I bought a 20Watt Knight amplifier at a garage sale a few years ago. When the capacitors died, I decided to rework it a little bit for musical instrument use. I've attached my final schematic, wondering if I'm going to blow up the 7189s by running them too hard. I'm keeping the original power transformer, but going to silicon diode rectifiers instead of the tube rectifier used before, and doubling my supply capacitance.
The power transformer output is rated at 320-0-320, so with the silicon diodes, I'm expecting 450V at the plates. Is this going to damage them, or will they just break up a little? I'm also counting on a little bit of fuzz from transformer saturation.
I apologize for the exceedingly light lines in the PDF - That's just the way GSchem did it years ago.
The power transformer output is rated at 320-0-320, so with the silicon diodes, I'm expecting 450V at the plates. Is this going to damage them, or will they just break up a little? I'm also counting on a little bit of fuzz from transformer saturation.
I apologize for the exceedingly light lines in the PDF - That's just the way GSchem did it years ago.
Schematic perfectly readable in linux version of adobe.
Now to the meat of the matter, your B+ is going to be way too high for the 7189 in this application and the only viable replacement - Russian 6P14P-EV is not going to live long at these voltages either. These are still only 12W tubes (pd) and at >400V are really 8W tubes.. I would keep the B+ below 400V in a guitar amp application, and I would suggest a tube rectifier with relatively little capacitance. I don't think you want a really stiff supply for a guitar amplifier, and I think you will also discover quickly that hifi opts don't saturate the way you probably want them to.
Good US made 7189 actually are worth some money as well..
Now to the meat of the matter, your B+ is going to be way too high for the 7189 in this application and the only viable replacement - Russian 6P14P-EV is not going to live long at these voltages either. These are still only 12W tubes (pd) and at >400V are really 8W tubes.. I would keep the B+ below 400V in a guitar amp application, and I would suggest a tube rectifier with relatively little capacitance. I don't think you want a really stiff supply for a guitar amplifier, and I think you will also discover quickly that hifi opts don't saturate the way you probably want them to.
Good US made 7189 actually are worth some money as well..
I just realized you have an electronic switch in the power supply that removes the plate voltage without removing the screen voltage - the first time you do this would be the last time for that set of 7189 - so in addition to fixing the B+ issue you need to make sure the screens aren't powered when the plates aren't.
Also that FQN1N60 may blow up the instant it is switched to operate if the gate to source voltage rating is exceeded - if there is an internal zener in this part (I haven't checked) the bias will get pulled down to the internal zener voltage. In fact you don't need this mosfet at all - take another look..
There should be a gate stopper resistor right at each mosfet gate to prevent oscillation.
No selenium rectifier for the bias supply. Use a silicon diode.
Put 1M resistor and 0.01uF/1KV ceramic disk across each 1N5408 to make sure noise is suppressed during transition to reverse bias and PIV ratings are not exceeded.
I would not power the output stage from Q1 as this point is shared with all of the low level circuitry as well.
Regulating the screens probably is not a bad idea as you probably don't want a lot of distortion in the bass amp..
In terms of limiting the voltage to something more reasonable a 5AR4 would be a good choice and would still offer good regulation under load or you could figure out a simple voltage regulator for the plate supply as well. OR you could use a mosfet buffered voltage divider.
Take a look at your bias adjustment pots, there is no current path in the grid circuit, the bias voltage cannot be adjusted as the pots are incorrectly connected as rheostats - I assume the intention is to make the bias adjustable. You need at minimum a couple of additional resistors. Note that you should configure the pots so that if a wiper goes open your output tube(s) do not get destroyed.
A grid stopper resistor on the input tubes is a great idea particularly since this signal is coming in on a long instrument cable which may not be that well shielded.
I also think this amplifier is going to be too small for duty as anything but a practice bass amp, unless you have phenomenally efficient bass speakers. As a learning experience however I think it is a good idea.
Don't get too discouraged, this is a workable design with a little more effort.
Does this bass have active electronics in it?
Also that FQN1N60 may blow up the instant it is switched to operate if the gate to source voltage rating is exceeded - if there is an internal zener in this part (I haven't checked) the bias will get pulled down to the internal zener voltage. In fact you don't need this mosfet at all - take another look..
