Don, isn't it that OPT reflected impedance and actual transformer ac primary impedance does not correspond?
The usual rule for spec'ing the low freq. end of an OT is the official rated Zpri = 2 pi f Lpri (f being the low end freq spec'd). The 10K OT has 1.4 times as many turns to double it's L versus the 5K OT as it should. Unfortunately, both OTs still fall short of this rule at the spec'd f = 40 Hz. More like f = 80 Hz. But as Michael said, you can always use a lower driving impedance than the spec'd Zpri (and Neg Fdbk too) to get better results, so this -can- fix the problem. But you paid for an OT that was supposed to meet the spec. Kinda like buying a used car with flat tires, sure, it -can- be fixed.... We have the technology, you pay for it.
The problem is that SE OTs are so severely constrained by the air gap and DC flux. P-P OT's, on the other hand, typically have 10 times the minimum L to meet the Lpri rule. This means that SE OTs also run into core saturation problems at their max power rating. The GXSE 15 at 8 Watts (40 Hz) is drawing 48 times the magnetizing current (and beginning to sat. current spike) of a CXPP 25 at 25 Watts (and at 20 Hz). Put 15 Watts into the GXSE 15 at 40 Hz and you will get square waves out or burned up tubes. The usual comments made about SE OTs having less crossover distortion is a sick joke. Try 10X to 48X. (applies thru the whole signal range, since the equiv. P-P OT typically has 10X the primary inductance)
The usual design rule of thumb for SE OTs is that they require a 3 to 4 X heavier core to meet the same minimum Lpri and power (still no reserve though) requirements of an equivalent power P-P OT. (would take a 40X bigger core to get the same reserve Lpri headroom as the P-P case) Note that the GXSE's are the same size as the equivalent power level (and freq.) GXPP's, clearly something is not going to be adding up correctly there.
The problem is that SE OTs are so severely constrained by the air gap and DC flux. P-P OT's, on the other hand, typically have 10 times the minimum L to meet the Lpri rule. This means that SE OTs also run into core saturation problems at their max power rating. The GXSE 15 at 8 Watts (40 Hz) is drawing 48 times the magnetizing current (and beginning to sat. current spike) of a CXPP 25 at 25 Watts (and at 20 Hz). Put 15 Watts into the GXSE 15 at 40 Hz and you will get square waves out or burned up tubes. The usual comments made about SE OTs having less crossover distortion is a sick joke. Try 10X to 48X. (applies thru the whole signal range, since the equiv. P-P OT typically has 10X the primary inductance)
The usual design rule of thumb for SE OTs is that they require a 3 to 4 X heavier core to meet the same minimum Lpri and power (still no reserve though) requirements of an equivalent power P-P OT. (would take a 40X bigger core to get the same reserve Lpri headroom as the P-P case) Note that the GXSE's are the same size as the equivalent power level (and freq.) GXPP's, clearly something is not going to be adding up correctly there.
Last edited:
First off, I always assumed the "G" in the part number meant "Guitar". Look at the freq. rang of these and it fits. They claim response down to 40Hz but I doubt it. But for Guitars that is OK because the lowest note "E" in standard running is about 80Hz. A guitar can only make sounds from about 80Hz up to maybe 10kHz.
If you are building a HiFi amp the "C" series is for you.
I strongly suspect the two listed transformers are in fact the same. The identical weight says a lot. How could one have more turns and still the same weight. Smaller diaper wire. If so then how can the power handing be the same? My goes is the 10K transformer works to spec and in the 5K application it is better than spec'd.
The 10K does indeed have smaller wire and 1.4X the turns. It ends up with 2X the DC wire resistance from the smaller and longer wire (I measured this and also the # of turns). Since the AC current draw at 10K is 1/2 that at 5K Ohms, that is OK audio power wise. Rating stays the same.
