My take on a Single Ended amp
- Thread starter kward
- Start date
Here's a close up pic of the square hole I had them drill for me before shipping (I don't have a square bit so it's really hard to drill this particular hole). This hole is for the IEC power inlet, and you can get an idea of the thickness of the plate from this picture. One good thing though, you can drive over it with a 1 ton pickup and and nothing will happen. 
Uhhh, I typo'd it. It's 3/16" thick. I didn't ask the fabricator for anything specific. This is just what he sent me.
The fabricator is IAG Audio -- www.iagaudio.com (no affiliation), if anyone is interested in using this supplier. But I didn't order through his website. I emailed him and discussed options directly.
The fabricator is IAG Audio -- www.iagaudio.com (no affiliation), if anyone is interested in using this supplier. But I didn't order through his website. I emailed him and discussed options directly.
Last edited:
I've also used those chassis- but I got them through their listings on eBay. Great chassis- extremely rigid, and well-finished.
I was able to make chassis holes for tube sockets, with a stepped bit and a hole saw. The material is thick, but not that hard- so, just take it slow, keep the bit lubed (WD40 works), and it's OK...
Regards,
Gordon.
I was able to make chassis holes for tube sockets, with a stepped bit and a hole saw. The material is thick, but not that hard- so, just take it slow, keep the bit lubed (WD40 works), and it's OK...
Regards,
Gordon.
I did same thing to accommodate larger chokes in my amp build you helped me a lot with.
Air gap in the bottom plate around chokes helps cooling them too.
I wound up spacing the whole bottom panel down, 3/16", to make my additional heater transformer fit properly in my build. Used some of the slugs that were left from hole-sawing out tube socket holes, as washers over the screws, between the bottom panel and chassis. That also added some ventilation, all the way around the bottom perimeter of the amp.
Regards,
Gordon.
Regards,
Gordon.
The Edcor XPWR155 power transformer was delivered today. I've had the impression from past experience with Edcor power transformers that they are built fairly stout. But what does "stout" mean in engineering terms?
I mocked up a test environment with dummy loads on each winding that will be used in the build, drawing as close to actual current from each winding as was feasible given the power resistors I had on hand.
First, I measured the unloaded voltages:
This test is running the HV winding at 230 mA, which is over driving it by 30 mA as per spec from Edcor. Estimated current consumption from the amp (from load line analysis not shown here) will be 253 mA, which includes steady state (DC) current draw plus dynamic (AC) current draw when running both channels at full power. This assumed a 1:1 ratio between current draw needed and AC secondary current availability. I know this is being aggressive, in reality I probably should be using a 1:1.5 ratio, meaning for every 1A current drawn from the load, I need 1.5A of AC current availability from the secondary (given a capacitor input power supply topology).
Also I know I am over taxing the VA (volt-amp) rating somewhat on this transformer overall. Transformer is spec'd for 186VA total (sum of all three secondary windings). I'll be requiring 202 VA at full power from this amp, which is a 16 VA overage.
In other words, the transformer will likely run a little hot. I probably should have used a larger rated transformer, but I think I can make it work.
===
Also I note again that in the above tests, windings fully loaded, the power transformer buzzed like a chain saw. I've come to expect this with Edcor power iron. I know how to fix it though. Remove the end bells, jam Popsicle sticks in between the bobbin and core, and hold them tight with silicon glue.
I mocked up a test environment with dummy loads on each winding that will be used in the build, drawing as close to actual current from each winding as was feasible given the power resistors I had on hand.
First, I measured the unloaded voltages:
- HV winding: 368-0-368
- Filament winding: 6.75V
- The HV winding was loaded with a 1.5K 100 Watt resistor between each end and the center tap.
- The filament winding was loaded with a 1.4Ω 100 Watt resistor.
- The 5V winding was left unloaded (since I will not be using this winding in the build).
- HV winding: 345-0-345 (with a 1.5K resistor between each leg and CT produces 230 mA current per leg)
- Filament winding: 6.0V (with a 1.4Ω resistor across the secondary produces 4.3A current)
- HV winding: 6.7%
- Filament winding: 12%
This test is running the HV winding at 230 mA, which is over driving it by 30 mA as per spec from Edcor. Estimated current consumption from the amp (from load line analysis not shown here) will be 253 mA, which includes steady state (DC) current draw plus dynamic (AC) current draw when running both channels at full power. This assumed a 1:1 ratio between current draw needed and AC secondary current availability. I know this is being aggressive, in reality I probably should be using a 1:1.5 ratio, meaning for every 1A current drawn from the load, I need 1.5A of AC current availability from the secondary (given a capacitor input power supply topology).
Also I know I am over taxing the VA (volt-amp) rating somewhat on this transformer overall. Transformer is spec'd for 186VA total (sum of all three secondary windings). I'll be requiring 202 VA at full power from this amp, which is a 16 VA overage.