There should be a gate stopper resistor right at each mosfet gate to prevent oscillation.
No selenium rectifier for the bias supply. Use a silicon diode.
Put 1M resistor and 0.01uF/1KV ceramic disk across each 1N5408 to make sure noise is suppressed during transition to reverse bias and PIV ratings are not exceeded.
I would not power the output stage from Q1 as this point is shared with all of the low level circuitry as well.
Regulating the screens probably is not a bad idea as you probably don't want a lot of distortion in the bass amp..
In terms of limiting the voltage to something more reasonable a 5AR4 would be a good choice and would still offer good regulation under load or you could figure out a simple voltage regulator for the plate supply as well. OR you could use a mosfet buffered voltage divider.
Take a look at your bias adjustment pots, there is no current path in the grid circuit, the bias voltage cannot be adjusted as the pots are incorrectly connected as rheostats - I assume the intention is to make the bias adjustable. You need at minimum a couple of additional resistors. Note that you should configure the pots so that if a wiper goes open your output tube(s) do not get destroyed.
A grid stopper resistor on the input tubes is a great idea particularly since this signal is coming in on a long instrument cable which may not be that well shielded.
I also think this amplifier is going to be too small for duty as anything but a practice bass amp, unless you have phenomenally efficient bass speakers. As a learning experience however I think it is a good idea.
Don't get too discouraged, this is a workable design with a little more effort.
Does this bass have active electronics in it?
Last edited:
Good to know - Thanks. This is the first thing I've ever done with tubes, apart from repairing old radios by replacing identical parts
The reason for this mosfet is that the switch is only rated to 120vac, so I don't trust it with 450vdc, even if there's practically no current there. Are you afraid of it blowing up during the split second the switch is open? I guess I could put a large resistor on gate-source, and that should fix it.
Granted, I'm a microelectronics engineer, but I haven't seen this issue before. Can you either elaborate or point me to a source of info on this topic?
The bias supply is identical to what was in the original amp. I just figured I'd leave it.
The 10M resistors are there for PIV equalization. Should they be reduced? I'll get some ceramics for noise suppression.
In the original circuit, the output stage was powered directly from the main cap (40uF, 400V) driven by the tube rectifier. What Q1 is now powering was powered by a resistor straight off the main cap and a 10uF cap of its own. The rest of the supplies were powered by the same resistors I have, only with 10uF caps instead of 22uF.
I'll probably get a second mosfet to make a small series regulator for the output stage. I'll just add another 24V on top of the 380V reference for the driver-stage for it.
That's actually supposed to be a volume control. As the resistance goes down, the impedance seen by the 6AV6 goes down, reducing the gate drive to the final stage.
Um... What? Are you referring to that first tapped potentiometer? The purpose of that is to act as a volume control and input location selector, so you can use the first half of the 12AX7 or not.
See http://www.diyaudio.com/forums/clas...riate-if-you-have-external-feedback-loop.html. Unfortunately, I got a bit thread-jacked, though the other topic is sort-of on topic with my question.
Thank you very much. I can't wait to see what I can do with this.
No. The bass I'm building it for is a passive Gibson Grabber with humbuckers (so very low-level drive).
Last edited:
One very important thing to include now are screen stopper resistors. About 220 ohms for each grid should be fine. When you regulated the screen supply, the impedance of the circuit providing power to the screens was lowered significantly, making the screen susceptible to transient arcing -- particularly now that the amplifier is seeing band use. The screen stoppers protect against this.
Protect the screens, operate them within limits, and keep the plate dissipation safely within limits as well. Then, the actual plate B+ takes on much less importance. 7189 tubes can safely operate with 500 volts on the plate, and with long tube life as well, as long as the dissipation ratings are respected, and the screens are taken care of.
Dave
Protect the screens, operate them within limits, and keep the plate dissipation safely within limits as well. Then, the actual plate B+ takes on much less importance. 7189 tubes can safely operate with 500 volts on the plate, and with long tube life as well, as long as the dissipation ratings are respected, and the screens are taken care of.
Dave
Not saying its wrong, but why the 1K bypassed by 3.3uF?