The problem with using the 10K for the 5K application is two fold. First the doubling of wire resistance is causing some AC losses as expected. The more serious problem is the DC current rating is only 0.7 that of the 5K OT due to having 1.4X as many turns. (ampere turns being the criteria for saturation). (both OTs have the same gap spacer thickness)
I was able to use a 10K version in the 5K application by adding an additional DC buck winding onto the core in the remaining core window space. This gets DC bucking current put through it from a LV CCS to cancel the sat. effect of the DC primary current. You can either use just enough bucking current to bring it back in spec, or go whole hog and cancel it out altogether (which I did, and then removed the air gap spacer to get 4X increased primary inductance, a properly interleave stacked P-P core will get more like 10X).
Or you could just buy a P-P OT and put a HV CCS on one side, easier. Or then replace the HV CCS with a class A anti-triode circuit and get twice the SE emulation power out for the same B+ power input.
And much better low freq. response and damping factor too with the P-P OT.
The problem with using the 10K for the 5K application is two fold. First the doubling of wire resistance is causing some AC losses as expected. The more serious problem is the DC current rating is only 0.7 that of the 5K OT due to having 1.4X as many turns. (ampere turns being the criteria for saturation). (both OTs have the same gap spacer thickness)
I was able to use a 10K version in the 5K application by adding an additional DC buck winding onto the core in the remaining core window space. This gets DC bucking current put through it from a LV CCS to cancel the sat. effect of the DC primary current. You can either use just enough bucking current to bring it back in spec, or go whole hog and cancel it out altogether (which I did, and then removed the air gap spacer to get 4X increased primary inductance, a properly interleave stacked P-P core will get more like 10X).
Or you could just buy a P-P OT and put a HV CCS on one side, easier. Or then replace the HV CCS with a class A anti-triode circuit and get twice the SE emulation power out for the same B+ power input.
And much better low freq. response and damping factor too with the P-P OT.
Last edited:
Check back to my previous posts, I believe pages 2-5. Your comment has already been covered and an explanation, more than once, of why the trannies do not meet the spec posted.
Cheers.
Last edited:
yes of course, but in this thread a question was asked about the difference between 2 transformers....
They are not transformers until the leads are connected. There is no frequency response until they are connected.
In the case at hand we are considering the transformers as possibly interchangeable and connected to either a 16 ohm or an 8 ohm speaker.
There is no reflected impedance to the amplifier without a connected load.
The real world is even more complicated than that. Look at the impedance vs frequency curve of a typical transducer, the higher the frequency the higher the impedance. What does that do to your calculations of frequency response?
My take is that connected load is important and needs to be included in a complete transformer model.
DT
I wonder how the cost compares between the above and buying the correct SE transformer. At some point the bucking current CCS might actually be cost effective. Remember how speaker field coils where used for power supply chokes. They did double duty. I wonder if a DC bucking winding could be used that way. Then you'd have "free" current to run it and a free choke too.
Last edited:
Hey guys, accept this humble advice.
In the design of a SET is useful Eq(41)
N = [L x 10^8 i(DC)] / (S B)
Starting point is the lower frequency roll-off, then fix L.
With core specs you can fix S, the air gap can be calculated
μ=(l μ)/(l+le μ)
l is the magnetic path length in cm, le is the air gap in cm
After that, you can check it measuring L.
This saves time and money.
In the design of a SET is useful Eq(41)
N = [L x 10^8 i(DC)] / (S B)
Starting point is the lower frequency roll-off, then fix L.
With core specs you can fix S, the air gap can be calculated
μ=(l μ)/(l+le μ)
l is the magnetic path length in cm, le is the air gap in cm
After that, you can check it measuring L.
This saves time and money.
"Remember how speaker field coils where used for power supply chokes. They did double duty. I wonder if a DC bucking winding could be used that way. Then you'd have "free" current to run it and a free choke too. "
Interesting idea if the hum issue can be controlled. Maybe use it for the 2nd L in an LCLC filter.
