I recently built a new phono preamp using LC coupling. The first and second gain stages are choke-loaded, and the line output stage uses a plate-to-line transformer.
The preamp sounds really nice in spite of the measured frequency response. As to be expected, the measured low-end response drops rapidly below around 40 Hz, due to the limits of the inductance within the plate chokes. The measured high-end response drops off rapidly above around 21 Khz, due to the parasitic capacitance of the plate chokes.
The real issue I am seeing is a serious 6.5 dB suck-out in the response centered at around 7 Khz. The amplitude response is otherwise reasonably flat from 40 Hz to 20 Khz, but within the region of 6 to 8 Khz it just dies. The suck-out occurs at the same frequency region in both channels.
There is obviously a resonance occurring within the circuit between one of the plate chokes and a capacitive node within the circuit.
Can someone simulate this circuit in PSpice or a similar program, and help me to determine where this resonance may be originating? Any help would be very much appreciated.
Here is the circuit: http://2.bp.blogspot.com/-RsqSLKy3XB0/TeH5wYH002I/AAAAAAAAAOw/bDbU-LvzAIo/s1600/OctalPreMk2.jpg
Thanks!
The preamp sounds really nice in spite of the measured frequency response. As to be expected, the measured low-end response drops rapidly below around 40 Hz, due to the limits of the inductance within the plate chokes. The measured high-end response drops off rapidly above around 21 Khz, due to the parasitic capacitance of the plate chokes.
The real issue I am seeing is a serious 6.5 dB suck-out in the response centered at around 7 Khz. The amplitude response is otherwise reasonably flat from 40 Hz to 20 Khz, but within the region of 6 to 8 Khz it just dies. The suck-out occurs at the same frequency region in both channels.
There is obviously a resonance occurring within the circuit between one of the plate chokes and a capacitive node within the circuit.
Can someone simulate this circuit in PSpice or a similar program, and help me to determine where this resonance may be originating? Any help would be very much appreciated.
Here is the circuit: http://2.bp.blogspot.com/-RsqSLKy3XB0/TeH5wYH002I/AAAAAAAAAOw/bDbU-LvzAIo/s1600/OctalPreMk2.jpg
Thanks!
Last edited:
My guess is that it's due to one of the nF caps and the leakage inductance of the anode choke.
For example, 32 nF + 16 mH is a 7033 KHz resonance. 16 mH is a rather low leakage inductance for a 800H choke.
What's the point of inductive load for such tiny power levels?
For example, 32 nF + 16 mH is a 7033 KHz resonance. 16 mH is a rather low leakage inductance for a 800H choke.
What's the point of inductive load for such tiny power levels?
45, thanks very much for the analysis and the reply.
I really wanted to try a topology that was very different from the conventional resistance-loaded plate circuits I have used in the past for my phono preamps, and this circuit looked interesting.
If your analysis is correct, changing the value of that 32 NF cap could be a problem, as it is a part of the RIAA bass lift network. Perhaps to maintain the same two time constants (3.18 and 3180 uS) associated with this part of the circuit, I could reduce the value of this cap, and raise the value of the two resistors in this part of the circuit, to shift that 7.2 Khz resonant frequency higher, and above say 25 Khz. What are your thoughts on this?
I would really like to see a PSpice or similar analysis of the resonance I have described, before I start to tear into this circuit.
I really wanted to try a topology that was very different from the conventional resistance-loaded plate circuits I have used in the past for my phono preamps, and this circuit looked interesting.
If your analysis is correct, changing the value of that 32 NF cap could be a problem, as it is a part of the RIAA bass lift network. Perhaps to maintain the same two time constants (3.18 and 3180 uS) associated with this part of the circuit, I could reduce the value of this cap, and raise the value of the two resistors in this part of the circuit, to shift that 7.2 Khz resonant frequency higher, and above say 25 Khz. What are your thoughts on this?