If already standing atop 68K unbypassed, is this small DC
drop meaningful? Also the 48Hz corner is maybe too close
to your low E or D. If that drop IS meaningful, the phase
shift around -3db corner frequency might matter too?
I think you can ditch both those extra parts to no effect
whatsoever. If you really need a drop to bias your splitter,
an LED is another option without -3db phase shift.
If already standing atop 68K unbypassed, is this small DC
drop meaningful? Also the 48Hz corner is maybe too close
to your low E or D. If that drop IS meaningful, the phase
shift around -3db corner frequency might matter too?
I think you can ditch both those extra parts to no effect
whatsoever. If you really need a drop to bias your splitter,
an LED is another option without -3db phase shift.
Last edited:
"Not saying its wrong, but why the 1K bypassed by 3.3uF?"
The voltage drop IS meaningful, as it creates the bias for the tube, since the grid resistor is returned to the bottom of the 1K resistor in question. The output to the bottom output tube coupling cap should also be taken from the bottom of the 1K resistor as well to maintain a balanced drive to the output stage.
Dave
The voltage drop IS meaningful, as it creates the bias for the tube, since the grid resistor is returned to the bottom of the 1K resistor in question. The output to the bottom output tube coupling cap should also be taken from the bottom of the 1K resistor as well to maintain a balanced drive to the output stage.
Dave
Actually my concern is with when you apply a negative supply to the source which exceeds the gate to source voltage rating which is usually less than 20V. You may or may not get away with this.
These mosfets are operating in the linear region, not as saturated switches so their transconductance in conjunction with external circuit constants like wiring inductance and stray capacitance + gate capacitance can (and usually will) conspire to create an oscillator. (electrostatic coupling and the phase shift introduced by external wiring.) A small resistor greatly reduces the q of the undesirable tuned circuit that results and also introduces an HF pole that reduces the gain above the break point. Google is your friend, I unfortunately don't have the time to explain in greater detail.
Perhaps a mark of inexperience, but selenium rectifiers have a defined service life which has by now been exceeded. Good practice dictates replacement whether or not it works. A bunch of bad things can happen:
- The smoke released when they fail is very toxic
- The output tubes get destroyed through loss of bias which results excessive plate and screen current - you can't count on the fuse blowing in time to save them
- The power transformer gets fried in some instances
Yes they should be reduced - they are there to swamp the reverse leakage current of the diodes thereby equalizing the voltage drop across them. To assure this happens the swamping current should be at least 10X the worst case leakage current of your diodes. 1M resistors should assure this in most cases. These diodes have fairly abrupt cut off when they become reverse biased which generates some emi which can be suppressed by small caps across them - these caps may also provide a small amount of protection against transient induced over-voltage as well. (Distant lightning strikes and the like)
The source impedance of your follower is undoubtedly a lot lower than the circuitry it replaced so this is mostly a matter of semantics, but having gone to the trouble of providing such good isolation I would use it. Note that the use of 10uF capacitors will allow more LF coupling between the individual stages which could lead to instability or not..
Sounds like a good solution to me.
Not such a good idea, this will grossly overload the phase splitter and result in very high distortion - in addition as you change the resistance at this point the LF -3dB will move all over the place. (Halve the resistance and the -3dB point doubles.) Making the bias adjustable for balance and to tailor the sound (running hot or cold) and trade off tube life for distortion performance is a useful feature IMO.
Yes I know exactly what that is having designed MI electronics professionally for a living.. That said you still need to add the grid stoppers at the grids of those tubes - I can't think of a better way to bring EMI directly into the amp than to leave them out.
Depends on the design of the class D amplifier, but external loops are used in some instances.
You may need additional gain, but I am not sure about this as I have not figured out your actual gain distribution.
Last edited:
As-is, I'm expecting -15V on the bias supply, so that should be well-within the 20V range. I could add a 12V zener and gate resistor there if need be.
Only the regulator mosfets should be operating non-saturated, so I'll add resistors there. 10K or so reasonable?
Very good to know. Thanks. I guess a 1N4002 or so would be reasonable.
Ok. I figured the 10M resistors would swamp the reverse leakage, but on further examination of the datasheet, maybe not. I'll get 1M for them.
