Home |
Search |
Today's Posts |
#1
|
|||
|
|||
Beating the Miller Effect
Lately I've been having fun finding small tube amps at estate sales
and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian |
#2
|
|||
|
|||
"Brian" wrote in message om... Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian Hi Brian - If i understand you correctly, it's not the Miller effect you're battlin' with - it's a simple case of an RC network, where the C (a couplin' cap or some such) satys the same, while the vol. pot it's feeding varies. It's one of the drawbacks of KISS - the volume also acts like a trle cut at lower-than max volumes... Hope this helps & I'm not totally off the mark. Schematics would help... -dim |
#3
|
|||
|
|||
In article , "Shiva"
wrote: Hi Brian - If i understand you correctly, it's not the Miller effect you're battlin' with - it's a simple case of an RC network, where the C (a couplin' cap or some such) satys the same, while the vol. pot it's feeding varies. It's one of the drawbacks of KISS - the volume also acts like a trle cut at lower-than max volumes... Hope this helps & I'm not totally off the mark. Schematics would help... Isn't the "C" in this case due to the Miller effect, or at least the real "C" gets greatly multiplied by the Miller effect? Not sure why you say it's not the Miller effect Brian is battling? Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
#4
|
|||
|
|||
Where are you putting the cathode-foller stage? Before/after the volume
control? Thanks "John Byrns" wrote in message ... In article , "Shiva" wrote: Hi Brian - If i understand you correctly, it's not the Miller effect you're battlin' with - it's a simple case of an RC network, where the C (a couplin' cap or some such) satys the same, while the vol. pot it's feeding varies. It's one of the drawbacks of KISS - the volume also acts like a trle cut at lower-than max volumes... Hope this helps & I'm not totally off the mark. Schematics would help... Isn't the "C" in this case due to the Miller effect, or at least the real "C" gets greatly multiplied by the Miller effect? Not sure why you say it's not the Miller effect Brian is battling? Regards, John Byrns Surf my web pages at, http://users.rcn.com/jbyrns/ |
#5
|
|||
|
|||
Put a cap between signal input and pot wiper. Ideally it should equal
the ratio of resistance with respect to the miller C, such that it forms a similar voltage divider to the pot, only of lower impedance at HF. Or use a 10 to 100k pot. Tim -- "That's for the courts to decide." - Homer Simpson Website @ http://webpages.charter.net/dawill/tmoranwms "Brian" wrote in message om... Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian |
#6
|
|||
|
|||
Brian wrote: Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian Use cheap speakers. ;-) LV |
#7
|
|||
|
|||
If i understand you correctly, it's not the Miller effect you're battlin'
with - it's a simple case of an RC network, where the C (a couplin' cap or The problem occurs when the wiper of a pot (500K to 1.3 meg) directly feeds a tride grid (12AU7, 6CG7, or 6SN7). I can see the rolloff when placing a 10X scope probe on the wiper. The worst rolloff occurs when the wiper is at the half-resistance point. It is not bad when near either end. By a process of elimination, I conclude that the rolloff is due to the Miller effect (there's nothing else around with nearly enough shunt capacitance). Brian |
#8
|
|||
|
|||
Not really, short of using a compensated stepped attenuator, as used in test
equipment. This problem is a side effect of using high plate circuit impedances to get lots of gain per stage.. This is one reason why many excellent preamps used lots of stages, at less gain per stage. Take a look at a McIntosh C-20, it even uses two volume stages per channel, partly to work around this type of problem! This is a very common problem, particularly in amps with lots of switching. All those shielded cables running to the front panel take their toll. Some very popular Heathkit products are among the worst offenders. The worst case I ever saw was a Heathkit AD-40 tape recorder. "Brian" wrote in message om... Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian |
#9
|
|||
|
|||
Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian Just another piece of information... It's common for the 'input' capacitance at the grid to form a low pass with the resistance of what ever the grid see to ground through the pot. It's known that as you increase the value at the wiper to ground (turning up the volume) that the low pass point shifts down, grid sees more resistance. A trick is to insert say a 100K-200K resistor between the wiper and grid to increase the resistance seen by the grid. This shifts the low pass point down. There is a little bit of signal dropped across the new resistance, but is minimal. |
