|
Audio Asylum Thread Printer Get a view of an entire thread on one page |
For Sale Ads |
65.202.22.70
In Reply to: RE: 6Y6GT/6922 Link to larger schematics and PDF file posted by jyourison on March 28, 2015 at 11:52:01
My first thought is that you can't get much voltage swing from the split-load inverter - if there's 97 on the gate, about 92 on the source, and the drain resistor will have the same drop, so the drain is at 220-92V, leaving about 36V across the transistor. Might JUST make twice the bias voltage peak-to peak. A little lower on the 6DJ8 plate would be better. The source and drain loads SHOULD be equal - they're not, so it's probably canceling 2F distortion from a non-ideal operating point. Also there would be an advantage to running the phase inverter at the same current as the gain stage: net current drawn by the two stages is constant, since the AC components are equal and opposite. I think more decoupling would be better too, given the large swings in supply current in deep AB.
All minor tweaks to a very good start. Have fun!
Follow Ups:
Hi, Tom,
I set the current for 6.5 mA through the 6922, reduced from 8 mA.
- The plate is now at 90VDC.
- The splitter rails are at 87VDC each.
- The B2+ (the splitter HV rail) has now "sunk" some to 215VDC.
- B+ has dropped as well, so this may represent an extra pull of current though the supply by the CCS.
- Sine waves look much better at equal resistances.
Do these numbers sound better to you?
Thanks,
Jeff
Hi, Tom,You wrote:
"Also there would be an advantage to running the phase inverter at the same current as the gain stage."1. Is the advantage you wrote of contained in your following statement ("net current drawn by the two stages is constant") or is it something else?
2. Also, would the IRFBC20 function well at 7 or 8 mA (the gain stage current)? It's running in the low 20s now.
Please also see my reply to Eli below, which is also addressed to you.
Thanks,
Jeff
Edits: 03/30/15
The FET will be just fine at the lower current. However, transconductance (g fs ) will decline. Win some, lose some. Compromises will forever be with us.
Eli D.
Two questions:
1. What is the advantage in having the currents through the voltage amplifier and the splitter be the same?
2. Given that the voltage across the MOSFET itself seemed to increase slightly (and current decrease) when I lowered the anode CCSs, how do I figure the change in resistor vales on the legs? With the drop in CCS current, the voltage drop across the leg resistors went from about 94 and 94 to 87 and 87. The B2+ voltage rail dropped some too (220 to 215). But that doesn't account for all of the shift in the leg measurements. So there seem to be some moving targets here.
Would anyone ever put a CCS on the splitter? That might force a matched current, but I have no idea if that would work or just be overbuilding.
Thanks for any insights.
--Jeff
Don't put a CCS on the splitter. Remember, FETs and pentodes are "constant" current devices. Do you want "head butting? I think not. Also, the volts needed to operate a CCS are not present.Go back and reread Tom's post. He gave you a rationale.
Altering the total resistance the FET's drain current flows through is how you control that current's value. DC coupling the FET to the 6922 anode turns it on "hard". The internal resistance of the turned an FET is so small that it can be ignored.
Eli D.
Edits: 04/01/15
Okay, then, 6.5mA across 215 VDC B2+ gives a total of about 33 KOhms, or 16.5K per splitter leg. The1.5 KOhm pot can still have some effect on the total for fine tuning. Maybe 16.5 KOhm fixed on the drain leg and 16 KOhm fixed plus the pot on the source leg for starters. Tweak the leg values on the bench using the actual B2+ rail voltage to attain 6.5 mA, matching the CCS current.
Does this seem good?
Thanks,
Jeff
Tom,The FET source resistance value is adjustable. That allows for "exact" mirroring of the 2 phases. Sometimes, a little asymmetry sounds better than "dead nuts". A lot of folks like a bit of 2nd order HD. The builder/owner gets to chose. :> D
"A little lower on the 6DJ8 plate would be better."