In other words, the transformer will likely run a little hot. I probably should have used a larger rated transformer, but I think I can make it work.
===
Also I note again that in the above tests, windings fully loaded, the power transformer buzzed like a chain saw. I've come to expect this with Edcor power iron. I know how to fix it though. Remove the end bells, jam Popsicle sticks in between the bobbin and core, and hold them tight with silicon glue.
Last edited:
Progress continues...
I've taken a hiatus for a month but I'm back at it now. Edcor power transformer is now painted black, bobbin secured with wood shims and silicon glue so it won't rattle when running. Additionally I'm gong to mount the PT on neoprene washers on the top of the chassis. Those two actions should keep that PT really quiet.
I'm starting to mount the hardware now. I didn't like having to cut a hole in the bottom plate to make the bias transformer fit, even though it was the shortest one (2 inches) I could find, because the chassis is a few 16th's of an inch less than 2" inside measurement. So I've ordered a 6V filament transformer (Triad F-13X) that will fit (1 3/8" high), and I'll run it backwards off of the unused 5V filament supply off of the main PT. When it's delivered I can get the rest of the parts mounted and start wiring. Should't be too long now till it's completed (famous last words...)
The back of the chassis is facing forward so you an see the IEC inlet, the fuse holder, speaker terminators, and RCA input jacks, with the tube sockets, bias test points, and bias pots mounted on the chassis.
I've taken a hiatus for a month but I'm back at it now. Edcor power transformer is now painted black, bobbin secured with wood shims and silicon glue so it won't rattle when running. Additionally I'm gong to mount the PT on neoprene washers on the top of the chassis. Those two actions should keep that PT really quiet.
I'm starting to mount the hardware now. I didn't like having to cut a hole in the bottom plate to make the bias transformer fit, even though it was the shortest one (2 inches) I could find, because the chassis is a few 16th's of an inch less than 2" inside measurement. So I've ordered a 6V filament transformer (Triad F-13X) that will fit (1 3/8" high), and I'll run it backwards off of the unused 5V filament supply off of the main PT. When it's delivered I can get the rest of the parts mounted and start wiring. Should't be too long now till it's completed (famous last words...)
The back of the chassis is facing forward so you an see the IEC inlet, the fuse holder, speaker terminators, and RCA input jacks, with the tube sockets, bias test points, and bias pots mounted on the chassis.
Is there a second speaker terminal strip going in?
Got it
Update.
I've finally got the amp far enough along to wheel it into my listening room and give it a try. The feedback components are alligator clipped in at the moment so I can easily change things if I feel the need.
The amp measured only -1 dB down at 20 KHz without any feedback whatsoever, so I thought I would listen to it this way as one of the listening tests performed. Without feedback, input sensitivity is about 1V, and with 9 dB feedback it's 2.8V input sensitivity.
I've got only about 90 minutes time on this amp so far, so not nearly enough to make concrete statements, but initial thinking is:
I was initially worried about the 12AT7 driving the somewhat difficult load of a 0.22 uF coupling cap into a 47K grid leak load on the KT120 output stage. But all seems to be fine there. I am running the 12AT7 just a bit hotter -- 4.5 mA quiescent current, -3.2V cathode bias, and 227V across the tube -- than previous schematic updates indicated. This provides just a bit more signal swing headroom that I think will help make the amp more compatible with either new full strength, or more well worn 12AT7 tubes.
I've finally got the amp far enough along to wheel it into my listening room and give it a try. The feedback components are alligator clipped in at the moment so I can easily change things if I feel the need.
The amp measured only -1 dB down at 20 KHz without any feedback whatsoever, so I thought I would listen to it this way as one of the listening tests performed. Without feedback, input sensitivity is about 1V, and with 9 dB feedback it's 2.8V input sensitivity.
I've got only about 90 minutes time on this amp so far, so not nearly enough to make concrete statements, but initial thinking is:
- I can readily tell the difference with and without feedback. Without feedback, amp sounds more open and forward, like I am sitting in the middle of the performers. This is a very captivating aspect of this amp's presentation, but at the same time the bass is sloppy. Bass drum kicks are round and not well defined.
- With or without feedback, the bass overall is deep. More so than I expected with a single ended amp. At reasonable listening levels it goes as deep as any of my push pull amps.
- So far this amp doesn't have the finesse of my decent push pull amps of the same output power category. But I know it's hot off the bench and will break in and come in to its own with a dozen or so more hours on it.
- As mentioned in a previous post way back on this thread, this amp outputs 12 watts per channel, both channels driven, with reasonably low distortion for an SE amp. It will be interesting to compare sound presentation of this amp to my 12 watt per channel 6V6 push pull amp (that I dubbed the "Modded out 'Moto"), which I have discussed elsewhere on this forum.