The original circuit had 10uF, but I plan to keep my 22s.
Alright. I guess I'll leave the bias resistors at 100K, but maybe put a pot on the bias voltage, and maybe make a voltage doubler off that transformer winding so I have a wider range to adjust with. Does that make any sense?
Ok. Being for a bass, would 47K make sense?
Note, the external loop is feeding current from the output back to the tube amp. The class D itself should have an internal loop to set voltage gain to 2.5.
If need be, I think gain can be boosted with an effect pedal of sorts. Otherwise, I could probably throw in a small jfet gain stage powered off the output bias winding.
Very good to know. Thanks.
Dare I try to use an MLX90614 to make an automatic bias adjuster for the output?
An interesting approach if you want the design exercise! For me though, I think I would just limit the bias range on the low voltage end of the adjustment range so that max quiescent dissipation could not be exceeded for the available B+.
Dave
The plate dissipation rating for a given service application of a power tube is a fixed figure. It doesn't change simply because plate voltage changes. Plate dissipation in watts is equal to plate volts X plate amps, so you can see that as voltage goes up, amps (mA in this case) must go down for a given dissipation rating to be maintained.
For the 6BQ5 class of tubes, dissipation is specified at 12 watts, but this is the figure as rated under the old Design Center rating system. Under the final rating system used (Design Maximum), this would convert to 14.4 watts. Still, it would be wise to stay under 13 watts, understanding that you are pushing for max safe operation at that level.
Dave
For the 6BQ5 class of tubes, dissipation is specified at 12 watts, but this is the figure as rated under the old Design Center rating system. Under the final rating system used (Design Maximum), this would convert to 14.4 watts. Still, it would be wise to stay under 13 watts, understanding that you are pushing for max safe operation at that level.
Dave
Was the amp originally designed for guitar or bass? Many guitar amps from the past had output transformers that were marginally big enough for guitar use (Most Fenders). Playing bass through one of these amps will push the transformer into saturation. Several things happen when the OPT saturates. Saturation creates a very unique sounding distortion which sounds really ugly in a HiFi amp but may be useful in MI applications. A side effect that could be fatal here is the fact that the tube current skyrockets when the transformer saturates. The tubes in this amp will be operating near maximum and saturation could blow them up. Back in the 60's there was a rumor that "playing bass through a Bandmaster will blow it up". One of my friends decided to play bass through his because it sounded cool. It took a while, but the fire gods danced all over the 6L6GC's one day and the amp died.
The real problem here is the fact that the term "7189" means a lot of different things. Even the original manufacturers differed a lot on the specs of the 7189 and the 7189A.
Sylvania shows the 7189 and the 7189A to be the same tubes with DIFFERENT PINOUTS. They are very stout tubes capable of 13.2 watt dissipation and 440 volts on the plate. The maximum screen voltage is 330 volts.
Tung Sol shows different ratings for the two versions, but I dont have them with me. I know one of them is 400 volts max on the plate and 300 on the screen. 12 watts dissipation.
GE shows the 7189 with a 300 volt screen rating and a 440 volt plate rating. The 7189A has a 440 volt plate rating and a 400 volt screen rating. The pinouts for the two tubes are different. 12 watt dissipation.
Clearly all of the old time 7189 type tubes are different, and using one in place of the other can lead to blown tubes. The situation gets worse because the old 7189's and 7189A's have become scarce and expensive. This had led to several people including some reputable importers to relable Russian 6P14P's as 7189A's and sell them. Avoid "7189A" tubes from Ebay. Using these will lead to fried parts quickly.
So what is a 7189, and which ones do you have? Can you get new ones for a reasonable price when yours finally die?
I offer this suggestion. I have been experimenting with EL84 - 6BQ5 - 7189 - 7189A's for use in my Simple P-P HiFi amps. Yes I have found several "7189A" tubes that will blow up when operated at power levels that shouldn't hurt ANY true 7189 tube. These are relabled Russian tubes marked as Ei or Tesla or several other names that I can't remember. Since the specs for all of the true 7189 types are different, I decided to set my amp up for the lowest common denominator (wimpiest specs) of 300 volts on the screens and 400 volts on the plates. I then tried dozens of different tubes to see what worked. Yes all of the old 7189's and 7189A's worked just fine. The surprise came when I discovered that many old 6BQ5's worked too including GE and Sylvania. Unfortunately these can be expensive too. But, wait theres more! I found that the current production JJ EL84's have no problem dealing with these voltages and they can be found for $10 to $12 each.