Another way would be to use a speaker that is aligned at one end of it's excursion, instead of at the center. Then a very low voltage (one volt maybe) constant V power supply in series with the speaker and OT secondary would provide the bucking current through both devices to center the speaker and null the DC in the OT. The constany V supply does not need any V compliance in this case, so can be very efficient. (current limited by the DC resistance of the OT secondary and the speaker V coil) Unfortunately, the speaker design would have to be coordinated with the amplifier design. Maybe just put an air pressure (or vacuum) pre-charge port on the sealed speaker box to provide an adjustable speaker offset bias pressure. Not too practical I guess. But with a vacuum behind the speaker, the box dimensions would just drop out of the enclosure design picture. A big advantage for bass frequencies.
Interesting idea if the hum issue can be controlled. Maybe use it for the 2nd L in an LCLC filter.
Another way would be to use a speaker that is aligned at one end of it's excursion, instead of at the center. Then a very low voltage (one volt maybe) constant V power supply in series with the speaker and OT secondary would provide the bucking current through both devices to center the speaker and null the DC in the OT. The constany V supply does not need any V compliance in this case, so can be very efficient. (current limited by the DC resistance of the OT secondary and the speaker V coil) Unfortunately, the speaker design would have to be coordinated with the amplifier design. Maybe just put an air pressure (or vacuum) pre-charge port on the sealed speaker box to provide an adjustable speaker offset bias pressure. Not too practical I guess. But with a vacuum behind the speaker, the box dimensions would just drop out of the enclosure design picture. A big advantage for bass frequencies.
Last edited:
On second thought, using the SE OT bucking winding as an L in the power supply won't work, because it has to have the full audio signal across its terminals. That was the reason for using a CCS to power it. If one had two channels, with opposite signals on them, then one could sum them to zero voltage. Admittedly, the CCS solution is rather power inefficient, doubling power consumption with no increase in output. The anti-triode circuit however turns that around to get twice the power out.
I have seen a Westinghouse stereo amplifier that used a main channel amplifier and a small difference channel amplifier. The difference channel OT had two secondaries. Each one got summed with the main channel secondary but with opposite phase to produce the two stereo outputs. If one used two main channels instead, and a difference amplifier, the main channels could be in opposite phase so that the bucking windings could sum to zero signal voltage. Looks more like the hard way to do things though.
I have seen a Westinghouse stereo amplifier that used a main channel amplifier and a small difference channel amplifier. The difference channel OT had two secondaries. Each one got summed with the main channel secondary but with opposite phase to produce the two stereo outputs. If one used two main channels instead, and a difference amplifier, the main channels could be in opposite phase so that the bucking windings could sum to zero signal voltage. Looks more like the hard way to do things though.
Last edited:
I have seen a Westinghouse stereo amplifier that used a main channel amplifier and a small difference channel amplifier. The difference channel OT had two secondaries. Each one got summed with the main channel secondary but with opposite phase to produce the two stereo outputs. If one used two main channels instead, and a difference amplifier, the main channels could be in opposite phase so that the bucking windings could sum to zero signal voltage. Looks more like the hard way to do things though and not likely SE sounding.
Then there is that strange amplifier/OT that uses P-P around the outside of the OT E laminations and SE through the center (gapped) leg. Uses the same amplifier for both channels, using P-P difference mode and SE common mode gains. Could put the stereo difference channel throught the SE channel. Not likely to end up sounding like dual SE overall though.
Ugh, the center lamination leg must be gapped or else the two outer opposing DC fluxes will go through it, causing it to saturate.
OOPs, somehow duplicated the previous post too.
Then there is that strange amplifier/OT that uses P-P around the outside of the OT E laminations and SE through the center (gapped) leg. Uses the same amplifier for both channels, using P-P difference mode and SE common mode gains. Could put the stereo difference channel throught the SE channel. Not likely to end up sounding like dual SE overall though.
Ugh, the center lamination leg must be gapped or else the two outer opposing DC fluxes will go through it, causing it to saturate.
OOPs, somehow duplicated the previous post too.
Last edited:
Don, where dos this usual rule come from? I think one needs to look at an actual feasable design or range of designs and look at the result to achieve the correct tradeoff. I think the RDH4 method yields correct results when the effective plate resistance of an actual amp is taken into account. Why not use that method?