I was initially worried about the 12AT7 driving the somewhat difficult load of a 0.22 uF coupling cap into a 47K grid leak load on the KT120 output stage. But all seems to be fine there. I am running the 12AT7 just a bit hotter -- 4.5 mA quiescent current, -3.2V cathode bias, and 227V across the tube -- than previous schematic updates indicated. This provides just a bit more signal swing headroom that I think will help make the amp more compatible with either new full strength, or more well worn 12AT7 tubes.
Last edited:
The final step in calling this amp complete is to perform stability testing. And that's where the fairy tale ends with these output transformers -- they really do not like global feedback around them. I was able to beat them into submission but at the expense of some other important performance criteria, which I will explain.
Low frequency stability was never at issue. Amp is rock stable at LF under any kind of load.
High frequency stability, on the other hand, had me standing one legged on a tightrope patting head and rubbing abdomen trying to get it stable. The root issue (other than obviously the output transformers themselves) was with the amount of feedback used (9 dB). With a cap only speaker load of anywhere from 0.22 uF to 0.47 uF, the amp would self-oscillate (at 90 KHz) at full power, with no input excitation at all. But any cap only load smaller than 0.22 uF or larger than 0.47 uF did not cause instability. Furthermore, none of the normal HF stability taming methods worked without totally killing frequency response -- HF step network on the input stage, screen to cathode feedback, Zobel on the output terminals, etc, etc. It's almost as if these output transformers defied being tamed.
After listening to the amp under open loop conditions, I just could not live with the sloppy bass it produced. I had to have some feedback. So to get it tamed, four important performance aspects were needed to be compromised:
1. Input sensitivity. The amp behaved like it had too much forward gain for the amount of feedback used. I ended up removing the 12AT7 cathode bypass cap all together, thus lowering stage gain and as a result significantly lowering input sensitivity, and therefore slightly lowering amount of feedback applied (to 6 dB). I was shooting for about 2V input sensitivity during the prototype stage, but ended up with a whopping 4V in the final build. And that's in units of RMS volts! I might try putting the 12AT7 cathode bypass caps back and lowering feedback amount down to the same 6 dB. That might get the same result but with increased sensitivity.
2. Distortion. A resistor only load across the speaker terminals was the only thing that ultimately tamed the instability. The resistor value used was a 62 ohm 20 watt wire wound. No amount of capacitance in series with this resistor worked to tame instability without totally killing HF response. The end result is mid-band distortion (from about 200 Hz to 5 KHz) is approximately 2% THD at full power output. If there is a silver lining, distortion at 20 KHz at full power output is approximately the same. However, mid band distortion is double give or take of what I was shooting for during prototyping.
3. Hotter bias. To somewhat counteract the higher distortion, I biased the output stage at 100 mA (instead of the targeted 90 mA during the prototype phase), while keeping the plate voltage at 450V. Plate dissipation is a little less than 45 watts, but that isn't concerning me greatly with KT120 tubes. Under these conditions, plate dissipation sits at 75% of design center max.
4. Not able to use 6550 or KT88 tubes. Bias is too hot to safely use those tubes.
But the result is....a perfectly stable amp with any speaker load you can throw at it, These compromises were painful to accept. Could there be some other approach to obtain a 100% stable amp that I have not considered or that I've overlooked? Maybe, but I sure couldn't find it.
I'll offer some more observations on sound presentation later, along with the final schematic, complete performance specifications, and some pictures.
Low frequency stability was never at issue. Amp is rock stable at LF under any kind of load.
High frequency stability, on the other hand, had me standing one legged on a tightrope patting head and rubbing abdomen trying to get it stable. The root issue (other than obviously the output transformers themselves) was with the amount of feedback used (9 dB). With a cap only speaker load of anywhere from 0.22 uF to 0.47 uF, the amp would self-oscillate (at 90 KHz) at full power, with no input excitation at all. But any cap only load smaller than 0.22 uF or larger than 0.47 uF did not cause instability. Furthermore, none of the normal HF stability taming methods worked without totally killing frequency response -- HF step network on the input stage, screen to cathode feedback, Zobel on the output terminals, etc, etc. It's almost as if these output transformers defied being tamed.
After listening to the amp under open loop conditions, I just could not live with the sloppy bass it produced. I had to have some feedback. So to get it tamed, four important performance aspects were needed to be compromised:
1. Input sensitivity. The amp behaved like it had too much forward gain for the amount of feedback used. I ended up removing the 12AT7 cathode bypass cap all together, thus lowering stage gain and as a result significantly lowering input sensitivity, and therefore slightly lowering amount of feedback applied (to 6 dB). I was shooting for about 2V input sensitivity during the prototype stage, but ended up with a whopping 4V in the final build. And that's in units of RMS volts! I might try putting the 12AT7 cathode bypass caps back and lowering feedback amount down to the same 6 dB. That might get the same result but with increased sensitivity.
