Go Back   Home > Forums > >

Headphone Systems Everything to do with Headphones

Designing my headphone amp
Designing my headphone amp
Please consider donating to help us continue to serve you.

Ads on/off / Custom Title / More PMs / More album space / Advanced printing & mass image saving
Reply
 
Thread Tools
Old 15th September 2021, 07:44 AM   #181
NickKUK is offline NickKUK  United Kingdom
diyAudio Member
 
Join Date: Dec 2019
Ok, that was nasty. There wasn't a single component but the interaction between a number of components. I simplified the output to a single tube on one side.

Output at the load:
Screenshot 2021-09-15 at 07.34.47.png

Starting from the front of the amp:
1. The driver tube outputs I used a 730 and then a 31K load, this was interacting with the couping caps behind it (0.22+1u)
2. The coupling caps were interacting with the bias and grid leak for the circlotron and the voltage divider I was using to reduce input level into the ecc99s.
3. The output tube, although I thought was isolated all the way to ground with even a B+ isolation was still coupled and seeing the opposite phase's non-isolated phase shift.
4. And all that was lovely playing with the ecc99 grid capacitance.

So the outcome to fix this:
a) switch to the driver cathode biasing the tube cascode.
b) remove the coupling caps entirely from the driver-cascode
c) remove the cascode input level dividers and grid leak.

I've used the same before without issue so I suspect that for circlotrons the same approaches have side effects.

I suspect that Atmasphere had similar phase shift issues hence biasing via the drivers

Anyway.. today is going to be a busy day.. so probably a slow down of updates
__________________
V1 - low gain tube HPA OTL, fully balanced cascode 12BH7A, 12BH7A CCS, 12BH7A drivers and ECC99-BJT hybrid circlotron. (build)
V2 - pure digital tube HPA, RF-tube DSD512 modulation with mosfet output stage. (initial design)

Last edited by NickKUK; 15th September 2021 at 07:47 AM.
  Reply With Quote
Old 15th September 2021, 07:55 AM   #182
NickKUK is offline NickKUK  United Kingdom
diyAudio Member
 
Join Date: Dec 2019
The current schematic:
Screenshot 2021-09-15 at 07.52.25.png

I noted that the driver input voltage divider is adding a very very small bass phase increase, so I will also have a look at that later - hence the big capitals.

I'm happy that the back end is feeling better, the cascode is, in reality, probably creating too much gain ... but I have a couple of ideas on that which do not use voltage dividers near caps but possibly use some local negative feedback to reduce signal level.
__________________
V1 - low gain tube HPA OTL, fully balanced cascode 12BH7A, 12BH7A CCS, 12BH7A drivers and ECC99-BJT hybrid circlotron. (build)
V2 - pure digital tube HPA, RF-tube DSD512 modulation with mosfet output stage. (initial design)
  Reply With Quote
Old 15th September 2021, 09:38 AM   #183
NickKUK is offline NickKUK  United Kingdom
diyAudio Member
 
Join Date: Dec 2019
You can see once a 100K resistor is added to connect the feedback the small bass increase and phase shift is amplified:

Screenshot 2021-09-15 at 09.37.28.png

Now to run a sim for the FFT and check the harmonics..
__________________
V1 - low gain tube HPA OTL, fully balanced cascode 12BH7A, 12BH7A CCS, 12BH7A drivers and ECC99-BJT hybrid circlotron. (build)
V2 - pure digital tube HPA, RF-tube DSD512 modulation with mosfet output stage. (initial design)
  Reply With Quote
Old 16th September 2021, 11:46 AM   #184
NickKUK is offline NickKUK  United Kingdom
diyAudio Member
 
Join Date: Dec 2019
I wrote a lovely post only to be told my token had expired..

That will do - given an LTSpice sim is about as precise and accurate as a game of pin the donkey at a state level. It does give a general picture of things working or not which can curtail the obvious and the oblivious.

Screenshot 2021-09-16 at 10.40.36.png

It's currently configured for my 104dBV headphones and CD player which I've measured 1.5V peak on some CDs. Let's be clear - I've seen this push 24V and 450mA just as happily as the values below (I have a 24V TVS across the load - an idea stolen from UltimateX86's schematic).