What I'm saying is not that you -can- fix the problem, but precisely that virtually any practical amplifier using this transformer will meet the -1 dB at 40 Hz spec. at the specified reflected Zpri because it will use a triode or feedback to achieve a reasonable damping factor, anything above DF=2.
So it's not exactly like buying a car with flat tires. I don't believe Edcor is misrepresenting the intended application or the spec. Sure It's not a high end transformer. They have the CX line for that which has better specs.
Here you hit on the head the real limitation to SE amp performance, and it's not the -1 dB small signal response. So in this sense Edcor could be more forthcoming with test data such as you just published
Thanks!
"Don, where dos this usual rule come from? I think one needs to look at an actual feasable design or range of designs and look at the result to achieve the correct tradeoff. I think the RDH4 method yields correct results when the effective plate resistance of an actual amp is taken into account. Why not use that method?"
Crowhurst mentions this on page 41 of "High Fidelity Circuit Design" as the point where the tubes begin to distort due to inductive current. He is talking about pentodes there, so the response drop is quoted as the 3 dB point (Rp = inf, so Rl = Xp).
Grossner (Transformers for Electronic Circuits 2nd ed.) has a table on page 283 that gives low frequency attenuation and phase shift for the general case (source impedance included). Maximum power transfer occurs at Rs = Rl. For Xp/R|| = 1.96 (approx. Xp = Rs = Rl) the dB drop is given as 1 dB with a 27 deg phase shift. Xp/R|| = 1.002 gives a 3 dB drop with a 45 deg. phase shift. Obviously one can fudge this dB factor by using very low Rp triodes, but then what universal triode do you use for specifying an OT? Zero Ohms will make anything work.
I quite often see Rl = Xp mentioned for simple models as the small signal LF cutoff (especially using transistors of course). I'm not sure one could call this an enforcable universal law as far as commercial products go though. But its the point where the tube delivers equal current to the core magnetizing effects as to the load, seems rather inefficient to go below that. The tube is getting stressed significantly also.
P-P OTs generally far exceed this rule with lots of inductance (like by 10X). Normally the limitation (power) for them is magn. saturation voltage at the low f end. The GXSE 15 is far into magn. saturation at 40 Hz and 15 Watts (with the DC current at max spec: 1/2 max flux), with the inductance completely collapsed due to the saturation, and the consequent magnetizing current spiking toward infinity. (If one had a zero Ohm Rp tube driving it, the copper would have melted and the insulation burned up.) (To continue the car analogy, sure it meets the 5 mph bumper requirement, see..., the occupants survived the crash!)
Crowhurst mentions this on page 41 of "High Fidelity Circuit Design" as the point where the tubes begin to distort due to inductive current. He is talking about pentodes there, so the response drop is quoted as the 3 dB point (Rp = inf, so Rl = Xp).
Grossner (Transformers for Electronic Circuits 2nd ed.) has a table on page 283 that gives low frequency attenuation and phase shift for the general case (source impedance included). Maximum power transfer occurs at Rs = Rl. For Xp/R|| = 1.96 (approx. Xp = Rs = Rl) the dB drop is given as 1 dB with a 27 deg phase shift. Xp/R|| = 1.002 gives a 3 dB drop with a 45 deg. phase shift. Obviously one can fudge this dB factor by using very low Rp triodes, but then what universal triode do you use for specifying an OT? Zero Ohms will make anything work.
I quite often see Rl = Xp mentioned for simple models as the small signal LF cutoff (especially using transistors of course). I'm not sure one could call this an enforcable universal law as far as commercial products go though. But its the point where the tube delivers equal current to the core magnetizing effects as to the load, seems rather inefficient to go below that. The tube is getting stressed significantly also.
P-P OTs generally far exceed this rule with lots of inductance (like by 10X). Normally the limitation (power) for them is magn. saturation voltage at the low f end. The GXSE 15 is far into magn. saturation at 40 Hz and 15 Watts (with the DC current at max spec: 1/2 max flux), with the inductance completely collapsed due to the saturation, and the consequent magnetizing current spiking toward infinity. (If one had a zero Ohm Rp tube driving it, the copper would have melted and the insulation burned up.) (To continue the car analogy, sure it meets the 5 mph bumper requirement, see..., the occupants survived the crash!)