I would really like to see a PSpice or similar analysis of the resonance I have described, before I start to tear into this circuit.
You could simply replace the anode choke with a CCS or even better with a gyrator. That would also be a sanity check about the origin of the resonance.
I could also take my capacitor decade box and parallel a cap across the existing 32 NF cap, and see if the resonant/suck-out frequency decreases. For example, paralleling another 32 NF across that cap should decrease the resonance to something on the order of 3.5 Khz.
You can see which stage causes the problem by inputting a small 7kHz sine and scoping the plate of the first tube as you vary the frequency a little. If the notch is not at the first stage's plate, move to the next stage's plate and repeat.
Hopefully it's not in the third (line) stage, which can be tested separately.
Last edited:
Rayma, that is true, but remember that each of the first two gain stages has split passive RIAA networks. The first stage has the high-cut network, and the second stage has the bass-boost network, so the frequency response anomalies would be only the deviation from the ideal for each network. If the RIAA networks were not included, it would be very easy to determine which of the two stages is creating the -6.5 dB suck-out I had described.
As such, I looked for that point of resonance at the output of the preamp.
I would be very much surprised if the resonance is occurring in the line output stage. I am of the opinion it is originating in the 6N7 second stage. I doubt if the 200 PF cap that is part of the RIAA high-cut network within the first stage is resonating with any value of inductance (leakage or otherwise) within the circuit, due to its very small value.
If the resonant circuit is being formed by the leakage inductance of the 800 hy plate choke that provides the plate load for the 6N7 second stage, and the 32 NF cap that is part of the 318 and 3180 uS RIAA bass boost poles, then the answer may be to scale the R and C values of that network to some resonant point beyond 20 Khz or so. For example, change the 32 NF to 8 NF, the 82K resistor to 320K, and the 10K resistor to 40K. The time constants for both poles would remain the same, but the resonant frequency would shift upwards by a factor of 4, to roughly 28 Khz, due to the capacitance reduction of the 32 NF cap. It can't do much damage to the sonics of the preamp at that frequency.
I'd like to keep the series resistance (currently 82K) as low as possible from a sonics standpoint, but I may not have much of choice here. A 6.5 dB suck-out within the passband of any audio amplifier is unacceptable.
As such, I looked for that point of resonance at the output of the preamp.
I would be very much surprised if the resonance is occurring in the line output stage. I am of the opinion it is originating in the 6N7 second stage. I doubt if the 200 PF cap that is part of the RIAA high-cut network within the first stage is resonating with any value of inductance (leakage or otherwise) within the circuit, due to its very small value.
If the resonant circuit is being formed by the leakage inductance of the 800 hy plate choke that provides the plate load for the 6N7 second stage, and the 32 NF cap that is part of the 318 and 3180 uS RIAA bass boost poles, then the answer may be to scale the R and C values of that network to some resonant point beyond 20 Khz or so. For example, change the 32 NF to 8 NF, the 82K resistor to 320K, and the 10K resistor to 40K. The time constants for both poles would remain the same, but the resonant frequency would shift upwards by a factor of 4, to roughly 28 Khz, due to the capacitance reduction of the 32 NF cap. It can't do much damage to the sonics of the preamp at that frequency.
I'd like to keep the series resistance (currently 82K) as low as possible from a sonics standpoint, but I may not have much of choice here. A 6.5 dB suck-out within the passband of any audio amplifier is unacceptable.
You could also localize the problem to the second stage by adding say 10nF
across the 32nF and see if the resonant frequency changes accordingly.
Rayma, Yes, I had made the same suggestion in my 3rd post above. I suggested that I could take my capacitor decade box, and parallel another value of capacitance across that 32 NF cap. Adding another 32 NF cap across it should reduce the resonant frequency to approximately one-half of the existing resonant frequency of around 7 Khz, if that is indeed the capacitive element that is part of the parasitic resonant network in this phono preamp.
There's an obvious bad-fit, and several unknowns.