#10
|
|||
|
|||
said above Isn't the "C" in this case due to the Miller effect, or at
least the real "C" gets greatly multiplied by the Miller effect? Not sure why you say it's not the Miller effect Brian is battling? That's what I thought but WTFDIK. The Miller capacitance is part of the input capacitance and "forms a lowpass filter with a corner frequency determined by the source resistance of the previous stage and the input capacitance." as described he http://www.aikenamps.com/MillerCapacitance.html Cheers Craig Ryder |
#11
|
|||
|
|||
"Gilbert Bates" wrote in message ... Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian Just another piece of information... It's common for the 'input' capacitance at the grid to form a low pass with the resistance of what ever the grid see to ground through the pot. Agreed, although there are multiple ways to see "to ground." It's known that as you increase the value at the wiper to ground (turning up the volume) that the low pass point shifts down, grid sees more resistance. Sort of / kind of / up to a point. Depends a lot on the impedance feeding the pot in relation to the resistance of the pot. A trick is to insert say a 100K-200K resistor between the wiper and grid to increase the resistance seen by the grid. Why? I thought that adding more resistance shifted the low pass DOWN. I want to shift it UP! This shifts the low pass point down. There is a little bit of signal dropped across the new resistance, but is minimal. Why should there be any drop if there is no grid current? Seems to me like a series resistor would just make the low-pass rolloff even lower! Call me- Confused in Connecticut. |
#12
|
|||
|
|||
Yes indeed. The OP is battling input capacitance; whether it is due to pure
stray/interelectrode capacitance or the Miller multiplication of this capacitance is a moot point. However, since this is (presumably) a stage with gain, there will be a great contribution from the Miller effect. "Craig" wrote in message om... said above Isn't the "C" in this case due to the Miller effect, or at least the real "C" gets greatly multiplied by the Miller effect? Not sure why you say it's not the Miller effect Brian is battling? That's what I thought but WTFDIK. The Miller capacitance is part of the input capacitance and "forms a lowpass filter with a corner frequency determined by the source resistance of the previous stage and the input capacitance." as described he http://www.aikenamps.com/MillerCapacitance.html Cheers Craig Ryder |
#13
|
|||
|
|||
Brian, The best way to beat this is to use a Kelvin-Varley divider. This can
be accomplished by using a dual linear 50 or 100 k pot. Wire them so that the tube and source see constant impedance. i.e.as one resistance is reducing the other is increasing and compensating. Carroll "Brian" wrote in message om... Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian |
#14
|
|||
|
|||
Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian A trick is to insert say a 100K-200K resistor between the wiper and grid to increase the resistance seen by the grid. Why? I thought that adding more resistance shifted the low pass DOWN. I want to shift it UP! Sorry, I should have read more into your words 'lowpass' and '7kHz'. What I stated does shift it down. I really haven't thought about the miller capacitance impacting circuit operation at such a low frequency. This shifts the low pass point down. There is a little bit of signal dropped across the new resistance, but is minimal. Why should there be any drop if there is no grid current? Seems to me like a series resistor would just make the low-pass rolloff even lower! I was talking AC response not DC. Call me- Confused in Connecticut. |
#15
|
|||
|
|||
Put a cap between signal input and pot wiper. Ideally it should equal
the ratio of resistance with respect to the miller C, such that it forms a similar voltage divider to the pot, only of lower impedance at HF. If I understand this right, it appears to compensate exactly at only one pot setting. Or use a 10 to 100k pot. This sounds good, but most of the problem pots are right at the amp input. What is the smallest input resistance I can get away with, assuming unspecified tube equipment drives the amp? Will just about anything use a cathode follower? Brian |
#16
|
|||
|
|||
The simplest ways to reduce Miller affect capacity is to either reduce
the resistance value of the volume control as is possible, go to a lower Cg-p tube such as 6BK7 instead of 6FQ7 (for example), or using feedback to the cathode of the stage after the volume control so that it runs at lower gain thus reduced miller capacity. Recording studio amplifiers used transformers to lower the Zout of a stage to ~600 ohms balanced, the attenuator maintains a constant impedance to the source and practically no treble loss occurs because of the low impedance. Jeff Goldsmith Brian wrote: Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian |
#17
|
|||
|
|||
Use cheap speakers. ;-)
LV Hey that's a very cheap answer !! ;-)) Ronald . |
#19
|
|||
|
|||
Brian wrote: If i understand you correctly, it's not the Miller effect you're battlin' with - it's a simple case of an RC network, where the C (a couplin' cap or The problem occurs when the wiper of a pot (500K to 1.3 meg) directly feeds a tride grid (12AU7, 6CG7, or 6SN7). I can see the rolloff when placing a 10X scope probe on the wiper. The worst rolloff occurs when the wiper is at the half-resistance point. It is not bad when near either end. By a process of elimination, I conclude that the rolloff is due to the Miller effect (there's nothing else around with nearly enough shunt capacitance). Brian Here's a trick worth giving a try; add a "speed-up" capacitor, for lack of a better descriptive term, between the top of the volume control and the wiper. Select it such that you get a flat response where the rolloff is the worst (ohmic center of pot's rotation). You're essentially curve-fitting at three points now, instead of just two. At the top end of the pot, input impedance is sufficiently low as to be ignored, and response is essentially flat. Same at the low end. And in the middle, it's been "tweaked." There will still be some variance in frequency response as you turn the pot, but it won't be nearly as pronounced as before. Cheers, Fred -- +--------------------------------------------+ | Music: http://www3.telus.net/dogstarmusic/ | | Projects, Vacuum Tubes & other stuff: | | http://www.dogstar.dantimax.dk | +--------------------------------------------+ |
#20
|
|||
|
|||
"Brian" wrote in message
om... If I understand this right, it appears to compensate exactly at only one pot setting. Yup. It won't be flat anywhere but at those three settings. I mean, you could rip a stepped attenuator out of an old scope and use that, but it'll be stepped... not continuous... Or you could use a variable cap instead, and get out the scope readjust it every single time you change the volume control, but somehow I get the feeling you'll settle for the easiest solution.... ;-) Or use a 10 to 100k pot. This sounds good, but most of the problem pots are right at the amp input. What is the smallest input resistance I can get away with, assuming unspecified tube equipment drives the amp? Will just about anything use a cathode follower? 220k. I've seen several preamps that showed a 12AX7 driving a line. Not a CF. :-/ I'm sure it can still handle 100k, which I use all the time. Of course, my signal sources are either SS or this preamp here which has CF outputs. Hopefully you'll have good equipment, in which case a 12AU7 or similar will be able to drive 10k no problem. Tim -- "That's for the courts to decide." - Homer Simpson Website @ http://webpages.charter.net/dawill/tmoranwms |
#21
|
|||
|
|||
Here's a trick worth giving a try; add a "speed-up" capacitor, for lack
of a better descriptive term, between the top of the volume control and the wiper. Select it such that you get a flat response where the rolloff is the worst (ohmic center of pot's rotation). Fred, this idea sounded good. I didn't have an amp handy on the bench so I modeled the circuit. But I found that when optimized for half scale, the capacitor does not compensate well above half scale. Worse, below half scale a very pronounced treble peak set in. I tried adjusting the compensating C value but I could not find anything satisfactory. I'll check the model once again, but it is so simple I don't think there are any errors. Brian |
#22
|
|||
|
|||
This is a very common problem, particularly in amps with lots of switching.
All those shielded cables running to the front panel take their toll. Some very popular Heathkit products are among the worst offenders. The worst case I ever saw was a Heathkit AD-40 tape recorder. For RF cables I use 30 pF/ft as an estimate for cable capacitance. Is there a similar rule-of-thumb for internal shielded audio wiring? Brian |
#23
|
|||
|
|||
"Brian" wrote in message om... This is a very common problem, particularly in amps with lots of switching. All those shielded cables running to the front panel take their toll. Some very popular Heathkit products are among the worst offenders. The worst case I ever saw was a Heathkit AD-40 tape recorder. For RF cables I use 30 pF/ft as an estimate for cable capacitance. Is there a similar rule-of-thumb for internal shielded audio wiring? ** Cheap, thin gauge shielded cable can be up to 200pF per ft. Time to get yourself a DMM with a capacitance range - very handy. ............ Phil |