Last edited:
Sorry to keep raining down bad news on these things, but another problem occurs when trying to use the GXSE 15-16-10K as a 5K:8 OT. The leakage inductance measures 84.6 mH on the primary side, which would be 9568 Ohms at 18 KHz. So the upper 3 dB band limit will drop to 9.4 KHz when used as a 5K:8 OT.
Just out of curiosity, I measured the leakage L with the gap spacer removed and pressure clamped. Surprisingly, leakage L does drop to 70.7 mH, but that means that my buck winding scheme for using it as a 5K:8 still only has an 11.2 KHz top end.
Just out of curiosity, I measured the leakage L with the gap spacer removed and pressure clamped. Surprisingly, leakage L does drop to 70.7 mH, but that means that my buck winding scheme for using it as a 5K:8 still only has an 11.2 KHz top end.
Last edited:
I am sensing perhaps too pleasure in the rain of “Bad News”.
It must be some sort of black art that any Single End Triode works at all without the wheels melting into a pile or smoking iron and copper on the table. If this is the premise, I am skeptical of any of the stated “Bad News”.
Due to the lack of reports of melted iron and copper there must be an alternate truth.
For grins, what is the impedance of that nominal 8 ohm tweeter at 18K Hz?
DT
That leakage inductance is too high for any of the two transformers.
I would suspect manufacturing defects such as wrong interleaved winding technique.
Just out of curiosity, how do you get adequate air gap again?
"It must be some sort of black art that any Single End Triode works at all without the wheels melting into a pile or smoking iron and copper on the table. If this is the premise, I am skeptical of any of the stated “Bad News”."
That was a hypothetical case of a zero Ohm source driving the OT well into magnetic saturation. Something like plugging a 120 VAC xfmr into a 240 VAC socket.
-----------------------
"That leakage inductance is too high for any of the two transformers.
I would suspect manufacturing defects such as wrong interleaved winding technique."
Well no, 84.6 mH gives 9568 Ohms reactance at 18 KHz, which is just below the 10 K Ohm primary Z, which it acts in series with at the high freq. end to give about a 3 dB drop as expected. Its just not good for a 5K primary.
And these OTs don't have any interleave winding, so it would be hard to mess that up any further.
-------------------------
"Just out of curiosity, how do you get adequate air gap again?"
There is an epoxy saturated paper spacer that I just put back in the gap. Maybe a few mils thickness.
-------------------------
I have managed to get the laminations totally out of one of the GXSE's and separated them into individual lams. I then restacked the core as a 1 to 1 interleave so I could make some more measurements. Rather surprising results. The following all measured with a low level EXTECH L/C meter:
The leakage L went down to 61 mH (interleaved stack) from 70.7 mH (butt gapped) and 82 mH (spacer gapped). Surprising that better iron would lower the leakage L, but I guess the better the effective permeability the core is, the better the coupling.
The primary L went up to 80 H (interleaved stack) from 45 H (butt gapped) and 13.2 H (spacer gapped). Not as much as I expected, but helpful.
(note: the primary L's without the spacer, being metal core related will increase by about 1.5X by higher AC levels than the low level L/C meter readings indicate; the leakage L's, being air related, will not directly, see below though)
Well, 61 mH leakage gets me to about 13.2 KHz top end using the interleaved stack. Better but not good enough. I am wondering if the improvement in leakage will go further with a larger AC level (the 1.5 X thing) since the permeability of the iron increases with higher excitation. I'm guessing about 54 mH leakage with a 1.5 X perm. boost. That would get it up to about 14.9 KHz top end. I guess I could just measure the thing with a load and signal generator.
That was a hypothetical case of a zero Ohm source driving the OT well into magnetic saturation. Something like plugging a 120 VAC xfmr into a 240 VAC socket.
-----------------------
"That leakage inductance is too high for any of the two transformers.
I would suspect manufacturing defects such as wrong interleaved winding technique."