Input -> Output
-10dB 447mV -> 290mV, 5.2mA
-3dB 1V -> 650mV, 11.1mA
1.5V -> 960mV, 17.5mA
+10dbV 3.16Vrms 4.5Vpeak -> 2.14V +3.6dB

Mark - thank you for the link and the graphs.

I feel like I've learnt a lot about tube amps, some solid state, and cascodes attempting this project.

At 6am I found these two interesting articles which got me thinking more and why people seem to have issue with cascodes - more importantly their misunderstanding of them. So perhaps these musings and learnings will help people interested:
* Nelson Pass's article: https://www.firstwatt.com/pdf/art_cas_amp.pdf
* Broskie's interesting adaptations: Retrograde Cascode

A Cascode:
* consists of a V-to-I converter then a V-to-I converter.
* use of constant votlage/ground on V-to-I means virtually no miller effect as there's a constant charge on the grid. So no capacitance issues means higher frequency response.
* linearises the tube's characteristics
* linear lines = less asymmetric distortion on the load line = less even harmonics.
* linear lines = more symmetrical distortion = more odd harmonics (but not much see next points)
* triode normal cascode has low power output so should be driving a high impedance load.
* cascode (unfolded) needs voltage headroom to convert I-to-V without clipping and adding odd harmonics or intermodulation.
* is a gain machine, with 17mu 12BH7As it's entirely possible to get 100Vpp output from under 2V for example. See my point about headroom.

The interesting piece from Broskie is essentially using the V-to-I from the bottom.. a cathode follower as a suggestion. Interesting but given I've not seen many like that.. I'm less inclined to waste time on that. However Nelson highlights the use of Dayton-Wright as a further improvement over the standard cascode that drops harmonics further. Interesting.. perhaps a change to play with in future.


So my musings this morning also covered the general structure of the amp. The use of a cascode, drivers and the output section plus the specifics of the tubes in use - the 12BH7A and the ecc99. The achilles heal of the ecc99 is the input voltage range, and to a lesser extent the the same issue occurs with the 12BH7A.
The 12BH7A can take 20Vpp but will be letting you know. The ecc99 will give you the middle finger so sticking with 10Vpp is probably recommended.
The cascode of the design can be made to provide anything from 2V to 100Vpp, with a driver requirement a good 2-5Vpeak input is useful to sit at -10V bias and have current available to drive with the BH7. However chaining a ecc99 with it's 10Vpp input is a bit like drinking a straw whilst pinching the straw. It doesn't make sense (being objective).
This is where the M60 would be driving the low-mu 6AS7 which would be very happy with 20-40Vpp or more input. So you can see the voltage gain through the amp. You could do exactly the same 6AS7 for headphones too as tube only but you would have higher Zout than this hybrid.

I've also learn a lot about cascode interactions with negative feedback and phase shift.. I can see why the M60 uses the driver output to bias the output stage due to the additional phase shift adding coupling caps would add.
I also see why the M60 uses 0.1uF coupling caps between the cascode and the driver (not just to decouple) but the value also introduces a rolloff under 10Hz so by the time it's below 1Hz it's well attenuated to prevent LF instability. Additionally the 4.73Kohm resistor on the input prevents HF noise from causing instability by rolling off - starting just before 10Khz. This makes the amp sound less forced at the top end but keeps a good airy nature.

The M60-inspired frontend is awesome and very flexible. Probably too awesome for a headphone amp (which is probably why they don't use it) and my backend is probably capable of simply taking a triode output and being just as good with all the H2 goodness you want.
However this is the achilles heel of running the same amp as a headphone and as a speaker amp:
* cascode + driver + tubes - be a swinger - free the voltage swing! The obvious point here is OTL or big output tubes with an OT.. those in a cyclotron could be run as SE..

* cascode + driver + hybrid backend with larger BJTs - this suffers from the ecc99's grid limitations that limit the voltage swing available without distortion.

* cascode + driver driving larger bjts - this would be an option, with a single driver being able to drive the larger base capacitance of the large BJTs. Simply parallel for additional wattage. This keeps with the philosophy of using the tube output to driver the speakers with the BJT assisting..