But first: the chokes are not resonating with the EQ caps because there are 150k and 82k resistors in the way.
By inspection (thumbs, not SPICE), the 200p should be near 550pFd. (60k+150k)||500k is about 150k, 150k with 200pFd is 5.3KHz; 150k with 550pFd is 1.9KHz; we really want 2.2KHz.
I believe you picked 200pFd because your chokes are throwing much more top-loss than you think.
Take out the EQ caps, pad your signal generator way down, and measure the actual response of the amplifier as a "flat amp". Is it flat? I suspect it is falling above a few KHz, due indeed to choke capacitance. "Correcting" with a shy top-corner is likely to be messy.
That said, I (SPICE) can not find any hint of 6dB suck-out @ 7KHz. While I don't have your tubes in my idiot, the 12AX7 is son of 6SC7, and modest differences in the 2nd stage mostly change gain, EQ is swamped by resistance. I had to model choke C or it bumps-up in the top of the audio band. I may have missed some detail, but this is what the idiot throws up:
Physical layout. Chokes throw and catch signal. Mutual inductance; also their large area makes a lot of mutual capacitance. While choke-to-choke coupling alone would only reduce the highs, combined with the interstage EQ we have at least 2 poles and marginal stability.
Try clipping a ground lead to a box-cover and put it between the chokes. Any change of suck-out?
There is an answer along these lines. Put EQ caps directly on the chokes, swamp-out the chokes' own strays. You get your two poles and a zero is easy. A couple swing-and-miss comes out +/-1dB 20Hz-20KHz, +/-0.25dB 50Hz-20KHz.
The problem is that it depends directly on the actual tube and part parameters. Changing the 1st cathode resistor to 500 threw the overall gain down except a big bump at 50Hz. You would have to pick "probably good tubes", spend a lot of time trimming the four areas of response, then listen. If the tube used sounds ugly, change it and do a re-calibration. It would be wise to put a few months break-in on the proposed tubes before you begin. IME tubes drift little after the first 1,000 hours, much of that in the first 10-100 hours.
But first: the chokes are not resonating with the EQ caps because there are 150k and 82k resistors in the way.
By inspection (thumbs, not SPICE), the 200p should be near 550pFd. (60k+150k)||500k is about 150k, 150k with 200pFd is 5.3KHz; 150k with 550pFd is 1.9KHz; we really want 2.2KHz.
I believe you picked 200pFd because your chokes are throwing much more top-loss than you think.
Take out the EQ caps, pad your signal generator way down, and measure the actual response of the amplifier as a "flat amp". Is it flat? I suspect it is falling above a few KHz, due indeed to choke capacitance. "Correcting" with a shy top-corner is likely to be messy.
That said, I (SPICE) can not find any hint of 6dB suck-out @ 7KHz. While I don't have your tubes in my idiot, the 12AX7 is son of 6SC7, and modest differences in the 2nd stage mostly change gain, EQ is swamped by resistance. I had to model choke C or it bumps-up in the top of the audio band. I may have missed some detail, but this is what the idiot throws up:
Physical layout. Chokes throw and catch signal. Mutual inductance; also their large area makes a lot of mutual capacitance. While choke-to-choke coupling alone would only reduce the highs, combined with the interstage EQ we have at least 2 poles and marginal stability.
Try clipping a ground lead to a box-cover and put it between the chokes. Any change of suck-out?
There is an answer along these lines. Put EQ caps directly on the chokes, swamp-out the chokes' own strays. You get your two poles and a zero is easy. A couple swing-and-miss comes out +/-1dB 20Hz-20KHz, +/-0.25dB 50Hz-20KHz.
The problem is that it depends directly on the actual tube and part parameters. Changing the 1st cathode resistor to 500 threw the overall gain down except a big bump at 50Hz. You would have to pick "probably good tubes", spend a lot of time trimming the four areas of response, then listen. If the tube used sounds ugly, change it and do a re-calibration. It would be wise to put a few months break-in on the proposed tubes before you begin. IME tubes drift little after the first 1,000 hours, much of that in the first 10-100 hours.