2. Distortion. A resistor only load across the speaker terminals was the only thing that ultimately tamed the instability. The resistor value used was a 62 ohm 20 watt wire wound. No amount of capacitance in series with this resistor worked to tame instability without totally killing HF response. The end result is mid-band distortion (from about 200 Hz to 5 KHz) is approximately 2% THD at full power output. If there is a silver lining, distortion at 20 KHz at full power output is approximately the same. However, mid band distortion is double give or take of what I was shooting for during prototyping.
3. Hotter bias. To somewhat counteract the higher distortion, I biased the output stage at 100 mA (instead of the targeted 90 mA during the prototype phase), while keeping the plate voltage at 450V. Plate dissipation is a little less than 45 watts, but that isn't concerning me greatly with KT120 tubes. Under these conditions, plate dissipation sits at 75% of design center max.
4. Not able to use 6550 or KT88 tubes. Bias is too hot to safely use those tubes.
But the result is....a perfectly stable amp with any speaker load you can throw at it, These compromises were painful to accept. Could there be some other approach to obtain a 100% stable amp that I have not considered or that I've overlooked? Maybe, but I sure couldn't find it.
I'll offer some more observations on sound presentation later, along with the final schematic, complete performance specifications, and some pictures.
Last edited:
After more fiddling and fussing, I discovered that a Zobel network was required to be strapped across the output terminals to make this amp unconditionally stable. The Zobel values used were 50 ohm resistor and 0.1 uF cap. Also a screen to cathode feedback cap was required, as well as the normal feedback cap strapped across the feedback resistor.
Only 6 DB feedback could be applied in a fully non-bypassed 12AT7 cathode stage. When bypassing the upper cathode resistor, as one would mormally do, the amp cannot be made 100% stable without killing HF response.
Atter all the HF stabilization discussed is applied, frequency response is flat to 19 KHz, input sensitivity is 4 volts, and output power is 12 watts, at slightly less than 2% THD at 1KHz at 12 watts.
I think this is reasonably good performance for a low feedback SE amp of this output power class.
So after all the fussing over the performance specs of this amp, the end result turned out pretty well. Only thing that bothers me still is the input sensitivity at 4 volts. Nothing I can do about that though without changing topology, so I will live with it, and note that a preamp is absolutely required to drive this amp.
I've got a few parts on order for the Zobel. When they arrive I will install them and then take a few pics, as well as post schematic with full performance criteria.
I'm just itching to make commentary about SE amps vs push pull amps of this output power class but I've got to wait on that until I have some time in the listening room.
Only 6 DB feedback could be applied in a fully non-bypassed 12AT7 cathode stage. When bypassing the upper cathode resistor, as one would mormally do, the amp cannot be made 100% stable without killing HF response.
Atter all the HF stabilization discussed is applied, frequency response is flat to 19 KHz, input sensitivity is 4 volts, and output power is 12 watts, at slightly less than 2% THD at 1KHz at 12 watts.
I think this is reasonably good performance for a low feedback SE amp of this output power class.
So after all the fussing over the performance specs of this amp, the end result turned out pretty well. Only thing that bothers me still is the input sensitivity at 4 volts. Nothing I can do about that though without changing topology, so I will live with it, and note that a preamp is absolutely required to drive this amp.
I've got a few parts on order for the Zobel. When they arrive I will install them and then take a few pics, as well as post schematic with full performance criteria.
I'm just itching to make commentary about SE amps vs push pull amps of this output power class but I've got to wait on that until I have some time in the listening room.
Gadget--it has crossed my mind. These are big transformers. As big as Dynakit 60 watt A431's. I'd have to replace them with something similar in footprint. These are the CXSE series and I just wonder if the smaller GXSE series would work better in the stability sense. It is odd because the transformers have very little resonance in square wave tests. Square waves look awesome and yet stability allows only a small amount of feedback.
6--The feedback signal from the screen is in phase with the signal at the 12AT7 cathode to produce a negative feedback signal, same as it is with the Dynaco style in their push pull amps. The Dynaco design may have needed to swap the primary leads off of the imverter to obtain that condition, I'm not sure though, as I haven't looked deeply at their design. What I do know for fact is in this amp, the screen signal has same polarity as the cathode signal of the driver stage to provide negative feedback.
6--The feedback signal from the screen is in phase with the signal at the 12AT7 cathode to produce a negative feedback signal, same as it is with the Dynaco style in their push pull amps. The Dynaco design may have needed to swap the primary leads off of the imverter to obtain that condition, I'm not sure though, as I haven't looked deeply at their design. What I do know for fact is in this amp, the screen signal has same polarity as the cathode signal of the driver stage to provide negative feedback.
