Since I have a function generator coming, I was thinking that the HP8903E might be a good distortion analyser. It has automatic notch filtering for the fundemental. Any experiences that says I should look at an A or B version instead?
The "B" version is supposed to be more reliable, and some of the parts used in the "A" version (processor, memory) are no longer available. The "B" version was made for about 15 years whille the "A" version maybe 5 years.
There is a brand new version that still has the numbers 8903 on it, a fancy color screen, but the SINAD reaadings it gives don't match the old units and Agilent has no clue.
When I got mine (7 or 8 years ago) the 8903B's were going for $600 to $1000. THe 8903A's were $450 to $700 in working condition. I got one electrically dead unit and one physically damaged unit for $75. I made one good unit out of the collection, but I have had to repair it twice since then.
There is a brand new version that still has the numbers 8903 on it, a fancy color screen, but the SINAD reaadings it gives don't match the old units and Agilent has no clue.
When I got mine (7 or 8 years ago) the 8903B's were going for $600 to $1000. THe 8903A's were $450 to $700 in working condition. I got one electrically dead unit and one physically damaged unit for $75. I made one good unit out of the collection, but I have had to repair it twice since then.
I saw an E version which doesn't have the signal generator. The notch filter auto tunes to the fundemental of the applied signal. Working units are in the $350-$400 range. Has all of the other fetures such as wideband voltmeter but not SINAD. I was thinking that that would be good to have with the DSO as tubemack suggested to improve resolution in a DSO, however the DSO won't read distortion values, just show them in the FFT.
What I do like about having dedicated units is that they take a higher input voltage without going poof.
I am not good enough to fix these units so I need a working one.
What I do like about having dedicated units is that they take a higher input voltage without going poof.
I am not good enough to fix these units so I need a working one.
Finally found the stability issue and closed the loop for feedback.
Even after implementing all the above measures it still would oscillate. Found that the 12bh7 I was using was the culprit. Even though it biased ok, I noticed it had alot of noise, tapped it and it went wild. Replaced it and problems disappeared.
The work was worthwhile as I have Zero Hum and Noise now.
Running 610R on the feedback loop. Will get some measurements out a little later.
Thanks for all the input so far as it REALLY made a difference.
Even after implementing all the above measures it still would oscillate. Found that the 12bh7 I was using was the culprit. Even though it biased ok, I noticed it had alot of noise, tapped it and it went wild. Replaced it and problems disappeared.
The work was worthwhile as I have Zero Hum and Noise now.
Running 610R on the feedback loop. Will get some measurements out a little later.
Thanks for all the input so far as it REALLY made a difference.
I have listened and played with the amp now for about a day. A couple of comments.
1) It likes the higher voltages. I am running 495VB+ with 62mA bias on my GEC KT's. The power is up and distortion down. I consider my measurements qualitative at this point as my test system runs >1% THD in loop back. I should have my signal generator next week. My current signal generator shows quite a bit of third order.
2) I prefer the CCS in the FET follower. Tried a resistor and found that thd went nuts at high power. Listening the amps sounds cleaner and more defined with the CCS.
3) 12bh7's are horrible in the micrphonic category. Tube dampners a must!
4) I cut the feedback quite abit. Mostly to match my rotel that I am using for the other channel, but also I like the sound alot better. Current resistor value is 1.47k.
5) Even with -110V on C-, the FETS run out of room. I am designing a compact power supply using George's suggestion of common mode choke. The idea is based on using two small 60mA transformers one as the isolation and the other as a choke. I hope to fit it on a small board. I would like to see the rails +/- 150V or so. Even with the limited headroom I measured 55W continuous (no distortion) and >85W peak (burst no distortion).
My current voltages are B+ = 495V B'+(ltp) = 480V B''+(driver) = 180V. The later being too low. Something to do later today as I am back on Step one of my previous list ;-)
I attached the schematic as built so far.
1) It likes the higher voltages. I am running 495VB+ with 62mA bias on my GEC KT's. The power is up and distortion down. I consider my measurements qualitative at this point as my test system runs >1% THD in loop back. I should have my signal generator next week. My current signal generator shows quite a bit of third order.
2) I prefer the CCS in the FET follower. Tried a resistor and found that thd went nuts at high power. Listening the amps sounds cleaner and more defined with the CCS.
3) 12bh7's are horrible in the micrphonic category. Tube dampners a must!
4) I cut the feedback quite abit. Mostly to match my rotel that I am using for the other channel, but also I like the sound alot better. Current resistor value is 1.47k.
5) Even with -110V on C-, the FETS run out of room. I am designing a compact power supply using George's suggestion of common mode choke. The idea is based on using two small 60mA transformers one as the isolation and the other as a choke. I hope to fit it on a small board. I would like to see the rails +/- 150V or so. Even with the limited headroom I measured 55W continuous (no distortion) and >85W peak (burst no distortion).
