HK A500 Output Tube Experiments (Again)
- Thread starter TriodeLuvr
- Start date
Thanks! The circuit as shown applies 20dB NFB. Sensitivity is 0.65V for full output. The last amplifier I built came in closer to 1V. The tube lineup in that amp exhibits roughly the same gain (triode EF86 - 6J6 LTP - 6L6), but the step-down ratio of the output transformer is higher (6.6K primary). Both amps work well with the preamp I'm using.
Jack
As a sanity thing for component numbers, you can group them so there is no overlap but without having to really keep track of it. 1xx parts for the audio chain and 2xx parts for power supply for example.
Or just drive everyone insane and dont number anything, which is what i usually do.
Or just drive everyone insane and dont number anything, which is what i usually do.
I used to keep it all sequential, but then I would change one thing in the amp or power supply, and every component in the other schematic would have to be updated. That's a good idea to use 1xx and 2xx numbers.
Jack
Thank you. I talked about AC heater wiring in a recent discussion here:
Heater Wiring
Also, I always elevate heater wiring electrically to +50V or so. That greatly reduces hum pickup within the tubes, especially when the cathodes aren't bypassed.
Jack
Earlier in this thread I mentioned that I had an A500 coming in from a client for work to be performed, and that I would take some measurements not only for this project, but also to show how far Jack's work went over the original design. Here's a few of the highlights from the results of those tests as based the original design, and measured from a loaded 16Ω tap:
1. The design provides 14.5 db of Positive Feedback from the cathode of the inverter to the cathode of the AF amplifier stage -- this to maximize the gain of the AF stage and therefore OL gain of the power amp section to achieve the desired sensitivity sought for this section after the Global NFB loop is closed.
2. With the PFB loop in place, 9 db of feedback is produced by the global NFB loop. I had originally estimated this at 6 db earlier -- fairly close.
3.The sensitivity of the power amp section is .32 vac (rms) to produce maximum power output (25 watts measured with only one channel operating).
4. A 10 kHz square wave is provided with the signal injected at the top of the balance control. This not only eliminates the effect of those controls and earlier stages, but also, accounts for the fact that the balance control (when centered) represents a 78KΩ series resistance at the AF amplifier's input grid (which is paralleled with a 2700 pF cap). This network is part of the power amplifier section, so its effects needed to be included with the testing. All the test results provided here were generated with the signal injected in this manner.
5. With 9 db of NFB, the internal output impedance of the amplifier is a whopping 8.37Ω, which is to be expected. With a pentode output stage and a limited amount of NFB applied, this would be the result -- not unlike that which guitar amplifier's produce. A benefit of this reduced NFB level however, is that the amplifier is completely stable whether loaded, unloaded, or into any kind of cap only load.
While this result seems rather extreme, the reduced DF produced (1.91) -- as calculated in the classic vintage manner -- is a quality that many folks favor, with its tendency being to add increased body to the sound. I have no doubt it is a significant reason why the A500 receives much of the praise it does. If the speaker's DC resistance is included in the calculation, the real DF then is even lower.
6. The output transformers are the golden nugget of the design that Jack took advantage of. Even with the limited amount of the NFB provided in the original design, the transformers could easily deliver 25 watts of power at 20 Hz (one channel at a time of course) without visable distortion, which is a real credit to HK's efforts with this amplifier. At 20 kHz, they deliver 22.5 watts under the same conditions.
7. I always measure output transformer impedance without loading the measured winding, since loading produces copper losses that distort the determination of turns ratio. Others have a different opinion on this which I respect. My own belief is that measuring the turns ratio in this manner delivers the most accurate turns ratio results, while the difference between that and the apparent increased turns ratio produced when loading the measured winding is more accurately a measure of the transformer's losses when in circuit. Whether it is really losses, or indicative of a higher actual reflected impedance, either way results in reduced power output versus that which a no load turns ratio test indicates. Measured using the method described, the output transformer reflected a 3,543Ω impedance from the 16Ω winding.
10. The real downfall of the original design is the biasing method used -- that being the use of the 12 volt heaters series connected in the three phono preamp and line stage tubes. With the quiecent heater voltage set at 29 vdc across the string as indicated on the schematic, it rises to a whopping 47 vdc with both channels at a full boil, with notable crossover distortion appearing at just over 14 watts in each channel. The 3 12AX7 heaters surely can't like that! But with an 18 volt increase in bias, there would surely be cross over distortion indeed! It's for that reason however that the B+ only drops 15 volts from quiescent to maximum power output in both channels, and the power output dropoff is so significant. The voltage doubler power supply is good -- but not that good! This design screams for at least zeners to be added across the heater string in the same way I published for them to be added to certain Fisher amplifiers using the same biasing approach. I'll be working on that soon.
Against the original design, I have no doubt that Jack's project produced very fine results indeed!
Dave
1. The design provides 14.5 db of Positive Feedback from the cathode of the inverter to the cathode of the AF amplifier stage -- this to maximize the gain of the AF stage and therefore OL gain of the power amp section to achieve the desired sensitivity sought for this section after the Global NFB loop is closed.