PRR, thanks very much for the PSpice simulation, and for your comments.
I agree with your comment that the chokes are not resonating with the EQ caps, as the choke values are way too high (800 Hys) for this, for a suck-out at 7 Khz.
However, please refer to 45's comment about the possibility of leakage inductance within the plate choke to the 6N7. If the leakage inductance is indeed around 16 Mhy as he suggests, that could be responsible for a potential circuit resonance/suck-out at that node.
For what it is worth, I am aware of leakage inductance as it applies to transformers, but I have never heard it applied to a choke, aside from common-mode chokes, but we are not dealing with common-mode chokes in this circuit...……..
I would be surprised if the issue was due to mutual coupling of the chokes, as the signal levels are very small within the chokes, and any coupling flux would be very low of course.
As far as I am aware, PSpice cannot model parasitic inductances or capacitances within an audio circuit, and I suspect that may the issue here.
Your further comments and suggestions would be very much welcomed!
What happens to the modeled response if the Zobel network across the first choke is eliminated, and only the 2.1 NF cap across the second choke remains?
BTW, the coupling capacitors I used in this circuit are 0.1 uf, and not the 0.22 uf indicated in the schematic. I doubt if this would impact your model, but you never know...……..
What about ringing in the 2nd plate choke with the 2.1 NF cap across it? I doubt if ringing would occur in the first plate choke, due to the presence of the Zobel network.
Also, please note that the 6SC7 input tube is paralleled, to further reduce the plate resistance (by one-half), as is the 6N7 stage for the same reason. The Ip through the paralleled 6SC7 is 4.2 Ma, and the Ip through the paralleled 6N7 is 7.8 Ma.
As such, using an inverse RIAA network, the measured amplitude response is actually quite flat from around 50 Hz to 19 Khz, with the exception of the -6.5 dB suck-out centered at around 7 Khz.
I agree with your comment that the chokes are not resonating with the EQ caps, as the choke values are way too high (800 Hys) for this, for a suck-out at 7 Khz.
However, please refer to 45's comment about the possibility of leakage inductance within the plate choke to the 6N7. If the leakage inductance is indeed around 16 Mhy as he suggests, that could be responsible for a potential circuit resonance/suck-out at that node.
For what it is worth, I am aware of leakage inductance as it applies to transformers, but I have never heard it applied to a choke, aside from common-mode chokes, but we are not dealing with common-mode chokes in this circuit...……..
I would be surprised if the issue was due to mutual coupling of the chokes, as the signal levels are very small within the chokes, and any coupling flux would be very low of course.
As far as I am aware, PSpice cannot model parasitic inductances or capacitances within an audio circuit, and I suspect that may the issue here.
Your further comments and suggestions would be very much welcomed!
That's a very interesting model! I'd be thrilled if I could get a measured amplitude response of 20 to 20 Khz, +/- 0.5 dB.……….
What happens to the modeled response if the Zobel network across the first choke is eliminated, and only the 2.1 NF cap across the second choke remains?
BTW, the coupling capacitors I used in this circuit are 0.1 uf, and not the 0.22 uf indicated in the schematic. I doubt if this would impact your model, but you never know...……..
What about ringing in the 2nd plate choke with the 2.1 NF cap across it? I doubt if ringing would occur in the first plate choke, due to the presence of the Zobel network.
The capacitor RIAA EQ values were calculated on the basis of taking the plate resistance of each tube into account, as well as the Miller capacitance of the following stage. This is a must for any RIAA passive EQ circuit, when calculating the first-cut circuit constants.
Also, please note that the 6SC7 input tube is paralleled, to further reduce the plate resistance (by one-half), as is the 6N7 stage for the same reason. The Ip through the paralleled 6SC7 is 4.2 Ma, and the Ip through the paralleled 6N7 is 7.8 Ma.
