For pretty obvious reasons the output caps are not needed. There is a bit of a WW missprint in this but 2x15w = 30w versions have been built.
The later designs over the original are refinements really. The route to perfection on all is still mostly matched power transistors.
A lot of people seem to miss one of JLH's main points - measure how much power you actually need. 15w is mostly mentioned due to a rather famous 15w high quality valve amp. As he was using one of those previously he measured how much power he was actually using. When HiFi kicked off and every man and his dog wanted it power was a status symbol not what was actually needed.
I think it still is with some.
Last edited:
That was then. Loudspeaker efficiencies have generally reduced as manufacturers have reduced distortion, so you may well need slightly more power today than back in 1969.
One valid reason for omitting the output capacitor is capacitor distortion. I suspect that modern SMPS high capacitance high current units with low ESR and ESL would have less distortion than the older types, but I've not attempted to measure any. One reason they are omitted in higher power designs is that it is actually simpler to use a split rail - and cheaper as lower voltage capacitors can be used.
But there are other effects. Each capacitor which is added in an amplifier chain will give rise to potential overshoots in a response to a signal. For n capacitors there are n-1 swings past zero (-consider a single RC filter - that never crosses zero in response to a step transient, only decaying exponentially) and the size and frequency of those crossings may give rise to some effects on the bass notes.
OldDIY has also mentioned the effect of capacitor-speaker interactions. That would depend on the speaker behaviour at lower frequencies, so there are several opportunities for "capacitor effects" which actually aren't specifically distortion related.
This does not mean that capacitor outputs should be avoided. I'm sure one or two on here will have measured modern capacitors for distortion, but I 've not searched for any reports. What does seem to be important is that the PSU should not droop either, as Ian discusses, or at least not on the centre rail if a capacitor is used. Generally I have found the best results with stabilised PSU's when using capacitors. At least that will avoid power supply ripple effects from a varying DC supply voltage from creating unwanted "bass" signals.
One valid reason for omitting the output capacitor is capacitor distortion. I suspect that modern SMPS high capacitance high current units with low ESR and ESL would have less distortion than the older types, but I've not attempted to measure any. One reason they are omitted in higher power designs is that it is actually simpler to use a split rail - and cheaper as lower voltage capacitors can be used.
But there are other effects. Each capacitor which is added in an amplifier chain will give rise to potential overshoots in a response to a signal. For n capacitors there are n-1 swings past zero (-consider a single RC filter - that never crosses zero in response to a step transient, only decaying exponentially) and the size and frequency of those crossings may give rise to some effects on the bass notes.
OldDIY has also mentioned the effect of capacitor-speaker interactions. That would depend on the speaker behaviour at lower frequencies, so there are several opportunities for "capacitor effects" which actually aren't specifically distortion related.
This does not mean that capacitor outputs should be avoided. I'm sure one or two on here will have measured modern capacitors for distortion, but I 've not searched for any reports. What does seem to be important is that the PSU should not droop either, as Ian discusses, or at least not on the centre rail if a capacitor is used. Generally I have found the best results with stabilised PSU's when using capacitors. At least that will avoid power supply ripple effects from a varying DC supply voltage from creating unwanted "bass" signals.
It's not that difficult to build an AC millivolt meter. The divider on the input for simplicity could use 100k per volt or some other number. If 100k then the resistance on the 1k range would be 100k. 10v 1m etc. 100v 10m. This is a simple directly coupled designTest your thoughts: As a DIY, I assume you have a multimeter, specifically an older moving coil type with an AC voltmeter range or, if you have access, an oscilloscope to observe the supply voltage whilst the amplifier is producing "strong, deep bass" into a load. Use a large, fixed 4-8 ohms resistor load of at least 10W rating because it will get hotImmersing it in clean water will help but be careful with electrical safety. One side (earth) of the load resistor should remain securely grounded.
http://electronic-projects.50webs.com/low-voltage-ac-voltmeter.htm
The advantage of that is that it can be calibrated with dc. It uses the meter resistance but a resistor can be added plus a meter.. Bandwidth can be 100'skhz. True RMS multimeters - take care. The bandwidth needs to be known and may just be 50 ~ 60Hz.