My current voltages are B+ = 495V B'+(ltp) = 480V B''+(driver) = 180V. The later being too low. Something to do later today as I am back on Step one of my previous list ;-)
I attached the schematic as built so far.
Last edited:
3) Did you ever consider another type, like 6CG7?
4) That value suggests 36.5dB closed loop gain. Overtakes what you mentioned as available OLG earlier. Maybe you had more than 35.3dB open loop gain originally?
4) That value suggests 36.5dB closed loop gain. Overtakes what you mentioned as available OLG earlier. Maybe you had more than 35.3dB open loop gain originally?
Yes. May try them at some point in the near future. At the time, I went with the 12bh7 as it had a slightly lower mu.
I am actually considering the 12b4 also at this point. Biases about 12mA more than I am running currently but would allow for even less feedback and a VERY low output impedance.
I just saw in my email that mouser finally shipped (or will ship Monday) the remaining parts for my Pmillet sound card interface. This will be a HUGE help in getting some good measurements.
I am actually considering the 12b4 also at this point. Biases about 12mA more than I am running currently but would allow for even less feedback and a VERY low output impedance.
I just saw in my email that mouser finally shipped (or will ship Monday) the remaining parts for my Pmillet sound card interface. This will be a HUGE help in getting some good measurements.
Last edited:
If you measure the OPT's secondary voltage on a 5.6R dummy load with the loop broken and 200mV input 1kHZ sine wave and then calculate (Vout/Vin) Log20 you will know your OLG in dB voltage gain. 12B4 is microphonic and weird for oscillations, not an easy tube. 5687 easier in that vein. My guess is that some spare gain to burn on NFB is always handy. 5687, 6CG7, and ECC80 even I would test at a point.
I agree the 12b4 has microphonic tendancies. IMHO less than than the 12bh7. My premap is a 12b4. The other tubes you indicate will go on the list of candidates for trials.
I think that before I toss the 12bh7 out I should first benchmark this amp and tweek the obvious for improvements and build the second channel to be able to really listen to it. I also need to apply what I have read and studied in regard to measurement techniques and equipment.
In no way does this thing sound bad or have any tendencies that make it immediately unlistenable. In fact in most ways I like it better than my SET. ( I am still a sucker for 2nd order I guess)
Into an 8ohm load it was running 45dB OLG. with a 100mV input @1kHz.
I think that before I toss the 12bh7 out I should first benchmark this amp and tweek the obvious for improvements and build the second channel to be able to really listen to it. I also need to apply what I have read and studied in regard to measurement techniques and equipment.
In no way does this thing sound bad or have any tendencies that make it immediately unlistenable. In fact in most ways I like it better than my SET. ( I am still a sucker for 2nd order I guess)
Into an 8ohm load it was running 45dB OLG. with a 100mV input @1kHz.
45dB OLG much explains why you can use 1.47k/22R. Cool.
12BH7 is closer to 12AU7, 5687 is more linear and LowZ. I have messed with 12B4 preamps myself too and I know what you mean. There is much variation between available 12B4s for stability and noises though. But I believe you already got a nice stash. 2 12B4s in a system's chain can stick out their tone more is one negative consideration too. A CCS loaded 12B4 will give a gain of 5-6 maybe, on a big amp that normally is used on rather demanding speakers, low final Zout for a good damping factor equals some NFB in the end. So some gain margin can be crucial. Your now 8.5dB NFB curiously is exactly how much I use on my trioded KT88 20W and I found it good between drive and tone.
P.S. Before you get it in stereo but mainly until compensated you don't know its tone for sure. Only hinting. Use silver mica for the TBD comp caps when finalized if you can find, or C0G/NP0.
P.S.2. From previous preamps experience I am having a suspicion about 10M45S used as an anode load on tubes like 5687, 12BH7, 12B4, some tend to get hissier and more microphonic than with just a resistor. If you continue to find your front end touchy even with other types or samples, keep that in the back of your mind. Try to bias at half current CCS than you got now, and if it gets more quiet on microphonics, its a hint.
12BH7 is closer to 12AU7, 5687 is more linear and LowZ. I have messed with 12B4 preamps myself too and I know what you mean. There is much variation between available 12B4s for stability and noises though. But I believe you already got a nice stash. 2 12B4s in a system's chain can stick out their tone more is one negative consideration too. A CCS loaded 12B4 will give a gain of 5-6 maybe, on a big amp that normally is used on rather demanding speakers, low final Zout for a good damping factor equals some NFB in the end. So some gain margin can be crucial. Your now 8.5dB NFB curiously is exactly how much I use on my trioded KT88 20W and I found it good between drive and tone.
P.S. Before you get it in stereo but mainly until compensated you don't know its tone for sure. Only hinting. Use silver mica for the TBD comp caps when finalized if you can find, or C0G/NP0.