Well no, 84.6 mH gives 9568 Ohms reactance at 18 KHz, which is just below the 10 K Ohm primary Z, which it acts in series with at the high freq. end to give about a 3 dB drop as expected. Its just not good for a 5K primary.
And these OTs don't have any interleave winding, so it would be hard to mess that up any further.
-------------------------
"Just out of curiosity, how do you get adequate air gap again?"
There is an epoxy saturated paper spacer that I just put back in the gap. Maybe a few mils thickness.
-------------------------
I have managed to get the laminations totally out of one of the GXSE's and separated them into individual lams. I then restacked the core as a 1 to 1 interleave so I could make some more measurements. Rather surprising results. The following all measured with a low level EXTECH L/C meter:
The leakage L went down to 61 mH (interleaved stack) from 70.7 mH (butt gapped) and 82 mH (spacer gapped). Surprising that better iron would lower the leakage L, but I guess the better the effective permeability the core is, the better the coupling.
The primary L went up to 80 H (interleaved stack) from 45 H (butt gapped) and 13.2 H (spacer gapped). Not as much as I expected, but helpful.
(note: the primary L's without the spacer, being metal core related will increase by about 1.5X by higher AC levels than the low level L/C meter readings indicate; the leakage L's, being air related, will not directly, see below though)
Well, 61 mH leakage gets me to about 13.2 KHz top end using the interleaved stack. Better but not good enough. I am wondering if the improvement in leakage will go further with a larger AC level (the 1.5 X thing) since the permeability of the iron increases with higher excitation. I'm guessing about 54 mH leakage with a 1.5 X perm. boost. That would get it up to about 14.9 KHz top end. I guess I could just measure the thing with a load and signal generator.
Last edited:
The only bad news is only that you can't really use the 10K model at 5K. They are optimized for a particular reflected impedance and won't work well outside that design center.
At the end of the day with these $30 transformers you'll get something like 40Hz-18KHz small signal response and 50 Hz at "full" power. They are the economy models after all. Even that will not seriously impair the sound of most bookshelf speakers. It's about what was used in good quality console amps of the 60s.
You can spend more with Edcor and get wider response, and you can drop about $200 per channel for a Lundahl that is even better. And so on. SE amps are quite practical but it's important to respect the balance of cost and performance and the transformer design center tradeoffs. The cheaper transformers have a much narrower operating window and less tolerance for use outside their design center and outside the design assumptions.
Any of us could spend a few minutes with google to find out the typical response of a particular tweeter. Why don't you do it and post the results?
Last edited:
Leakage inductance is what is left over after the less than perfect coupling between the primary and secondary windings. LL testing is difficult on a well-equipped manufacturer’s bench. I am curious how this is done in the DIY world. How was the reported LL testing done, at a single test frequency and the frequency responses extrapolated by the Chrowhurst rule of thumb; 2*Pi*F*L?
The question regarding a tweeters impedance was rhetorical and to beg the question as to how the LL was tested. The transformer secondary leads are not just left dangling for the test.
For grins: http://0071c19.netsolhost.com/support/articles/55/Measuring Leakage Inductance (104-105).pdf
A typical tweeter has a nominal impedance somewhere near twice the resonate frequency. The impedance rises smoothly from nominal due to the inductance of the voice coil to perhaps 4 times nominal at 18K Hz.
Yes connect a real world tweeter and do a frequency response sweep!
Caution the magic smoke will escape the tweeter voice coil before the transformer!
DT
The question regarding a tweeters impedance was rhetorical and to beg the question as to how the LL was tested. The transformer secondary leads are not just left dangling for the test.
For grins: http://0071c19.netsolhost.com/support/articles/55/Measuring Leakage Inductance (104-105).pdf
A typical tweeter has a nominal impedance somewhere near twice the resonate frequency. The impedance rises smoothly from nominal due to the inductance of the voice coil to perhaps 4 times nominal at 18K Hz.
Yes connect a real world tweeter and do a frequency response sweep!
Caution the magic smoke will escape the tweeter voice coil before the transformer!
DT
- Status
- This old topic is closed. If you want to reopen this topic, contact a moderator using the "Report Post" button.