So, I would scare up a pair of JJ EL84's and experiment with them. That way you can be sure that you aren't going to blow up your 7189's. I have cranked the plate voltage on these to 425 volts and I think that they may be pushed a little more. As with the 7189's these tubes are very sensitive to screen grid voltage, and I wouldn't go beyond 320 volts, probably stay closer to 300 in a guitar amp. Carefully watch the guts inside the tube through the little hole in the plate. If the grid wires inside the tube are glowing at idle you are approaching meltdown. If the wires aren't glowing but start glowing when the amp is played loudly, you are still too close to meltdown. Ditto if the glow from within the tube gets brighter when you play it loud. Back off on the screen voltage until this goes away.
As mentioned before, most power tubes will fry quickly if the plate voltage is removed with the screen voltage present. The EL84 - 6BQ5 - 7189 type tubes are very sensitive to this and will blow up quickly. Ask me how I know this!
It's actually a PA amp. It has 4/8/16 ohm and 25/70v outputs on the OPT.
Were the amps not designed with enough headroom that the 6L6GCs could handle full plate voltage? I thought that these amps generally shorted the speaker output when nothing was plugged in, and that doing so would have a similar effect.
I'm using 7189s because I have them, the transformers, and a nice-enough chassis. If this entire concept works well, I'm thinking of trying to reproduce it with new parts (probably 6L6GCs and Hammond transformers or some such), though EL84s may be a better idea, as they're similar and easily sourced.
As I don't really want to spend too much money on this, would I miss much going with Sovtek EL84s, or are they no good? I guess I would only save $6 or so vs JJ, but I don't want to spend frivolously here.
I'm making a revised schematic now, which I'll probably post in a few hours when I finish it.
Done revising.
Notes, in no particular order:
Notes, in no particular order:
- Q2 has an internal Source-Gate zener, so no need for an external part.
- I don't necessarily need proper phantom power on the XLR, as I'm really intending just to drive a J201-based preamp built-in to the cable.
- R21 is there as a current monitor
- Considering replacing R14/C7 with a 2V LED
- Considering replacing R8 with a 2V LED
- RV1 is an internal trimmer
- RV2 is externally-accessible
- RV3 is externally-accessible
- As-designed, screen grids will get 375~380V. This can (should be?) lowered by selecting different parts for D5/D7.
Since we gonna talk splitter balance... Needs 220K in parallel with top 68K?
To load balance against 220K to virtual ground (back to V2's cathode).
Yeah bias... My bad not seein' that Concertina grid weren't direct couple.
If you' just thinking the grid might see too much current during warmup?
Could forward bias a sandy diode to cathode to keep the v3 grid from
taking full v2 plate current all by itself. May not matter, lessn 1mA thru
that 470K above V2's plate worst case anyway... If so, why not direct
couple, ditch 2 caps and a 1k?
If we just determined NO to direct couple, 3.3uF under the cathode is
still silly. You just do not need a cap here to smooth DC bias? The cap
from V2's plate will accumulate the DC bias... So it ends up being 69k,
big deal... Like I said, you also got 220K in parallel, so the top half will
have to be tweaked to match anyhow...
Since you are taking a feedback from the OPT, I'd think you want as
few capacitive shifts in that loop as you can get rid of... The caps to
the output grids are prolly the only ones you need inside the loop.
I ain't looked the revised schematic yet, so still commenting first one.
---
OK, I've looked now, comments still relevant.
Last edited:
Care to explain what a sandy diode is? Do you mean silicon?
The cap was there in the original amp, though after a little thought, I can see how it's unnecessary. I will probably bump C7 up to 0.1u to account for it, though. Most of that circuit is straight out of the original amp, including the bias mismatch on the phase splitter.
Any value in bumping up the output caps to reduce phase shift?
Also, on a different matter, can I think of a pentode or beam-tetrode as a cascode pair of JFETs?
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.