The sums for that type of op amp circuit are a little different to gain calcs. Still simple. Say 1k input resistance, With 1v going 1ma flows so the feedback has to provide 1ma to balance it out. The op amp gain takes care of the diode drops.
Just in case some don't know this. i suppose most do.Resistors for testing. I did use a 6ohm hollow ceramic tube coil resistor rated at lots of watts. Not sure where it is so bought 2 8ohm 200w resistors. They are metal cased so need heatsink but can take a fair amount of power without and are designed to get very hot. I used sourcingmap on amazon. Checking sizes against industrial sources they are actually larger and those sources provide data. I have found that Amazon seller tends to sell good stuff.
I think it's a rather over exaggerated subject. Take kef B139 and it's 87db spl. AR speakers said to be highly inefficient. Compared with post war stuff they were but I know for a fact they don't need loads of power to drive them yet my dad still bought high power amp. The assumtion was that the amp he used with his warfdales wasn't powerful enough.There is a good potted history of hifi here
http://ukhhsoc.torrens.org/other/Papers/David_Lord_June_2017/History_of_DIY_HiFi_in_UK.html
Anyway I agree with Hood - don't assume measure what you need to use.
Last edited:
You have to put things in context. A dual supply amplifier with decent PSRR will be largely unaffected by what the PSU capacitors are doing.
In a single rail supply with OP caps, the older designs running from unstabilised rails could see the centre rail moving up and down in sympathy with the PSU rail but at a low rate and out of phase as (we presume) it will have some filtering. If a single rail supply with OP cap is run from a stabilised PSU this problem is also significantly reduced.
...But in the JLH operating in Class A this should be reasonably well controlled too, though a filtered supply would still be worth it. If I were to build up a stereo I would be tempted to use choke filters. Pretty reliable and effective, if costly these days, but only a small choke would work well with modern large capacitors.
.
What centre rail? They are generally AC coupled at both ends as there is no centre rail. Dual supplies have a centre rail but supply ripple will have much the same effect as on single rail.
I have seen that sort of arrangement used on dml panels along with a subwoofer. Just a simple cross over but tied in with the cross over used on the subwoofer to provide a smooth freq response through the transition region which will be over a range of frequencies.
An easier answere could be do it at the imput and use separate amps. One for sub and 2 for for stereo and the low freq range of that ideally needs to be as low as possible.
You can test how components affect the sound, including power supply components such as diodes and capacitors of a dual-strand power supply, by connecting channel-separated, "double-mono" power supplies. I interpret already once ahead - Decades do not remain also you:
- Same components modulate the signal differently.
- Ergo: Components modulate the signal.
- Ergo: Components modulate the signal even if they are not in "signal path".
Further tests would show:
- Components modulate the signal in the same way, whether they are in the "signal path" or not.
- Components have individual sound characters.
Further tests:
- Material and structure and diameter and mass (and more?) of parts shape the sound character of components.
I talk about sound characters of parts most time. THESE we do hear much more than viewed thd, tmd... viewed sine or square waves.
;-)
Surely (don't call me Shirley!) the DC voltage across the output cap keeps it polarised so there won't be measurable* distortion from the AC superimposed?
Agreed that it should be a deal bigger than 2mF, not just for bandwidth but so there's no signal voltage across it at the frequencies we care about.
*...yes, I know... 😄
Agreed that it should be a deal bigger than 2mF, not just for bandwidth but so there's no signal voltage across it at the frequencies we care about.
*...yes, I know... 😄
The centre rail on a capacitor coupled amplifier is the one in the middle - the output rail, if you prefer.
DId you read what I wrote or just wanted to object?
An amplifier operating on dual rail supplies with good PSRR will be largely unaffected by the PSU voltage.
A single ended amplifier (AC coupled if you prefer) operating from an unregulated supply with a simple RC filter to set the centre rail (the rail in the middle) voltage will suffer the centre rail (the output rail) voltage moving up or down as the PSU voltage moves up or down, but out of phase and to a lesser extent based on the degree of filtering.
That is particularly true of an AB amplifier but the JLH class A should be less affected if the PSU delivers a more or less constant voltage thanks to the mostly constant current drain. Though both circuit types would benefit from regulated supplies.
I would not advise using small capacitors at the output to set the frequency response of an amplifier because that will force the amplifier to operate with a reactive load at lower frequencies. The output transistors will have to dissipate the power instead. Although the impedance will be increasing at lower frequencies, and the current demand will reduce, that depends on the music/signals applied. In a crossover, such capacitors are part of a network which aim to set the overall impedance more or less to the equivalent of a resistive load or close to it.
I would have to repeat my earlier advice to split the signals with a preamp because using a small capacitor on the input of a JLH will give rise to a high impedance which will increase noise at the lower frequencies particularly. That might not be an issue to you - easy to test whether a small capacitor increases the noise. It may also affect stability BTW.
The preamp won't have to be very complicated, perhaps even just a buffer so that the frequency components can be selected to work at a lower overall impedance than 50k or so.
I would have to repeat my earlier advice to split the signals with a preamp because using a small capacitor on the input of a JLH will give rise to a high impedance which will increase noise at the lower frequencies particularly. That might not be an issue to you - easy to test whether a small capacitor increases the noise. It may also affect stability BTW.
The preamp won't have to be very complicated, perhaps even just a buffer so that the frequency components can be selected to work at a lower overall impedance than 50k or so.
That's certainly true and in DIY style but I think it wise to consider the total cost and usefulness to someone who possibly already has, or has access to a cheap multimeter but not likely to have a suitable panel meter with some form of mounting and a drawer full of components to choose from. A cheap multimeter that will suffice can be cheaper than US$5 and has a lot more uses. Try and beat that with a simple panel meter calibrated only in one linear DC scale and a handful of components from your local (i.e. expensive) parts distributor.
The post (#9307) was some days ago, but It could need some clarification for the power needed to reproduce audio generally. JLH is only 10W in an 8R system so it is soon compromised by bass demands with typically low-sensitivity hi-fi speakers in medium to large rooms. Conversely, that makes it quite popular as a tweeter driver for electronic crossover speakers, where 10W of high frequencies can be plenty but matching to midrange and bass speakers might then require 30W and 100W power respectively.
Simply, the energy to shift the volume of air necessary for the same SPL at all audio frequencies increases as the frequency falls from 20kHz to 20Hz.
Correspondingly, when we look at the energy spectrum of well recorded songs, we find SPL typically falls from around -12dB at 50Hz to -55dB at 20 kHz.
Image credit to Solderdude (Netherlands)
Last edited:
Generally, I would say not, for three reasons.
1. High voltage transistors use lightly doped epitaxial collector regions which will give rise to moderately high quasi-saturation voltages. (Although modern audio devices share this concept, high voltage transistors for 500V+ push this further).
2. They typically have a safe operating area which has a steep reduction slope above the SOA voltage limit point. That is generally in the range of 20-30V.
3. They usually have low gain, like 10 for example.
The (possibly) higher quasi -saturation voltage can be ameliorated by operating the JLH at a higher than recommended voltage PSU to offset the value. This is only a problem near clipping, but may introduced premature clipping and distortion near clipping, or, in other words, not 10W of undistorted output at (say) 27V (the recommended voltage for 8 ohms).
The generally worse SOA lines would limit the devices in Class AB but in the JLH at typically 30V DC it might be possible to use the 2SC4237
The low gain would be a problem as the open loop gain will be low, distortion higher, and driver dissipation would be increased. That would be my overall expectation of the biggest problem.
Best advice - check the data sheet!!! Then see if it might be a good fit.
These are generally easy to obtain on line.