2. With the PFB loop in place, 9 db of feedback is produced by the global NFB loop. I had originally estimated this at 6 db earlier -- fairly close.
3.The sensitivity of the power amp section is .32 vac (rms) to produce maximum power output (25 watts measured with only one channel operating).
4. A 10 kHz square wave is provided with the signal injected at the top of the balance control. This not only eliminates the effect of those controls and earlier stages, but also, accounts for the fact that the balance control (when centered) represents a 78KΩ series resistance at the AF amplifier's input grid (which is paralleled with a 2700 pF cap). This network is part of the power amplifier section, so its effects needed to be included with the testing. All the test results provided here were generated with the signal injected in this manner.
5. With 9 db of NFB, the internal output impedance of the amplifier is a whopping 8.37Ω, which is to be expected. With a pentode output stage and a limited amount of NFB applied, this would be the result -- not unlike that which guitar amplifier's produce. A benefit of this reduced NFB level however, is that the amplifier is completely stable whether loaded, unloaded, or into any kind of cap only load.
While this result seems rather extreme, the reduced DF produced (1.91) -- as calculated in the classic vintage manner -- is a quality that many folks favor, with its tendency being to add increased body to the sound. I have no doubt it is a significant reason why the A500 receives much of the praise it does. If the speaker's DC resistance is included in the calculation, the real DF then is even lower.
6. The output transformers are the golden nugget of the design that Jack took advantage of. Even with the limited amount of the NFB provided in the original design, the transformers could easily deliver 25 watts of power at 20 Hz (one channel at a time of course) without visable distortion, which is a real credit to HK's efforts with this amplifier. At 20 kHz, they deliver 22.5 watts under the same conditions.
7. I always measure output transformer impedance without loading the measured winding, since loading produces copper losses that distort the determination of turns ratio. Others have a different opinion on this which I respect. My own belief is that measuring the turns ratio in this manner delivers the most accurate turns ratio results, while the difference between that and the apparent increased turns ratio produced when loading the measured winding is more accurately a measure of the transformer's losses when in circuit. Whether it is really losses, or indicative of a higher actual reflected impedance, either way results in reduced power output versus that which a no load turns ratio test indicates. Measured using the method described, the output transformer reflected a 3,543Ω impedance from the 16Ω winding.
10. The real downfall of the original design is the biasing method used -- that being the use of the 12 volt heaters series connected in the three phono preamp and line stage tubes. With the quiecent heater voltage set at 29 vdc across the string as indicated on the schematic, it rises to a whopping 47 vdc with both channels at a full boil, with notable crossover distortion appearing at just over 14 watts in each channel. The 3 12AX7 heaters surely can't like that! But with an 18 volt increase in bias, there would surely be cross over distortion indeed! It's for that reason however that the B+ only drops 15 volts from quiescent to maximum power output in both channels, and the power output dropoff is so significant. The voltage doubler power supply is good -- but not that good! This design screams for at least zeners to be added across the heater string in the same way I published for them to be added to certain Fisher amplifiers using the same biasing approach. I'll be working on that soon.
Against the original design, I have no doubt that Jack's project produced very fine results indeed!
Dave
Dave, many thanks for posting this information. I know it will be helpful to others who restore and upgrade these amplifiers. My own work was too far from the factory design to be of much assistance in that regard.
The original driver is an economy design that minimizes the number of tubes. As you note, this necessitates a compromise regarding global feedback that impacts distortion and damping. Despite this, frequency response is remarkably good, due in no small part to the exceptional quality of the transformers. I would not attempt changes in this area if I was restoring one of these amplifiers.
Regarding output bias, I am inclined to excuse the shortcomings of the method HK used. Everything you've pointed out is certainly true, but I think the critique falls outside the context of 1960s design criteria. For example, the continuous power measurements that we find desirable in characterizing an amplifier are not representative of typical musical program material. HK engineers, like many of their day, probably felt that the increase in bias voltage at full output would be mostly transient in nature and therefore not overly significant in practice. In addition, the movement toward Class B at high output levels may be beneficial with respect to the dissipation rating of the 7355s.
The upside to all this is that HK saved the cost of additional tubes, as well as fixed bias and DC filament supplies. That allowed presenting a very high quality product to the market that was affordable for many audiophiles.
I'm sure the unit you're restoring will prove to be an excellent performer. Please continue to share your work!
Jack
My TA5000x transformers measured 3.47Ohms unloaded and 8.5dB of NFB. Very close to Dave's measurements.
I implemented Zener mod with gadget's suggestion and added the Bias-Balance network. Amp runs a LOT cooler when it runs; it has been very cold on a dusty shelf for the last 2 years.
I implemented Zener mod with gadget's suggestion and added the Bias-Balance network. Amp runs a LOT cooler when it runs; it has been very cold on a dusty shelf for the last 2 years.