#24
|
|||
|
|||
prick
"Phil Allison" wrote in message ... |
#25
|
|||
|
|||
** Cheap, thin gauge shielded cable can be up to 200pF per ft.
Great heads-up, Phil. I checked the thin 7" cable attached to the volume control wiper in one of my problem amps. This short piece measured 80 pF! I found some Belden 8254 (RG-62/U, 93 ohms) in my junk box. This stuff measured only 13.5 pF/ft. After replacing the thin cable with RG-62, I no longer needed the shunt resistors I had added across the 500k pot. I think I still have a Miller-effect problem with another amp that uses a 1.3 meg volume control with a nonshielded wire feeding a 6SN7 with its relatively high Cgp. But this cable capacitance problem was a real eye-opener. Brian |
#26
|
|||
|
|||
Brian wrote: Lately I've been having fun finding small tube amps at estate sales and restoring them. But in more than one case, I've found that the frequency response varies with the volume setting. This turns out to be due to the Miller effect, where the triode grid-plate capacitance gets multiplied by the tube gain and forms a lowpass filter with the high-resistance volume control feeding the grid. In one amp the rolloff was 3 dB down at just 7 kHz with the pot at midsetting. I've been able to improve the response by shunting the pot sections with resistors, or by feeding back voltage to the formerly grounded pot arm, which effectively lowers the pot resistance. But these methods never completely flatten the response, and they have side effects. Adding a cascode stage ought to fix the problem, but I really don't want to do any metalwork modifications. Converting to a pentode also should work, but for one reason or another this hasn't been feasible. Is there a simple way to beat the Miller effect? Brian We are left to assume you have a 1 meg volume pot feeding the grid of a 12AX7, which might have a typical gain of say 60. The output Z of the pot in mid position is about 250k, and the input C of the tube is Cin of the tube x gain = approx 100pF, and hence you have an RC filter with 250k and 100pF, which has a pole at 6.36 kHz. So to reduce the bandwidth limiting of the miller effect, you either have to reduce tube gain and or reduce the input R. It depends where the volume pot occurs, and if it is after a passive network for tone controls, then lowing the value of the volume pot will adversly affect the tone control operation. But perhaps a 500k pot might be sneaked in without ruining the circuit performance, and then series R would reduce to 125k. Using a 12AU7 as a gain tube with gain =12 might be OK, and Cin would be reduced 3 times, so the 125k and 33pF would give a pole at 38 kHz. Pasive tone control pots should never be more than 500k imho, and perhaps even a 250k gain pot could be used. Food for thought. Patrick Turner. |
#27
|
|||
|
|||
"Brian" Phil A wrote: ** Cheap, thin gauge shielded cable can be up to 200pF per ft. Great heads-up, Phil. I checked the thin 7" cable attached to the volume control wiper in one of my problem amps. This short piece measured 80 pF! I found some Belden 8254 (RG-62/U, 93 ohms) in my junk box. This stuff measured only 13.5 pF/ft. After replacing the thin cable with RG-62, I no longer needed the shunt resistors I had added across the 500k pot. ** Cool. I think I still have a Miller-effect problem with another amp that uses a 1.3 meg volume control with a nonshielded wire feeding a 6SN7 with its relatively high Cgp. But this cable capacitance problem was a real eye-opener. ** See if you can track down some Belden 83265 ( RG178B/U ) co-ax - this is 50 ohm, teflon insulated, silver plated copper wire and only 5/64 inch (1.9mm ) diameter ! Nice stuff. About 29 pF per foot. ............ Phil |
Reply |
Thread Tools | |
Display Modes | |
|
|
Similar Threads | ||||
Thread | Forum | |||
Kroomel, Like Arthur Miller, Is-Irrelevant-In The Long Run | Audio Opinions | |||
Some Recording Techniques | Pro Audio | |||
What is media player WOW effect? How's it work? | Tech | |||
Megatron (Transformers) effect? Which vocoder (VST/DX) and settings/preset? | Pro Audio |