater circuit balance. I considered switching to a different tube with one device per bottle but settled on this approach for now since I already had the tubes on hand and didn’t want to introduce yet another variable at that point. Something I might experiment with later.
Edit: Fixed broken image link
I added three holes to the bottom cover below the U terminal board. The board had a small hole between each pair of terminals. I used four of those holes for mounting and enlarged all the rest for ventilation. If I were doing this again, I would probably opt for a wider U terminal board (maybe four additional terminal pairs) to allow spreading the components out a little more.
I am sure there is plenty of room for further improvement, which I may explore later. For now, I am very pleased with the sound quality and enjoying the amp in my primary system.
This schematic reflects the configuration I have settled on for now.
It was interesting to see how the distortion generated in different stages can interact to produce a wide range of results. The black line in this graph is THD (1 KHz, first eight harmonics) for this current configuration, as reported by the Pico scope. The blue and red lines are based on my reading of the second and third harmonic levels in the Pico spectrum analyzer view. I don't know how accurate the absolute values are, but I think the relative values are meaningful regardless (which was my main interest in collecting this data). I believe the dominant second harmonic at higher distortion levels is due to the input stage approaching a limit before the other stages. The resulting asymmetry in the signal prevents symmetrical clipping downstream. If anyone is interested, I have similar distortion graphs for some other configurations I tried.
Frequency and phase response (8 Ohm resistive load, 8 Ohm tap).
Power into 8 Ohm resistive load, 8 Ohm tap, ~125 Volt line voltage, both channels driven:
~16 Watts/ch @ 1% THD
~22 Watts/ch @ 5% THD
Hum: 84/85 dB (L/R) below 22 Watts/ch
I switched to a set of new JJ output tubes (vs used JJ tubes that came with the amp), which increased output power a tad. Also, I decided to take the variac out of my system and return it to test bench duty.
- I was already altering the power supply dropping resistors and chose the new values based on my typical home line voltages (~125 Volts).
- The modern capacitors on the cap board have sufficient voltage ratings to handle the higher secondary voltages.
- The audio heater circuits run at about 6.8 Volts (about 8% above nominal) with my typical home line voltage. The data sheets I have for the tubes I am currently using only specify the nominal heater voltage, but I have read from multiple sources that a tolerance of +/-10% on the heater voltage is typically acceptable. Let me know if anyone disagrees or has any other thoughts on that. I could add a small series resistance if it is truly needed.
The board and sockets are mounted on a stainless steel plate with an anodized aluminum cover plate (SendCutSend).
The perforated steel piece was something I just made from scrap to make the assembly more rigid (reduce flexing when inserting tubes). Those holes to either side of the oval vents are for probe jacks that double as the bias measurement points and mounts for the output stage cathode resistors. The original bias measurement terminals are now used for the driver power supply stages.
There is an additional tube now because I opted to use a separate tube for the phase splitter of each channel instead of sharing one. This allowed for a cleaner layout and restored the original left/right heater circuit balance. I considered switching to a different tube with one device per bottle but settled on this approach for now since I already had the tubes on hand and didn’t want to introduce yet another variable at that point. Something I might experiment with later.
Edit: Fixed broken image link
I added three holes to the bottom cover below the U terminal board. The board had a small hole between each pair of terminals. I used four of those holes for mounting and enlarged all the rest for ventilation. If I were doing this again, I would probably opt for a wider U terminal board (maybe four additional terminal pairs) to allow spreading the components out a little more.
I am sure there is plenty of room for further improvement, which I may explore later. For now, I am very pleased with the sound quality and enjoying the amp in my primary system.
Last edited:
Similar threads