* cascade + driver + mosfet output stage - let it go.. simply drive the mosfet gate with the cascode or driver depending on the capacitance. High impedance input makes that possible.. this is just like UltimateX86's circlotron.

Given the size of the chassis, the power supply requirements and the additional supporting components real estate.. I'm tempted to keep it HP for now but test it out later with speakers..
__________________
V1 - low gain tube HPA OTL, fully balanced cascode 12BH7A, 12BH7A CCS, 12BH7A drivers and ECC99-BJT hybrid circlotron. (build)
V2 - pure digital tube HPA, RF-tube DSD512 modulation with mosfet output stage. (initial design)
  Reply With Quote
Old 17th September 2021, 03:52 AM   #185
NickKUK is offline NickKUK  United Kingdom
diyAudio Member
 
Join Date: Dec 2019
So I’ve been thinking about slew rates and rise/fall times.

A 20V 20KHz signal needs a minimum of about 5V/uS according to some, however a sine wave at zero crossing is near vertical thus I would expect far higher. Thus any caps you use, to my thinking, would need to charge and discharge (5RC etc) within that time or transitional distortion and general symetrical distortions would rise (ie odd harmonics).

WIMA publish their V/uS for their caps. The metalised poly caps are around 5V/uS vs the likes of the MKP10 at a good 500V/uS to the FKP1 with a blinding 11000V/uS. I have no idea what the nichicon Muse ES are however given the ESR measured, it’s capacitance (330uF in this case) it’s making me wonder about the V/uS and 5RC time. Probably glacial compared to the FKP but from listening tests parallel muses seem a little murky compared to the FKP (no surprise).
I’m actually glad the pair of decoupling caps were removed - I now have a pair of MKP10 and FKP1s from each channel free for recycling into the new backend..

However that learning is for another day.
__________________
V1 - low gain tube HPA OTL, fully balanced cascode 12BH7A, 12BH7A CCS, 12BH7A drivers and ECC99-BJT hybrid circlotron. (build)
V2 - pure digital tube HPA, RF-tube DSD512 modulation with mosfet output stage. (initial design)
  Reply With Quote
Old 17th September 2021, 03:50 PM   #186
Mark Johnson is offline Mark Johnson  United States
diyAudio Donor
diyAudio Member
 
Mark Johnson's Avatar
 
Join Date: May 2011
Location: Silicon Valley
Designing my headphone amp
Why not use good old mathematics? The "slew rate" of a given voltage waveform is its maximum dV/dt. Calculating that for a sine wave y(t) = A * sin(2*pi*f*t), we find
  • dy/dt = 2*pi*f * A * cos(2*pi*f*t)
dy/dt reaches its maximum when the cos() expression equals +1, thus
  • max dy/dt = slew rate = 2*pi*f * A
Working through some numerical examples:
for a sine wave with A = 5V and f = 20 kHz, slew rate = 6.28E+5 volts/second {0.628 volts per microsecond}

for a sine wave with A = 15V (max amplitude an opamp can produce when running from 18V supply rails) and f = 140 kHz (7th harmonic of 20k), slew rate = 1.3E+7 volts/second {13 volts per microsecond}
Some designers employ a rule of thumb which includes their own personal integer "N":
One of my design goals is to make my amplifier's slew rate be at least a factor of N times higher than the slew rate of a rail-to-rail 20 kHz sine wave.
Then they choose a value of N which gives them enough margin-of-safety to achieve peace of mind.
  Reply With Quote
Old 17th September 2021, 04:10 PM   #187
hpasternack is offline hpasternack
diyAudio Member
 
Join Date: Sep 2002
Location: New England
Quote:
A 20V 20KHz signal needs a minimum of about 5V/uS according to some however a sine wave at zero crossing is near vertical thus I would expect far higher.
3.55 V/uS for 20 VRMS.

Quote:
Thus any caps you use, to my thinking, would need to charge and discharge (5RC etc) within that time or transitional distortion and general symmetrical distortions would rise (ie odd harmonics).
Slewing is a nonlinearity designers need to watch out for, but it's not really a function of capacitor ESR.