As such, using an inverse RIAA network, the measured amplitude response is actually quite flat from around 50 Hz to 19 Khz, with the exception of the -6.5 dB suck-out centered at around 7 Khz.
That would be the textbook case with the leakage inductance Ls in series with L.
However a suck-out makes me think that here he has got a series LsC circuit in parallel with the load.
Leakage flux is just flux that doesn't link and it can do strange things.
It could also be the choke parasitic capacitance coupling with some parasitic inductance. I would replace the choke and see what happens.
You simply insert them as added components. SPICE has no way to know that they are "mandatory options". I added 220p (post #11, C17 C18) as a wild-guess of the stray C of a lump like I am guessing those chokes are; trimmed some because you reported the high-end roll-off.
IMHO, leakage inductance by the conventional definition is an oxymoron in a choke.
Yes, a wound structure can have part of its inductance resonate with part of its capacitance. We see this in output transformers as peak/dip, usually in the 30KHz-200KHz range. For it to happen 2 octaves lower seems odd. HowEVER, if you would just test the tubes and caps alone (without EQ), this would be clear.
I grant that my "~~550pFd" must be corrected for grid capacitance; obviously. Much depends on cathode bypass (shown grey on your plan??) and tube type. Wait, you really mean 6N7 from 1932?? (I had assumed some Russian type.) I had to squint German to find 1.5pF grid-plate. Gain like 20, we have 30pF; 60pF for doubled. Add a bit for strays. Still 200+60+20 is maybe half what I would expect. However that would cause bump-up above 2KHz, not the reported suckage.
Leakage inductance in this case might make sense being two separate coils connected in series on one core. I am guessing the 800 H choke is something like the LL1667/5mA. Anyway there is for sure leakage flux as in any core-based inductor.
One more thought is that maybe it's just the inductance of the choke resonating with a parasitic capacitance. At 7 KHz the inductance will not certainly be 800H because of the core losses. It will be much less than that. How much is hard to tell as one should know the specs of the core used. Anyway, just to figure out what sort of thing might be going on lets assume it is 20 H. In order to resonate at 7 KHz then it needs to form a series with some 26 pF capacitance. This sort of capacitance can come from everywhere, including wiring.....
One more thought is that maybe it's just the inductance of the choke resonating with a parasitic capacitance. At 7 KHz the inductance will not certainly be 800H because of the core losses. It will be much less than that. How much is hard to tell as one should know the specs of the core used. Anyway, just to figure out what sort of thing might be going on lets assume it is 20 H. In order to resonate at 7 KHz then it needs to form a series with some 26 pF capacitance. This sort of capacitance can come from everywhere, including wiring.....
Hi!
This schematic/circuit comes from Thomas Mayer. Maybe you could ask him advice about your issue?
He used to comment on this forum in the past.
Best regards,
M.
This schematic/circuit comes from Thomas Mayer. Maybe you could ask him advice about your issue?
He used to comment on this forum in the past.
Best regards,
M.
45, thanks once again for your reply and guidance. Much appreciated!
Yes, the plate chokes are indeed the Lundahl type LL1667, gapped for 800 hys of inductance at 5 Ma of current.
Later this week, I intend to start making some additional frequency response measurements with respect to the resonance issue at 7 Khz. I'm first going to go after that 32 NF cap in the bass-boost pole of the circuit. If the point of resonance changes with a change in the value of that cap, we'll know that is mostly likely where this issue is coming from.
Yes, the plate chokes are indeed the Lundahl type LL1667, gapped for 800 hys of inductance at 5 Ma of current.
Later this week, I intend to start making some additional frequency response measurements with respect to the resonance issue at 7 Khz. I'm first going to go after that 32 NF cap in the bass-boost pole of the circuit. If the point of resonance changes with a change in the value of that cap, we'll know that is mostly likely where this issue is coming from.
- Status
- Not open for further replies.