P.S.2. From previous preamps experience I am having a suspicion about 10M45S used as an anode load on tubes like 5687, 12BH7, 12B4, some tend to get hissier and more microphonic than with just a resistor. If you continue to find your front end touchy even with other types or samples, keep that in the back of your mind. Try to bias at half current CCS than you got now, and if it gets more quiet on microphonics, its a hint.
These kt88 sound good don't they?
I don't have any capacitance in the feedback loop yet as I don't have a good square wave source. I also have not put in the RC network on the voltage driver tube anode. I assume that it picks up the HF response, however I see some increase of gain at 20kHz as it is.
My 12b4 is resistive loaded with the CCS in the cathode. I havn't wanted to change it other than cleaning up my p2p wiring. Why it doesn't hum or oscillate, I don't know.
I don't have any capacitance in the feedback loop yet as I don't have a good square wave source. I also have not put in the RC network on the voltage driver tube anode. I assume that it picks up the HF response, however I see some increase of gain at 20kHz as it is.
My 12b4 is resistive loaded with the CCS in the cathode. I havn't wanted to change it other than cleaning up my p2p wiring. Why it doesn't hum or oscillate, I don't know.
With the CCS at the cathode they do OK. On anode they sing. Yes KT88 are OK. I use Genalex Russian new edition. That second RC network is a pole zero, can impair dynamic loop behavior, don't try it together if the Rf Ccomp alone will do the job decidedly.
Thanks Salas, I just caught your PS's above. Will try your hints.
I do hear more hiss than I want when the tube is warming up. It quites down to inaududible when its hot.
I also agree that the true character can't be heard without the other channel. Unfortunatley I don't want to finish that until this channel is close to were I want it.
You can hear the heater expand when it is warming up.... The 5687 is almost a drop in replacement pin wise. Will have to reset some resistors to get it to work though. 6cg7 does look interesting as all it would require is a pin out change
I do hear more hiss than I want when the tube is warming up. It quites down to inaududible when its hot.
I also agree that the true character can't be heard without the other channel. Unfortunatley I don't want to finish that until this channel is close to were I want it.
You can hear the heater expand when it is warming up.... The 5687 is almost a drop in replacement pin wise. Will have to reset some resistors to get it to work though. 6cg7 does look interesting as all it would require is a pin out change
Last edited:
Okay time for an update. I am still running the 12bh7. My hiss is from my preamp and I will have to deal with that seperatly. I have the 6CG7 and 5687 in hand now but will hold off based on the measurements I have so far. My audio work has to slow down now that it is starting to warm up outside so I can finish my wife's kitchen cabinets. Oh Well.
I have finally recieved my scope and function generator which has made a world of difference.
The amp puts out 65.3Watts before clip. freq range is between 20Hz f3 and >100kHz. No phase issues.
For feedback, I have used 1.47k for the resistor and 680pF for the cap. I started at 10pF and kept going till the square wave stabilized. 1000pF was too much. Attached are the photo of the square waves at both 1kHz and 10kHz with 680pF. THD is <0.5% at 30W with a nice cascade of the harmonics. 2nd being the highest.
Now to fire up the second channel.!!! and build a chassis for them.
I have finally recieved my scope and function generator which has made a world of difference.
The amp puts out 65.3Watts before clip. freq range is between 20Hz f3 and >100kHz. No phase issues.
For feedback, I have used 1.47k for the resistor and 680pF for the cap. I started at 10pF and kept going till the square wave stabilized. 1000pF was too much. Attached are the photo of the square waves at both 1kHz and 10kHz with 680pF. THD is <0.5% at 30W with a nice cascade of the harmonics. 2nd being the highest.
Now to fire up the second channel.!!! and build a chassis for them.
Attachments
Okay here it is at 2kHz. I decreased the time scale to focus on the leading edge of the square wave and am wondering what I can do to make it better. What am I looking for? Any more capacitance and it rounds over, any less and it shows ringing.
Ignore the 33.2kHz on the display, that is related to the cursors window. (I havn't figured out yet how to turn it off!!!!)
I took the measurement at 1.25 Watts power using a 10x probe. edge doesn't change at higher or lower power.
Ignore the 33.2kHz on the display, that is related to the cursors window. (I havn't figured out yet how to turn it off!!!!)
I took the measurement at 1.25 Watts power using a 10x probe. edge doesn't change at higher or lower power.
Attachments
Better a bit of a knee than a bit of an edge. You will be surprised of how audible compensation adjustment is. I use 2x360=720p Silver Mica on 1k2 on my KT88PP, many different things and parts but interestingly where about you move.🙂 Did you try 10kHz square waves? I mainly tune there and usually lower frequency does better too. What type are the Cdom Rf caps right now?
- Status
- Not open for further replies.