Quote:
WIMA publish their V/uS for their caps. The metalised poly caps are around 5V/uS vs the likes of the MKP10...
According to the MKP10 data sheet, the worst-case dV/dt spec is 30 V/uS for the 22-47 uF 100V range. That corresponds to a charging current of 660-1,410 Amperes. You're unlikely to have a problem with this in any practical audio circuit design.
  Reply With Quote
Old 17th September 2021, 06:29 PM   #188
NickKUK is offline NickKUK  United Kingdom
diyAudio Member
 
Join Date: Dec 2019
I did look at the maths, although the 2pi*f rate gives a time (x axis) per second and the Vpeak gives the y axis, which in my mind is a triangle to the peak rather than the steepest differential of the sine curve. I’ll have to think on that more but I can see some validity for the N integer idea (like the rule of 10x sensing minimum for measurement at 1 unit precision).

I think a larger role is the speed of charge/discharge of devices connected. The second formulae was SR=Voltage gain*Isat/C from memory.
__________________
V1 - low gain tube HPA OTL, fully balanced cascode 12BH7A, 12BH7A CCS, 12BH7A drivers and ECC99-BJT hybrid circlotron. (build)
V2 - pure digital tube HPA, RF-tube DSD512 modulation with mosfet output stage. (initial design)
  Reply With Quote
Old 17th September 2021, 06:31 PM   #189
Bonsai is offline Bonsai  Europe
diyAudio Member
 
Bonsai's Avatar
 
Join Date: Jul 2003
Location: Europe
Designing my headphone amp
The other ‘general’ rule of thumb is 1V/us/Vopk~pk

So if the peak to peak output voltage of your amp is say 10 V, design for 10 V/ us, if it’s 140 V ok~pk (a big amp), design for 140V/us
__________________
Bonsai X-Altra MC/MM preamp http://hifisonix.com/x-altra-phono-eq-preamp/ Ground Loops http://hifisonix.com/ground-loops/
  Reply With Quote
Old 17th September 2021, 08:12 PM   #190
hpasternack is offline hpasternack
diyAudio Member
 
Join Date: Sep 2002
Location: New England
Quote:
Originally Posted by NickKUK View Post
I did look at the maths, although the 2pi*f rate gives a time (x axis) per second and the Vpeak gives the y axis, which in my mind is a triangle to the peak rather than the steepest differential of the sine curve. Ill have to think on that more but I can see some validity for the N integer idea (like the rule of 10x sensing minimum for measurement at 1 unit precision).

I think a larger role is the speed of charge/discharge of devices connected. The second formulae was SR=Voltage gain*Isat/C from memory.
The formula for the instantaneous rate of change with respect to time of a voltage

v(t) = A * sin(wt)

is

v'(t) = A * w * cos(wt)

This is all you need to know. There is no guessing, no mystery.

The dV/dt figures on the capacitor data sheet are not relevant to the slew rate of the amplifier you might design using the capacitors.
  Reply With Quote

Reply


Designing my headphone ampHide this!Advertise here!
Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off

Forum Jump

Similar Threads
Thread Thread Starter Forum Replies Last Post
I'm designing a TPA6120 based headphone amp, is there anyone can give me some advice shadownee Headphone Systems 13 31st August 2012 05:50 AM
C&C request: designing a headphone amp Thalyn Chip Amps 0 14th April 2012 01:12 PM
Designing a super low distortion headphone amp. Rainwulf Chip Amps 41 1st April 2011 01:24 AM
designing horn for compression driver & tweeters jayam000 Multi-Way 0 15th April 2009 10:26 AM
Designing for guys who like designing Sch3mat1c Tubes / Valves 14 31st August 2003 12:46 PM


New To Site? Need Help?

All times are GMT. The time now is 12:44 AM.


Search Engine Optimisation provided by DragonByte SEO (Pro) - vBulletin Mods & Addons Copyright © 2021 DragonByte Technologies Ltd.
Resources saved on this page: MySQL 13.64%
vBulletin Optimisation provided by vB Optimise (Pro) - vBulletin Mods & Addons Copyright © 2021 DragonByte Technologies Ltd.
Copyright ©1999-2021 diyAudio
Wiki