Do you think changing the CCS load to 7 mA. would be enough to get that tweak done?
"I think more decoupling would be better too, given the large swings in supply current in deep AB."
Diode check valves are already part of the decoupling scheme. Increasing the decoupling cap. value to say 22 muF. would provide extra energy storage for those times when the check valve operates. What say you about this?
Do you have any thoughts about lowering the end to end impedance of the O/P "iron" primary, in order to increase power O/P?
Eli D.
Edits: 03/30/15
Oops! The 5 muF film capacitors shown on the power supply schematics are the bypass caps for a larger capacitor that should also be shown. In my case, there was a F&T 32 + 32 muF dual electrolytic on hand, à la El Cheapo's B2+ supply. This larger cap was left out of the schematic!
The 22 muF film caps recommended above might be better than the F&T with film bypass. I have used metalized film Solen caps with success.
For motor run capacitor lovers, 22-24 mF "oilers" should work well, but space will probably be a factor in casing up this amp, given the CCSs, bias supply, adjustment pots, etc.
--Jeff
Eli (and Tom),Here's what I experienced on the scope in adjusting the AC balance pot. I was using dual inputs to show both sides of the splitter. The images are for 1 KHz at 2Vrms with 10 V per division I believe (please pardon the dust!):
1. At equal resistances, one side of the sinus form had a flat spot. That disappeared upon raising the overall resistance of the Source to Ground leg slightly.
2. Further increase in the lower leg "evened out" the top and bottom of the sinus form, so that their amplitudes matched (well, as much as I could get it to match).
So I thought that I was reducing second order HD by making the form symmetrical. Is that the correct idea? It did require slightly mismatching the resistances.There is still a minor mismatch in amplitude of top and bottom, but note that the scope had drifted slightly in vertical position, so it's a little better than the reticle grid shows.
What is causing the flat spot in the first image with equal resistances on both sides of the splitter? Is bias insufficient in the 6922? Peak voltage at 2Vrms probably exceeds my 6922 measured bias at idle. Or is this a cutoff at the MOSFET? Or a sag? If this flat spot symptom could be corrected at the cause, would I be able to match the resistances and not have much distortion?
Thanks,
Jeff
Edits: 03/30/15 03/30/15
Tom pointed out that we are on the razor's edge in the FET's voltage swing dept. Perhaps what you see is related to that.
While we are waiting for Mr. Bavis to post again, would it be possible to try the 6922 CCS at 7 or even 6.5 mA. Let's see what happens.
Eli D.
The results are positive. No flat top at 2Vrms input with slightly less current through the 6922. The splitter legs are symmetrical, too. See the first image below. This is 1KHz at 2Vrms,10V per division. The resistances are equal for each leg.
NOTE: With both sides equal in resistance, the waveforms for top and bottom legs looked good.One glitch was a Source to Ground leg measurement, which showed some oscillation. I have had this before, and it may be a probe issue, although it does show up on both channels at times, and with both of my probes. It sometimes disappears. It could be the lanky breadboard layout too. Any tips for checking into that would be helpful. One positive note is that it disappears at 1.86Vrms. So it would only appear at full volume with the volume pot turned all the way up.
See the next image and the "bubble" of possible oscillation. Anyone know what this represents and why it might be occasional?
Here is the same wave form from Source to Ground, with the input slightly reduced by 0.14V, or at 1.86V. (The blurred curve is my fault, not the trace):
Over all I am more than happy with the shape this is taking.
--Jeff
Edits: 03/31/15 03/31/15
A tot of rum for Mr. Bavis! The splitter was clipping, but not any more. :> ) Even with the MOSFET "concertina", things are darned tight. No way would a triode "concertina" have worked.
Eli D.
Post a Followup:
FAQ |
Post a Message! |
Forgot Password? |
|
||||||||||||||
|
This post is made possible by the generous support of people like you and our sponsors: