New 183 build

[ijedouglas]

I found this discussion of caps with "slow" and fast diode rectifiers: https://ampgarage.com/forum/viewtopic.p ... 70#p277970

Great find. I think I saw a similar discussion elsewhere but this lays it out clearly. Thank you

[bepone]

Maybe someone can help me out here?

put some load on this circuit. immitate the amp current (lets say 150mA-200mA), put 50uF after the rectifier, and then check what will be..if using fast rectifier no need snubber caps, dont worry about

[Stephen1966]

Thank you, bepone. Regarding the load... I used the 100k/2W resistor as seen in the photo. The decision to use this value was based on the power ratings of resistors I have lying around. And anyway, this test circuit is a full wave bridge rectifier, by necessity, as the transformer I am using for the experiment doesn't have a centre tap. The 183's rectifier with slugging caps on the four diodes is a type of full wave rectifier but consists of two half wave rectifiers in parallel: a two-phase full wave rectifier. Incidentally, MrD used the same type of rectifier in the 124 as well but with six diodes for added security. The power transformer used though (in both amps) has a CT and so allows for this type of circuit. So anyway, the load, as I see it, isn't so important, as long as the transformer sees some load! A 100k resistor should be able to handle 1mA of current and 0.1W of power whereas a 500R resistor (which would emulate 200mA of current) would require a power rating of 20W. I tested the circuit using 100V, and using 10V I could reduce the power to 2W with a 50R resistor and this would give me 200mA but, I wanted as clean a sine as possible and experience tells me the propensity for noise in the mains supply is much more pronounced at lower voltages. I also figured that a reservoir filter in the circuit would only add another layer of information to the resulting traces when really, all I am looking for are the reverse recovery transients when the diodes switch.

In all then, I could tinker with the experimental set up and get something like an in circuit/in amp test bed but as I'm really only interested in the moments the diodes switch on and off, and the effect the slugging caps have on this process, I kept it simple.

Thank you, cys. I agree, a great find... I didn't come across this when searching on the topic. Teemuk said:

For example, diode switching noise is dependent on transformer's leakage inductance: When diode becomes reverse biased it will still conduct for a brief period of reverse recovery time, during this time leakage inductance of transformer will store energy, which will subsequently be released in a quick transient when diode eventually turns non-conductive.

There are basically two ways to fight this: One is to use very fast diodes, like schottkys. Another one is to introduce parallel capacitance to diodes. These methods can not be used interchangeably, they depend on what type of diode is used. Fast diodes do not need caps, in fact, the caps would hinder their operation. Slow diodes (like silicon rectifiers) "need" caps. A parallel capacitor will actually slow down the diode even further (this is why you do not want to fit them to circuits with fast diodes) and while the added sluggishness will not prevent the aforementioned phenomenon it happens to decrease amplitude of the resulting transients.

This, goes at least some way to describing the process that occurs when adding slugging caps to fast recovery diodes. The "critical" factor here is clearly leakage inductance of the power transformer - an unknown at this point in time*, but even with the little reverb transformer I am using it can be seen in the "cliff-faced" peaks of the traces I recorded earlier. For this reason, Merlin points out, a more appropriate name for the phenomena is 'rectifier switching noise' because it takes into account the whole system at work; it isn't just a phenomenon of the diodes in isolation. Though strictly speaking, I guess we should also be calling the diodes, "rectifiers." Anyway - semantics! But if the leakage inductance is preventing the diode switching off, generating a flyback voltage this sounds to me, much like the effect the slugging caps would induce as well. I disagree that there are just 'basically two ways to fight this' though; because there is a third, which is to use a snubbing net. More of that, shortly. *(I have an LCR meter and may be able to get an indication of the leakage inductance under no load conditions. Watch this space!)

So the problem with caps on fast recovery diodes is that in the effort to maintain an equal share of voltage across the switching process, they slow the switching process way down. Like leakage inductance, they are part of the system whereby the harmonics, the Fourier components, of the mains supply are augmented as well. I realise the effects of slugging caps, of leakage inductance and the increase in Fourier components are all essentially different from one another but they seem to confer a similar set of problems on the audio system. The presence of high frequency Fourier components, as Merlin says, '... can couple into the audio circuit via stray capacitance, leading to irritating switching spikes appearing in the signal path.' The leakage inductance of the transformer prolongs the switching process, and slugging caps, would be only dissipating this energy out over a longer period/transposing these into lower (and arguably more audible) frequencies.

From a design perspective I don't think there isn't much we can do about the leakage inductance of the transformer that appears across its coils - as it's a property of its construction - and similarly, there isn't much we can do about the non-linear loads present in the Fourier components of the grid supply either (move to a different room perhaps ). The use of different types of diodes, slugging caps and snubbing nets are within our control though and might help to mitigate, if not entirely remove, the unpleasant ringing of a poorly regulated supply. I'm still figuring this out as I go along, but I think I see more clearly now that fast diodes in the rectifier circuit are not suited to slugging caps because of a "prolongation" of the already present Fourier components the caps would store and then release. A superfast 500ns (RGP10M) diode would be prevented from switching as fast as it would like and during that null phase where the AC voltage passes through 0V and there would be a residual charge in the cap that might in effect, amplify the transients upon more of the higher frequency harmonics, simply by remaining "on" longer than it wants or needs.

A comparison of the oscillograms in Merlin's book shows that the 1N4007s transients are much greater in amplitude than the UF4007 he used. It looks like all the transients peaks and troughs are still there, but as noted by teemuk (above) at a much lower amplitude. By way of reference, the reverse recovery times of those are:

• 1N4007 - 1500ns

• RGP10M - 500ns

• UF4007 - 75ns

I don't really know what I can expect to find by way of difference with the RGP10M as compared to the 1N4007, the statement that slugging caps on faster diodes 'would hinder their operation' doesn't mean they will when the RGP10M is only three times faster than the 1N4007. However, I think the phenomena is much more likely to be pronounced when UF4007s, which are twenty times faster, are used. None of this takes into account the stray capacitance that lead dress in the amp is sure to produce and is sure to affect the final result but at the component level of the rectifier circuit it's nice to have an understanding of the contributing factors introduced by different topologies.

Whereas slugging caps are not advised for fast diodes then, a snubbing network of a 10nF + 1k5 (in Merlin's example) between the HT rails seems suited to both fast and slow diodes. With 1N4007s the negative peak voltage is much reduced and when used with UF4007s it appears to have been removed entirely. The net would look something like this:

Rectifiers - FWB and Bipolar.jpg

The value of the cap is intended only to 'swamp' stray capacitance across the transformer secondary and junction capacitance of the diodes and a ceramic 10nF, Merlin suggests, is large enough for that. The only complication is in finding the optimum value of the resistor - between 1k and 4k7, 1/2W is suggested. The aim being to 'burn off the oscillatory energy (as heat) as quickly as possible' It should be possible to tune the resistor for optimum snubbing with the scope. 1k5 is shown to be effective though so this seems as good a place as any to start. The best resistors would be non-inductive/Carbon comp and the circuit board would need to be designed so that the snubbing net is physically as close as possible to the rectifying part to minimize any stray capacitance there. [Edit: a quick look around reveals that CC resistors between 1k - 5k are pretty thin on the ground but KOA PCF 1W series is quite well stocked in Mouser https://cz.mouser.com/c/passive-compone ... &instock=y these ceramic resistors are non-inductive and might also be suited for the dropping resistors in the power supply.]

Another design mod which might be worth considering here is to increase the over voltage capacity of the half-phases of the bipolar rectifier with 3 diodes in series on each leg of the HT, as was done in the 124. That will incur a slightly higher cost in terms of voltage drop but is practically meaningless when the rectifier is putting out close to 450Vdc

Some references used;

Fourier_Analysis_for_Harmonic_Signals_in_Electrica.pdf

snubber.pdf

[cys]

I think the following pdf is a fascinating analysis of transformer/rectifier/snubber interactions: http://www.hagtech.com/pdf/snubber.pdf

[cys]

I'm going to go out on a limb and suggest that the reason the ultrafast UF4007 diodes shouldn't have a slugging cap is not becasue their operation is hindered by such a cap, but is instead because UF4007's are "soft recovery" and have less switching noise as a result -- making a slugging cap unnecessary.

[Stephen1966]

I'm going to go out on a limb and suggest that the reason the ultrafast UF4007 diodes shouldn't have a slugging cap is not becasue their operation is hindered by such a cap, but is instead because UF4007's are "soft recovery" and have less switching noise as a result -- making a slugging cap unnecessary.

It could be. I haven't been able to find any documentation which states they are soft recovery but the recovery time is so fast that - I go out on a limb now - the lower amplitude transient they exhibit is limited by the smaller amount of space charge they can accumulate. Ultrafast is therefore, "more equivalent" to a slower diode with a more abrupt transient damped by a cap in parallel.

I would go out a little further, and suggest that the slugging cap approach chiefly affects the reverse recovery of the diode, almost only in relation to the diode itself. However, the snubbing circuit chiefly affects the leakage inductance of the transformer and so, can be used with any type of diode - standard 1n4007s, superfast and ultrafast diode rectifiers.

I'm considering another experimental approach and I think I am going to start a new thread over in the technical discussion part of the site about using the LCR meter to test the power transformer properties and record some data. I've already done it with the output transformer earlier in this thread but the utility of the LCR meter is now starting to grow as I pick up on new ways to collect useful data. I know others may find it helpful outside the narrow confines of this build.

I'm ordering some UF4007s in the meantime though so if anyone can offer any more insight into them, please do chime in.

[rccolgan]

I've searched a bit on the topic but couldn't find anything scientific about the relationship between rectifier diodes and "tone." I used UF4007 diodes in my JM Sig and I believe the radiation noise was lower than a 1N4007 being so close to the 1MA gain pot. Some have chimed in and said that the diode would change the tone & feel. I don't want to hijack the thread but I think it would be helpful to have a refresher article link (or spin-off on a separate discussion on how to measure scientifically or A/B diodes). I swapped the diodes out back and forth on the JM sig and couldn't tell a difference... I'm open minded about the difference

[cys]

The UF4007's have references in some data sheets to being soft recovery. However the ratio of tb/ta would put it on quantitative ground if it could be found.

Screen Shot 2023-01-09 at 3.45.09 PM.png

.
I can't post the full pdf of the reference as I'm getting it with restrictions from my institution, but below are two figures from it illustrating respectively "snappy" and "soft" recovery of a "FRD" (fast recovery diode). All fast recovery diodes are not soft. The paper also notes: "... a soft recovery diode has a higher loss than a snappy diode but requires very small or no snubbing circuitry."

Screen Shot 2023-01-09 at 3.44.31 PM.png

Screen Shot 2023-01-09 at 3.44.41 PM.png

see: Shammas, N; Chamund, D and Taylor, P. "Forward and Reverse Recovery Behavior of Diodes in Power Converter Applications," Proc 24th International Conf on Microelectronics.



Here is a paper with real world testing of diodes and the UF4004 did well:

Linear Audio - Soft Recovery Diodes Lower Transformer Ringing by 10-20X.pdf

.
Lastly, the "snubber.pdf" paper posted previously has covered what occurs when a slugging capacitor is introduced:

Screen Shot 2023-01-09 at 4.09.37 PM.png

[Stephen1966]

I've searched a bit on the topic but couldn't find anything scientific about the relationship between rectifier diodes and "tone." I used UF4007 diodes in my JM Sig and I believe the radiation noise was lower than a 1N4007 being so close to the 1MA gain pot. Some have chimed in and said that the diode would change the tone & feel. I don't want to hijack the thread but I think it would be helpful to have a refresher article link (or spin-off on a separate discussion on how to measure scientifically or A/B diodes). I swapped the diodes out back and forth on the JM sig and couldn't tell a difference... I'm open minded about the difference

Sounds good. Where would you start with the measurement parameters?

[Stephen1966]

The UF4007's have references in some data sheets to being soft recovery. However the ratio of tb/ta would put it on quantitative ground if it could be found.

Screen Shot 2023-01-09 at 3.45.09 PM.png


.
I can't post the full pdf of the reference as I'm getting it with restrictions from my institution, but below are two figures from it illustrating respectively "snappy" and "soft" recovery of a "FRD" (fast recovery diode). All fast recovery diodes are not soft. The paper also notes: "... a soft recovery diode has a higher loss than a snappy diode but requires very small or no snubbing circuitry."
Screen Shot 2023-01-09 at 3.44.31 PM.png
Screen Shot 2023-01-09 at 3.44.41 PM.png
see: Shammas, N; Chamund, D and Taylor, P. "Forward and Reverse Recovery Behavior of Diodes in Power Converter Applications," Proc 24th International Conf on Microelectronics.



Here is a paper with real world testing of diodes and the UF4004 did well:

Linear Audio - Soft Recovery Diodes Lower Transformer Ringing by 10-20X.pdf



.
Lastly, the "snubber.pdf" paper posted previously has covered what occurs when a slugging capacitor is introduced:
Screen Shot 2023-01-09 at 4.09.37 PM.png

Cool! Last thing first though... I think part of the reason Hagerman put that paper together was to challenge the idea that following conventional wisdom and using a cap alone (without a resistor) doesn't have the snubbing effect you might expect. The topology of the Spice diagram gives more information:

Figure 7.jpg

In this the cap is Cx.

I think it's worth digesting Haberman's paper as whole because in another trace you can see the effect in more context:

Figure 10.jpg

Looking at the time divisions here supports what Merlin said about the lower frequency of the transients. Hagerman, doesn't call the snubber with a cap only, a slugging cap, but its what our two traces describe: a net with an RC snubber, and a net with a slugging cap.

When you bring it together with Merlin's section on fast rectifiers I think it is easier to appreciate these differences of approach.

Switching Noise and Fast Rectifiers - Blencowe.pdf

[Stephen1966]

I particularly appreciate this paragraph from the 'Linear Audio' paper.

Intro - Linear Audio.jpg

[Cys, do you have another reference to the text, is it from a textbook or is it someone's paper? I'd like to get a copy.]

I like it because it is a much better description of the thinking my own thoughts were coalescing around. I totally agree, there are much more important things to discover and to be concerned about when it comes to the tone of the circuit. Do I think it affects tone? Maybe not in a way that makes any significant sense. The noise of the fundamental transients would be very low in amplitude, even though they fall in the audio range 120Hz/100Hz and the subsequent Fourier components become increasingly harder to hear. It really isn't a level of noise that is meaningful in itself. It is however, the high frequency capacitive coupling which I consider unnecessary - [edit: and possibly more problematic]. More important factors to the contribution of tone are likely to be the ESR of reservoir caps and noise properties of different types of dropping resistors all feeding into the signal path. I do think it is debatable whether there is any difference from a slugged or snubbed rectifier. Some people do seem to hear a difference here but it's just as likely some will hear no difference between the ubiquitous 1N4007 which we have been using with fine results, seemingly forever, and a fancy new ultrafast rectifier with swish snubbing and a very linear supply.

I do think, however, that when we put these amps together our assumptions of the supply coming from the rectifier are predicated on a "linear" model. Like a lot of very small effects, it is often simpler to just ignore them. This is instructive for the purists as well because if MrD didn't do it, then it's for sure he had his reasons. The last couple of sentences in that opening paragraph, say it all for me.

[cys]

That pdf was just something I ran into after some exhaustive searching. Some more sleuthing turned up that it was a part of the following "bookzine" available from Amazon: https://www.amazon.com/Linear-Audio-Vol ... 975&sr=8-1. A better description can be seen here; Blencowe has an article in the volume as well: https://linearaudio.net/volumes/2235

[Stephen1966]

That pdf was just something I ran into after some exhaustive searching. Some more sleuthing turned up that it was a part of the following "bookzine" available from Amazon: https://www.amazon.com/Linear-Audio-Vol ... 975&sr=8-1. A better description can be seen here; Blencowe has an article in the volume as well: https://linearaudio.net/volumes/2235

Thank you Cys, that's really helpful.

This discussion is starting to turn technical and armed with an LCR meter I'm starting to see applications beyond the rectifier circuit. I think it's a bit late to divest this thread of the rectifier design discussion, it can be a bit frustrating trying to pin down a coherent piece of useful advice when it is distributed piecemeal throughout the site but I've just created a new post over in the Technical pages for the LCR methodology and its attendant maths. The data I post there can be used here as well, but in condensed form so readers don't have to slog through the mathematical process we have used to reach those findings.

https://ampgarage.com/forum/viewtopic.p ... 01#p448001

As I wait now for the delivery of the UF4007s I just want to say a quick thank you for the Linear audio paper on the rectifier diodes testing, and to bepone for the succinct commentary on the kind of circuit to use, to run our own tests. I've been thinking about the latter and considering the different platforms they used as a test bed in both the Linear Audio and Hagerman papers and I'm rethinking my own now. I see there is going to be a better set of qualitative (if not empirically accurate) results if the test circuit is designed to maximise the transients produced in the diodes and for that, using the components I have on hand, I need to emulate the current the amp is going to draw. I'm not set up to run laboratory scale experiments, but with the blunt set of tools I have at my disposal I think we can achieve some strong indicators for a more practical application in the design.

[WhopperPlate]

Having my curiosity sparked by this topic , I found this old thread that brings up some first hand experience and food for thought .

https://ampgarage.com/forum/viewtopic.php?t=19513

[Stephen1966]

Having my curiosity sparked by this topic , I found this old thread that brings up some first hand experience and food for thought .

https://ampgarage.com/forum/viewtopic.php?t=19513

Thanks for that! This is one of the posts I was referring to earlier in this post - but had lost track of. It's good to bring this information together. The caps in parallel with diodes describes slugging caps in the two-phase full wave rectifier and slugging caps can go between the HT rails as well. The oscillation is what gets my interest as well and I want to be able to reproduce it to try and figure what's actually happening there. But its true, these ultrafast diodes are sometimes described in the datasheets as not needing a snubber. No explanation why though. I would like to understand the mechanism better.